Reviewer #1 (Public review):

Batra, Cabrera and Spence et al. present a model which integrates histone posttranslational modification (PTM) data across cell models to predict gene expression with the goal of using this model to better understand epigenetic editing. This gene expression prediction model approach is useful if a) it predicts gene expression in specific cell lines b) it predicts expression values rather than a rank or bin, c) if it helps us to better understand the biology of gene expression or d) it helps us to understand epigenome editing activity. Problematically for point a) and b) it is easier to directly measure gene expression than to measure multiple PTMs and so the real usefulness of this approach mostly relates to c) and d).

Other approaches have been published that use histone PTM to predict expression (e.g. PMID 27587684, 36588793). Is this model better in some way? No comparisons are made although a claim is made that direct comparisons are difficult. I appreciate that the authors have not used the histone PTM data to predict gene expression levels of an "average cell" but rather that they are predicting expression within specific cell types or for unseen cell types. Approaches that predict expression levels are much more useful whereas some previous approaches have only predicted expressed or not expressed or a rank order or bin-based ranking. The paper does not seem to have substantial novel insights into understanding the biology of gene expression.

The approach of using this model to predict epigenetic editor activity on transcription is interesting and to my knowledge novel although only examined in the context of a p300 editor. As the author point out the interpretation of the epigenetic editing data is convoluted by things like sgRNA activity scoring and to fully understand the results likely would require histone PTM profiling and maybe dCas9 ChIP-seq for each sgRNA which would be a substantial amount of work.

Furthermore from the model evaluation of H3K9me3 is seems the model is performing modestly for other forms of epigenetic or transcriptional editing- e.g. we know for the best studied transcriptional editor which is CRISPRi (dCas9-KRAB) that recruitment to a locus is associated with robust gene repression across the genome and is associated with H3K9me3 deposition by recruitment of KAP1/HP1/SETDB1 (PMID: 35688146, 31980609, 27980086, 26501517).

One concern overall with this approach is that dCas9-p300 has been observed to induce sgRNA independent off target H3K27Ac (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8349887/ see Figure S5D) which could convolute interpretation of this type of experiment for the model.

Reviewer #2 (Public review):

Summary:

The authors build a gene expression model based on histone post-translational modifications, and find that H3K27ac is correlated with gene expression. They proceed to perturb H3K27ac at 13 gene promoters in two cell types, and measure gene expression changes to test their model.

Strengths:

The combination of multiple methods to model expression, along with utilizing 6 histone datasets in 13 cell types allowed the authors to build a model that correlates between 0.7-0.79 with gene expression. They use dCas9-p300 fusions to perturb H3K27ac and monitor gene expression to test their model. Ranked correlations of the HEK293 data showed some support for the predictions after perturbation of H3K27ac.

Weaknesses:

The perturbation of 5 genes in K562 with perturb-seq data shows a modest correlation of ~0.5 and isn't included in the main figures. The authors are then left to speculate reasons why the outcome of epigenome editing doesn't fit their predictions, which highlights the limited value in the current version of this method.

As mentioned before, testing genes that were not expressed being most activated by dCas9-p300 weaken the correlations vs. looking at a broad range of different gene expression as the original model was trained on.<br /> If the authors want this method to be used to predict outcomes of epigenome editing, expanding to dCas9-KRAB and other CRISPRa methods (SAM and VPR) would be useful. Those datasets are published and could be analyzed for this manuscript.<br /> The authors don't compare their method to other prediction methods.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Batra, Cabrera, Spence et al. present a model which integrates histone posttranslational modification (PTM) data across cell models to predict gene expression with the goal of using this model to better understand epigenetic editing. This gene expression prediction model approach is useful if a) it predicts gene expression in specific cell lines b) it predicts expression values rather than a rank or bin, c) it helps us to better understand the biology of gene expression, or d) it helps us to understand epigenome editing activity. Problematically for points a) and b) it is easier to directly measure gene expression than to measure multiple PTMs and so the real usefulness of this approach mostly relates to c) and d).

We thank the reviewer for their comment and we agree that directly measuring gene expression (e.g., by performing RNA-seq) is easier than performing multiple PTMs in a new cell line. We designed our approach keeping in mind that the primary use case is to understand how epigenome editing would affect gene expression.

Other approaches have been published that use histone PTM to predict expression (e.g. 27587684, 36588793). Is this model better in some way? No comparisons are made. The paper does not seem to have substantial novel insights into understanding the biology of gene expression. The approach of using this model to predict epigenetic editor activity on transcription is interesting and to my knowledge novel but I doubt given the variability of the predictions (Figures 6 and S7&8) that many people will be interested in using this in a practical sense. As the authors point out, the interpretation of the epigenetic editing data is convoluted by things like sgRNA activity scoring and to fully understand the results likely would require histone PTM profiling and maybe dCas9 ChIP-seq for each sgRNA which would be a substantial amount of work.

We thank the reviewer for this insightful comment. We have included citations for a series of papers (PMIDs: 27587684, 30147283, 36588793) that performed gene expression prediction using histone PTM data. However, each of these methods perform classification of gene expression as opposed to predicting the actual gene expression value via regression. Additionally, the referenced studies all work with Roadmap Epigenomics read depth data as opposed to p-values obtained from the ENCODE pipelines, making it difficult to make direct comparisons.

We outline in the Discussion section that by creating a comprehensive dataset of epigenome editing outcomes, which include quantification of histone PTMs before and after in situ perturbations, will improve our understanding of the effects of dCas9-p300 on gene expression and assist in the design of gRNAs for achieving fine-tuned control over gene expression levels. 

Furthermore from the model evaluation of H3K9me3 it seems the model is not performing well for epigenetic or transcriptional editing- e.g. we know for the best studied transcriptional editor which is CRISPRi (dCas9-KRAB) that recruitment to a locus is associated with robust gene repression across the genome and is associated with H3K9me3 deposition by recruitment of KAP1/HP1/SETDB1 (PMID: 35688146, 31980609, 27980086, 26501517). However, it seems from Figures 2&4 that the model wouldn't be able to evaluate or predict this.

We thank the reviewer for their comment. We have included a supplementary figure, Figure 4 – figure supplement 1, that quantifies how sensitive the trained gene expression model is to perturbations in H3K9me3. Indeed our data suggests that the model predictions are sensitive to perturbations in H3K9me3. For instance, there is a clear decrease and a gradual increase as the position where the perturbation is performed moves from upstream to downstream of the TSS. Additionally, the magnitude of the predicted fold-change is a function of how much the H3K9me3 is perturbed and hence the magnitude of change would be even higher if the perturbation magnitude is increased. However, this precise magnitude is hard to estimate In the absence of experimental perturbation data for H3K9me3.

The model seems to predict gene expression for endogenous genes quite well although the authors sometimes use expression and sometimes use rank (e.g. Figure 6) - being clearer with how the model predicts expression rather than using rank or fold change would be very useful.

We thank the reviewer for this important suggestion. We have added text in the revised manuscript to clarify that the model predicts gene expression values, which can be interpreted as rank or fold change, depending on the use case.

One concern overall with this approach is that dCas9-p300 has been observed to induce sgRNA-independent off-target H3K27Ac (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8349887/ see Figure S5D) which could convolute interpretation of this type of experiment for the model.

This is an excellent point and indeed, we and others have observed that dCas9-p300 can result in off-target H3K27ac levels (both increased and suppressed) across the genome. However, p300 is one of the few known proteins that can catalyze H3K27ac in the human genome, and H3K27ac remains a proxy for active genomic regulatory elements. Nevertheless, dCas9-p300 off target activity could certainly convolute our approach. We have included language to address this caveat in our discussion. Interestingly, even though dCas9-p300 (and other epigenome editing enzymes) can lead to off-target chromatin modifications, these effects often occur without coincident disruptions to the transcriptome. This suggests that many chromatin modifications, while “supportive” or “instructive” of/for transcription, may be insufficient (either alone or in the context of dCas9-based fusions) for transcriptional effects.

Figure 2

It seems this figure presents known rather than novel findings from the authors' description. Please comment on whether there are any new findings in this figure. Please comment on differences in patterns of repressive and activating histone PTMs between cell lines (e.g. H1-Esc H3K27me3 green 25-50% is more enriched than red 0-25%).

Thank you for pointing out this issue. We have revised the text in both the Results and Discussion sections to better articulate that the goal of this figure is to validate the hypothesis that there are consistent patterns of histone PTMs with respect to gene expression across different human cell types.

In Figure 2, which illustrates the raw histone marks data, the non-monotonic behavior of H3K27me3 in H1-hESC cells is indicative of a real biological phenomenon. This interpretation is supported by the relatively low Pearson correlation for the H3K27me3 mark observed in these cells, as documented in Figure 1b of another study: https://www.biorxiv.org/content/10.1101/2024.03.29.587323v1.

Figure 3&4

There are a number of approaches including DeepChrome and TransferChrome that predict endogenous gene expression from histone PTMs. I appreciate that the authors have not used the histone PTM data to predict gene expression levels of an "average cell" but rather that they are predicting expression within specific cell types or for unseen cell types. But from what is presented it isn't clear that the author's model is better or enabling beyond other approaches. The authors should show their model is better than other approaches or make clear why this is a significant advance that will be enabling for the field. For example is it that in this approach they are actually predicting expression levels whereas previous approaches have only predicted expressed or not expressed or a rank order or bin-based ranking?

We thank the reviewer for this comment. We have added text to clarify the difference between our approach and existing approaches. There are two key differences between our model and other approaches. First, the gene expression model that we have trained here predicts gene expression values instead of gene expression levels as either high or low. Second, we have trained our models on ENCODE p-value data instead of read depths obtained from the Roadmap Epigenomics Consortium.

Figure 5

From the methods, it seems gene activation is measured by qpcr in hek293 transfected with individual sgRNAs and dCas9-p300. The cells aren't selected or sorted before qPCR so how are we sure that some of the variability isn't due to transfection efficiency associated with variable DNA quality or with variable transfection efficiency?

This is a good question. All DNA preps were generated using high-quality reagents and consistent protocols. In addition, the only variable that changed with respect to transfection efficiency was the gRNA-encoding vector used in qPCR assays. We have added new data which demonstrates that transfection efficiency is shared across experiments (Figure 5 – figure supplement 1). We have also added additional experimental data as well as computational analysis analyzing a new dCas9-p300 based Perturb-seq dataset to the manuscript (Figure 6 – figure supplement 1), which use lentiviral transduction and RNA-seq as readouts and thus, are buffered against the variances mentioned by the Reviewer.

Figure 6

The use of rank in 6D and 6E is confusing. In 6D a higher rank is associated with higher expression while in 6E a higher rank seems to mean a lower fold change e.g. CYP17A1 has a low predicted fold-change rank and qPCR fold-change rank but in Figure 5 a very high qPCR fold change. Labeling this more clearly or explaining it in the text further would be useful.

We thank the reviewer for their suggestion. We have made relevant changes to the caption of Figure 6 to clarify this.

Reviewer #2 (Public Review):

Summary:

The authors build a gene expression model based on histone post-translational modifications and find that H3K27ac is correlated with gene expression. They proceed to perturb H3K27ac at 8 gene promoters, and measure gene expression changes to test their model.

Strengths:

The combination of multiple methods to model expression, along with utilizing 6 histone datasets in 13 cell types allowed the authors to build a model that correlates between 0.7-0.79 with gene expression. This group also utilized a tool they are experts in, dCas9-p300 fusions to perturb H3K27ac and monitor gene expression to test their model. Ranked correlations showed some support for the predictions after the perturbation of H3K27ac.

Weaknesses:

The perturbation of only 8 genes, and the only readout being qPCR-based gene expression, as opposed to including H3K27ac, weakened their validation of the computational model. Likewise, the use of six genes that were not expressed being most activated by dCas9-p300 might weaken the correlations vs. looking at a broad range of different gene expressions as the original model was trained on.

We thank the reviewer for their comments. We have added additional experimental data as well as computational analysis analyzing a new dCas9-p300 based Perturb-seq dataset to the manuscript. We observe that the models we have developed are able to predict the fold-change rank across genes reasonably well (Figure 6 – figure supplement 1), similar to what we observe in Figure 6E.

Reviewer #1 (Recommendations For The Authors):

The authors should comment on how their model is different from or better than other models that use histone PTM data to predict gene expression.

We thank the reviewer for this insightful suggestion. We have included citations for a series of papers (PMIDs: 27587684, 30147283, 36588793) that performed gene expression prediction using histone PTM data. However, each of these methods perform classification of gene expression as opposed to predicting the actual gene expression value via regression. Additionally, the referenced studies all work with Roadmap Epigenomics read depth data as opposed to p-values obtained from the ENCODE pipelines, making it difficult to make direct comparisons.

The authors need to make clear whether their model will apply to other common epigenetic or transcriptional editors such as CRISPRi/H3K9me3 which is widely used.

In this study, we focus on the histone changes induced by p300. However, future studies may use the framework described in our manuscript and apply it to other transcriptional editors as well.

The authors need to be clearer about where they are predicting expression and where they are using rank. Ideally, show both.

We thank the reviewer for this important suggestion. We have added text in the revised manuscript to clarify that the model predicts gene expression values, which can be interpreted as rank or fold change, depending on the use case.

The authors should ideally show a case where they use the model to make a prediction of genes that can and can not be activated by dCas9-p300 or other epigenetic editors and then prove this with experiments.

Thank you for the excellent suggestion. While it is indeed relevant, exploring this would extend beyond the scope of our current study. We consider it a valuable topic for future research.

Reviewer #2 (Recommendations For The Authors):

The y-axis in 5C needs to be labeled. The authors state it is "relative mRNA" but these numbers correlated with fold changes shown in Table S2.

We have clarified the definition of the Y-axis in the caption for Figure 5C.

This note says that its Julia Atomics Manifesto

These are last two point!

Some notes on blocking for different purposes in GEMM operations.

scron= minachting & gratitude = dankbaarheid

Zelfstigma is het proces waarbij een persoon negatieve stereotyperingen, vooroordelen, en stigma's uit de maatschappij over zichzelf internaliseert.

Vrouwen worden vaker gediagnosticeerd met angststoornissen dan mannen, wat deels kan worden verklaard door fluctuaties in oestrogeen tijdens hun levensloop. Bijvoorbeeld, in periodes met lage oestrogeenspiegels (zoals tijdens de menopauze) kan extinctie van angst minder effectief zijn, wat bijdraagt aan een verhoogd risico op angststoornissen over tijd.

Pre-ACQ (Pre-Acquisition): Hier gebeurt nog niets; het is de baseline zonder enige conditionering.

Uninstructed ACQ (Uninstructed Acquisition): De proefpersoon ziet een gezicht, direct gevolgd door een harde schreeuw. Er wordt niets uitgelegd, waardoor de angstreactie spontaan ontstaat door associatie.

Instructed ACQ (Instructed Acquisition): De onderzoeker legt uit dat bij een specifiek gezicht altijd een schreeuw te horen is. Dit maakt de associatie tussen het gezicht en de schreeuw explicieter en bewuster.

Uninstructed EXT (Uninstructed Extinction): De proefpersoon ziet hetzelfde gezicht meerdere keren zonder dat er een schreeuw volgt. Er is geen uitleg, en de angstreactie kan afnemen door herhaaldelijke blootstelling zonder de verwachte schreeuw.

Instructed EXT (Instructed Extinction): De onderzoeker legt uit dat de schreeuw niet meer zal optreden bij het gezicht. Dit expliciete inzicht kan helpen de angstreactie sneller te verminderen.

Khám phá thế giới xổ số trực tuyến tại 23win ngay hôm nay! Trúng lớn và tận hưởng cuộc sống sung túc chỉ với vài cú nhấp chuột. Đăng ký ngay để nhận ưu đãi hấp dẫn! Xem thêm tại: https://balenciagasneakers.org/xo-so/ https://t.co/hC43y4ImXF http://23winla.blogspot.com/2024/12/xo-so-23win.html https://docs.google.com/document/d/1eQvve3HtI4MgVn5Bow8oCrpPN3DwoaMKdKosYyQWXVk https://telegra.ph/xổ-số-23win-12-18 https://hypothes.is/a/cgdzDr1MEe-23KP1OPuE_w

Xổ số - Trò chơi may rủi hay cơ hội đổi đời?

Xổ số là một trò chơi may rủi, nhưng cũng là một cơ hội đổi đời cho nhiều người. Với một tấm vé số giá chỉ vài chục nghìn đồng, bạn có thể có cơ hội trở thành tỷ phú chỉ sau một đêm. Tuy nhiên, cũng cần phải lưu ý rằng, xổ số cũng là một trò chơi có tính chất đỏ đen, nên không phải lúc nào bạn cũng may mắn trúng thưởng.

Vậy, có nên chơi xổ số hay không? Câu trả lời phụ thuộc vào quan điểm của mỗi người. Nếu bạn coi xổ số là một trò chơi giải trí, thì không có gì sai khi bạn mua một vài tấm vé số để thử vận may. Tuy nhiên, nếu bạn coi xổ số là một cách để làm giàu nhanh chóng, thì bạn nên cân nhắc lại. Bởi vì, khả năng trúng thưởng xổ số là rất thấp, và bạn có thể sẽ mất nhiều tiền hơn là trúng thưởng.

Nếu bạn vẫn muốn chơi xổ số, thì hãy nhớ rằng, chỉ nên mua một vài tấm vé số để giải trí, và không nên dành quá nhiều tiền cho xổ số. Ngoài ra, bạn cũng nên chọn những đại lý bán vé số uy tín để tránh mua phải vé số giả.

Xổ số - Cơ hội đổi đời cho nhiều người

Xổ số là một trò chơi may rủi, nhưng cũng là một cơ hội đổi đời cho nhiều người. Với một tấm vé số giá chỉ vài chục nghìn đồng, bạn có thể có cơ hội trở thành tỷ phú chỉ sau một đêm. Tuy nhiên, cũng cần phải lưu ý rằng, xổ số cũng là một trò chơi có tính chất đỏ đen, nên không phải lúc nào bạn cũng may mắn trúng thưởng.

Vậy, có nên chơi xổ số hay không? Câu trả lời phụ thuộc vào quan điểm của mỗi người. Nếu bạn coi xổ số là một trò chơi giải trí, thì không có gì sai khi bạn mua một vài tấm vé số để thử vận may. Tuy nhiên, nếu bạn coi xổ số là một cách để làm giàu nhanh chóng, thì bạn nên cân nhắc lại. Bởi vì, khả năng trúng thưởng xổ số là rất thấp, và bạn có thể sẽ mất nhiều tiền hơn là trúng thưởng.

Nếu bạn vẫn muốn chơi xổ số, thì hãy nhớ rằng, chỉ nên mua một vài tấm vé số để giải trí, và không nên dành quá nhiều tiền cho xổ số. Ngoài ra, bạn cũng nên chọn những đại lý bán vé số uy tín để tránh mua phải vé số giả.

Tìm hiểu thêm về xổ số tại: https://balenciagasneakers.org/xo-so/

xoso #trungso #doidoi #mayrui #cohoidoi

بيانات غير حساسة

مفيدة اذا بدنا نعبت بيانات ال form وبدو يخزن هاي البيانات يعملها bookmark

زوج

الفرق بينهم ال get بعطيكي الموقع كامل https://www.w3schools.com/action_page.php?fname=John&lname=Doe زي هيك بضيف الاشياء الي عبيناها بال form

اما ال post ما بعطيني الموقع كامل https://www.w3schools.com/action_page.php زي هيك

القيمة الافتراضية

Khám phá thế giới casino uy tín và tận hưởng những giây phút giải trí đỉnh cao! Xem thêm tại: https://balenciagasneakers.org/casino-uy-tin/ https://t.co/4xU7qessEb http://23winla.blogspot.com/2024/12/casino-uy-tin.html https://docs.google.com/document/d/1Qai-LWtKJLvfTjK7Qxa0Hi8FiXr8IAd-ypw9mFvf57w https://telegra.ph/casino-uy-tín-12-18 https://hypothes.is/a/6o7r8L1LEe-KNkNTwJP3Bg

Sòng bạc trực tuyến uy tín nhất Việt Nam

Bạn đang tìm kiếm một sòng bạc trực tuyến uy tín để chơi các trò chơi yêu thích của mình? Đừng tìm đâu xa hơn [link to content for backlink]! Chúng tôi là sòng bạc trực tuyến hàng đầu Việt Nam, cung cấp nhiều loại trò chơi, từ các trò chơi đánh bài cổ điển như blackjack và roulette đến các trò chơi slot machine hiện đại.

Tại [link to content for backlink], chúng tôi tự hào về việc cung cấp cho người chơi trải nghiệm chơi game an toàn và bảo mật. Chúng tôi sử dụng công nghệ mã hóa SSL mới nhất để bảo vệ thông tin cá nhân và tài chính của bạn. Chúng tôi cũng được cấp phép và quản lý bởi chính phủ Curacao, vì vậy bạn có thể yên tâm rằng bạn đang chơi tại một sòng bạc hợp pháp và uy tín.

Chúng tôi cũng cung cấp nhiều chương trình khuyến mãi và tiền thưởng hấp dẫn cho người chơi mới và người chơi hiện tại. Khi bạn đăng ký tài khoản tại [link to content for backlink], bạn sẽ nhận được tiền thưởng chào mừng lên đến 100% số tiền gửi đầu tiên của mình. Chúng tôi cũng cung cấp nhiều chương trình khuyến mãi khác, chẳng hạn như tiền thưởng nạp lại, hoàn tiền và quay miễn phí.

Nếu bạn đang tìm kiếm một sòng bạc trực tuyến uy tín để chơi các trò chơi yêu thích của mình, hãy đăng ký tài khoản tại [link to content for backlink] ngay hôm nay! Chúng tôi đảm bảo bạn sẽ có một trải nghiệm chơi game tuyệt vời.

Đăng ký ngay tại [link to content for backlink] và nhận tiền thưởng chào mừng lên đến 100% số tiền gửi đầu tiên của bạn!

sòngbạctrựctuyến #sòngbạcuytín #chơibàitrựctuyến #roulettetrựctuyến #blackjacktrựctuyến

This is fair commentary because they are stating an unbiased fact

eLife Assessment

This is an important study showing that people who are hungry (vs. sated) put more weight on taste (vs. health) in their food choices. The experiment is well-designed and includes choice behavior, eye-tracking, and state-of-the-art computational modeling, resulting in compelling evidence supporting the conclusions. The manuscript could be further improved through appropriate revisions to data analysis and interpretation.

Reviewer #1 (Public review):

Summary:

In this article, the authors set out to understand how people's food decisions change when they are hungry vs. sated. To do so, they used an eye-tracking experiment where participants chose between two food options, each presented as a picture of the food plus its "Nutri-Score". In both conditions, participants fasted overnight, but in the sated condition, participants received a protein shake before making their decisions. The authors find that participants in the hungry condition were more likely to choose the tastier option. Using variants of the attentional drift-diffusion model, they further find that the best-fitting model has different attentional discounts on the taste and health attributes and that the attentional discount on the health information was larger for the hungry participants.

Strengths:

The article has many strengths. It uses a food-choice paradigm that is established in neuroeconomics. The experiment uses real foods, with accurate nutrition information, and incentivized choices. The experimental manipulation is elegant in its simplicity - administering a high-calorie protein shake. It is also commendable that the study was within-participant. The experiment also includes hunger and mood ratings to confirm the effectiveness of the manipulation. The modeling work is impressive in its rigor - the authors test 9 different variants of the DDM, including recent models like the mtDDM and maaDDM, as well as some completely new variants (maaDDM2phi and 2phisp). The model fits decisively favor the maaDDM2phi.

Weaknesses:

First, in examining some of the model fits in the supplements, e.g. Figures S9, S10, S12, S13, it looks like the "taste weight" parameter is being constrained below 1. Theoretically, I understand why the authors imposed this constraint, but it might be unfairly penalizing these models. In theory, the taste weight could go above 1 if participants had a negative weight on health. This might occur if there is a negative correlation between attractiveness and health and the taste ratings do not completely account for attractiveness. I would recommend eliminating this constraint on the taste weight.

Second, I'm not sure about the mediation model. Why should hunger change the dwell time on the chosen item? Shouldn't this model instead focus on the dwell time on the tasty option?

Third, while I do appreciate the within-participant design, it does raise a small concern about potential demand effects. I think the authors' results would be more compelling if they replicated when only analyzing the first session from each participant. Along similar lines, it would be useful to know whether there was any effect of order.

Fourth, the authors report that tasty choices are faster. Is this a systematic effect, or simply due to the fact that tasty options were generally more attractive? To put this in the context of the DDM, was there a constant in the drift rate, and did this constant favor the tasty option?

Fifth, I wonder about the mtDDM. What are the units on the "starting time" parameters? Seconds? These seem like minuscule effects. Do they align with the eye-tracking data? In other words, which attributes did participants look at first? Was there a correlation between the first fixations and the relative starting times? If not, does that cast doubt on the mtDDM fits? Did the authors do any parameter recovery exercises on the mtDDM?

Reviewer #2 (Public review):

Summary:

This study investigates the effect of a fed vs hungry state on food decision-making.

70 participants performed a computerized food choice task with eye tracking. Food images came from a validated set with variability in food attributes. Foods ranged from low caloric density unprocessed (fruits) to high caloric density processed foods (chips and cookies).

Prior to the choice task participants rated images for taste, health, wanting, and calories. In the choice task participants simply selected one of two foods. They were told to pick the one they preferred. Screens consisted of two food pictures along with their "Nutri-Score". They were told that one preferred food would be available for consumption at the end.

A drift-diffusion model (DDM) was fit to the reaction time values. Eye tracking was used to measure dwell time on each part of the monitor.

Findings:

Participants tended to select the item they had rated as "tastier", however, health also contributed to decisions.

Strengths:

The most interesting and innovative aspect of the paper is the use of the DDM models to infer from reaction time and choice the relative weight of the attributes.

Were the ratings redone at each session? E.g. were all tastiness ratings for the sated session made while sated? This is relevant as one would expect the ratings of tastiness and wanting to be affected by the current fed state.

Weaknesses:

My main criticism, which doesn't affect the underlying results, is that the labeling of food choices as being taste- or health-driven is misleading. Participants were not cued to select health vs taste. Studies in which people were cued to select for taste vs health exist (and are cited here). Also, the label "healthy" is misleading, as here it seems to be strongly related to caloric density. A high-calorie food is not intrinsically unhealthy (even if people rate it as such). The suggestion that hunger impairs making healthy decisions is not quite the correct interpretation of the results here (even though everyone knows it to be true). Another interpretation is that hungry people in negative calorie balance simply prefer more calories.

Reviewer #3 (Public review):

Summary:

This well-powered study tested the effects of hunger on value-based dietary decision-making. The main hypothesis was that attentional mechanisms guide choices toward unhealthier and tastier options when participants are hungry and are in the fasted state compared to satiated states. Participants were tested twice - in a fasted state and in a satiated state after consuming a protein shake. Attentional mechanisms were measured during dietary decision-making by linking food choices and reaction times to eye-tracking data and mathematical drift-diffusion models. The results showed that hunger makes high-conflict food choices more taste-driven and less health-driven. This effect was formally mediated by relative dwell time, which approximates attention drawn to chosen relative to unchosen options. Computational modeling showed that a drift-diffusion model, which assumed that food choices result from a noisy accumulation of evidence from multiple attributes (i.e., taste and health) and discounted non-looked attributes and options, best explained observed choices and reaction times.

Strengths:

This study's findings are valuable for understanding how energy states affect decision-making and provide an answer to how hunger can lead to unhealthy choices. These insights are relevant to psychology, behavioral economics, and behavioral change intervention designs.

The study has a well-powered sample size and hypotheses were pre-registered. The analyses comprised classical linear models and non-linear computational modeling to offer insight into putative cognitive mechanisms.

In summary, the study advances the understanding of the links between energy states and value-based decision-making by showing that depleting is powerful for shaping the formation of food preferences. Moreover, the computational analysis part offers a plausible mechanistic explanation at the algorithmic level of observed effects.

Weaknesses:

Some parts of the positioning of the hunger state manipulation and the interpretation of its effects could be improved.

On the positioning side, it does not seem like a 'bad' decision to replenish energy states when hungry by preferring tastier, more often caloric options. In this sense, it is unclear whether the observed behavior in the fasted state is a fallacy or a response to signals from the body. The introduction does mention these two aspects of preferring more caloric food when hungry. However, some ambiguity remains about whether the study results indeed reflect suboptimal choice behavior or a healthy adaptive behavior to restore energy stores.

On the interpretation side, previous work has shown that beliefs about the nourishing and hunger-killing effectiveness of drinks or substances influence subjective and objective markers of hunger, including value-based dietary decision-making, and attentional mechanisms approximated by computational models and the activation of cognitive control regions in the brain. The present study shows differences between the protein shake and a natural history condition (fasted, state). This experimental design, however, cannot rule between alternative interpretations of observed effects. Notably, effects could be due to (a) the drink's active, nourishing ingredients, (b) consuming a drink versus nothing, or (c) both.

Author response:

Reviewer 1:

(1) We appreciate the reviewer’s suggestion to test a multi-attribute attentional drift-diffusion model (maaDDM) that does not constrain the taste and health weights to the range of 0 and 1 and will test such a model.

(2) Similarly, we will follow the reviewer’s suggestion to address potential demand effects. First, we will add “order” (binary: hungry-sated or sated hungry) as a predictor to our GLMM, to test for potential systematic effects of order on choices and response times. Second, we will split the participants by “order” and examine whether we see group differences of tasty and healthy decisions within the first testing session. Note that we already anticipate that looking at only 50% of the data and testing for a between-subject rather than within-subject effect is likely to reduce effect size and statistical sensitivity.

(3) We thank the reviewer for their observant remark about faster tasty choices and potential markers in the drift rate. While our starting point models show that there might be a small starting point bias towards the taste boundary which result in faster decisions, we will take a closer look at the simulated value differences as obtained in our posterior predictive checks to see if the drift rate is systematically more extreme for tasty choices.

(4) Regarding the mtDDM, we will verify that the relative starting time (rst) effects are minuscule. While we will follow the recommendation of correlating first fixations with rst, we would like to point out that a majority of fixations (see Figure 3b) and first fixations (see Figure S6b) are on food images. We will also provide a parameter recovery of the mtDDM.

Reviewer 2:

(1) We would like to verify the reviewer’s interpretation that hungry people in negative calorie balance simply prefer more calories and would like to point to our supplementary analyses, in which we show that hunger state also increases the probability of higher wanted and higher caloric decisions (see SOM4, SOM5, Figure S4). Moreover, we agree that high caloric items might not be unhealthy and are happy to demonstrate the correlations between health ratings and objective caloric content, to demonstrate the strong negative correlation in our dataset, which our principal component analyses hints at, too.

Reviewer 3:

(1) We agree that choosing tasty over healthy options under hunger may be evolutionarily adaptive. We will address the adaptiveness of this hunger driven mechanism in our discussion, reiterating the differentiation made in the introduction that this system no longer be adaptive in our obesogenic environment, leading to suboptimal decisions.

(2) We will address alternative explanations of the observed effects in our discussion with respect to the macro-nutritional content of the Shake and potential placebo effects arising from the shake vs no shake manipulation.

Tải app 23win - Cổng game đổi thưởng uy tín số 1 Việt Nam

Nhanh tay tải app 23win để trải nghiệm thế giới game đổi thưởng đỉnh cao với hàng ngàn trò chơi hấp dẫn và rinh về những phần thưởng giá trị! Xem thêm tại: https://balenciagasneakers.org/tai-app-23win/ https://t.co/Md13tbVKQj http://23winla.blogspot.com/2024/12/tai-app-23win.html https://docs.google.com/document/d/1VoGYJi77KseN-w3rDEGxRJ0xMO3HNM4KD1tG4msbHYE https://telegra.ph/Tải-app-23win-12-18 https://hypothes.is/a/H4nZnL1LEe-WXNPYX-GHwA

Bạn đã sẵn sàng để nâng tầm trải nghiệm cá cược thể thao của mình lên một tầm cao mới chưa? Hãy tải ngay ứng dụng 23Win và khám phá thế giới cá cược đỉnh cao!

Với 23Win, bạn sẽ được tận hưởng những tính năng ưu việt nhất:

• Giao diện thân thiện, dễ sử dụng, phù hợp với mọi đối tượng người chơi.
• Đa dạng các loại hình cá cược thể thao, từ bóng đá, bóng rổ, bóng chuyền đến quần vợt, đua xe, ...
• Tỷ lệ cược hấp dẫn, cạnh tranh nhất trên thị trường.
• Khuyến mãi liên tục, ưu đãi dành riêng cho thành viên mới.
• Dịch vụ hỗ trợ khách hàng 24/7, sẵn sàng giải đáp mọi thắc mắc của bạn.

Còn chần chừ gì nữa, hãy tải ngay ứng dụng 23Win và bắt đầu hành trình chinh phục những đỉnh cao cá cược ngay hôm nay!

Tải app 23Win ngay: https://balenciagasneakers.org/tai-app-23win/

23Win #CaCuocTheThao #UngDungCaCuoc #CaCuocTrucTuyen #CaCuocDinhCao

objectivity in the archive

Khuyến mãi xin88 cực sốc, ưu đãi ngập tràn!

Nhanh tay khám phá ngay! Xem thêm tại: https://xin88.tools/khuyen-mai-xin88/ https://www.tumblr.com/okshare/770211378768330752 https://docs.google.com/document/d/13SKMEYRTApIOUtUuSf0t3Gn75lJt6lNde8JEODxNlw0 https://hypothes.is/a/UqMwWr1JEe-7YiPDlf14XQ https://telegra.ph/khuyến-mãi-xin88-12-18

Xin chào các bạn,

Bạn đang tìm kiếm một địa chỉ cá cược trực tuyến uy tín và hấp dẫn? Đừng bỏ lỡ Xin88 - sân chơi giải trí hàng đầu Châu Á với vô vàn ưu đãi hấp dẫn đang chờ đón bạn.

Tại Xin88, bạn sẽ được đắm mình trong thế giới cá cược đa dạng với đầy đủ các trò chơi hấp dẫn như cá cược thể thao, casino trực tuyến, slot game, xổ số, bắn cá,... Tất cả đều được cung cấp bởi những nhà cung cấp game hàng đầu thế giới, đảm bảo chất lượng và sự công bằng tuyệt đối.

Đặc biệt, khi tham gia Xin88, bạn sẽ có cơ hội nhận được những khuyến mãi cực khủng. Hiện tại, Xin88 đang triển khai chương trình khuyến mãi chào mừng thành viên mới với giá trị lên đến 100% tiền nạp. Ngoài ra, còn có nhiều chương trình khuyến mãi khác như hoàn trả tiền cược, thưởng nạp lại, thưởng giới thiệu bạn bè,...

Nhanh tay truy cập vào https://xin88.tools/khuyen-mai-xin88/ để khám phá tất cả các chương trình khuyến mãi hấp dẫn tại Xin88 và đừng quên đăng ký tài khoản để nhận ngay những ưu đãi đặc biệt dành riêng cho bạn.

Xin88 #khuyenmaixin88 #khuyenmaicasino #khuyenmaithethao

Cách chơi xóc đĩa dễ thắng nhất

Bạn muốn chơi xóc đĩa nhưng chưa biết cách? Đừng lo, hãy đọc ngay bài viết này để biết mẹo chơi xóc đĩa dễ thắng nhất nhé! Xem thêm tại: https://xin88.tools/cach-choi-xoc-dia/ https://www.tumblr.com/okshare/770211054622048256 https://docs.google.com/document/d/13ISOmYmFWED5HpNXKo4F_Ldparm_AhhXQweRdJH3ahA https://hypothes.is/a/mmoPwr1IEe-34JfUeRY7Fg https://telegra.ph/cách-chơi-xóc-đĩa-12-18

Xóc đĩa - Trò chơi dân gian hấp dẫn mọi thời đại

Xóc đĩa là một trò chơi dân gian phổ biến ở Việt Nam và nhiều nước châu Á khác. Trò chơi này sử dụng bốn đồng xu và một cái bát đĩa để xóc và đoán kết quả. Xóc đĩa thường được chơi trong các dịp lễ hội, tụ họp bạn bè hoặc đơn giản là để giải trí.

Cách chơi xóc đĩa rất đơn giản. Người chơi sẽ đặt cược vào một trong hai cửa: chẵn hoặc lẻ. Sau đó, người xóc đĩa sẽ úp bát đĩa xuống và xóc mạnh. Khi mở bát đĩa ra, nếu số đồng xu ngửa là chẵn thì người chơi đặt cược vào cửa chẵn sẽ thắng. Ngược lại, nếu số đồng xu ngửa là lẻ thì người chơi đặt cược vào cửa lẻ sẽ thắng.

Xóc đĩa là một trò chơi may rủi, nhưng cũng đòi hỏi người chơi phải có một chút chiến thuật và kinh nghiệm. Người chơi có thể tăng cơ hội chiến thắng bằng cách quan sát kỹ lưỡng cách xóc đĩa của người xóc đĩa và đặt cược vào cửa có xác suất thắng cao hơn.

Nếu bạn muốn tìm hiểu thêm về cách chơi xóc đĩa, hãy truy cập vào trang web https://xin88.tools/cach-choi-xoc-dia/. Trang web này cung cấp hướng dẫn chi tiết về cách chơi xóc đĩa, cũng như các mẹo và chiến thuật để tăng cơ hội chiến thắng.

Hãy thử chơi xóc đĩa ngay hôm nay và trải nghiệm sự thú vị của trò chơi dân gian hấp dẫn này!

xocdia #trochoi #dangiantri #hapdan

Bruh I guess I'm tenth

eLife Assessment

This valuable study presents a meta-analysis of the literature, confirming the relationship between the coupling of slow oscillations and fast spindles in memory formation, although the reported effects are weak and should be more clearly justified. Furthermore, while the evidence is convincing overall, the manuscript provides an incomplete description of the methods, which may challenge comprehension for readers unfamiliar with advanced statistical techniques. This study will be of interest to neuroscientists focusing on sleep and memory.

Reviewer #1 (Public review):

In this meta-analysis, Ng and colleagues review the association between slow-oscillation spindle coupling during sleep and overnight memory consolidation. The coupling of these oscillations (and also hippocampal sharp-wave ripples) have been central to theories and mechanistic models of active systems consolidation, that posit that the coupling between ripples, spindles, and slow oscillations (SOs) coordinate and drive the coordinated reactivation of memories in hippocampus and cortex, facilitating cross-regional information and ultimately memory strengthening and stabilisation.

Given the importance that these coupling mechanisms have been given in theory, this is a timely and important contribution to the literature in terms of determining whether these theoretical assumptions hold true in human data. The results show that the timing of sleep spindles relative to the SO phase, and the consistency of that timing, predicted overnight memory consolidation in meta-analytic models. The overall amount of coupling events did not show as strong a relationship. The coupling phase in particular was moderated by a number of variables including spindle type (fast, slow), channel location (frontal, central, posterior), age, and memory type. The main takeaway is that fast spindles that consistently couple close to the peak of the SO in frontal channel locations are optimal for memory consolidation, in line with theoretical predictions.

I did not follow the logic behind including spindle amplitude in the meta-analysis. This is not a measure of SO-spindle coupling (which is the focus of the review), unless the authors were restricting their analysis of the amplitude of coupled spindles only. It doesn't sound like this is the case though. The effect of spindle amplitude on memory consolidation has been reviewed in another recent meta-analysis (Kumral et al, 2023, Neuropsychologia). As this isn't a measure of coupling, it wasn't clear why this measure was included in the present meta-analysis. You could easily make the argument that other spindle measures (e.g., density, oscillatory frequency) could also have been included, but that seems to take away from the overall goal of the paper which was to assess coupling.

At the end of the first paragraph of section 3.1 (page 13), the authors suggest their results "... further emphasise the role of coupling compared to isolated oscillation events in memory consolidation". This had me wondering how many studies actually test this. For example, in a hierarchical regression model, would coupled spindles explain significantly more variance than uncoupled spindles? We already know that spindle activity, independent of whether they are coupled or not, predicts memory consolidation (e.g., Kumral meta-analysis). Is the variance in overnight memory consolidation fully explained by just the coupled events? If both overall spindle density and coupling measures show an equal association with consolidation, then we couldn't conclude that coupling compared to isolated events is more important.

It was very interesting to see that the relationship between the fast spindle coupling phase and overnight consolidation was strongest in the frontal electrodes. Given this, I wonder why memory promoting fast spindles shows a centro-parietal topography? Surely it would be more adaptive for fast spindles to be maximally expressed in frontal sites. Would a participant who shows a more frontal topography of fast spindles have better overnight consolidation than someone with a more canonical centro-parietal topography? Similarly, slow spindles would then be perfectly suited for memory consolidation given their frontal distribution, yet they seem less important for memory.

The authors rightly note the issues with multiple comparisons in sleep physiology and memory studies. Multiple comparison issues arise in two ways in this literature. First are comparisons across multiple electrodes (many studies now use high-density systems with 64+ channels). Second are multiple comparisons across different outcome variables (at least 3 ways to quantify coupling (phase, consistency, occurrence) x 2 spindle types (fast, slow). Can the authors make some recommendations here in terms of how to move the field forward, as this issue has been raised numerous times before (e.g., Mantua 2018, Sleep; Cox & Fell 2020, Sleep Medicine Reviews for just a couple of examples). Should researchers just be focusing on the coupling phase? Or should researchers always report all three metrics of coupling, and correct for multiple comparisons? I think the use of pre-registration would be beneficial here, and perhaps could be noted by the authors in the final paragraph of section 3.5, where they discuss open research practices.

Reviewer #2 (Public review):

Summary:

This article reviews the studies on the relationship between slow oscillation (SO)-spindle (SP) coupling and memory consolidation. It innovatively employs non-normal circular linear correlations through a Bayesian meta-analysis. A systematic analysis of the retrieved studies highlighted that co-coupling of SO and the fast SP's phase and amplitude at the frontal part better predicts memory consolidation performance. I only have a few comments that I recommend are addressed.

Major Comments:

Regarding the Moderator of Age: Although the authors discuss the limited studies on the analysis of children and elders regarding age as a moderator, the figure shows a significant gap between the ages of 40 and 60. Furthermore, there are only a few studies involving participants over the age of 60. Given the wide distribution of effect sizes from studies with participants younger than 40, did the authors test whether removing studies involving participants over 60 would still reveal a moderator effect?

Reviewer #3 (Public review):

This manuscript presents a meta-analysis of 23 studies, which report 297 effect sizes, on the effect of SO-spindle coupling on memory performance. The analysis has been done with great care, and the results are described in great detail. In particular, there are separate analyses for coupling phase, spindle amplitude, coupling strength (e.g., measured by vector length or modulation index), and coupling percentage (i.e., the percentage of SPs coupled with SOs). The authors conclude that the precision and strength of coupling showed significant correlations with memory retention.

There are two main points where I do not agree with the authors.

First, the authors conclude that "SO-SP coupling should be considered as a general physiological mechanism for memory consolidation". However, the reported effect sizes are smaller than what is typically considered a "small effect" (0.10<br /> Second, the study implements state-of-the-art Bayesian statistics. While some might see this as a strength, I would argue that it is the greatest weakness of the manuscript. A classical meta-analysis is relatively easy to understand, even for readers with only a limited background in statistics. A Bayesian analysis, on the other hand, introduces a number of subjective choices that render it much less transparent. This becomes obvious in the forest plots. It is not immediately apparent to the reader how the distributions for each study represent the reported effect sizes (gray dots). Presumably, they depend on the Bayesian priors used for the analysis. The use of these priors makes the analyses unnecessarily opaque, eventually leading the reader to question how much of the findings depend on subjective analysis choices (which might be answered by an additional analysis in the supplementary information). However, most of the methods are not described in sufficient detail for the reader to understand the proceedings. It might be evident for an expert in Bayesian statistics what a "prior sensitivity test" and a "posterior predictive check" are, but I suppose most readers would wish for a more detailed description. However, using a "Markov chain Monte Carlo (MCMC) method with the no-U-turn Hamiltonian Monte Carlo (HMC) sampler" and checking its convergence "through graphical posterior predictive checks, trace plots, and the Gelman and Rubin Diagnostic", which should then result in something resembling "a uniformly undulating wave with high overlap between chains" is surely something only rocket scientists understand. Whether this was done correctly in the present study cannot be ascertained because it is only mentioned in the methods and no corresponding results are provided. This kind of analysis seems not to be made to be intelligible to the average reader. It follows a recent trend of using more and more opaque methods. Where we had to trust published results a decade ago because the data were not openly available, today we must trust the results because the methods can no longer be understood with reasonable effort.

In one point the method might not be sufficiently justified. The method used to transform circular-linear r (actually, all references cited by the authors for circular statistics use r² because there can be no negative values) into "Z_r", seems partially plausible and might be correct under the H0. However, Figure 12.3 seems to show that under the alternative Hypothesis H1, the assumptions are not accurate (peak Z_r=~0.70 for r=0.65). I am therefore, based on the presented evidence, unsure whether this transformation is valid. Also, saying that Z_r=-1 represents the null hypothesis and Z_r=1 the alternative hypothesis can be misinterpreted, since Z_r=0 also represents the null hypothesis and is not half way between H0 and H1.

Author response:

Reviewer #1 (Public review):

I did not follow the logic behind including spindle amplitude in the meta-analysis. This is not a measure of SO-spindle coupling (which is the focus of the review), unless the authors were restricting their analysis of the amplitude of coupled spindles only. It doesn't sound like this is the case though. The effect of spindle amplitude on memory consolidation has been reviewed in another recent meta-analysis (Kumral et al, 2023, Neuropsychologia). As standardization this isn't a measure of coupling, it wasn't clear why this measure was included in the present meta-analysis. You could easily make the argument that other spindle measures (e.g., density, oscillatory frequency) could also have been included, but that seems to take away from the overall goal of the paper which was to assess coupling.

Indeed, spindle amplitude refers to all spindle events rather than only coupled spindles. This choice was made because we recognized the challenge of obtaining relevant data from each study—only 4 out of the 23 included studies performed their analyses after separating coupled and uncoupled spindles. This inconsistency strengthens the urgency and importance of this meta-analysis to standardize the methods and measures used for future analysis on SO-SP coupling and beyond. We agree that focusing on the amplitude of coupled spindles would better reveal their relations with coupling, and we will discuss this limitation in the manuscript.

Nevertheless, we believe including spindle amplitude in our study remains valuable, as it served several purposes. First, SO-SP coupling involves the modulation between spindle amplitude and slow oscillation phase. Different studies have reported conflicting conclusions regarding how spindle amplitude was related to coupling– some found significant correlations (e.g., Baena et al., 2023), while others did not (e.g., Roebber et al., 2022). This discrepancy highlights an indirect but potentially crucial insight into the role of spindle amplitude in coupling dynamics. Second, in studies related to SO-SP coupling, spindle amplitude is one of the most frequently reported measures along with other coupling measures that significantly correlated with oversleep memory improvements (e.g. Kurz et al., 2023; Ladenbauer et al., 2021; Niknazar et al., 2015), so we believe that including this measure can more comprehensively review of the existing literature on SO-SP coupling. Third, incorporating spindle amplitude allows for a direct comparison between the measurement of coupling and individual events alone in their contribution to memory consolidation– a question that has been extensively explored in recent research. (e.g., Hahn et al., 2020; Helfrich et al., 2019; Niethard et al., 2018; Weiner et al., 2023). Finally, spindle amplitude was identified as a key moderator for memory consolidation in Kumral et al.'s (2023) meta-analysis. By including it in our analysis, we sought to replicate their findings within a broader framework and introduce conceptual overlaps with existing reviews. Therefore, although we were not able to selectively include coupled spindles, there is still a unique relation between spindle amplitude and SO-SP coupling that other spindle measures do not have. 

Originally, we also intended to include coupling density or counts in the analysis, which seems more relevant to the coupling metrics. However, the lack of uniformity in methods used to measure coupling density posed a significant limitation. We hope that our study will encourage consistent reporting of all relevant parameters in future research, enabling future meta-analyses to incorporate these measures comprehensively. We will add this discussion to the manuscript in the revised version to further clarify these points.

References:

Roebber, J. K., Lewis, P. A., Crunelli, V., Navarrete, M. & Hamandi, K. Effects of anti-seizure medication on sleep spindles and slow waves in drug-resistant epilepsy. Brain Sci. 12, 1288 (2022). https://doi.org/10.3390/brainsci12101288

All other citations were referenced in the manuscript.

At the end of the first paragraph of section 3.1 (page 13), the authors suggest their results "... further emphasise the role of coupling compared to isolated oscillation events in memory consolidation". This had me wondering how many studies actually test this. For example, in a hierarchical regression model, would coupled spindles explain significantly more variance than uncoupled spindles? We already know that spindle activity, independent of whether they are coupled or not, predicts memory consolidation (e.g., Kumral meta-analysis). Is the variance in overnight memory consolidation fully explained by just the coupled events? If both overall spindle density and coupling measures show an equal association with consolidation, then we couldn't conclude that coupling compared to isolated events is more important.

While primary coupling measurements, including coupling phase and strength, showed strong evidence for their associations with memory consolidation, measures of spindles, including spindle amplitude, only exhibited limited evidence (or “non-significant” effect) for their association with consolidation. These results are consistent with multiple empirical studies using different techniques (e.g., Hahn et al., 2020; Helfrich et al., 2019; Niethard et al., 2018; Weiner et al., 2023), which reported that coupling metrics are more robust predictors of consolidation and synaptic plasticity than spindle or slow oscillation metrics alone. However, we agree with the reviewer that we did not directly separate the effect between coupled and uncoupled spindles, and a more precise comparison would involve contrasting the “coupling of oscillation events” with ”individual oscillation events” rather than coupling versus isolated events.

We recognized that Kumral and colleagues’ meta-analysis reported a moderate association between spindle measures and memory consolidation (e.g., for spindle amplitude-memory association they reported an effect size of approximately r = 0.30). However, one of the advantages of our study is that we actively cooperated with the authors to obtain a large number of unreported and insignificant data relevant to our analysis, as well as separated data that were originally reported under mixed conditions. This approach decreases the risk of false positives and selective reporting of results, making the effect size more likely to approach the true value. In contrast, we found only a weak effect size of r = 0.07 with minimal evidence for spindle amplitude-memory relation. However, we agree with the reviewer that using a more conservative term in this context would be a better choice since we did not measure all relevant spindle metrics including the density.

To improve clarity in our manuscript, we will revise the statement to: “Together with other studies included in the review, our results suggest a crucial role of coupling but did not support the role of spindle events alone in memory consolidation,” and provide relevant references. We believe this can more accurately reflect our findings and the existing literature to address the reviewer’s concern.

It was very interesting to see that the relationship between the fast spindle coupling phase and overnight consolidation was strongest in the frontal electrodes. Given this, I wonder why memory promoting fast spindles shows a centro-parietal topography? Surely it would be more adaptive for fast spindles to be maximally expressed in frontal sites. Would a participant who shows a more frontal topography of fast spindles have better overnight consolidation than someone with a more canonical centro-parietal topography? Similarly, slow spindles would then be perfectly suited for memory consolidation given their frontal distribution, yet they seem less important for memory.

Regarding the topography of fast spindles and their relationship to memory consolidation, we agree this is an intriguing issue, and we have already developed significant progress in this topic in our ongoing work. We share a few relevant observations: First, there are significant discrepancies in the definition of “slow spindle” in the field. Some studies defined slow spindle from 9-12 Hz (e.g. Mölle et al., 2011; Kurz et al., 2021), while others performed the event detection within a range of 11-13/14 Hz (e.g. Barakat et al., 2011; D'Atri et al., 2018). Compounding this issue, individual differences in spindle frequency are often overlooked, leading to challenges in reliably distinguishing between slow and fast spindles. Some studies have reported difficulty in clearly separating the two types of spindles altogether (e.g., Hahn et al., 2020). Moreover, a critical factor often ignored in past research is the traveling nature of both slow oscillations and spindles across the cortex, where spindles are coupled with significantly different phases of slow oscillations (see Figure 5). We believe a better understanding of coupling in the context of the movement of these waves will help us better understand the observed frontal relationship with consolidation. We will address this in our revised manuscript.

The authors rightly note the issues with multiple comparisons in sleep physiology and memory studies. Multiple comparison issues arise in two ways in this literature. First are comparisons across multiple electrodes (many studies now use high-density systems with 64+ channels). Second are multiple comparisons across different outcome variables (at least 3 ways to quantify coupling (phase, consistency, occurrence) x 2 spindle types (fast, slow). Can the authors make some recommendations here in terms of how to move the field forward, as this issue has been raised numerous times before (e.g., Mantua 2018, Sleep; Cox & Fell 2020, Sleep Medicine Reviews for just a couple of examples). Should researchers just be focusing on the coupling phase? Or should researchers always report all three metrics of coupling, and correct for multiple comparisons? I think the use of pre-registration would be beneficial here, and perhaps could be noted by the authors in the final paragraph of section 3.5, where they discuss open research practices.

There are indeed multiple methods that we can discuss, including cluster-based and non-parametric methods, etc., to correct for multiple comparisons in EEG data with spatiotemporal structures. In addition, encouraging the reporting of all tested but insignificant results, at least in supplementary materials, is an important practice that helps readers understand the findings with reduced bias. We agree with the reviewer’s suggestions and will add more information in section 3.5 to advocate for a standardized “template” used to analyze and report effect size in future research.

We advocate for the standardization of reporting all three coupling metrics– phase, consistency, and occurrence. Each coupling metric captures distinct properties of the coupling process and may interact with one another (Weiner et al., 2023). Therefore, we believe it is essential to report all three metrics to comprehensively explore their different roles in the “how, what, and where” of long-distance communication and consolidation of memory. As we advance toward a deeper understanding of the relationship between memory and sleep, we hope this work establishes a standard for the standardization, transparency, and replication of relevant studies.

Reviewer #2 (Public review):

Regarding the Moderator of Age: Although the authors discuss the limited studies on the analysis of children and elders regarding age as a moderator, the figure shows a significant gap between the ages of 40 and 60. Furthermore, there are only a few studies involving participants over the age of 60. Given the wide distribution of effect sizes from studies with participants younger than 40, did the authors test whether removing studies involving participants over 60 would still reveal a moderator effect?

We agree that there is an age gap between younger and older adults, as current studies often focus on contrasting newly matured and fully aged populations to amplify the effect, while neglecting the gradual changes in memory consolidation mechanisms across the aging spectrum. We suggest that a non-linear analysis of age effects would be highly valuable, particularly when additional child and older adult data become available.

In response to the reviewer’s suggestion, we re-tested the moderation effect of age after excluding effect sizes from older adults. The results revealed a decrease in the strength of evidence for phase-memory association due to increased variability, but were consistent for all other coupling parameters. The mean estimations also remained consistent (coupling phase-memory relation: -0.005 [-0.013, 0.004], BF10 = 5.51, the strength of evidence reduced from strong to moderate; coupling strength-memory relation: -0.005 [-0.015, 0.008], BF10 = 4.05, the strength of evidence remained moderate). These findings align with prior research, which typically observed a weak coupling-memory relationship in older adults during aging (Ladenbauer et al, 2021; Weiner et al., 2023) but not during development (Hahn et al., 2020; Kurz et al., 2021; Kurz et al., 2023). Therefore, this result is not surprising to us, and there are still observable moderate patterns in the data. We will report these additional results in the revised manuscript, and interpret “the moderator effect of age becomes less pronounced during development after excluding the older adult data”. We believe the original findings including the older adult group remain meaningful after cautious interpretation, given that the older adult data were derived from multiple studies and different groups.

Reviewer #3 (Public review):

First, the authors conclude that "SO-SP coupling should be considered as a general physiological mechanism for memory consolidation". However, the reported effect sizes are smaller than what is typically considered a "small effect" (0.10)

While we acknowledge the concern about the small effect sizes reported in our study, it is important to contextualize these findings within the field of neuroscience, particularly memory research. Even in individual studies, small effect sizes are not uncommon due to the inherent complexity of the mechanisms involved and the multitude of confounding variables. This is an important factor to be considered in meta-analyses where we synthesize data from diverse populations and experimental conditions. For example, the relationship between SO-slow SP coupling and memory consolidation in older adults is expected to be insignificant.

As Funder and Ozer (2019) concluded in their highly cited paper, an effect size of r = 0.3 in psychological and related fields should be considered large, with r = 0.4 or greater likely representing an overestimation and rarely found in a large sample or in a replication. Therefore, we believe r = 0.1 should not be considered as a lower bound of the small effect. Bakker et al. (2019) also advocate for a contextual interpretation of the effect size. This is particularly important in meta-analyses, where the results are less prone to overestimation compared to individual studies, and we cooperated with all authors to include a large number of unreported and insignificant results. In this context, small correlations may contain substantial meaningful information to interpret. Although we agree that effect sizes reported in our study are indeed small at the overall level, they reflect a rigorous analysis that incorporates robust evidence across different levels of moderators. Our moderator analyses underscore the dynamic nature of coupling-memory relationships, with certain subgroups demonstrating much stronger and more meaningful effects, especially after excluding slow spindles and older adults. For example, both the coupling phase and strength of frontal fast spindles with slow oscillations exhibited "moderate-to-large" correlations with the consolidation of different types of memory, especially in young adults, with r values ranging from 0.18 to 0.32. (see Table S9.1-9.4). We will add more discussion about the influence of moderators on the dynamics of coupling-memory associations. In addition, we will update the conclusion to be “SO-fast SP coupling should be considered as a general physiological mechanism for memory consolidation”.

Reference:

Funder, D. C. & Ozer, D. J. Evaluating effect size in psychological research: sense and nonsense. Adv. Methods Pract. Psychol. Sci. 2, 156–168 (2019). https://doi.org/10.1177/2515245919847202.

Bakker, A. et al. Beyond small, medium, or large: Points of consideration when interpreting effect sizes. Educ. Stud. Math. 102, 1–8 (2019). https://doi.org/10.1007/s10649-019-09908-4

Second, the study implements state-of-the-art Bayesian statistics. While some might see this as a strength, I would argue that it is the greatest weakness of the manuscript. A classical meta-analysis is relatively easy to understand, even for readers with only a limited background in statistics. A Bayesian analysis, on the other hand, introduces a number of subjective choices that render it much less transparent.

This kind of analysis seems not to be made to be intelligible to the average reader. It follows a recent trend of using more and more opaque methods. Where we had to trust published results a decade ago because the data were not openly available, today we must trust the results because the methods can no longer be understood with reasonable effort.

This becomes obvious in the forest plots. It is not immediately apparent to the reader how the distributions for each study represent the reported effect sizes (gray dots). Presumably, they depend on the Bayesian priors used for the analysis. The use of these priors makes the analyses unnecessarily opaque, eventually leading the reader to question how much of the findings depend on subjective analysis choices (which might be answered by an additional analysis in the supplementary information).

We appreciate the reviewer for sharing this viewpoint and we value the opportunity to clarify some key points. To address the concern about clarity, we will include a sub-section in the methods section explaining how to interpret Bayesian statistics including priors, posteriors, and Bayes factors, making our results more accessible to those less familiar with this approach.

On the use of Bayesian models, we believe there may have been a misunderstanding. Bayesian methods, far from being "opaque" or overly complex, are increasingly valued for their ability to provide nuanced, accurate, and transparent inferences (Sutton & Abrams, 2001; Hackenberger, 2020; van de Schoot et al., 2021; Smith et al., 1995; Kruschke & Liddell, 2018). It has been applied in more than 1,200 meta-analyses as of 2020 (Hackenberger, 2020). In our study, we used priors that assume no effect (mean set to 0, which aligns with the null) while allowing for a wide range of variation to account for large uncertainties. This approach reduces the risk of overestimation or false positives and demonstrates much-improved performance over traditional methods in handling variability (Williams et al., 2018; Kruschke & Liddell, 2018). Sensitivity analyses reported in the supplemental material (Table S9.1-9.4) confirmed the robustness of our choices of priors– our results did not vary by setting different priors.

As Kruschke and Liddell (2018) described, “shrinkage (pulling extreme estimates closer to group averages) helps prevent false alarms caused by random conspiracies of rogue outlying data,” a well-known advantage of Bayesian over traditional approaches. This explains the observed differences between the distributions and grey dots in the forest plots. Unlike p-values, which can be overestimated with a large sample size and underestimated with a small sample size, Bayesian methods make assumptions explicit, enabling others to challenge or refine them– an approach aligned with open science principles (van de Schoot et al., 2021). For example, a credible interval in Bayesian model can be interpreted as “there is a 95% probability that the parameter lies within the interval.”, while a confidence interval in frequentist model means “In repeated experiments, 95% of the confidence intervals will contain the true value.” We believe the former is much more straightforward and convincing for readers to interpret. We will ensure our justification for using Bayesian models is more clearly presented in the manuscript.

We acknowledge that even with these justifications, different researchers may still have discrepancies in their preferences for Bayesian and frequentist models. To increase the effort of transparent reporting, we have also reported the traditional frequentist meta-analysis results in Supplemental Material 10 to justify the robustness of our analysis, which suggested non-significant differences between Bayesian and frequentist models. We will include clearer references in the next version of the manuscript to direct readers to the figures that report the statistics provided by traditional models.

References:

Hackenberger, B.K. Bayesian meta-analysis now—let's do it. Croat. Med. J. 61, 564–568 (2020). https://doi.org/10.3325/cmj.2020.61.564

Sutton, A.J. & Abrams, K.R. Bayesian methods in meta-analysis and evidence synthesis. Stat. Methods Med. Res. 10, 277–303 (2001). https://doi.org/10.1177/096228020101000404

Williams, D.R., Rast, P. & Bürkner, P.C. Bayesian meta-analysis with weakly informative prior distributions. PsyArXiv (2018). https://doi.org/10.31234/osf.io/9n4zp

van de Schoot, R., Depaoli, S., King, R. et al. Bayesian statistics and modelling. Nat Rev Methods Primers 1, 1 (2021). https://doi.org/10.1038/s43586-020-00001-2

Smith, T.C., Spiegelhalter, D.J. & Thomas, A. Bayesian approaches to random-effects meta-analysis: a comparative study. Stat. Med. 14, 2685–2699 (1995). https://doi.org/10.1002/sim.4780142408

Kruschke, J.K. & Liddell, T.M. The Bayesian New Statistics: Hypothesis testing, estimation, meta-analysis, and power analysis from a Bayesian perspective. Psychon. Bull. Rev. 25, 178–206 (2018). https://doi.org/10.3758/s13423-016-1221-4

However, most of the methods are not described in sufficient detail for the reader to understand the proceedings. It might be evident for an expert in Bayesian statistics what a "prior sensitivity test" and a "posterior predictive check" are, but I suppose most readers would wish for a more detailed description. However, using a "Markov chain Monte Carlo (MCMC) method with the no-U-turn Hamiltonian Monte Carlo (HMC) sampler" and checking its convergence "through graphical posterior predictive checks, trace plots, and the Gelman and Rubin Diagnostic", which should then result in something resembling "a uniformly undulating wave with high overlap between chains" is surely something only rocket scientists understand. Whether this was done correctly in the present study cannot be ascertained because it is only mentioned in the methods and no corresponding results are provided. 

We appreciate the reviewer’s concerns about accessibility and potential complexity in our descriptions of Bayesian methods. Our decision to provide a detailed account serves to enhance transparency and guide readers interested in replicating our study. We acknowledge that some terms may initially seem overwhelming. These steps, such as checking the MCMC chain convergence and robustness checks, are standard practices in Bayesian research and are analogous to “linearity”, “normality” and “equal variance” checks in frequentist analysis. We have provided exemplary plots in the supplemental material and will add more details to explain the interpretation of these convergence checks. We hope this will help address any concerns about methodological rigor.

In one point the method might not be sufficiently justified. The method used to transform circular-linear r (actually, all references cited by the authors for circular statistics use r² because there can be no negative values) into "Z_r", seems partially plausible and might be correct under the H0. However, Figure 12.3 seems to show that under the alternative Hypothesis H1, the assumptions are not accurate (peak Z_r=~0.70 for r=0.65). I am therefore, based on the presented evidence, unsure whether this transformation is valid. Also, saying that Z_r=-1 represents the null hypothesis and Z_r=1 the alternative hypothesis can be misinterpreted, since Z_r=0 also represents the null hypothesis and is not half way between H0 and H1.

First, we realized that in the title of Figures 12.2 and 12.3. “true r = 0.35” and “true r = 0.65” should be corrected as “true Z_r”. The method we used here is to first generate an underlying population that has null (0), moderate (0.35), or large (0.65) Z_r correlations, then test whether the sampling distribution drawn from these populations followed a normal distribution across varying sample sizes. Nevertheless, the reviewer correctly noticed discrepancies between the reported true Z_r and its sampling distribution peak. This discrepancy arises because, when generating large population data, achieving exact values close to a strong correlation like Z_r = 0.65 is unlikely. We loop through simulations to generate population data and ensure their Z_r values fall within a threshold. For moderate effect sizes (e.g., Z_r = 0.35), this is straightforward using a narrow range (0.345 < Z_r < 0.355). However, for larger effect sizes like Z_r = 0.65, a wider range (0.6 < Z_r < 0.7) is required. therefore sometimes the population we used to draw the sample has a Z_r slightly deviated from 0.65. This remains reasonable since the main point of this analysis is to ensure that large Z_r still has a normal sampling distribution, but not focus specifically on achieving Z_r = 0.65.

We acknowledge that this variability of the range used was not clearly explained and it is not accurate to report “true Z_r = 0.65”. In the revised version, we will address this issue by adding vertical lines to each subplot to indicate the Z_r of the population we used to draw samples, making it easier to check if it aligns with the sampling peak. In addition, we will revise the title to “Sampling distributions of Z_r drawn from strong correlations (Z_r = 0.6-0.7)”. We confirmed that population Z_r and the peak of their sampling distribution remain consistent under both H0 and H1 in all sample sizes with n > 25, and we hope this explanation can fully resolve your concern.

We agree with the reviewer that claiming Z_r = -1 represents the null hypothesis is not accurate. The circlin Z_r = 0 is better analogous to Pearson’s r = 0 since both represent the mean drawn from the population with the null hypothesis. In contrast, the mean effect size under null will be positive in the raw circlin r, which is one of the important reasons for the transformation. To provide a more accurate interpretation, we will update Table 6 to describe the following strength levels of evidence: no effect (r < 0), null (r = 0), small (r = 0.1), moderate (r = 0.3), and large (r = 0.5).

iacore/featherwiki-deno-server

from

to

eLife Assessment

This manuscript describes a fundamental investigation of the functioning of Cas9 and in particular on how variant xCas9 expands DNA targeting ability by an increase-flexibility mechanism. The authors provide compelling evidence to support their mechanistic models and the relevance of flexibility and entropy in recognition. This work can be of interest to a broad community of structural biophysicists, computational biologists, chemists, and biochemists.

Joint Public Review:

Summary:

Hossain and coworkers investigate the mechanisms of recognition of xCas9, a variant of Cas9 with expanded targeting capability for DNA. They do so by using molecular simulations and combining different flavors of simulation techniques, ranging from long classical MD simulations, to enhanced sampling, to free energy calculations of affinity differences. Through this, the authors are able to develop a consistent model of expanded recognition based on the enhanced flexibility of the protein receptor.

Strengths:

The paper is solidly based on the ability of the authors to master molecular simulations of highly complex systems. In my opinion, this paper shows no major weaknesses. The simulations are carried out in a technically sound way. Comparative analyses of different systems provide valuable insights, even within the well-known limitations of MD. Plus, the authors further investigate why xCas9 exhibits improved recognition of the TGG PAM sequence compared to SpCas9 via well-tempered metadynamics simulations focusing on the binding of R1335 to the G3 nucleobase and the DNA backbone in both SpCas9 and xCas9. In this context, the authors provide a free-energy profiling that helps support their final model.

The implementation of FEP calculations to mimic directed evolution improvement of DNA binding is also interesting, original and well-conducted.

Overall, my assessment of this paper is that it represents a strong manuscript, competently designed and conducted, and highly valuable from a technical point of view.

Weaknesses:

To make their impact even more general, the authors may consider expanding their discussion on entropic binding to other recent cases that have been presented in the literature recently (such as e.g. the identification of small molecules for Abeta peptides, or the identification of "fuzzy" mechanisms of binding to protein HMGB1). The point on flexibility helping adaptability and expansion of functional properties is important, and should probably be given more evidence and more direct links with a wider picture.

القيمة الافتراضية انو ينبعت على نفس الصفحة

صفحة ال server الصفحة الي رح ارسلها ال form

للاشخاص كبار السن الي بعانو يضغطو ع الاشياء الصغيرة

زر ع شكل مربع

زر

القيمة الافتراضية عي text

القيمة الافتراضية لل input هي text

Dried bonga fillets, a delicacy appreciated by seafood enthusiasts around the world, are a staple in various cuisines. This guide will teach you how to cook dried bonga fillets to perfection while exploring the cultural significance and versatility of this delicious fish. Let’s dive into the best methods, tips, and recipes to make the most of your dried bonga fish from www.topropics.co.uk bonga fillets

Planning a trip that aligns with your passion for adventure, a romantic getaway, or the desire for solo exploration? jaazpackages.com adventure offers curated travel experiences designed to create memories of a lifetime. With seamless bookings, unique destinations, and custom itineraries, the platform caters to every travel enthusiast’s dream.

Seorang agen Kiss918 download fasilitas pelanggan yang baik mesti bisa menjawab pertanyaan bersama cepat, dan juga menambahkan solusi atas persoalan yang terjadi.

Discover why hiring MERN Stack developers is a smart choice for your next web project. Learn about its cost-effectiveness, seamless maintenance, and flexibility. https://adrians-newsletter-e9c134.beehiiv.com/p/top-reasons-to-hire-mern-stack-developers-for-your-next-web-project-800d

here is a comment

Vertrouwensbeginsel, gelijkheidsbeginsel, verbod op detournement de pouvoir, legaliteitsbeginsel, onschuldpresumptie

Art. 345 lid 3 Sv.

How to notarize your app for secure Apple Store approval. Explore its importance, benefits, challenges, and a step-by-step notarization process. https://www.cmarix.com/blog/notarize-on-ios-app-with-apple-store/

eLife Assessment

This study reports valuable findings that highlight the importance of data quality and data representation for ligand-based virtual screening experiments. The authors' claims are supported by solid evidence, although the conclusions have been inferred from only two datasets. The work would gain much impact if additional datasets were used. The main findings will be of interest to cheminformaticians and medicinal chemists working in QSAR modeling, and possibly in other areas related to machine learning.

Reviewer #1 (Public review):

Summary:

The work provides more evidence of the importance of data quality and representation for ligand-based virtual screening approaches. The authors have applied different machine learning (ML) algorithms and data representation using a new dataset of BRAF ligands. First, the authors evaluate the ML algorithms, and demonstrate that independently of the ML algorithm, predictive and robust models can be obtained in this BRAF dataset. Second, the authors investigate how the molecular representations can modify the prediction of the ML algorithm. They found that in this highly curated dataset the different molecule representations are adequate for the ML algorithms since almost all of them obtain high accuracy values, with Estate fingerprints obtaining the worst performing predictive models and ECFP6 fingerprints producing the best classificatory models. Third, the authors evaluate the performance of the models on subsets of different composition and size of the BRAF dataset. They found that given a finite number of active compounds, increasing the number of inactive compounds worsens the recall and accuracy. Finally, the authors analyze if the use of "less active" molecules affect the model's predictive performance using "less active" molecules taken from ChEMBl Database or using decoys from DUD-E. As results, they found that the accuracy of the model falls as the number of "less active" examples in the training dataset increases while the implementation of decoys in the training set generates results as good as the original models or even better in some cases. However, the use of decoys in the training set worsens the predictive power in the test sets that contain active and inactive molecules.

Strengths:

This is a highly relevant topic in medicinal chemistry and drug discovery. The manuscript is well-written, with a clear structure that facilitates easy reading, and it includes up-to-date references. The hypotheses are clearly presented and appropriately explored. The study provides valuable insights into the importance of deriving models from high-quality data, demonstrating that, when this condition is met, complex computational methods are not always necessary to achieve predictive models. Furthermore, the generated BRAF dataset offers a valuable resource for medicinal chemists working in ligand-based virtual screening.

Weaknesses:

While the work highlights the importance of using high-quality datasets to achieve better and more generalizable results, it does not present significant novelty, as the analysis of training data has been extensively studied in chemoinformatics and medicinal chemistry. Additionally, the inclusion of "AI" in the context of data-centric AI is somewhat unclear, given that the dataset curation is conducted manually, selecting active compounds based on IC50 values from ChEMBL and inactive compounds according to the authors' criteria.

Moreover, the conclusions are based on the analysis of only two high-quality datasets. To generalize these findings, it would be beneficial to extend the analysis to additional high-quality datasets (at least 10 datasets for a robust benchmarking exercise).

A key aspect that could be improved is the definition of an "inactive" compound, which remains unclear. In the manuscript, it is stated:

• "The inactives were carefully selected based on the fact that they have no known pharmacological activity against BRAF."<br /> Does the lack of BRAF activity data necessarily imply that these compounds are inactive?<br /> • "We define a compound as 'inactive' if there are no known pharmacological assays for the said compound on our target, BRAF."<br /> However, in the authors' response, they mention:<br /> • "We selected certain compounds that we felt could not possibly be active against BRAF, such as ligands for neurotransmitter receptors, as inactives."

Given that the definition of "inactive" is one of the most critical concepts in the study, I believe it should be clearly and consistently explained.

Lastly, while statistical comparison is not always common in machine learning, it would greatly enhance the value of this work, especially when comparing models with small differences in accuracy.

Reviewer #2 (Public review):

Summary:

The authors explored the importance of data quality and representation for ligand-based virtual screening approaches. I believe the results could be of potential benefit to the drug discovery community, especially to those scientists working in the field of machine learning applied to drug research. The in silico design is comprehensive and adequate for the proposed comparisons.

This manuscript by Chong A. et al describes that it is not necessary to resort to the use of sophisticated deep learning algorithms for virtual screening, since based on their results considering conventional ML may perform exceptionally well if feeded by the right data and molecular representations.

The article is interesting and well-written. The overview of the field and the warning about dataset composition are very well thought-out and should be of interest to a broad segment of the AI in drug discovery readership. This article further highlights some of the considerations that need to be taken into consideration for the implementation of data-centric AI for computer-aided drug design methods.

Strengths:

This study contributes significantly to the field of machine learning and data curation in drug discovery. The paper is, in general, well-written and structured. However, in my opinion, there are some suggestions regarding certain aspects of the data analyses.

Weaknesses:

The conclusions drawn in the study are based on the analysis of a two dataset. The authors chose BRAF as an example in this study, and expanded with BACE-1 dataset; however a benchmark with several targets would be suitable to evaluate reproducibility or transferability of the method. One concern could be the applicability of the method in other targets.

Reviewer #3 (Public review):

Summary:

The authors presented a data-centric ML approach for virtual ligand screening. They used BRAF as an example to demonstrate the predictive power of their approach.

Strengths:

The performance of predictive models in this study is superior (nearly perfect) with respect to exiting methods.

Comments on revisions:

In the revised manuscript, the presented approach has been robustly tested and can be very useful for ligand prediction.

Author response:

The following is the authors’ response to the original reviews.

We thank the Editors and reviewers for their candid evaluation of our work. While it was suggested that we should demonstrate the validity of our approach with maybe 10 different datasets but we felt that this would place an undue burden on our resources. Generally, it takes about 4 to 6 months for us to build a dataset and this does not include the time taken to train and test our AI models. This would mean that it would take us another 3 to 5 years to complete this research project if we chose to provide 10 different datasets. Publishing a research on one dataset is definitely not unheard of: for example, Subramanian et al. (2016) published their widely-cited benchmark dataset for just BACE1 inhibitors. However, we hoped that the additional work where we showed that we were able to improve the benchmark dataset for BACE1 inhibitors and achieve the same high level of predictive performance for this dataset would convince the readers (and reviewers) of the reproducibility of our approach. Furthermore, we also showed that our approach is robust and does not rely on a large volume of data to achieve this near-perfect accuracy. As can be seen in the Supplemental section, even our AI models trained on ONLY 250 BRAF actives and 250 inactives could achieve 96.3% accuracy! Logically, if the model is robust then we would expect the model to be reproducible. As such, we do not feel it is necessary for us to test our approach on 10 different datasets. 

It was also suggested that we expand this study to other types of molecular representations to give a better idea of generalizability. We would like to point out that we tested, in total, 55 single fingerprints and paired combinations. Our goal was to create an approach that could give superior performance for virtual screening and we believe that we have achieved this. Based on the results of our study, we are of the opinion that molecular representations do not, in general, have an oversized effect on AI virtual screening. Although it is important to be aware that certain molecular representations may give SLIGHTLY better performance but we can see that with the exception of the 79-bit E-State fingerprint (which could still achieve an impressive 85% accuracy for the SVM model), nearly all molecular fingerprints and paired combinations that we used were able to achieve an accuracy of above 97%. Therefore, we do not share the reviewers' concern that our approach may not be useful when applied with other types of molecular representations.

It is true that our work involved manual curation of the datasets but the goal of this paper is to lay down some  ground rules for the future development of a data-centric AI approach. Although manual curation is a routine practice in AI/ML, but it should be recognised that there is good manual curation and bad manual curation, and rules need to be established to ensure we have good manual curation. Without these rules, we would also not be able to establish and train a data-centric AI. All manual curation involves a level of subjectiveness but that subjectiveness comes from one's experience and domain knowledge of the field in which the AI is being applied. For example, in the case of this study, we relied on our knowledge and understanding of pharmacology to determine whether a compound is pharmacologically inactive or active. This may seem somewhat arbitrary to the uninitiated but it is anything but arbitrary. It is through careful thought and assessment of the chemical compounds that we choose these compounds for training the AI. Unfortunately, this sort of subjective assessment cannot be easily or completely explained but we do show where current practices have failed when building a dataset for training an AI for virtual screening.

Project Liberty Summit Day 1

eLife Assessment

This important study used an automated system to collect eggs laid over the course of multiple days by individual female Drosophila to successfully reveal a robust yet noisy circadian rhythm of egg-laying. Their results show that the neural control of this rhythm is entirely different from the one that controls locomotor activity rhythmicity. Preliminary connectome-based analyses provide evidence for connections between the relevant clock neurons and neurons involved in oviposition. The evidence provided is solid, although using an independent tool for targeted knockdown of clock genes and including the time series of representative individuals for all genotypes tested would help interpret the results.

Joint Public Review:

Riva et al uncovered the neural substrate underlying the oviposition rhythm in Drosophila melanogaster using a novel device that automates egg collection from individual mated females over the course of multiple days. By systematically knocking down the clock gene period in specific clock neurons the authors show that three cryptochrome (cry) positive dorso-lateral neurons (LNds) present in each hemisphere of the fly brain are critical to generating a female, sex-specific rhythm in oviposition. Interestingly, these neurons are not essential for freerunning locomotor activity. By contrast, the LNvs (lateral ventral neurons), which are essential for freerunning locomotor activity rhythmicity, were not involved in controlling the circadian rhythmicity of oviposition. Thus, this work has identified the first truly sex-specific circadian circuit in Drosophila. Using available Drosophila hemibrain connectome data they identify bidirectional connections between cry-expressing LNd and oviposition-related neurons.

Strengths:

This paper established a new semi-automatic device to register egg-laying activity, in Drosophila and found a specific role for a subset of clock neurons in the control of a female-specific circadian behavior. They also lay the groundwork for understanding how these neurons are connected to the neurons that control egg laying.

Weaknesses:

(1) Controls for the genetic background are incomplete, leaving open the possibility that the observed oviposition timing defects may be due to targeted knockdown of the period (per) gene but from the GAL4, Gal80, and UAS transgenes themselves. To resolve this issue the authors should determine the egg-laying rhythms of the relevant controls (GAL4/+, UAS-RNAi/+, etc); this only needs to be done for those genotypes that produced an arrhythmic egg-laying rhythm.

(2) Reliance on a single genetic tool to generate targeted disruption of clock function leaves the study vulnerable to associated false positive and false negative effects: a) The per RNAi transgene used may only cause partial knockdown of gene function, as suggested by the persistent rhythmicity observed when per RNAi was targeted to all clock neurons. This could indicate that the results in Fig 2C-H underestimate the phenotypes of targeted disruption of clock function. b) Use of a single per RNAi transgene makes it difficult to rule out that off-target effects contributed significantly to the observed phenotypes. We suggest that the authors repeat the critical experiments using a separate UAS-RNAi line (for period or for a different clock gene), or, better yet, use the dominant negative UAS-cycle transgene produced by the Hardin lab (https://doi.org/10.1038/22566).

(3) The egg-laying profiles obtained show clear damping/decaying trends which necessitates careful trend removal from the data to make any sense of the rhythm. Further, the detrending approach used by the authors is not tested for artefacts introduced by the 24h moving average used.

(4) According to the authors the oviposition device cannot sample at a resolution finer than 4 hours, which will compel any experimenter to record egg laying for longer durations to have a suitably long time series which could be useful for circadian analyses.

(5) Despite reducing the interference caused by manually measuring egg-laying, the rhythm does not improve the signal quality such that enough individual rhythmic flies could be included in the analysis methods used. The authors devise a workaround by combining both strongly and weakly rhythmic (LSpower > 0.2 but less than LSpower at p < 0.05) data series into an averaged time series, which is then tested for the presence of a 16-32h "circadian" rhythm. This approach loses valuable information about the phase and period present in the individual mated females, and instead assumes that all flies have a similar period and phase in their "signal" component while the distribution of the "noise" component varies amongst them. This assumption has not yet been tested rigorously and the evidence suggests a lot more variability in the inter-fly period for the egg-laying rhythm.

(6) This variability could also depend on the genotype being tested, as the authors themselves observe between their Canton-S and YW wild-type controls for which their egg-laying profiles show clearly different dynamics. Interestingly, the averaged records for these genotypes are not distinguishable but are reflected in the different proportions of rhythmic flies observed. Unfortunately, the authors also do not provide further data on these averaged profiles, as they did for the wild-type controls in Figure 1, when they discuss their clock circuit manipulations using perRNAi. These profiles could have been included in Supplementary figures, where they would have helped the reader decide for themselves what might have been the reason for the loss of power in the LS periodogram for some of these experimental lines.

(7) By selecting 'the best egg layers' for inclusion in the oviposition analyses an inadvertent bias may be introduced and the results of the assays may not be representative of the whole population.

(8) An approach that measures rhythmicity for groups of individual records rather than separate individual records is vulnerable to outliers in the data, such as the inclusion of a single anomalous individual record. Additionally, the number of individual records that are included in a group may become a somewhat arbitrary determinant for the observed level of rhythmicity. Therefore, the experimental data used to map the clock neurons responsible for oviposition rhythms would be more convincing if presented alongside individual fly statistics, in the same format as used for Figure 1.

(9) The features in the experimental periodogram data in Figures 3B and D are consistent with weakened complex rhythmicity rather than arrhythmicity. The inclusion of more individual records in the groups might have provided the added statistical power to demonstrate this. Graphs similar to those in 1G and 1I, might have better illustrated qualitative and quantitative aspects of the oviposition rhythms upon per knockdown via MB122B and Mai179; Pdf-Gal80.

Wider context:

The study of the neural basis of oviposition rhythms in Drosophila melanogaster can serve as a model for the analogous mechanisms in other animals. In particular, research in this area can have wider implications for the management of insects with societal impact such as pests, disease vectors, and pollinators. One key aspect of D. melanogaster oviposition that is not addressed here is its strong social modulation (see Bailly et al.. Curr Biol 33:2865-2877.e4. doi:10.1016/j.cub.2023.05.074). It is plausible that most natural oviposition events do not involve isolated individuals, but rather groups of flies. As oviposition is encouraged by aggregation pheromones (e.g., Dumenil et al., J Chem Ecol 2016 https://link.springer.com/article/10.1007/s10886-016-0681-3) its propensity changes upon the pre-conditioning of the oviposition substrates, which is a complication in assays of oviposition rhythms that periodically move the flies to fresh substrate.

Author response:

(1) Controls for the genetic background are incomplete, leaving open the possibility that the observed oviposition timing defects may be due to targeted knockdown of the period (per) gene but from the GAL4, Gal80, and UAS transgenes themselves. To resolve this issue the authors should determine the egg-laying rhythms of the relevant controls (GAL4/+, UAS-RNAi/+, etc); this only needs to be done for those genotypes that produced an arrhythmic egg-laying rhythm.

We agree with this objection, and in the corrected version we plan to provide the assessment of the egg laying rhythms for the missing GAL4 controls as recommended only for Figure 3.

(2) Reliance on a single genetic tool to generate targeted disruption of clock function leaves the study vulnerable to associated false positive and false negative effects: a) The per RNAi transgene used may only cause partial knockdown of gene function, as suggested by the persistent rhythmicity observed when per RNAi was targeted to all clock neurons. This could indicate that the results in Fig 2C-H underestimate the phenotypes of targeted disruption of clock function. b) Use of a single per RNAi transgene makes it difficult to rule out that off-target effects contributed significantly to the observed phenotypes. We suggest that the authors repeat the critical experiments using a separate UAS-RNAi line (for period or for a different clock gene), or, better yet, use the dominant negative UAS-cycle transgene produced by the Hardin lab (https://doi.org/10.1038/22566).

We have recently acquired mutant flies with a dominant negative-cycle transgene (UAS-cycDN, Tanoue et al. 2004), and we plan to repeat our experiments with these mutants, in order to confirm our results.

(3) The egg-laying profiles obtained show clear damping/decaying trends which necessitates careful trend removal from the data to make any sense of the rhythm. Further, the detrending approach used by the authors is not tested for artefacts introduced by the 24h moving average used.

In the revised version we will show that the detrending approach used does not introduce any artefacts. The analysis of numerical simulations with an aperiodic stochastic signal superposed to a decaying signal shows that the detrending method used does not result in a spurious periodic signal. Furthermore, we can show that when the underlying signal is rhythmic, the correct period is obtained even when the moving average is a few hours larger or smaller than 24 h.

(4) According to the authors the oviposition device cannot sample at a resolution finer than 4 hours, which will compel any experimenter to record egg laying for longer durations to have a suitably long time series which could be useful for circadian analyses.

We apologize for not being clear enough. The device can in principle sample at any desired resolution. Notice, however, that the variable we are analyzing (number of eggs laid by a single female) has only a few possible values, which is one of the features that render the assessment of rhythmicity a particularly difficult task. If egg laying is sampled more often (say, at 2 h intervals) more time points will be available, but the values available for each time point will be much less. We will show an example where we compare both rates (2h and 4h). Even though the 2h sampling reveals the rhythmicity of the time series, the significance of the peaks obtained is less than when sampling at 4h intervals. We have found that a 4h sampling seems to provide the best compromise between frequency of the sampling and discreteness of the variable.

On the other hand, it is important to stress that sampling frequency and longer durations are not very correlated (see e.g. Cohen et al. Journal of Theoretical Biology 314, pp 182 [2012]). It has been shown that the best way to make accurate predictions of the period of a rhythmic signal is to have a series spanning many cycles, irrespective of the sampling frequency. In other words, it is not true that with a 2h sampling it would be possible to analyze shorter series than with 4h sampling. Unfortunately, egg laying records are usually less than 5 cycles long, which is one of the reasons for the difficulties in the assessment of their rhythmicity.

(5) Despite reducing the interference caused by manually measuring egg-laying, the rhythm does not improve the signal quality such that enough individual rhythmic flies could be included in the analysis methods used. The authors devise a workaround by combining both strongly and weakly rhythmic (LSpower > 0.2 but less than LSpower at p < 0.05) data series into an averaged time series, which is then tested for the presence of a 16-32h "circadian" rhythm. This approach loses valuable information about the phase and period present in the individual mated females, and instead assumes that all flies have a similar period and phase in their "signal" component while the distribution of the "noise" component varies amongst them. This assumption has not yet been tested rigorously and the evidence suggests a lot more variability in the inter-fly period for the egg-laying rhythm.

The assumption is difficult to test rigorously, since for individual flies the records seem to be so noisy that no information can be extracted. As shown in the paper, it is even very difficult to assess the presence of rhythmicity at the individual level. We consider that the appearance of a rhythm after averaging several records shows the presence of this rhythm at the individual level. But it could be argued that the presence of rhythmicity in the average record could be due to only a few (or even a single) rhythmic individuals. In order to show that this is probably not the case, in the revised version we will show that, when the individuals that are rhythmic are left out, the average of the remaining flies still shows a rhythm (albeit a weaker one, as was to be expected).

Regarding our assumption that all flies have the “same” period, the results on Fig. 1 F cannot really rule out this possibility, because with so few cycles, the determination of the period is not very accurate (see e.g. Cohen et al. Journal of Theoretical Biology 314, pp 182 [2012]). In our case, the error for the period is related to the width of the corresponding peak in the periodogram, which is typically 4 hs. In any case, in the revised version we will try to show, by using numerical simulations, that when the individual periods are not the same, but are distributed approximately as in Fig 1F, the average series is still rhythmic with the correct period.

(6) This variability could also depend on the genotype being tested, as the authors themselves observe between their Canton-S and YW wild-type controls for which their egg-laying profiles show clearly different dynamics. Interestingly, the averaged records for these genotypes are not distinguishable but are reflected in the different proportions of rhythmic flies observed. Unfortunately, the authors also do not provide further data on these averaged profiles, as they did for the wild-type controls in Figure 1, when they discuss their clock circuit manipulations using perRNAi. These profiles could have been included in Supplementary figures, where they would have helped the reader decide for themselves what might have been the reason for the loss of power in the LS periodogram for some of these experimental lines.

Even though we think that the individual records are in general too noisy to be really informative, we will provide all the individual egg profiles in the Supplementary Material of the revised version, in order to let the reader, check this for herself/himself.

(7) By selecting 'the best egg layers' for inclusion in the oviposition analyses an inadvertent bias may be introduced and the results of the assays may not be representative of the whole population.

We agree that this may introduce some bias in the results. But in our opinion this bias is very difficult to avoid, since for females that lay very few eggs, rhythmicity can even be difficult to define (some females can spend a whole day without laying a single egg). On the other hand, even when the results may not be representative of the whole population, they would be representative of the flies that lay most of the eggs in a population, which seems to be very relevant in ecological terms.

(8) An approach that measures rhythmicity for groups of individual records rather than separate individual records is vulnerable to outliers in the data, such as the inclusion of a single anomalous individual record. Additionally, the number of individual records that are included in a group may become a somewhat arbitrary determinant for the observed level of rhythmicity. Therefore, the experimental data used to map the clock neurons responsible for oviposition rhythms would be more convincing if presented alongside individual fly statistics, in the same format as used for Figure 1.

The question of possible rhythmic outliers has been addressed above, in question 5, where we discuss why we think that such outliers are not “determinant for the observed level of rhythmicity”. As also mentioned above, even though we think that they are too noisy to be informative, we plan to include all individual profiles in the Supplementary Material.

(9) The features in the experimental periodogram data in Figures 3B and D are consistent with weakened complex rhythmicity rather than arrhythmicity. The inclusion of more individual records in the groups might have provided the added statistical power to demonstrate this. Graphs similar to those in 1G and 1I, might have better illustrated qualitative and quantitative aspects of the oviposition rhythms upon per knockdown via MB122B and Mai179; Pdf-Gal80.

We assume that the features mentioned refer to the appearance in the periodograms of two small peaks under the significance lines. We are aware that in the studies of the rhythmicity of locomotor activity such features are usually interpreted as “complex rhythms”, i.e. as evidence of the existence of two different mechanisms producing two different rhythms in the same individual. In our case, however, at least two other possibilities should be taken into account. Since the periodograms we show assess the rhythmicity of the average time series of several individuals, the two small peaks could correspond to the periods of two different subpopulations. Another possibility could be that such peaks are simply an artifact of the method in the analysis of time series that consist of very few cycles (as explained above) and also few points per cycle. A cursory examination of the individual profiles, that will be provided in the new version, do not seem to support any of the first two possibilities mentioned. On the other hand, we will show evidence that the analysis of series that are perfectly random sometimes result in periodograms with some small peaks.

eLife Assessment

This important study demonstrates the ability for high-throughput recording and categorization of unconstrained and stimulus-based behaviors across a very large population of marmosets (n = 120 animals across 36 family units). The authors implement an analytical approach to identify "outlier" behavior that could be key in the development of next-generation precision psychiatry. While the strength of evidence appears solid overall, many key methodological details are incomplete.

Reviewer #1 (Public review):

Summary:

The authors demonstrate a fully unsupervised, high throughput (meaning very low human interaction required) approach to quantifying marmoset behavior in unconstrained environments.

Strengths:

The authors provide an approach that is scalable, easy to implement at face value, and highly robust. Currently, most behavioral quantification approaches do not work well on marmosets, or the published examples that do look promising do not scale towards high throughput as demonstrated by the authors.

While marmosets can certainly be a useful translational research model devoid of free behavior quantification, the authors make a compelling point about how this approach can be useful in the study of treatments of emerging marmoset disease models.

Overall this is a very exhaustive manuscript that overcomes significant shortcomings in previous work and speaks highly to the use of marmosets for unconstrained behavioral and neural assessment.

Weaknesses:

Recording marmoset behavior with a 60Hz frame rate is a significant limitation to the approach which is hopefully easily alleviated in the future through better cameras/reconstruction pipelines. Marmosets (in the reviewers' experience) have a lot of motion energy above the 30Hz nyquist limit imposed by this system and are agile to a degree requiring higher frame rates.

The manuscript neglects recent approaches to non-human primate behavioral quantification from other groups that should be included. Simians are simians after all.

As a minor weakness, this reviewer would have liked to see code shared for the reviewers to evaluate, especially pertaining to the high throughput and robustness of the approach.

Reviewer #2 (Public review):

In this manuscript, Menegas et al. classify the "control" behavior of captive marmosets. They combine behavioral screening from video recordings with audio and neural recordings (from the striatum) to better define what can be considered a typical behavioral repertoire for captive marmoset monkeys. A range of analyses is presented, investigating various aspects of behavior, such as social interactions and the detection of atypical individuals.

The manuscript is compelling in many respects, especially due to the richness of the dataset and the breadth of analyses presented. However, a significant issue with the manuscript lies in its writing: the results are conveyed in an overly succinct and superficial manner, and the "Methods" section is nearly absent. Key concepts are often undefined, and the mathematical details underlying the figures are not explained, leaving readers to guess the authors' approach.

Another issue is the vague use of the term "natural behavior." All data presented here appear to have been collected in small cages with limited climbing opportunities and enrichment. Thus, the authors should refrain from using "natural" to describe these conditions.

Below, we elaborate further on the lack of methodological detail. Based on these issues, we believe the manuscript, in its current form, does not meet the scientific standards necessary for proper review. We strongly encourage the authors to undertake an extensive revision.

Major Revision Points:

The methods and results require significantly more detail. A scientific publication should provide readers with enough information to reproduce the study. Here, the detail level is far too low to fully understand, or reproduce, the study, and in many instances, readers are left to guess how the figure panels were produced. Below is a non-exhaustive list of examples illustrating these issues:

(1) "we temporarily placed horizontal cage dividers to reduce the total cage size during data collection": What were the resulting (and initial) cage dimensions?

(2) "After training the network, we hierarchically clustered the latent space": What is the latent space? Based on Figure 2a, it appears related to the network's recurrent layer, but this is not clarified in the text.

(3) Alpha and perplexity parameters: Please define these terms. Since these concepts appear fundamental, readers should not have to consult external references.

(4) "We then traced cluster identities across hierarchical levels": What are hierarchical levels?

(5) "To understand how the input time series data was weighed in the bottleneck layer of the model": What is the bottleneck layer?

(6) "we measured the average attention allocation to previous time points": The authors should define "attention allocation."

(7) "we compared each neuron's firing rate distribution to shuffled data based on the overall frequency of each behavior during the session": This description is insufficient to understand the analysis.

(8) "we hierarchically clustered neurons according to their firing rate enrichment maps": No mathematical explanation is provided for neuron clustering, nor is the concept of a "firing rate enrichment map" clarified.

(9) "Cluster 4 showed higher activity when neurons were 'alone' or 'active'": This is vague and uses unclear jargon (e.g., "neurons alone"). Additionally, no mathematical explanation is provided for assigning neuronal activity to behavioral states.

(10) Figure 3f, right-side panels: The analysis seems to involve cage mate positioning, yet no description is provided.

(11) "we used motion watches to measure activity across all hours": Are these motion-sensitive watches physically attached to the animals? The methodology should be described, including data analysis details.

This list could continue, but we trust the authors understand the point. There is a wealth of analyses and information in this study, but the descriptions are too superficial. We understand that fully describing each analysis may require significant rewriting, including supplementary figures, and will likely make the manuscript longer. This is entirely acceptable, as the ideas presented here are worth the added rigor.

"Natural behavior": Typically, the term "natural" suggests that the dataset reflects the range of behaviors exhibited by animals in the wild. Here, however, recordings were made in a small cage with limited climbing opportunities and enrichment. Under these conditions, it's hard to justify describing the behavior as "natural". In a project aimed at classifying the behavioral repertoire of marmoset monkeys and making this dataset accessible to other laboratories, it would be helpful to include more detailed information about the animals' housing conditions. This might include cage sizes, temperature, humidity, and details on food quantities, quality, and feeding times.

Correlation versus causation: In the section titled "Large-scale data collection reveals variability across days and correlation between cagemates," the authors conclude: "Overall, these results indicate that measurements of animals' behavioral traits depend heavily on their social environment." This interpretation seems incorrect. We know that animal behavior varies throughout the day, with activity peaks typically occurring in the morning and afternoon. Such factors, or other external influences, could induce correlations between animals that are not caused by social interactions.

Figure 4g: What are we intended to conclude from this analysis?

Figure 5: Please specify the type of calls analyzed. For example, did you analyze only long-distance calls (aka 'loud phees' or 'shrills')? In "We split the audio data into 5-minute (non-continuous) segments and found that the average call rate in these segments varied from 0 calls per minute to 60 calls per minute (Fig. 5d-e)," does the call rate refer to individual animals or the entire cage?

"This implies that a high rate of calls in a room can interrupt animals during social resting states and cause them to preferentially exhibit more active/attentive states." Does it? This could simply indicate that more active animals produce more calls.

"We recorded neural activity in the striatum because it is known to contain diverse signals related to movement and social interactions." While I understand that the authors intend to publish neural data separately, a brief discussion of the striatum's role here would be helpful.

Author response:

We would like to thank the editors and reviewers for taking the time to help improve our manuscript. We appreciate the feedback and will definitely increase the level of methodological detail in a revised submission.

Here is a brief summary of our plan to address the points raised by the reviewers. We will respond to the comments in a point-by-point manner when we resubmit a revised manuscript.

Reviewer 1

This reviewer raised a question about the 60 Hz frame rate for recording. We agree that increasing the number of cameras and frame rate would improve the tracking quality, but this would come at the cost of scalability. In the current study (and other concurrent studies in the lab), we recorded from 10-20 families simultaneously to try to sample the distribution of behavioral responses to stimuli observed in animals in our colony. This was only possible logistically because of the lightweight equipment design allowing us to record data from animals without large disruptions to their home-cage environment.

One strategy for acquiring higher-resolution data is to build a small number of enclosures that are fully surrounded by cameras, and to cycle animals through these enclosures (1). However, this strategy limits throughput by reducing the number of animals per day that can be studied. If the size and cost of cameras and computers decreases in the future, then this recording strategy will be scalable to the whole-colony level. For our current study and analysis, we are limited by the resolution of our dataset. We do believe that our data (although not a perfect 3d reconstruction or an extremely high frame rate) is sufficient to label behavioral states with high accuracy. We will add a figure to more clearly show that behavioral state data can be accurately inferred from this imperfect data, which has also been recently highlighted by other groups (2).

Additionally, with recent progress in the application of deep learning to animal pose tracking, new models can infer 3d pose dynamics from 2d data (3) and leverage spatiotemporal structure to clean up noisy data (4). We believe that other groups will be able to use these types of approaches to extract much more value from this dataset. So, in summary, we do understand the concern related to reconstruction quality and will 1) more clearly define the usefulness of our current models, 2) release our data and code so that others can build upon it or repurpose it, and 3) plan future experiments with higher camera count and frame rate as permitted by logistical constraints. 

Reviewer 2

This reviewer asked for an increased level of methodological detail. We will try to address this in a few ways:

(1) Code and data sharing. We believe that many of the questions related to the methodology will be best answered by sharing the data and code directly. Because there is a large amount of code associated with this manuscript, it is impractical to list every step and every parameter in the paper. Along with our revised manuscript, we will make our data and code publicly available. That said, we will improve our description of key parameters in the paper as the reviewer suggested.

(2) More detailed Methods section. The reviewer asked us to provide more methodological detail. We understand that this is currently a weakness of our manuscript, and we will focus on addressing it. For instance, the reviewer rightly points out that we did not describe the motion watches used to generate the data in Figure S7. We will address this.

(3) Simplify the manuscript. The paper currently has 22 figures, and further analysis could be done based on the results shown in any of them. For instance, this reviewer asked us to add a comparison across females and males (similar to our comparison of juveniles and adults). While we plan to add that analysis, we recognize that there are several figures/panels that are not closely related to our intended goal of describing the patterns we found in our large dataset. We will simplify the manuscript by removing some excess figures/panels and focus on describing the parts of the analysis that are crucial to our conclusions in greater detail.

(4) More careful language. This reviewer pointed out that there were some inaccuracies with our descriptive language. For instance, we used the term "natural" behavior to describe the behavior of animals in captivity, which may more accurately be described as their home-cage behavior. We will be more careful to align our language to the standard for the field. For instance, several studies refer to unrestrained behavior in a laboratory setting as "spontaneous" behavior rather than "natural" behavior (5). In our case, the data consists of both spontaneously occurring behavior and responses to a set of stimuli. We will make sure that the descriptions are more precise in the revised manuscript.

(1) Bala, P. C. et al. Automated markerless pose estimation in freely moving macaques with OpenMonkeyStudio. Nat Commun 11, (2020).

(2) Weinreb, C. et al. Keypoint-MoSeq: parsing behavior by linking point tracking to pose dynamics. bioRxiv (2023) doi:10.1101/2023.03.16.532307.

(3) Gosztolai, A. et al. LiftPose3D, a deep learning-based approach for transforming two-dimensional to three-dimensional poses in laboratory animals. Nat Methods 18, 975–981 (2021).

(4) Wu, A. et al. Deep Graph Pose: a semi-supervised deep graphical model for improved animal pose tracking. Adv Neural Inf Process Syst 33, 6040–6052 (2020).

(5) Levy, D. R. et al. Mouse spontaneous behavior reflects individual variation rather than estrous state. Curr Biol 33, 1358-1364.e4 (2023).

Знак номера у меня оторван, очень желательно не отрывать.

В конце строки.

Тут тоже некрасиво последние буквы у слов оторвались.

У меня некрасиво разбито для переноса, абы как.

Оторвано.

Нет линейки.

В конце строки.

Это выглядит очень плохо. Может, переделать в простой текст? Никакой наглядности не получается, удобства тоже.

После списка предложение практически дублирует это. Может быть, тут убрать?

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reply to the Reviewers

We thank all the reviewers for their time and their constructive criticism. We are encouraged by the overall positive and enthusiastic responses from the reviewers. We have taken all comments and suggestions seriously and revised the manuscript. These revisions include adding more explanation for the meaning of synaptic learning rules, language definitions, and model characteristics and limitations with more detailed figure legends. We are confident that we have addressed all the reviewer’s concerns by incorporating the reviewer’s suggestions into the revised manuscript. All changes are indicated in red font in the revised manuscript. The point-by-point response to all concerns raised by the reviewers follows. The line numbers indicated here refer to those in the revised manuscript.

Reviewer #1

Major comments:

1. Introduction, line 64 and further: An important omission in the introduction is that several studies have shown that sleep deprivation, i.e., extended wakefulness, results in a loss of spines in some brain regions such as the hippocampus, which is directly opposing the SHY hypothesis (for review, see Raven et al. Sleep Med Rev 39: 3-11, 2018).

Response:

We appreciate the reviewer’s valuable comment. Indeed, as correctly pointed out, several studies have reported synaptic weakening in the hippocampus and cortical regions following sleep deprivation, which appears to contradict the SHY.

We have incorporated this point into the introduction section (lines 64-67), adding several articles, including Raven et al., the reviewer suggested.

1. Introduction, line 85-87: A short explanation of what exactly the anti-Hebbian and anti-STDP rules are, is important here. It may seem obvious to the authors, but it is best to spell it out for the potential broad readership interested in this paper.

Response:

We appreciate the reviewer’s important suggestion.

Previous studies reported that Anti-Hebbian plasticity, which leads to depression when synapses are presented with correlated activity, serves critical functions in the discrimination of specific spike sequences in the cortico-striatal synapses (G. Vignoud et al., Commun. Biol, 2024) and the detection of novel stimuli in mormyrid fish (P. D. Roberts et al., Biol. Cybern, 2008; P. D. Roberts et al, Front. Comput. Neurosci, 2010).

We have added the explanations for Anti-Hebbian and Anti-STDP rules into the introduction section (lines 87-89).

1. Results, line 116, 129/130, 333, 395, 400, figure captions: Pleases explain what is meant with the terms 'pre-neuronal synapse' and 'post-neuronal synapses'.

Response:

We appreciate the reviewer’s advice. We have replaced ‘pre-neuronal synapse’ and ‘post-neuronal synapse’ with ‘pre-synaptic’、’post-synaptic’, respectively, for readability in the Results section (lines 118-119, 131-133, 368, 371, 432, 436 and 437) and Figure legends.

1. Results, line 121-124 say that synaptic efficacy became higher in sleep-like states than in wake-like states under Hebbian and STDP learning rules and opposite results were observed with anti-Hebbian and anti-STDP learning rules. While these relative differences are indeed visible in Figure 1H, the figure also suggests that synaptic efficacy during sleep was largely independent of the average firing frequency. In other words, synaptic efficacy seems to be dependent on firing frequency only during wakefulness. Is that correct?

Response:

The reviewer raised an important point. As shown in Fig. 1H, synaptic efficacy during sleep appears to be largely independent of mean firing rates. Here, the firing rates were adjusted by varying Down-state durations. Regarding the relationship between firing patterns and synaptic efficacy, synaptic efficacy is influenced not only by firing frequency but also by how firing patterns are generated. When firing rates are adjusted by changing ISI, synaptic efficacy during sleep also increases with higher firing rates as wake-like patterns (Fig. 5). In Fig. 2D and E, we demonstrated that the synaptic efficacy during sleep becomes higher than during wakefulness regardless of whether the spike patterns were generated with changing Down-state duration or ISI, assuming the same mean firing rates during the sleep-like and wake-like states. We have clarified this point by adding the explanation in the Discussion section (lines 318-323).

1. Results, line 199 and down model the effect of differences in mean firing rate between sleep and waking, which is a crucial addition and more realistic approach for most brain regions that have lower average firing rates during sleep. It is interesting that in this case the relative effects of sleep and wakefulness can change direction, depending on the average firing frequency. Would the authors argue that this may even result in opposite effects in different brain regions after waking or sleep deprivation?

Response:

We appreciate the reviewer raising the interesting point. Our model predicted that the direction of synaptic changes depends on learning rules and firing rates. This prediction indicated that different brain regions may exhibit synaptic changes even in opposite directions after prolonged wakefulness or sleep deprivation. For example, under Hebbian and STDP, our model predicted that brain regions with firing rates increased during wakefulness or sleep deprivation compared to sleep would follow SHY, while brain regions where firing rates remain unchanged or decreased compared to sleep would follow WISE. The experimental validation of these predictions, focusing on brain regions with different activation states during wakefulness, is an interesting future work. We have clarified this point into the Discussion section (lines 260-262).

1. Figure 1: The caption needs more details to help understand the different panels. some work. (B) What is a post-neuronal synapse? (C) How exactly is synaptic efficacy defined? (E) Not totally clear what the colored top panels represent.

Response:

We sincerely appreciate the reviewer’s thoughtful feedback. We agreed that Figure 1 required a more thorough explanation. In response, we have expanded the figure legend to provide more detailed information for readers to easily understand.

1. Figure 5B. Since this appears to be a graphical abstract and unified framework for all the modelled parameters and learning rules, should this not be a separate figure?

__Response: __We thank the reviewer for the helpful suggestion. We have renumbered Figure 5B as Figure 6.

1. Figures captions: The information provided in the figure captions is in many cases quite minimal and does not reflect the complexity of some of the figure panels. This often makes it hard for a reader to extract all the relevant information without thumbing back and forth between figures, captions and main text. I strongly suggest to add more detail to the figure captions to make them more stand-alone and self-explanatory.

__Response: __We sincerely appreciate the reviewer’s significant feedback. We have added detailed explanations in the figure legends, including Supplementary Figures, for readers to understand easily.

Reviewer #2

Major comments:

1. I am not qualified to review this manuscript because I'm not sufficiently familiar with the type of modelling performed here and the specific use of terms. For example, without providing any explanation, I cannot reconstruct whether the estimates of synaptic efficacy (eq.1) are valid and applicable to the questions asked. I do have 2 general comments. I do find the premise of WISE intriguing and understand the attractiveness of the idea of opposing 'WISE' to SHY. Nevertheless, SHY is a theory that does not discount the occurrence of synaptic strengthening during sleep. It is rather that during sleep there is a net down-scaling. Therefore, the assumptions, as they are presented here, are confusing the issue.

Response:

We are deeply grateful that the reviewer found WISE intriguing and appreciate the insightful comment. We agree that SHY does not deny the occurrence of synaptic strengthening during sleep, but rather proposes a net downward scaling under the assumption of the overall synaptic homeostasis. In the present study, we assumed that SHY describes a net downscaling during sleep (and does not deny the occurrence of synaptic strengthening of some synapses during sleep) while WISE describes a net upscaling during sleep (and does not deny the occurrence of synaptic weakening of some synapses during sleep). Both SHY and WISE fulfill synaptic homeostasis. For example, SHY upscales synaptic strength during wakefulness and downscales during sleep to achieve synaptic homeostasis. On the other hand, WISE upscales synaptic strength during sleep and downscales during wakefulness __to achieve synaptic homeostasis. Our study demonstrated that __WISE is compatible with Hebbian and STDP learning rules when average neuron firing frequency is similar between sleep and wakefulness, and SHY is not compatible with Hebbian and STDP learning rules, but rather compatible with Anti-Hebbian and __Anti-STDP __learning rules.

We agreed with the reviewer that the lack of an explicit definition of SHY and WISE in the context of the present study could cause confusion for readers. Therefore, we have added a sentence to clarify SHY and WISE in the present study in the first paragraph of the Results section (lines 127-128), specifically defining them in terms of relative net synaptic changes within local neural network.

1. SHY was, in part, inspired by a type of plasticity that is not considered here, namely synaptic homeostasis. Would adding such a mechanism to the model alter any of the predictions?"

__Response: __

We appreciate the reviewer raising an important point on synaptic homeostasis. In this study, we did not explicitly include synaptic homeostasis in the preposition but consider synaptic homeostasis in the definitions of SHY and WISE. For example, we assume that SHY upscales synaptic strength during wakefulness and downscales during sleep to achieve synaptic homeostasis while WISE upscales synaptic strength during sleep and downscales during wakefulness to achieve synaptic homeostasis. Importantly, since both SHY and WISE can achieve synaptic homeostasis, there are two types of synaptic homeostasis. In our study, WISE-type synaptic homeostasis is compatible with Hebbian and STDP learning rules when average neuron firing frequency is similar between sleep and wakefulness, and SHY-type synaptic homeostasis is compatible with Anti-Hebbian and __Anti-STDP __learning rules. Since our studies already consider two types of synaptic homeostasis, adding the further mechanism of synaptic homeostasis in the preposition would not alter our predictions. We described these points in the Model characteristics and limitations part in the Discussion section (lines 332-339).

Reviewer #3

Major comments:

1. This is a well-written manuscript that is easily to follow and amply illustrated. The study seems very exciting but unfortunately I am not a mathematician so I cannot attest to the veracity or originality of the model. Assuming it is robust, it does appear to account for a quite a few anomalies (and inaccuracies depicted in textbooks). It would be helpful to discuss the limitations of other models that have been suggested to synaptic functions of sleep.

__Response: __

We appreciated the reviewer’s constructive suggestions. Some computational studies have investigated synaptic changes in neural networks under STDP protocols using Ca2+-based plasticity models (M. Graupner et al., PNAS, 2012; G. Chindemi et al., Nat. Commun, 2022), while other studies have examined how SWO affects synaptic plasticity under STDP conditions (T. Tadros et al., J.Neurosci, 2022). However, these previous studies were limited to a single synaptic learning rule or firing pattern. Our study is the first to comprehensively investigate synaptic dynamics during the sleep-wake cycle by integrating a Ca2+-based plasticity model to represent various types of synaptic learning rules and various simulated sleep-wake firing patterns.

We have added the sentences related to the reviewer’s comments in the Model characteristics and limitations part in the Discussion section (lines 306-312).

1. Much of the neurophysiological data comes from recordings in rodents, so the model is simulating rat EEG signatures-how readily applicable is this to the human condition? Indeed, how readily can they compare between mouse and rat? The authors should expand on this in the discussion section.

Another potential weakness or limitation is the unanswered question of the model can account for sleep/wake changes in other areas of the cortex or thalamus etc.

Does this model apply equally to males and females?

__Response: __

We appreciate the reviewer for raising this significant point. As the reviewer pointed out, we generated firing patterns using parameters derived from rat firing patterns (B. O. Watson et al., Neuron, 2016), such as ISI, Up-state duration, and Down-state duration. While we started our simulations from those parameter sets, we tested a range of different values for each parameter and found consistent results (detailed in Supplementary Materials, Generation of sleep and wake-like firing patterns). The ranges of Up-state and Down-state durations during SWO in mice, rats, and cats are approximately 100-500 milliseconds (M. Steriade et al., J. Neurophysiol, 2001; V. Crunelli et al., Pflugers Arch, 2012), while in humans, Up-state durations range from 250-1000 milliseconds (B. A. Riedner et al., Sleep, 2007), all of which fall within the ranges examined in Figs. 2 D and E. Similarly, wake-state ISI across various species typically range from 2-100 milliseconds (M. Steriade et al., J. Neurophysiol, 2001; G. Maimon et al., Neuron, 2009), mostly within the scope covered in Fig. 2E. Therefore, we suppose our finding in the present study captured universal aspects of synaptic dynamic in the sleep and wake cycles regardless of species, brain region, or sex.

We have added the description in the Model characteristics and limitations part in the Discussion section (lines 312-331).

Minor comments:

Minor typo: ref. 24 is missing page and volume numbers.

__Response: __

Thank you for pointing out this typo. We corrected this by adding the page and volume numbers in Ref. 28 in the revised manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Understanding the functions of sleep has and remains a key question in neuroscience. A popular hypothesis is that sleep is fundamental to learning and memory and that this can be detected and measured at the level of neural networks and connections as increased synaptic weights across waking states and reduced synaptic weights or depression during sleep states. However, there are many contradictions in the literature and while it is accepted that sleep plays a role in memory consolidation, the molecular/cellular basis of this is far from clear. As considerable experimental data on synaptic function have been collected during sleep and wake states, here the authors turned to modelling how manipulating the rules of synaptic plasticity can illuminate the problem. In this manuscript, the authors report the outcomes of these simulations neuronal oscillations, firing, and synaptic plasticity across sleep-like and wake-like neural states. They report that their simulations can account for several irregularities and highlight differential involvement of spike-firing dependent plasticity (STDP) and anti-STDP in wake and NREM sleep. In particular they note that under Hebbian and STDP rules, firing patterns associated with wake lead to decreased synaptic weights, while sleep-like patterns bolster synaptic weights and collectively they describe this tendency as WISE. They also note that under Anti-Hebbian and Anti-STDP rules, synaptic depression was observed under NREM. The chief strength of this study is shows how simulation can aid in bringing together disparate observations into a well-worked study space.

This is a well-written manuscript that is easily to follow and amply illustrated. The study seems very exciting but unfortunately I am not a mathematician so I cannot attest to the veracity or originality of the model. Assuming it is robust, it does appear to account for a quite a few anomalies (and inaccuracies depicted in textbooks). It would be helpful to discuss the limitations of other models that have been suggested to synaptic functions of sleep.

Much of the neurophysiological data comes from recordings in rodents, so the model is simulating rat EEG signatures-how readily applicable is this to the human condition? Indeed how readily can they compare between mouse and rat? The authors should expand on this in the discussion section.

Another potential weakness or limitation is the unanswered question of the model can account for sleep/wake changes in other areas of the cortex or thalamus etc.

Does this model apply equally to males and females? Minor typo: ref. 24 is missing page and volume numbers.

Significance

As noted above, there are discrepancies in the literature regarding synaptic plasticity and its mechanisms across the sleep-wake cycle. This model appears to answer some of the reasons for these and provides a framework for further experimental research to interrogate these mechanisms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

I am not qualified to review this manuscript because I'm not sufficiently familiar with the type of modelling performed here and the specific use of terms. For example, without providing any explanation, I cannot reconstruct whether the estimates of synaptic efficacy (eq.1) are valid and applicable to the questions asked.

I do have 2 general comments. I do find the premise of WISE intriguing and understand the attractiveness of the idea of opposing 'WISE' to SHY. Nevertheless, SHY is a theory that does not discount the occurrence of synaptic strengthening during sleep. It is rather that during sleep there is a net down-scaling. Therefore, the assumptions, as they are presented here, are confusing the issue. SHY was, in part, inspired by a type of plasticity that is not considered here, namely synaptic homeostasis. Would adding such a mechanism to the model alter any of the predictions?"

Significance

I do find the premise of WISE intriguing and understand the attractiveness of the idea of opposing 'WISE' to SHY. Nevertheless, SHY is a theory that does not discount the occurrence of synaptic strengthening during sleep. It is rather that during sleep there is a net down-scaling. Therefore, the assumptions, as they are presented here, are confusing the issue. SHY was, in part, inspired by a type of plasticity that is not considered here, namely synaptic homeostasis. Would adding such a mechanism to the model alter any of the predictions?"

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

While the function of sleep is still an unresolved mystery, some of the most influential theories propose that sleep serves a crucial role in regulating neuronal plasticity and synaptic strength. However, the exact way how synaptic strength is affected by sleep and impaired by sleep deprivation is a topic of much controversy and ongoing debate in the field of sleep research (SHY vs WISE). Using computation models, the manuscript illustrates that opposite effects of sleep on synaptic efficacy can be found, depending on the firing patterns and learning rules. Specifically, sleep promotes synaptic strength and efficacy under Hebbian and spike-timing dependent plasticity rules and it resulted in synaptic depression under anti-Hebbian and anti-STDP rules.

Major comments:

Introduction, line 64 and further: An important omission in the introduction is that several studies have shown that sleep deprivation, i.e., extended wakefulness, results in a loss of spines in some brain regions such as the hippocampus, which is directly opposing the SHY hypothesis (for review, see Raven et al. Sleep Med Rev 39: 3-11, 2018).

Introduction, line 85-87: A short explanation of what exactly the anti-Hebbian and anti-STDP rules are, is important here. It may seem obvious to the authors, but it is best to spell it out for the potential broad readership interested in this paper.

Results, line 116, 129/130, 333, 395, 400, figure captions: Pleases explain what is meant with the terms 'pre-neuronal synapse' and 'post-neuronal synapses'.

Results, line 121-124 say that synaptic efficacy became higher in sleep-like states than in wake-like states under Hebbian and STDP learning rules and opposite results were observed with anti-Hebbian and anti-STDP learning rules. While these relative differences are indeed visible in Figure 1H, the figure also suggests that synaptic efficacy during sleep was largely independent of the average firing frequency. In other words, synaptic efficacy seems to be dependent on firing frequency only during wakefulness. Is that correct?

Results, line 199 and down model the effect of differences in mean firing rate between sleep and waking, which is a crucial addition and more realistic approach for most brain regions that have lower average firing rates during sleep. It is interesting that in this case the relative effects of sleep and wakefulness can change direction, depending on the average firing frequency. Would the authors argue that this may even result in opposite effects in different brain regions after waking or sleep deprivation?

Figure 1: The caption needs more details to help understand the different panels. some work. (B) What is a post-neuronal synapse? (C) How exactly is synaptic efficacy defined? (E) Not totally clear what the colored top panels represent.

Figure 5B. Since this appears to be a graphical abstract and unified framework for all the modelled parameters and learning rules, should this not be a separate figure?

Figures captions: The information provided in the figure captions is in many cases quite minimal and does not reflect the complexity of some of the figure panels. This often makes it hard for a reader to extract all the relevant information without thumbing back and forth between figures, captions and main text. I strongly suggest to add more detail to the figure captions to make them more stand-alone and self-explanatory.

Significance

This paper addresses a major controversy in the field of sleep research: does sleep strengthen neuronal connections in the brain or does it downscale and weaken them (Raven et al. 2018)? Using computation models, the current paper shows that both options are possible and it does an admirable job in bridging the different views on sleep and synaptic strength. As such, the conceptual value of this paper can hardly be overestimated and provides an important framework for future experimental studies.

This paper is of interest for most everybody interested in sleep and brain function, as well as neuroscientist with a broader interest in brain plasticity.

eLife Assessment

MGPfactXMBD is a novel computational method for investigating cell evolutionary trajectory for scRNA-seq samples. It is important, with several potential future applications. The authors benchmarked this method using synthetic and real-world samples and showed superior performance for some of the tasks in cell trajectory analysis compared to other methods with compelling evidence.

Reviewer #1 (Public review):

Summary:

Ren et al developed a novel computational method to investigate cell evolutionary trajectory for scRNA-seq samples. This method, MGPfact, estimates pseudotime and potential branches in the evolutionary path through explicitly modeling the bifurcations in a Gaussian process. They benchmarked this method using synthetic as well as real world samples and showed superior performance for some of the tasks in cell trajectory analysis. They further demonstrated the utilities of MGPfact using single cell RNA-seq samples derived from microglia or T cells and showed that it can accurately identify the differentiation timepoint and uncover biologically relevant gene signatures.

Strengths:

Overall I think this is a useful new tool that could deliver novel insights for the large body of scRNA-seq data generated in the public domain. The manuscript is written is a logical way and most parts of the method are well described.

Comments on revisions:

In this revision, the authors have sufficiently addressed all of my concerns. I don't have any follow-up comments.

Reviewer #2 (Public review):

Summary of the manuscript:

Authors present MGPfactXMBD, a novel model-based manifold-learning framework designed to address the challenges of interpreting complex cellular state spaces from single-cell RNA sequences. To overcome current limitations, MGPfactXMBD factorizes complex development trajectories into independent bifurcation processes of gene sets, enabling trajectory inference based on relevant features. As a result, it is expected that the method provides a deeper understanding of the biological processes underlying cellular trajectories and their potential determinants.

MGPfactXMBD was tested across 239 datasets, and the method demonstrated similar to slightly superior performance in key quality-control metrics to state-of-the-art methods. When applied to case studies, MGPfactXMBD successfully identified critical pathways and cell types in microglia development, validating experimentally identified regulons and markers. Additionally, it uncovered evolutionary trajectories of tumor-associated CD8+ T cells, revealing new subtypes with gene expression signatures that predict responses to immune checkpoint inhibitors in independent cohorts.

Overall, MGPfactXMBD represents a relevant tool in manifold-learning for scRNA-seq data, enabling feature selection for specific biological processes and enhancing our understanding of the biological determinants of cell fate.

Summary of the outcome:

The novel method addresses core state-of-the-art questions in biology related to trajectory identification. The design and the case studies are of relevance.

Comments on revisions:

The authors have addressed all my previous comments to satisfaction.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1:

Comment#1: Ren et al developed a novel computational method to investigate cell evolutionary trajectory for scRNA-seq samples. This method, MGPfact, estimates pseudotime and potential branches in the evolutionary path by explicitly modeling the bifurcations in a Gaussian process. They benchmarked this method using synthetic as well as real-world samples and showed superior performance for some of the tasks in cell trajectory analysis. They further demonstrated the utilities of MGPfact using single-cell RNA-seq samples derived from microglia or T cells and showed that it can accurately identify the differentiation timepoint and uncover biologically relevant gene signatures. Overall I think this is a useful new tool that could deliver novel insights for the large body of scRNA-seq data generated in the public domain. The manuscript is written in a logical way and most parts of the method are well described.

Thank you for reviewing our manuscript and for your positive feedback on MGPfact. We are pleased that you find it useful for identifying differentiation timepoints and uncovering gene signatures. We will continue to refine MGPfact and explore its applications across diverse datasets. Your insights are invaluable, and we appreciate your support.

Comment#2: Some parts of the methods are not clear. It should be outlined in detail how pseudo time T is updated in Methods. It is currently unclear either in the description or Algorithm 1.

Thanks to the reviewers' comments. We've added a description of how pseudotime T is obtained between lines 138 and 147 in the article. In brief, the pseudotime of MGPfact is inferred through Gaussian process regression on the downsampled single-cell transcriptomic data. Specifically, T is treated as a continuous variable representing the progression of cells through the differentiation process. We describe the relationship between pseudotime and expression data using the formula:

Where f(T) is a Gaussian Process (GP) with covariance matrix S, and represents the error term. The Gaussian process is defined as:

Where is the variance set to 1e-6.

During inference, we update the pseudotime by maximizing the posterior likelihood. Specifically, the posterior distribution of pseudotime T can be represented as:

Where is the likelihood function of the observed data Y*, and is the prior distribution of the Gaussian process. This posterior distribution integrates the observed data with model priors, enabling inference of pseudotime and trajectory simultaneously. Due to the high autocorrelation of  in the posterior distribution, we use Adaptive Metropolis within Gibbs (AMWG) sampling (Roberts and Rosenthal, 2009; Tierney, 1994). Other parameters are estimated using the more efficient SLICE sampling technique (Neal, 2003).

Comment#3: There should be a brief description in the main text of how synthetic data were generated, under what hypothesis, and specifically how bifurcation is embedded in the simulation.

Thank you for the reviewers' comments. We have added descriptions regarding the synthetic dataset in the methods section. The revised content is from line 487 to 493:

The synthetic datasets were generated using four simulators: dyngen (Saelens et al., 2019), dyntoy (Saelens et al., 2019), PROSSTT (Papadopoulos et al., 2019), and Splatter (Zappia et al., 2017), each modeling different trajectory topologies such as linear, branching, and cyclic. Splatter simulates branching events by setting expression states and transition probabilities, dyntoy generates random expression gradients to reflect dynamic changes, and dyngen focuses on complex branching structures within gene regulatory networks.

Comment#4: Please explain what the abbreviations mean at their first occurrence.

We appreciate the reviewers' feedback. We have thoroughly reviewed the entire manuscript and made sure that all abbreviations have had their full forms provided upon their first occurrence.

Comment#5: In the benchmark analysis (Figures 2/3), it would be helpful to include a few trajectory plots of the real-world data to visualize the results and to evaluate the accuracy.

We appreciate the reviewer's feedback. To more clearly demonstrate the performance of MGPfact, we selected three representative cases from the dataset for visual comparison. These cases represent different types of trajectory structures: linear, bifurcation, and multifurcation. The revised content is between line 220 and 226.

As shown in Supplementary Fig. 5, it is evident that MGPfact excels in capturing main developmental paths and identifying key bifurcation points. In the linear trajectory structure, MGPfact accurately predicted the linear structure without bifurcation events, showing high consistency with the ground truth (overall\=0.871). In the bifurcation trajectory structure, MGPfact accurately captured the main bifurcation event ( ). In the multifurcation trajectory structure, although MGPfact predicted only one bifurcation point, its overall structure remains close to the ground truth, as evidenced by its high overall score (overall\=0.566). Overall, MGPfact demonstrates adaptability and accuracy in reconstructing various types of trajectory structures.

Comment#6: It is not clear how this method selects important genes/features at bifurcation. This should be elaborated on in the main text.

Thanks to the reviewers' comments. To enhance understanding, we've added detailed descriptions of gene selection in the main text and appendix, specifically from lines 150 to 161. In brief, MGPfact employs a Gaussian process mixture model to infer cell fate trajectories and identify independent branching events. We calculate load matrices using formulas 1 and 14 to assess each gene's contribution to the trajectories. Genes with an absolute weight greater than 0.05 are considered predominant in specific branching processes. Subsequently, SCENIC (Aibar et al., 2017; Bravo González-Blas et al., 2023) analysis was conducted to further infer the underlying regulons and annotate the biological processes of these genes.

Comment#7: It is not clear how survival analysis was performed in Figure 5. Specifically, were critical confounders, such as age, clinical stage, and tumor purity controlled?

To evaluate the predictive and prognostic impacts of the selected genes, we utilized the Cox multivariate regression model, where the effects of relevant covariates, including age, clinical stage, and tumor purity, were adjusted. We then conducted the Kaplan-Meier survival analysis again to ensure the reliability of the results. The revisions mainly include the following sections:

(1) We modified the description of adjusting for confounding factors in the survival analysis, from line 637 to 640:

To adjust for possible confounding effects, the relevant clinical features including age, sex and tumor stage were used as covariates. The Cox regression model was implemented using R-4.2 package “survival”. And we generated Kaplan-Meier survival curves based on different classifiers to illustrate differences in survival time and report the statistical significance based on Log-rank test.

(2) We updated the images in the main text regarding the survival analysis, including Fig. 5a-b, Fig. 6c, and Supplementary Fig. 8e.

Comment#8: I recommend that the authors perform some sort of 'robustness' analysis for the consensus tree built from the bifurcation Gaussian process. For example, subsample 80% of the cells to see if the bifurcations are similar between each bootstrap.

We appreciate the reviewers' feedback. We performed a robustness analysis of the consensus tree using 100 training datasets. This involved sampling the original data at different proportions, and then calculating the topological similarity between the consensus trajectory predictions of MGPfact and those without sampling, using the Hamming-Ipsen-Mikhailov ( ) metric. A higher score indicates greater robustness. The relevant figure is in Supplementary Fig. 4, and the description is in the main text from line 177 to 182.

The results indicate that the consensus trajectory predictions based on various sampling proportions of the original data maintain a high topological similarity with the unsampled results (HIM_mean=0.686). This demonstrates MGPfact’s robustness and generalizability under different data conditions, hence the capability of capturing bifurcative processes in the cells’ trajectory.

Reviewer #2:

Comment#1: The authors present MGPfact^XMBD, a novel model-based manifold-learning framework designed to address the challenges of interpreting complex cellular state spaces from single-cell RNA sequences. To overcome current limitations, MGPfact^XMBD factorizes complex development trajectories into independent bifurcation processes of gene sets, enabling trajectory inference based on relevant features. As a result, it is expected that the method provides a deeper understanding of the biological processes underlying cellular trajectories and their potential determinants. MGPfact^XMBD was tested across 239 datasets, and the method demonstrated similar to slightly superior performance in key quality-control metrics to state-of-the-art methods. When applied to case studies, MGPfact^XMBD successfully identified critical pathways and cell types in microglia development, validating experimentally identified regulons and markers. Additionally, it uncovered evolutionary trajectories of tumor-associated CD8+ T cells, revealing new subtypes with gene expression signatures that predict responses to immune checkpoint inhibitors in independent cohorts. Overall, MGPfact^XMBD represents a relevant tool in manifold learning for scRNA-seq data, enabling feature selection for specific biological processes and enhancing our understanding of the biological determinants of cell fate.

Thank you for your thoughtful review of our manuscript. We are thrilled to hear that you find MGPfact^XMBD beneficial for exploring cellular evolutionary paths in scRNA-seq data. Your insights are invaluable, and we look forward to incorporating them to further enrich our study. Thank you once again for your support and constructive feedback.

Comment#2: How the methods compare with existing Deep Learning based approaches such as TIGON is a question mark. If a comparison would be possible, it should be conducted; if not, it should be clarified why.

We appreciate the reviewer's comments. We have added a comparison with the sctour (Li, 2023) and TIGON methods (Sha, 2024).

It is important to note that the encapsulation and comparison of MGPfact are based on traditional differentiation trajectory construction. Saelens et al. established a systematic evaluation framework that categorizes differentiation trajectory structures into topological subtypes such as linear, bifurcation, multifurcation, graph, and tree, focusing on identifying branching structures in the cell differentiation process (Saelens et al., 2019). The sctour and TIGON methods mentioned by the reviewer are primarily used for estimating RNA velocity, focusing on continuous temporal evolution rather than explicit branching structures, and do not explicitly model branches. Therefore, we considered the predictions of these two methods as linear trajectories and compared them with MGPfact. While scTour explicitly estimates pseudotime, TIGON uses the concept of "growth," which is analogous to pseudotime, so we made the necessary adaptations.

Author response image 1 show that within this framework, compared to scTour (overall_mean=0.448) and TIGON (overall_mean=0.263), MGPfact still maintains a relatively high standard (overall_mean=0.534). This indicates that MGPfact has a significant advantage in accurately capturing branching structures in cell differentiation, especially in applications where explicit modeling of branches is required.

Author response image 1.

Comparison of MGPfact with scTour and TIGON in trajectory inference performance across 239 test datasets. a. Overall scores；b.F1_branches； c.HIM；e. wcor_features. All results are color-coded based on the trajectory types, with the black line representing the mean value. The “Overall” assessment is calculated as the geometric mean of all four metrics.

Comment#3: Missing Methods:

- The paper lacks a discussion of Deep Learning approaches for bifurcation analysis. e.g. scTour, Tigon.

- I am missing comments on methods such CellRank, and alternative approaches to delineate a trajectory.

We thank the reviewer for these comments.

(1) As mentioned in response to Comments#2, the scTour and TIGON methods are primarily used for estimating RNA velocity, focusing on continuous temporal evolution rather than explicit branching structures, and they do not explicitly model branches. We consider the predictions of these two methods as linear trajectories and compare them with MGPfact. The relevant description and discussion have been addressed in the response.

(2) We have added a description of RNA velocity estimation methods (scTour, TIGON, CellRank) in the introduction section. The revised content is from line 66 to 71:

Moreover, recent studies based on RNA velocity has provided insights into cell state transitions. These methods measure RNA synthesis and degradation rates based on the abundance of spliced and unspliced mRNA, such as CellRank (Lange et al., 2022). Nevertheless, current RNA velocity analyses are still unable to resolve cell-fates with complex branching trajectory. Deep learning methods such as scTour (Li, 2023) and TIGON (Sha, 2024) circumvent some of these limitations, offering continuous state assumptions or requiring prior cell sampling information.

Comment#4: Impact of MURP:

The rationale for using MURP is well-founded, especially for trajectory definition. However, its impact on the final results needs evaluation.

How does the algorithm compare with a random subselection of cells or the entire cell set?

Thank you for the comments. We fully agree that MURP is crucial in trajectory prediction. As a downsampling method, MURP is specifically designed to address noise issues in single-cell data by dividing the data into several subsets, thereby maximizing noise reduction while preserving the main structure of biological variation (Ren et al., 2022). In MGPfact, MURP typically reduces the data to fewer than 100 downsampled points, preserving the core biological structure while lowering computational complexity. To assess MURP's impact, we conducted experiments by randomly selecting 20, 40, 60, 80, and 100 cells for trajectory inference. These results were mapped back to the original data using the KNN graph structure for final predictions, which were then compared with the MURP downsampling results. Supplementary results can be found in Supplementary Fig. 3, with additional descriptions in the main text from line 170 to 176.

The results indicate that trajectory inference using randomly sampled cells has significantly lower prediction accuracy compared to that using MURP. This is particularly evident in branch assignment (F1_branches) and correlation cor_dist, where the average levels decrease by 20.5%-64.9%. In contrast, trajectory predictions using MURP for downsampling show an overall score improvement of 5.31%-185%, further highlighting MURP's role in enhancing trajectory inference within MGPfact.

Comment#5: What is the impact of the number of components selected?

Thank you for the comments. In essence, MGPfact consists of two main steps: 1) trajectory inference; 2) calculation of factorized scores and identification of high-weight genes. After step 1, MGPfact estimates parameters such as pseudotime T and bifurcation points B.  In step 2, we introduce a rotation matrix to obtain factor scores  for each trajectory  by rotating Y*.

For all trajectories,

where  is the error term for the -th trajectory. The number of features in Y* must match the dimensions of the rotation matrix R to ensure the factorized score matrix W contains factor scores for  trajectories, achieving effective feature representation and interpretation in the model.

Additionally, to further illustrate the impact of the number of principal components (PCs) on model performance in step 1, we conducted additional experiments. We used 3 PCs as the default and adjusted the number to evaluate changes from this baseline. As shown in Author response image 2, setting the number of PCs to 1 significantly decreases the overall performance score (overall_mean=0.363), as well as the wcor_features and cor_dist metrics.  In contrast, increasing the number of PCs does not significantly affect the metrics. It ought to be mentioned that number of components used should be determined by the intrinsic biological characteristics of the cell fate-determination. Our experiment based on a limited number of datasets may not represent more complex scenarios in other cell types.

Author response image 2.

Robustness testing of the number of MURP PCA components on 100 training datasets. With the number of principal components (PCs) set to 3 by default; we tested the impact of different number of components (1-10) on the prediction results. In all box plots, the asterisk represents the mean value, while the whiskers extend to the farthest data points within 1.5 times the interquartile range. Significance is denoted as follows: not annotated indicates non-significant; * P < 0.05; ** P < 0.01; *** P < 0.001; two-sided paired Student’s T-tests.

Comment#6: Please comment on the selection of the kernel functions (rbf and polynomial) and explain why other options were discarded.

Thank you for the comments. We have added a description regarding the selection of radial basis functions and polynomial kernels in lines 126-130. As the reviewers mentioned, the choice of kernel functions is crucial in the MGPfact analysis pipeline for constructing the covariance matrix of the Gaussian process. We selected the radial basis function (RBF) kernel and the polynomial kernel to balance capturing data complexity and computational efficiency. The RBF kernel is chosen for its ability to effectively model smooth functions and capture local variations in the data, making it well-suited to the continuous and smooth characteristics of biological processes; its hyperparameters offer modeling flexibility. The polynomial kernel is used to capture more complex nonlinear relationships between input features, with its hyperparameters also allowing further customization of the model. In contrast, other complex kernels, such as Matérn or spectral kernels, were omitted due to their interpretability challenges and the risk of overfitting with limited data. However, as suggested by the reviewers, we will consider and test the impact of other kernel functions on the covariance matrix of the Gaussian process and their role in trajectory inference in our subsequent phases of algorithm design.

Comment#7: What is the impact of the Pseudotime method used initially? This section should be expanded with clear details on the techniques and parameters used in each analysis.

We are sorry for the confusion. We've added a description of how pseudotime  is obtained between line 138 and 147 in the main text. And the specific hyperparameters involved in the model and their prior settings are detailed in the supplementary information.

In brief, the pseudotime and related topological parameters of the bifurcative trajectories in MGPfact are inferred by Gaussian process regression from downsampled single-cell transcriptomic data (MURP). Specifically,  is treated as a continuous variable representing the progression of cells through the differentiation process. We describe the relationship between pseudotime and expression data as:

where  is a Gaussian Process (GP) with covariance matrix , and  represents the error term. The Gaussian process is defined as:

where  is the variance set to 1e-6. During inference, we update the pseudotime by maximizing the posterior liklihood. Specifically, the posterior distribution of pseudotime is obtained by combining the observed data Y* with the prior distribution of the Gaussian process model.

We use the Markov Chain Monte Carlo method for parameter estimation, particularly employing the adaptive Metropolis-within-Gibbs (AMWG) sampling to handle the high autocorrelation of pseudotime.

Comment#8: Enhancing Readability: For clarity, provide intuitive descriptions of each evaluation function used in simulated and real data. The novel methodology performs well for some metrics but less so for others. A clear understanding of these measurements is essential.

To address the concern of readability, we have added descriptions of 5 evaluation metrics in the methodology section (Benchmarking MGPfact to state-of-the-art methods) in line 494 to 515. Additionally, we have included a summary and discussion of these metrics in the conclusion section in line 214-240 to help the readers better understand the significance and impact of these measurements.

(1) In brief, the Hamming-Ipsen-Mikhailov ( ) distance measures the similarity between topological structures, combining the normalized Hamming distance and the Ipsen-Mikhailov distance, which focus on edge length differences and degree distribution similarity, respectively. The  is used to assess the accuracy of a model's branch assignment via Jaccard similarity between branch pairs. In trajectory inference, cor_dist quantifies the similarity of inter-cell distances between predicted and true trajectories, evaluating the accuracy of cell ordering. The wcor_features assesses the similarity of key features through weighted Pearson correlation, capturing biological variation. The score is calculated as the geometric mean of these metrics, providing an assessment of overall performance.

(2) For MGPfact and the other seven methods included in the comparison, each has its own focus. MGPfact specializes in factorizing complex cell trajectories using Gaussian process mixture models, making it particularly capable of identifying bifurcation events. Therefore, it excels in the accuracy of branch partitioning and similarity of trajectory topology. Among other methods, scShaper (Smolander et al., 2022) and TSCAN(Ji and Ji, 2016) are more suited for generating linear trajectories and excel in linear datasets, accurately predicting pseudotime. The Monocle series, as typical representatives of tree methods, effectively capture complex topologies and are suitable for analyzing cell data with diversified differentiation paths.

Comment#9: Microglia Analysis:In Figures 3A-C, the genes mentioned in the text for each bifurcation do not always match those shown in the panels. Please confirm this.

Thank you for pointing this out. We have carefully reviewed the article and corrected the error where the genes shown in the figures did not correspond to the descriptions in the article. The specific corrections have been made between line 257 and 264:

The first bifurcation determines the differentiated cell fates of PAM and HM, which involves a set of notable marker genes of both cell types, such as Apoe, Selplg (HM), and Gpnmb (PAM). The second bifurcation determines the proliferative status, which is crucial for the development and function of PAM and HM (Guzmán, n.d.; Li et al., 2019). The genes affected by the second bifurcation are associated with cell cycle and proliferation, such as Mki67, Tubb5, Top2a. The third bifurcation influences the development and maturity of microglia, of which the highly weighted genes, such as Tmem119, P2ry12, and Sepp1 are all previously annotated markers for establishment of the fates of microglia (Anderson et al., 2022; Li et al., 2019) (Supplementary Table 4).

Comment#10: Regulons:

- The conclusions rely heavily on regulons. The Methods section describes using SCENIC, GENIE3, RCisTarget, and AUCell, but their relation to bifurcation analysis is unclear.

- Do you perform trajectory analysis on all MURP-derived cells or within each identified trajectory based on bifurcation? This point needs clarification to make the outcomes comprehensible. The legend of Figure 4 provides some ideas, but further clarity is required.

Thank you for the comments.

(1) To clarify, we used the tools like SCENIC to annotate the highly weighted genes (HWG) resulted from the bifurcation analysis for transcription factor regulation activity and possible impacts on biological processes. We have added descriptions to the analysis of our microglial data. The revised content is between line 265 and 266:

Moreover, we retrieved highly active regulons from the HWG by MGPfact, of which the significance is quantified by the overall weights of the member genes.

(2) We apologize for any confusion caused by our description. It is important to clarify that we performed an overall trajectory analysis on all MURP results, rather than analyzing within each identified trajectory. Specifically, we first used MURP to downsample all preprocessed cells, where each MURP subset represents a group of cells. We then conducted trajectory inference on all MURP subsets and identified bifurcation points. This process generated multiple independent differentiation trajectories, encompassing all MURP subsets. To clearly convey this point, we have added descriptions in the legend of Figure 4. The revised content is between line 276 and 283:

“Fig. 4. MGPfact reconstructed the developmental trajectory of microglia, recovering known determinants of microglia fate. a-c. The inferred independent bifurcation processes with respect to the unique cell types (color-coded) of microglia development, where phase 0 corresponds to the state before bifurcation; and phases 1 and 2 correspond to the states post-bifurcation. Each colored dot represents a metacell of unique cell type defined by MURP. The most highly weighted regulons in each trajectory were labeled by the corresponding transcription factors (left panels). The HWG of each bifurcation process include a set of highly weighted genes (HWG), of which the expression levels differ significantly among phases 1, 2, and 3 (right panels).”

Comment#11: CD8+ T Cells: The comparison is made against Monocle2, the method used in the publication, but it would be beneficial to compare it with more recent methods. Otherwise, the added value of MGPfact is unclear.

Per your request, we have expanded our comparative analysis to include not only Monocle2 but also more recent methods such as Monocle3 (Cao et al., 2019) and scFates Tree (Faure et al., 2023). We used adjusted R-squared values to evaluate each method's ability to explain trajectory variation. The results have been added to Table 2 and Supplementary Table 6. The revised content is between line 318 and 326:

We assessed the goodness-of-fit (adjusted R-square) of the consensus trajectory derived by MGPfact and three methods (Monocle 2, Monocle 3 and scFates Tree) for the CD8+ T cell subtypes described in the original studies (Guo et al., 2018; Zhang et al., 2018). The data showed that MGPfact significantly improved the explanatory power for most CD8+ T cell subtypes over Monocle 2, which was used in the original studies (P < 0.05, see Table 2 and Supplementary Table 6), except for the CD8-GZMK cells in the CRC dataset. Additionally, MGPfact demonstrated better explanatory power in specific cell types when compared to Monocle 3 and scFates Tree. For instance, in the NSCLC dataset, MGPfact exhibited higher explanatory power for CD8-LEF1 cells (Table 2, R-squared = 0.935), while Monocle 3 and scFates Tree perform better in other cell types.

Comment#12: Consensus Trajectory: A panel explaining how the consensus trajectory is generated would be helpful. Include both visual and textual explanations tailored to the journal's audience.

Thank you for the comments. Regarding how the consensus trajectory is constructed, we have illustrated and described this in Figure 1 and the supplementary methods. Taking the reviewers' suggestions into account, we have added more details about the generation process of the consensus trajectory in the methods section to enhance the completeness of the manuscript. The revised content is from line 599 to 606:

Following MGPfact decomposition, we obtained multiple independent bifurcative trajectories, each corresponds to a binary tree within the temporal domain. These trajectories were then merged to construct a coherent diffusion tree, representing the consensus trajectory of cells’ fate. The merging process involves initially sorting all trajectories by their bifurcation time. The first (earliest) bifurcative trajectory is chosen as the initial framework, and subsequent trajectories are integrated to the initial framework iteratively by adding the corresponding branches at the bifurcation timepoints. As a result, the trajectories are ultimately merged into a comprehensive binary tree, serving as the consensus trajectory.

Comment#13: Discussion:

- Check for typos, e.g., line 382 "pseudtime.".

- Avoid considering HVG as the entire feature space.

- The first three paragraphs are too similar to the Introduction. Consider shortening them to succinctly state the scenario and the implications of your contribution.

Thank you for pointing out the typos.

(1) We conducted a comprehensive review of the document to ensure there are no typographical errors.

(2) We restructured the first three paragraphs of the discussion section to clarify the limitations in the use of current manifold-learning methods and removed any absolute language regarding treating HVGs as the entire feature space. The revised content is from line 419 to 430:

Single-cell RNA sequencing (scRNA-seq) provides a direct, quantitative snapshot of a population of cells in certain biological conditions, thereby revealing the actual cell states and functions. Although existing clustering and embedding algorithms can effectively reveal discrete biological states of cells, these methods become less efficient when depicting continuous evolving of cells over the temporal domain. The introduction of manifold learning offers a new dimension for discovery of relevant biological knowledge in cell fate determination, allowing for a better representation of continuous changes in cells, especially in time-dependent processes such as development, differentiation, and clonal evolution. However, current manifold learning methods face major limitations, such as the need for prior information on pseudotime and cell clustering, and lack of explainability, which restricts their applicability. Additionally, many existing trajectory inference methods do not support gene selection, making it difficult to annotate the results to known biological entities, thereby hindering the interpretation of results and subsequent functional studies.

Comment#14: Minor Comments:

(1) Review the paragraph regarding the "current manifold-learning methods are faced with two major challenges." The message needs clarification.

(2) Increase the quality of the figures.

(3) Update the numbering of equations from #(.x) to (x).

We thank the reviewer for these detailed suggestions.

(1) We have thoroughly revised the discussion section, addressing overly absolute statements. The revised content is from line 426 to 428:

However, current manifold learning methods face major limitations, such as the need for prior information on pseudotime and cell clustering, and lack of explainability, which restricts their applicability.

(2) We conducted a comprehensive review of the figures in the article to more clearly present our results.

(3) We have meticulously reviewed the equations in the article to ensure there are no display issues with the indices.

Reference

Aibar S, González-Blas CB, Moerman T, Huynh-Thu VA, Imrichova H, Hulselmans G, Rambow F, Marine J-C, Geurts P, Aerts J, van den Oord J, Atak ZK, Wouters J, Aerts S. 2017. SCENIC: single-cell regulatory network inference and clustering. Nat Methods 14:1083–1086. doi:10.1038/nmeth.4463

Anderson SR, Roberts JM, Ghena N, Irvin EA, Schwakopf J, Cooperstein IB, Bosco A, Vetter ML. 2022. Neuronal apoptosis drives remodeling states of microglia and shifts in survival pathway dependence. Elife 11:e76564.

Bravo González-Blas C, De Winter S, Hulselmans G, Hecker N, Matetovici I, Christiaens V, Poovathingal S, Wouters J, Aibar S, Aerts S. 2023. SCENIC+: single-cell multiomic inference of enhancers and gene regulatory networks. Nat Methods. doi:10.1038/s41592-023-01938-4

Cao J, Spielmann M, Qiu X, Huang X, Ibrahim DM, Hill AJ, Zhang F, Mundlos S, Christiansen L, Steemers FJ, Trapnell C, Shendure J. 2019. The single-cell transcriptional landscape of mammalian organogenesis. Nature 566:496–502. doi:10.1038/s41586-019-0969-x

Faure L, Soldatov R, Kharchenko PV, Adameyko I. 2023. scFates: a scalable python package for advanced pseudotime and bifurcation analysis from single-cell data. Bioinformatics 39:btac746. doi:10.1093/bioinformatics/btac746

Guo X, Zhang Y, Zheng L, Zheng C, Song J, Zhang Q, Kang B, Liu Z, Jin L, Xing R, Gao R, Zhang L, Dong M, Hu X, Ren X, Kirchhoff D, Roider HG, Yan T, Zhang Z. 2018. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nat Med 24:978–985. doi:10.1038/s41591-018-0045-3

Guzmán AU. n.d. Single-cell RNA sequencing of spinal cord microglia in a mouse model of neuropathic pain.

Ji Z, Ji H. 2016. TSCAN: Pseudo-time reconstruction and evaluation in single-cell RNA-seq analysis. Nucleic Acids Res 44:e117–e117. doi:10.1093/nar/gkw430

Lange M, Bergen V, Klein M, Setty M, Reuter B, Bakhti M, Lickert H, Ansari M, Schniering J, Schiller HB, Pe’er D, Theis FJ. 2022. CellRank for directed single-cell fate mapping. Nat Methods 19:159–170. doi:10.1038/s41592-021-01346-6

Li Q. 2023. scTour: a deep learning architecture for robust inference and accurate prediction of cellular dynamics. Genome Biology.

Li Q, Cheng Z, Zhou L, Darmanis S, Neff NF, Okamoto J, Gulati G, Bennett ML, Sun LO, Clarke LE, Marschallinger J, Yu G, Quake SR, Wyss-Coray T, Barres BA. 2019. Developmental Heterogeneity of Microglia and Brain Myeloid Cells Revealed by Deep Single-Cell RNA Sequencing. Neuron 101:207-223.e10. doi:10.1016/j.neuron.2018.12.006

Neal RM. 2003. Slice sampling. The annals of statistics 31:705–767.

Papadopoulos N, Gonzalo PR, Söding J. 2019. PROSSTT: probabilistic simulation of single-cell RNA-seq data for complex differentiation processes. Bioinformatics 35:3517–3519. doi:10.1093/bioinformatics/btz078

Ren J, Zhang Q, Zhou Y, Hu Y, Lyu X, Fang H, Yang J, Yu R, Shi X, Li Q. 2022. A downsampling method enables robust clustering and integration of single-cell transcriptome data. Journal of Biomedical Informatics 130:104093. doi:10.1016/j.jbi.2022.104093

Roberts GO, Rosenthal JS. 2009. Examples of adaptive MCMC. Journal of computational and graphical statistics 18:349–367.

Saelens W, Cannoodt R, Todorov H, Saeys Y. 2019. A comparison of single-cell trajectory inference methods. Nat Biotechnol 37:547–554. doi:10.1038/s41587-019-0071-9

Sha Y. 2024. Reconstructing growth and dynamic trajectories from single-cell transcriptomics data 6.

Smolander J, Junttila S, Venäläinen MS, Elo LL. 2022. scShaper: an ensemble method for fast and accurate linear trajectory inference from single-cell RNA-seq data. Bioinformatics 38:1328–1335. doi:10.1093/bioinformatics/btab831

Tierney L. 1994. Markov chains for exploring posterior distributions. the Annals of Statistics 1701–1728.

Zappia L, Phipson B, Oshlack A. 2017. Splatter: simulation of single-cell RNA sequencing data. Genome Biol 18:174. doi:10.1186/s13059-017-1305-0

Zhang L, Yu X, Zheng L, Zhang Y, Li Y, Fang Q, Gao R, Kang B, Zhang Q, Huang JY, Konno H, Guo X, Ye Y, Gao S, Wang S, Hu X, Ren X, Shen Z, Ouyang W, Zhang Z. 2018. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature 564:268–272. doi:10.1038/s41586-018-0694-x

侧旁裂口 /侧叉 /摆衩 /下摆开衩

紧身的

食堂

adj. 缺乏的，不足的；稀有的，少见的；故意缺席的

达到最糟糕的程度

严重偏向于前者

分界线

卷心白菜

一小片属地

说到

才能 /天资 /能力 /自然倾向

长相

看重

eLife Assessment

This useful work identifies a key role for Tachykinin-1 parasubthalamic neurons in avoidance learning. At present, the evidence for the conclusions regarding fiber photometry, viral transfection, reporting of behavioral outcomes, and pathway-specificity is incomplete. This work will be of interest to neuroscientists studying neural mechanisms for avoidance and aversion.

Reviewer #1 (Public review):

This study is focused on a population of neurons in the mouse parasubthalamic nucleus (pSTN) that express Tackhykinin1 (Tac1). This gene has been used before to target pSTN for functional circuit studies because it is fairly selective for pSTN in this region, though it targets only a subset of pSTN neurons. Prior work has shown that activity in these neurons can impact motivated behaviors, including feeding and drinking behaviors, and that their activity is associated with aversion or avoidance behaviors. While not breaking much new ground, this study adds to that work by making use of a 2-way active avoidance assay, where a CS predicts a US (footshock), that the mice can escape. Using fiber photometry the authors show convincing evidence that Tac1 neurons in pSTN increase their activity in response to a US footshock, and that after some pairings the neurons will start responding to the CS too, though to a lesser extent than the US. Their most important data shows that either ablation or optogenetic inhibition of these cells can hugely block the active avoidance (escape) behavior, suggesting these neurons are key for the performance of this task, which they interpret as key for learning the task (but see more below). They show that optogenetic stimulation is aversive in a real-time place assay, and when paired with footshock can enhance active avoidance behavior. Finally, they show that Tac1 pSTN axons in PVT recapitulate these effects while showing that axons in CEA or PBN may only recapitulate some of these effects (more below). Overall I think the data is solid and shows that the activity of Tac1 pSTN neurons in the 2 way active avoidance task is causally related to avoidance behavior in the direction that would be predicted by recent literature. However, I think the authors overstate the conclusions in the title, abstract, and text. I do not think the data make a strong case for a role for these cells in learning, at least in any classical sense, as used in the title and abstract and elsewhere. Also the statement in the abstract that the pSTN mediates its effects 'differentially' through its downstream targets is not convincingly supported by data.

Major concerns:

(1) The authors infer that the activity in the Tac1 pSTN neurons is necessary for aversive or avoidance 'learning'. But this is not well defined, what exactly does that mean and what types of evidence would support or falsify such a hypothesis? Moreover, the authors show convincingly, and in line with prior reports, that these cells are activated by aversive stimuli (here footshock), and that activation of these cells is sufficient to induce avoidance behavior. Because manipulation of these cells can serve as a primary negative reinforcer, it becomes even more challenging and important to explain how experiments that manipulate these cells while measuring behavior/performance can discriminate between changes in: (1) primary aversion, (2) motivation to avoid, (3) associative learning, or (4) memory/retrieval. The authors seem to favor #3, but they don't make a clear case for this point of view or else what they mean by 'avoidance learning'. In my opinion, the data do not well discriminate between possibilities 1 through 3. The authors should clarify their logic and temper their conclusions throughout.

(2) Abstract line 37 is not well supported. The authors focus mostly on pSTN projections to PVT and show that the measurements or manipulation of these axons recapitulates the effects seen with pSTN cell bodies. The authors do fewer studies of axons in CeA and PBN, but do find that they can recapitulate the effects with opsin inhibition, but detect no effects with opsin stimulation. However, the lack of effect with opsin stimulation in Figure S7a-e proves very little on its own. It could be technical, due to inadequate expression or functional efficacy. It is not supported by histological and functional evidence that the manipulation was effective. Overall I can only conclude that the projections to these regions might be very similar (based on the inhibition data), or might be a little different. The data are thus inadequate to support the authors' claim that the pSTN mediates learning differentially through its downstream targets.

Other concerns:

(3) Line 93 is not adequately supported by data in Figure 1b. Additional data is needed that shows expression across cases, including any spread that may be visible when zooming out from pSTN. Additional methods are needed to indicate what exclusion criteria were applied and how many mice were excluded. These data could help support the statement on line 93 that expression was largely restricted within pSTN.

(4) From the results and methods it is not clear where the GFP signal would come from in the mice expressing Casp3 for the ablation studies. It is therefore not clear if the absence of GFP should be taken as evidence of cell loss. For example, it is not clear if multiple vectors were used, if volumes and titers were carefully matched between control groups, or if competition/occlusion between AAVs could be ruled out. It is also not clear how this was quantified, that is how many sections/subjects and how counting was done. It is not clear how long was waited between the AAV infusion, behavior, and euthanasia, perhaps especially important for the ablation done after avoidance learning occurred.

(5) The authors should consider showing individual measurements and not just mean/sem wherever feasible, for example, to support the statement on line 141 that 'all ablated mice showed...'.

(6) S3 is an important control for interpreting data in Figure 2d-i. Something similar is needed to support the inferences made in 2j-u. The very strong effect showing a lack of active avoidance in response to CS or the US when pSTN Tac1 neurons are inhibited during CS or during US suggests that something gross may be going on, such as a gross motor or sensory response that supersedes the effect of footshock. The authors do not comment on whether there are any gross behavioral responses to the inhibition, but an experiment as in S3 is needed, for example, to show that behavior is intact during pSTN inhibition if delivered after the mice already learned to associate CS with US.

(7) The authors use 100 shocks of 0.8 mA for 7 days. I think this is quite strong and in the pSTN inhibition experiments it seems to be functionally 'inescapable' and could thus produce behaviors similar to 'learned helplessness'. Can the authors consider whether this might contribute to the striking findings they observed in their opsin inhibition assays?

(8) The description of the experiment in S5 is inadequate. What are the adjacent areas? Where do the authors see spread? The use of the word 'case' in figure S5 implies an individual case, but the legend says 5 mice were used for 'case 1' and 3 mice were used for 'case 2'. The use of the word 'off-target in the figure implies that the expression was of the intended target. But the text of results and methods implies it was intentional targeting of unnamed and unshown adjacent regions. This should be clarified.

(9) The authors suggest the CPA study is divergent from Serra et al 2023. Though I think this could be due to how the conditioning was done, it would be helpful for the authors to include less processed data. This would aid in possible interpretations for any divergences across studies. Can the authors include raw data (in seconds of time spent) in each compartment for each group across baseline and test days?

Reviewer #2 (Public review):

Summary:

The manuscript by Hu et. al presents a clearly-designed examination of the role of tachykinin1-expressing neurons in the parasubthalamic nucleus of the lateral posterior hypothalamus (PTSN) in active avoidance learning. These glutamatergic neurons have previously been implicated in responding to negative stimuli. This manuscript expands the current understanding of PTSNTac1 neurons in learned responses to threats by showing their role in encoding and mediating the active avoidance response. The authors first use bulk fiber photometry imaging to show the encoding of the active avoidance procedure, followed by cell-type specific manipulations of PTSNTac1 neurons during active avoidance. Finally, they show that encoding and mediation of active avoidance in a downstream target of PTSNTac1 neurons, the PVT/intermediodorsal nuclei of the dorsal thalamus (IMD), has the same effect as what was discovered in the cell body. This contrasts other output regions of the PTSN, such as the PBN and CeA, which were not found to promote active avoidance learning. The experiments presented were well-designed to support the conclusions of the authors, however, the manuscript is missing several key control experiments and supplemental information to support their main findings.

Strengths:

The manuscript provides information on a brain region and downstream target that mediates active avoidance learning. The manuscript provides valuable information via necessity and sufficiency experiments to show the role of the population of interest (PTSNTac1 neurons) in active avoidance learning. The authors also performed most behavior experiments in male and female mice, with adequate power to address potential sex differences in the control of active avoidance by PTSNTac1 neurons. Finally, the manuscript provides valuable information about the specificity of the PTSNTac1 downstream target in regulating active avoidance learning, identifying the PVT/intermediodorsal nuclei of the dorsal thalamus as the key target and ruling out the PBN and CeA.

Weaknesses:

However, several main conclusions of the paper must be interpreted carefully due to missing or inadequate control experiments and histological verification.

(1) Inadequate presentation of viral localization. The authors state that expression was "largely restricted within PSTN" however there is no quantification of the amount of viral expression beyond the target region. Given that Tac1 is expressed in neighboring regions, it is critical to show the viral expression and fiber implant location data for all animals included in the figures. Furthermore, criteria for inclusion and exclusion based on mistargeting should be delineated. This should also be clearly outlined for the experiments in Figure S5, where "behavioral effects of activation of sparsely Tac1-expressing neurons in two adjacent areas of PSTN" was tested but the location of viral expression in those cases is unclear.

(2) Lack of motion artifact correction with isosbestic signal for GCamp recordings. It is appreciated that the authors included a separate EGFP-expressing group to compare to the GCamp-expressing group, however, additional explanation is required for the methods used to analyze the raw fluorescent signal. Namely, were fluorescent signals isosbestic-corrected prior to calculating ΔF/F? If no isosbestic signal was used to correct motion artifacts within a recording session, additional explanation is needed to explain how this was addressed. The lack of motion artifacts in the EGFP signal in a separate cohort is inadequate to answer this caveat as motion artifacts are within-animal.

(3) Missing control experiment demonstrating intact locomotor performance in caspase ablation experiments. The authors use caspase ablation of PTSNTac1 neurons prior to active avoidance learning to appraise the necessity of this cell population. However, a control experiment showing intact locomotor ability in ablated mice was not performed.

(4) Missing control experiment demonstrating [lack of] valence with PTSN silencing manipulations. The authors performed a real-time and conditioned place preference experiments for ChR2-expressing mice (Fig 3M) and found stimulation to be negatively-valenced and generate an aversive memory, respectively. Absent this control experiment with silencing, an alternative conclusion remains possible that optogenetic silencing via GtACR2 created nonspecific location preferences in the active avoidance apparatus, confounding the interpretation of those results.

(5) Incomplete analysis of sex differences. Data in female mice is conspicuously missing from inhibition experiments. The rationale for exclusion from this dataset would be useful for the interpretation of the other noted sex differences.

Reviewer #3 (Public review):

Summary:

This study by Hu et al. examined the role of tachykinin1 (Tac1)-expressing neurons in the para subthalamic nucleus (PSTH) in active avoidance of electric shocks. Bulk recording of PSTH Tac1 neurons or axons of these neurons in PVT showed activation of a shock-predicting tone and shock itself. Ablation of these neurons or optogenetic manipulation of these neurons or their projection to PVT suggests the causality of this pathway with the learning of active avoidance.

Strengths:

This work found an understudied pathway potentially important for active avoidance of electric shocks. Experiments were thoroughly done and the presentation is clear. The amount of discussion and references are appropriate.

Weaknesses:

Critical control experiments are missing for most experiments, and statistical tests are not clear or not appropriate in most parts. Details are shown below.

(1) There are some control experiments missing. Notably, optogenetic manipulation is not verified in any experiments. It is important to verify whether neural activation with optogenetic activation is at the physiological level or supra-physiological level, and whether optogenetic inhibition does not cause unwanted activity patterns such as rebound activation at the critical time window.

(2) Neural ablation with caspase was confirmed by GFP expression. However, from the present description, a different virus to express EITHER caspase or GFP was injected, and then the numbers of GFP-expressing neurons were compared. It is not clear how this can detect ablation.

(3) In many places, statistical approaches are not clear from the present figures, figure legends, and Methods. It seems that most statistics were performed by pooling trials, but it is not described, or multiple "n" are described. For example, it is explicitly mentioned in Figure 4H, "n = 3 mice, n = 213 avoidance trials and n = 87 failure trials". The authors should not pool trials, but should perform across-animal tests in this and other figures, and "n" for statistical tests should be clearly described in each plot.

(4) It is also unclear how the test types were selected. For example, in Figure 1K and O with similar datasets, one is examined by a paired test and the other is by an unpaired test. Since each animal has both early vs late trials, and avoidance vs failure trials, paired tests across animals should be performed for both.

(5) It is also strange to show violin plots for only 6 animals. They should instead show each dot for each animal, connected with a line to show consistent increases of activity in late vs early trials and avoidance vs failure trials.

(6) To tell specificity in avoidance learning, it is better to show escape in the current trials with optogenetic manipulation.

(7) For place aversion, % time decrease across days was tested. It is better to show the original number before normalization, as well.

(8) For anatomical results in Figure S6, it is important to show images with lower magnification, too.

(9) Inactivation of either pathway from PSTH to PBN or to CeA also inhibits active avoidance, but the authors conclude that these effects are "partial" compared to the inactivation of PSTH to PVT. It is not clear how the effects were compared since the effects of PSTH-CeA inactivation are quite strong, comparable to PSTH-PVT inactivation by eye. They should quantify the effects to conclude the difference.

(10) Supplementary table 1: as mentioned above, n for statistical tests should be clearer.

Author response:

Reviewer #1 (Public review):

This study is focused on a population of neurons in the mouse parasubthalamic nucleus (pSTN) that express Tackhykinin1 (Tac1). This gene has been used before to target pSTN for functional circuit studies because it is fairly selective for pSTN in this region, though it targets only a subset of pSTN neurons. Prior work has shown that activity in these neurons can impact motivated behaviors, including feeding and drinking behaviors, and that their activity is associated with aversion or avoidance behaviors. While not breaking much new ground, this study adds to that work by making use of a 2-way active avoidance assay, where a CS predicts a US (footshock), that the mice can escape. Using fiber photometry, the authors show convincing evidence that Tac1 neurons in pSTN increase their activity in response to a US footshock, and that after some pairings the neurons will start responding to the CS too, though to a lesser extent than the US. Their most important data shows that either ablation or optogenetic inhibition of these cells can hugely block the active avoidance (escape) behavior, suggesting these neurons are key for the performance of this task, which they interpret as key for learning the task (but see more below). They show that optogenetic stimulation is aversive in a real-time place assay, and when paired with footshock can enhance active avoidance behavior. Finally, they show that Tac1 pSTN axons in PVT recapitulate these effects while showing that axons in CEA or PBN may only recapitulate some of these effects (more below). Overall I think the data is solid and shows that the activity of Tac1 pSTN neurons in the 2 way active avoidance task is causally related to avoidance behavior in the direction that would be predicted by recent literature. However, I think the authors overstate the conclusions in the title, abstract, and text. I do not think the data make a strong case for a role for these cells in learning, at least in any classical sense, as used in the title and abstract and elsewhere. Also, the statement in the abstract that the pSTN mediates its effects 'differentially' through its downstream targets is not convincingly supported by data.

We are very pleased that Reviewer 1 thought our data is solid.

Major concerns:

(1) The authors infer that the activity in the Tac1 pSTN neurons is necessary for aversive or avoidance 'learning'. But this is not well defined, what exactly does that mean and what types of evidence would support or falsify such a hypothesis? Moreover, the authors show convincingly, and in line with prior reports, that these cells are activated by aversive stimuli (here footshock), and that activation of these cells is sufficient to induce avoidance behavior. Because manipulation of these cells can serve as a primary negative reinforcer, it becomes even more challenging and important to explain how experiments that manipulate these cells while measuring behavior/performance can discriminate between changes in: (1) primary aversion, (2) motivation to avoid, (3) associative learning, or (4) memory/retrieval. The authors seem to favor #3, but they don't make a clear case for this point of view or else what they mean by 'avoidance learning'. In my opinion, the data do not well discriminate between possibilities 1 through 3. The authors should clarify their logic and temper their conclusions throughout.

Thank you Reviewer 1 for providing us insightful suggestions. Based on our fiber photometry data that the activities of PSTN Tac1+ neurons show a significant increase in CS-evoked calcium fluorescent signals in late trials relative to those in early trials (Figure 1H-K) and our optogenetic inhibition experiments during CS (Figure 2N-Q), these results illustrate that the activities of PSTN Tac1+ neurons are modulated by learning and are required for active avoidance learning. Moreover, PSTN Tac1+ neurons are activated by footshock and activation of these cells is sufficient to induce avoidance behavior. These findings demonstrate that PSTN Tac1+ neurons encode aversive information. Together, our current data support that PSTN Tac1+ neurons encode both aversive event and its predicting cue. We will clarify our conclusions in the revised manuscript.

(2) Abstract line 37 is not well supported. The authors focus mostly on pSTN projections to PVT and show that the measurements or manipulation of these axons recapitulates the effects seen with pSTN cell bodies. The authors do fewer studies of axons in CeA and PBN, but do find that they can recapitulate the effects with opsin inhibition, but detect no effects with opsin stimulation. However, the lack of effect with opsin stimulation in Figure S7a-e proves very little on its own. It could be technical, due to inadequate expression or functional efficacy. It is not supported by histological and functional evidence that the manipulation was effective. Overall, I can only conclude that the projections to these regions might be very similar (based on the inhibition data), or might be a little different. The data are thus inadequate to support the authors' claim that the pSTN mediates learning differentially through its downstream targets.

In the revised version of manuscript, we will provide more histological and functional evidence for the PSTN-to-CeA and PSTN-to-PBN circuits to support our conclusion on the functional roles of these downstream targets. Similar with our anterograde experiment that the PSTN densely projects to CeA and PBN (Figure S6), optogenetic activation and inhibition experiments showed dense axonal terminals in the CeA and PBN from the PSTN and this line of data will be included in the revised manuscript. In addition, we will further examine these circuits by investigating the functional roles of CeA-projecting or PBN-Projecting PSTN neurons during 2-way active avoidance task.

Other concerns:

(3) Line 93 is not adequately supported by data in Figure 1b. Additional data is needed that shows expression across cases, including any spread that may be visible when zooming out from pSTN. Additional methods are needed to indicate what exclusion criteria were applied and how many mice were excluded. These data could help support the statement on line 93 that expression was largely restricted within pSTN.

In the revised version of manuscript, we will provide larger example images containing pSTN and its adjacent areas to demonstrate that the viral expression is well restricted into this brain area. Moreover, we will provide detailed information on the exclusion criteria and the number of mice excluded in the Method section.   

(4) From the results and methods it is not clear where the GFP signal would come from in the mice expressing Casp3 for the ablation studies. It is therefore not clear if the absence of GFP should be taken as evidence of cell loss. For example, it is not clear if multiple vectors were used, if volumes and titers were carefully matched between control groups, or if competition/occlusion between AAVs could be ruled out. It is also not clear how this was quantified, that is how many sections/subjects and how counting was done. It is not clear how long was waited between the AAV infusion, behavior, and euthanasia, perhaps especially important for the ablation done after avoidance learning occurred.

I totally agree with Reviewer 1’s concerns. We will perform immunohistochemistry or in situ hybridization for Tachykinin-1 itself and then measure colocalization of GFP with Tachykinin-1 inside and outside of the PTSN, and the degree of absence of Tachykinin-1 in Casp mice. In addition, we will provide more detailed experimental information in the revised manuscript.

(5) The authors should consider showing individual measurements and not just mean/sem wherever feasible, for example, to support the statement on line 141 that 'all ablated mice showed...'.

Thank you Reviewer 1 for this suggestion. We will re-plot the data as individual measurements in the revised manuscript.

(6) S3 is an important control for interpreting data in Figure 2d-i. Something similar is needed to support the inferences made in 2j-u. The very strong effect showing a lack of active avoidance in response to CS or the US when pSTN Tac1 neurons are inhibited during CS or during US suggests that something gross may be going on, such as a gross motor or sensory response that supersedes the effect of footshock. The authors do not comment on whether there are any gross behavioral responses to the inhibition, but an experiment as in S3 is needed, for example, to show that behavior is intact during pSTN inhibition if delivered after the mice already learned to associate CS with US.

Thank you Reviewer 1 for this insightful suggestion. During the review process, we have performed this line of experiment as in Figure S3. We measured the behavioral responses during pSTN optogenetic inhibition after the mice already learned to associate CS with US and found most GtACR-expressing mice showed unaffected avoidance learning. This data will be included in the revised manuscript.

(7) The authors use 100 shocks of 0.8 mA for 7 days. I think this is quite strong and in the pSTN inhibition experiments it seems to be functionally 'inescapable' and could thus produce behaviors similar to 'learned helplessness'. Can the authors consider whether this might contribute to the striking findings they observed in their opsin inhibition assays?

I agree with the Reviewer 1’s comment on the string findings in the optogenetic inhibition results. Indeed, based on the results on days 1 and 2, optogenetic inhibition of PSTN tac1+ neurons has significantly blocked GtACR-expressing animals’ behavioral performance during 2-way active avoidance task. To examine whether the effect by optogenetic inhibition of these neurons could possibly decline with prolonged training, we conducted additional 5-day training. We will discuss and add this comment in the revised manuscript.

(8) The description of the experiment in S5 is inadequate. What are the adjacent areas? Where do the authors see spread? The use of the word 'case' in figure S5 implies an individual case, but the legend says 5 mice were used for 'case 1' and 3 mice were used for 'case 2'. The use of the word 'off-target in the figure implies that the expression was of the intended target. But the text of results and methods implies it was intentional targeting of unnamed and unshown adjacent regions. This should be clarified.

We will add histological images and clarify these comments in the revised manuscript. The purpose of this experiment is to illustrate that even slightly spreading ChR2 viruses into Tac1+ neurons of the adjacent areas of the PSTN did not result in behavioral changes and this will indirectly support the main behavioral function caused by the PSTN tac1+ neurons rather than its neighboring areas. Because Tac1+ neurons outside the PSTN are sparsely expressed, it is quite difficult to completely restrict the viral expression in the PSTN from the anterior to the posterior. Thus, we will provide detailed information on the exclusion criteria and the number of mice excluded in the Method section.   

(9) The authors suggest the CPA study is divergent from Serra et al 2023. Though I think this could be due to how the conditioning was done, it would be helpful for the authors to include less processed data. This would aid in possible interpretations for any divergences across studies. Can the authors include raw data (in seconds of time spent) in each compartment for each group across baseline and test days?

We will follow Reviewer 1’s suggestion to include raw data (in seconds of time spent) in each compartment for each group across baseline and test days in the revised manuscript.

Reviewer #2 (Public review):

Summary:

The manuscript by Hu et. al presents a clearly-designed examination of the role of tachykinin1-expressing neurons in the parasubthalamic nucleus of the lateral posterior hypothalamus (PTSN) in active avoidance learning. These glutamatergic neurons have previously been implicated in responding to negative stimuli. This manuscript expands the current understanding of PTSNTac1 neurons in learned responses to threats by showing their role in encoding and mediating the active avoidance response. The authors first use bulk fiber photometry imaging to show the encoding of the active avoidance procedure, followed by cell-type specific manipulations of PTSNTac1 neurons during active avoidance. Finally, they show that encoding and mediation of active avoidance in a downstream target of PTSNTac1 neurons, the PVT/intermediodorsal nuclei of the dorsal thalamus (IMD), has the same effect as what was discovered in the cell body. This contrasts other output regions of the PTSN, such as the PBN and CeA, which were not found to promote active avoidance learning. The experiments presented were well-designed to support the conclusions of the authors, however, the manuscript is missing several key control experiments and supplemental information to support their main findings.

Strengths:

The manuscript provides information on a brain region and downstream target that mediates active avoidance learning. The manuscript provides valuable information via necessity and sufficiency experiments to show the role of the population of interest (PTSNTac1 neurons) in active avoidance learning. The authors also performed most behavior experiments in male and female mice, with adequate power to address potential sex differences in the control of active avoidance by PTSNTac1 neurons. Finally, the manuscript provides valuable information about the specificity of the PTSNTac1 downstream target in regulating active avoidance learning, identifying the PVT/intermediodorsal nuclei of the dorsal thalamus as the key target and ruling out the PBN and CeA.

We highly appreciate that Reviewer 2 thought that our experiments presented were well-designed to support the conclusions and provided valuable information in several aspects.

Weaknesses:

However, several main conclusions of the paper must be interpreted carefully due to missing or inadequate control experiments and histological verification.

(1) Inadequate presentation of viral localization. The authors state that expression was "largely restricted within PSTN" however there is no quantification of the amount of viral expression beyond the target region. Given that Tac1 is expressed in neighboring regions, it is critical to show the viral expression and fiber implant location data for all animals included in the figures. Furthermore, criteria for inclusion and exclusion based on mistargeting should be delineated. This should also be clearly outlined for the experiments in Figure S5, where "behavioral effects of activation of sparsely Tac1-expressing neurons in two adjacent areas of PSTN" was tested but the location of viral expression in those cases is unclear.

Similar with questions 3 and 8 of Reviewer 1. We will provide the viral expression and fiber implant location data for all animals included in the figures and histological images in Figure S5 in the revised manuscript. Moreover, we will provide detailed information on the exclusion criteria and the number of mice excluded in the Method section.  

2) Lack of motion artifact correction with isosbestic signal for GCamp recordings. It is appreciated that the authors included a separate EGFP-expressing group to compare to the GCamp-expressing group, however, additional explanation is required for the methods used to analyze the raw fluorescent signal. Namely, were fluorescent signals isosbestic-corrected prior to calculating ΔF/F? If no isosbestic signal was used to correct motion artifacts within a recording session, additional explanation is needed to explain how this was addressed. The lack of motion artifacts in the EGFP signal in a separate cohort is inadequate to answer this caveat as motion artifacts are within-animal.

We will follow Reviewer 2’s suggestion and perform isosbestic-correction for fluorescent signals prior to calculating ΔF/F. We will re-plot related figures and add this information in the revised manuscript.

(3) Missing control experiment demonstrating intact locomotor performance in caspase ablation experiments. The authors use caspase ablation of PTSNTac1 neurons prior to active avoidance learning to appraise the necessity of this cell population. However, a control experiment showing intact locomotor ability in ablated mice was not performed.

We will follow Reviewer 2’s suggestion to perform a control experiment showing intact locomotor ability in caspase 3-ablated mice and will include this data in the revised manuscript.

(4) Missing control experiment demonstrating [lack of] valence with PTSN silencing manipulations. The authors performed a real-time and conditioned place preference experiments for ChR2-expressing mice (Fig 3M) and found stimulation to be negatively-valenced and generate an aversive memory, respectively. Absent this control experiment with silencing, an alternative conclusion remains possible that optogenetic silencing via GtACR2 created nonspecific location preferences in the active avoidance apparatus, confounding the interpretation of those results.

Thank you Reviewer 2 for this useful suggestion. We will examine the valence with PTSN silencing manipulations by using a RTPP test and add this data in the revised manuscript.

(5) Incomplete analysis of sex differences. Data in female mice is conspicuously missing from inhibition experiments. The rationale for exclusion from this dataset would be useful for the interpretation of the other noted sex differences.

Thank you Reviewer 2 for this useful suggestion. During the review process, we have performed ablation and inhibition experiments in females, demonstrating similar behavioral effects as those in males. We will add these data in the revised manuscript.

Reviewer #3 (Public review):

Summary:

This study by Hu et al. examined the role of tachykinin1 (Tac1)-expressing neurons in the para subthalamic nucleus (PSTH) in active avoidance of electric shocks. Bulk recording of PSTH Tac1 neurons or axons of these neurons in PVT showed activation of a shock-predicting tone and shock itself. Ablation of these neurons or optogenetic manipulation of these neurons or their projection to PVT suggests the causality of this pathway with the learning of active avoidance.

Strengths:

This work found an understudied pathway potentially important for active avoidance of electric shocks. Experiments were thoroughly done and the presentation is clear. The amount of discussion and references are appropriate.

We are very pleased to have Reviewer 3’s positive comments on the manuscript.

Weaknesses:

Critical control experiments are missing for most experiments, and statistical tests are not clear or not appropriate in most parts. Details are shown below.

(1) There are some control experiments missing. Notably, optogenetic manipulation is not verified in any experiments. It is important to verify whether neural activation with optogenetic activation is at the physiological level or supra-physiological level, and whether optogenetic inhibition does not cause unwanted activity patterns such as rebound activation at the critical time window.

Thank you Reviewer 3 for this useful suggestion. We will perform in vitro slice recording experiments to verify optogenetic manipulations and add this line of evidence in the revised manuscript.

(2) Neural ablation with caspase was confirmed by GFP expression. However, from the present description, a different virus to express EITHER caspase or GFP was injected, and then the numbers of GFP-expressing neurons were compared. It is not clear how this can detect ablation.

Similar with question 4 of Reviewer 1. We will perform immunohistochemistry or in situ hybridization for Tachykinin-1 itself and then measure colocalization of GFP with Tachykinin-1 inside and outside of the PTSN, and the degree of absence of Tachykinin-1 in Casp-ablated mice. In addition, we will provide more detailed experimental information in the revised manuscript.

(3) In many places, statistical approaches are not clear from the present figures, figure legends, and Methods. It seems that most statistics were performed by pooling trials, but it is not described, or multiple "n" are described. For example, it is explicitly mentioned in Figure 4H, "n = 3 mice, n = 213 avoidance trials and n = 87 failure trials". The authors should not pool trials, but should perform across-animal tests in this and other figures, and "n" for should be clearly described in each plot.

We have provided all statistical information in the Supplementary Table 1. In the revised manuscript, we will perform across-animal tests, re-plot new figures and provide clear statistical information.

(4) It is also unclear how the test types were selected. For example, in Figure 1K and O with similar datasets, one is examined by a paired test and the other is by an unpaired test. Since each animal has both early vs late trials, and avoidance vs failure trials, paired tests across animals should be performed for both.

Following Reviewer 3’s suggestion, we will perform across-animal tests. In the first version of our manuscript, for fiber photometry experiments, we pooled trial data of each animal and performed statistics tests across trials. Because avoidance and failure trials were different, we thus selected an unpaired test for this kind of dataset.

(5) It is also strange to show violin plots for only 6 animals. They should instead show each dot for each animal, connected with a line to show consistent increases of activity in late vs early trials and avoidance vs failure trials.

Similar with question 4 of Reviewer 3, we pooled trial data of each animal and performed statistics tests across trials. We will perform across-animal tests and re-plot figures by connecting with a line to show consistent increases of activity in late vs early trials and avoidance vs failure trials for each animal.

(6) To tell specificity in avoidance learning, it is better to show escape in the current trials with optogenetic manipulation.

Thank you Reviewer 3 for this useful suggestion. We will follow this suggestion and add this analysis in the revised manuscript.

(7) For place aversion, % time decrease across days was tested. It is better to show the original number before normalization, as well.

Similar with question 9 of Reviewer 1, we will show the original number before normalization in the revised manuscript.

(8) For anatomical results in Figure S6, it is important to show images with lower magnification, too.

We will follow this suggestion and provide histological images with lower magnification in the revised manuscript.

(9) Inactivation of either pathway from PSTH to PBN or to CeA also inhibits active avoidance, but the authors conclude that these effects are "partial" compared to the inactivation of PSTH to PVT. It is not clear how the effects were compared since the effects of PSTH-CeA inactivation are quite strong, comparable to PSTH-PVT inactivation by eye. They should quantify the effects to conclude the difference.

We will quantify the effects of different downstream targets of the PSTN to make a precise conclusion.

(10) Supplementary table 1: as mentioned above, n for statistical tests should be clearer.

As mentioned above, we will perform across-animal tests and provide clear statistical information in the figure legends and supplementary table 1.

A better way to collaborate

Sign up for MarkUp.io, upload your content, drop a comment, and share for review. Yep, it’s that simple.

One price. Infinite users.

MarkUp.io is built for individuals, creative teams, and enterprise organizations. We scale with you.

Vibby – Annotate YouTube & Vimeo Videos on Your Own Site

secondary creator economy

vimeo

Curie Law magnetism with temp changes

articlepost

Hoptraveler.com Travel Lifestyle In today’s world, travel isn’t just about visiting new places; it’s about embracing a lifestyle that blends adventure, luxury, and personal growth. Hoptraveler.com travel lifestyle stands out as a beacon for modern travelers seeking not only unforgettable destinations but also curated experiences that align with their unique preferences. This article delves deep into the pillars of Hoptraveler.com and how it redefines the way we explore the world.

Hoptraveler.com Travel Lifestyle

Best Way To Manage Multiple Business Finances

Coyyn.com Business

Holiday Events and Festivities in Towns County, Georgia

У меня в конце строки.

eLife Assessment

This study uses a large dataset from both recent isolates and genomes in databases to provide an important analysis of the population structure of the pathogen Salmonella gallinarum. The authors present convincing results regarding the regional adaptation and the evolutionary trajectory of the resistome and mobilome, even though some issues regarding the genomic analysis could be improved. This work will interest microbiologists and researchers working on genomics, evolution, and antimicrobial resistance.

Reviewer #1 (Public review):

Summary:

The investigators in this study analyzed the dataset assembly from 540 Salmonella isolates, and those from 45 recent isolates from Zhejiang University of China. The analysis and comparison of the resistome and mobilome of these isolates identified a significantly higher rate of cross-region dissemination compared to localized propagation. This study highlights the key role of the resistome in driving the transition and evolutionary history of S. Gallinarum.

Strengths:

The isolates included in this study were from 16 countries in the past century (1920 to 2023). While the study uses S. Gallinarun as the prototype, the conclusion from this work will likely apply to other Salmonella serotypes and other pathogens.

Weaknesses:

While the isolates came from 16 countries, most strains in this study were originally from China.

Comments on revisions:

This reviewer is happy with the detailed responses from the authors regarding revising this manuscript. I do not have further comments.

Reviewer #2 (Public review):

Summary:

The authors sequence 45 new samples of S. Gallinarum, a commensal Salmonella found in chickens, which can sometimes cause disease. They combine these sequences with around 500 from public databases, determine the population structure of the pathogen, and coarse relationships of lineages with geography. The authors further investigate known anti-microbial genes found in these genomes, how they associate with each other, whether they have been horizontally transferred, and date the emergence of clades.

Strengths:

- It doesn't seem that much is known about this serovar, so publicly available new sequences from a high burden region are a valuable addition to the literature.<br /> - Combining these sequences with publicly available sequences is a good way to better contextualise any findings.<br /> - The genomic analyses have been greatly improved since the first version of the manuscript, and appropriately analyse the population and date emergence of clades.<br /> - The SNP thresholds are contextualised in terms of evolutionary time.<br /> - The importance and context of the findings are fairly well described.

Weaknesses:

- There are still a few issues with the genomic analyses, although they no longer undermine the main conclusions:

(1) Although the SNP distance is now considered in terms of time, the 5 SNP distance presented still represents ~7yrs evolution, so it is unlikely to be a transmission event, as described. It would be better to use a much lower threshold or describe the interpretation of these clusters more clearly. Bringing in epidemiological evidence or external references on the likely time interval between transmissions would be helpful.

(2) The HGT definition has not fundamentally been changed and therefore still has some issues, mainly that vertical evolution is still not systematically controlled for. Using a 5kb window is not sufficient, as LD may extend across the entire genome. As the authors have now run gubbins correctly, they could use the results from this existing analysis to find recent HGT. To definite mobilisation, perhaps a standard pipeline such (e.g. https://github.com/EBI-Metagenomics/mobilome-annotation-pipeline) would be more convincing.

(3) The invasiveness index is better described, but the authors still did not provide convincing evidence that the small difference is actually biologically meaningful (there was no statistical difference between the two strains provided in response Figure 6). What do other Salmonella papers using this approach find, and can their links be brought in? If there is still no good evidence, a better description of this difference would help make the conclusions better supported.

In summary, the analysis is broadly well described and feels appropriate. Some of the conclusions are still not fully supported, although the main points and context of the paper now appear sound.

Author response:

The following is the authors’ response to the original reviews.

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

The investigators in this study analyzed the dataset assembly from 540 Salmonella isolates, and those from 45 recent isolates from Zhejiang University of China. The analysis and comparison of the resistome and mobilome of these isolates identified a significantly higher rate of cross-region dissemination compared to localized propagation. This study highlights the key role of the resistome in driving the transition and evolutionary 

Thank you for summarizing our work. According to your comments, we carefully considered and responded to them and made corresponding revisions to the text. Additionally, to fully contextualize the background knowledge and clarify the major points in this study, we add some references.

Upon further review of our initial manuscript, we realized that the original submission did not strictly follow the lineage order proposed by Zhou et al. (Natl Sci Rev. 2023 Sep 2;10(10):nwad228). To avoid confusion and keep the uniform knowledge in the typing system, we have adjusted the lineage nomenclature along the revised manuscript to reflect the corrected order as follows:

Author response table 1.

To ensure consistency with previous studies, we have revised the nomenclature for the different lineages of bvSP.

Strengths: 

The isolates included in this study were from 16 countries in the past century (1920 to 2023). While the study uses S. Gallinarun as the prototype, the conclusion from this work will likely apply to other Salmonella serotypes and other pathogens. 

Thanks for the constructive comments and the positive reception of the manuscript.

Weaknesses: 

While the isolates came from 16 countries, most strains in this study were originally from China. 

We appreciate the reviewer's observation regarding the sampling distribution of isolates in this study. We acknowledge that while the isolates were collected from 15 different countries, with a significant proportion originated from China (Author response image 1). This focus is due to several reasons:

Author response image 1.

Geographic distribution of 580 S. Gallinarum. Different colors indicate the countries of origin for the 580 S. Gallinarum strains in the dataset. Darker shades represent higher numbers of strains.

(1) As once a globally prevalent pathogen across the 20th century, S. Gallinarum was listed by the World Organization for Animal Health (WOAH) due to its economic importance. After 30 years of implementation of the National Poultry Improvement Plan in the US, it was almost eradicated in high-income countries, and interestingly, it became an endemic pathogen with sporadic outbreaks in most low- or middle-income countries like China and Brazil. Given the vast expanse of China's land area and the country's economic factors, implementing the same measures remains challenging.  

(2) S. Gallinarum is an avian-specific pathogen, particularly affecting chickens, and its distribution is closely linked to chicken meat production in different countries. There are more frequent reports of fowl typhoid in some high chicken-producing developing countries. Data from the United States Department of Agriculture (USDA) on annual chicken meat production for 2023/2024 show that the global distribution of S. Gallinarum aligns closely with the overall chicken meat production of these countries (https://fas.usda.gov/data/production/commodity/0115000).

Author response image 2.

The United States Department of Agriculture (USDA) data on annual chicken meat production for 2023/2024 across different countries globally.

(3) Our primary objective was to investigate the localized resistome adaptation of S. Gallinarum in regions. Being a region with significant disease burden, China has reported numerous outbreaks (Sci Data. 2022 Aug 13;9(1):495; Sci Data. 2024 Feb 27;11(1):244) and a high AMR prevalence of this serovar (Natl Sci Rev. 2023 Sep 2;10(10):nwad228; mSystems. 2023 Dec 21;8(6):e0088323), making it an excellent example for understanding localized resistance mechanisms.

(4) As China is the primary country of origin for the strains in this study, it is necessary to ensure that the strains from China are consistent with the local geographic characteristics of the country. Therefore, we conducted a correlation analysis between the number of strains from different provinces in China and the total GDP/population size of those provinces (Author response image 3). The results show that most points fall within the 95% confidence interval of the regression line. Although some points exhibit relative unbalance in the number of S. Gallinarum strains, most data points for these regions have a small sample size (n < 15). Overall, we found that the prevalence of S. Gallinarum in different regions of China is consistent with the overall nationwide trend.

Author response image 3.

Correlation analysis between the number of S. Gallinarum collected from different provinces in China and the total GDP/population size. The figure depicts a series of points representing individual provinces. The x-axis indicates the number of S. Gallinarum included in the dataset, while the y-axis displays the values for total GDP and total population size, respectively.

Nevertheless, a search of nearly a decade of literature on PubMed and a summary of the S. Gallinarum genome available on public databases indicate that the dataset used is the most complete. Furthermore, focusing on a specific region within China allowed us to conduct a detailed and thorough analysis. However, we highly agree that expanding the study to include more isolates from other countries would enhance the generalizability of our findings, and we are actively collecting additional S. Gallinarum genome data. In the revised manuscript, we have further emphasized the limitations as follow:

Lines 427-429: “However, the current study has some limitations. Firstly, despite assembling the most comprehensive WGS database for S. Gallinarum from public and laboratory sources, there are still biases in the examined collection. The majority (438/580) of S. Gallinarum samples were collected from China, possibly since the WGS is a technology that only became widely available in the 21st century. This makes it impractical to sequence it on a large scale in the 20th century, when S. Gallinarum caused a global pandemic. So, we suspect that human intervention in the development of this epidemic is the main driving force behind the fact that most of the strains in the data set originated in China. In our future work, we aim to actively gather more data to minimize potential biases within our dataset, thereby improving the robustness and generalizability of our findings.”

Reviewer #2 (Public review): 

Summary: 

The authors sequence 45 new samples of S. Gallinarum, a commensal Salmonella found in chickens, which can sometimes cause disease. They combine these sequences with around 500 from public databases, determine the population structure of the pathogen, and coarse relationships of lineages with geography. The authors further investigate known anti-microbial genes found in these genomes, how they associate with each other, whether they have been horizontally transferred, and date the emergence of clades. 

Thank you for your constructive suggestions, which are valuable and highly beneficial for improving our paper. According to your comments, we carefully considered and responded to them and made corresponding revisions to the text. Furthermore, to fully contextualize the background knowledge and clarify the major points in this study, we add some references to support our findings and policy implications.

Upon further review of our initial manuscript, we realized that the original submission did not strictly follow the lineage order proposed by Zhou et al. (Natl Sci Rev. 2023 Sep 2;10(10):nwad228). To avoid confusion in the typing system, we have adjusted the lineage nomenclature in the revised manuscript to reflect the corrected order (see Author response table 1).

Strengths: 

(1) It doesn't seem that much is known about this serovar, so publicly available new sequences from a high-burden region are a valuable addition to the literature. 

(2) Combining these sequences with publicly available sequences is a good way to better contextualise any findings. 

Thank you so much for your thorough review and constructive comments on the manuscript.

Weaknesses: 

There are many issues with the genomic analysis that undermine the conclusions, the major ones I identified being: 

(1) Recombination removal using gubbins was not presented fully anywhere. In this diversity of species, it is usually impossible to remove recombination in this way. A phylogeny with genetic scale and the gubbins results is needed. Critically, results on timing the emergence (fig2) depend on this, and cannot be trusted given the data presented. 

We sincerely thank you for pointing out this issue. In the original manuscript, we aimed to present different lineages of S. Gallinarum within a single phylogenetic tree constructed using BEAST. However, in the revised manuscript, we have addressed this issue by applying the approach recommended by Gubbins to remove recombination events for each lineage defined by FastBAPs. Additionally, to better illustrate the removal of recombination regions in the genome, we have included a figure generated by Gubbins (New Supplementary Figure 12). 

Our results indicate that recombination events are relatively infrequent in Lineage 1, followed by Lineage 3, but occur more frequently in Lineage 2. In the revised manuscript, we have included additional descriptions in the Methods section to clarify this analysis. We hope these modifications adequately address the reviewer’s concerns and enhance the trustworthiness of our findings.

(2) The use of BEAST was also only briefly presented, but is the basis of a major conclusion of the paper. Plot S3 (root-to-tip regression) is unconvincing as a basis of this data fitting a molecular clock model. We would need more information on this analysis, including convergence and credible intervals. 

Thank you very much for raising this issue. We decided to reconduct separate BEAST analyses for each lineage, accurately presenting the evolutionary scale based on the abovementioned improvements. The implementation of individual lineage for BEAST analysis was conducted based on the following steps:

(1) Using R51 as the reference, a reference-mapped multiple core-genome SNP sequence alignment was created, and recombination regions were detected and removed as described above.

(2) TreeTime was used to assess the temporal structure by performing a regression analysis of the root-to-tip branch distances within the maximum likelihood tree, considering the sampling date as a variable (New Supplementary Figures 6). However, the root-to-tip regression analysis presented in New Supplementary Figures 6 was not intended as a basis for selecting the best molecular clock model; its purpose was to clean the dataset with appropriate measurements.

(3) To determine the optimal model for running BEAST, we tested a total of six combinations in the initial phase of our study. These combinations included the strict clock, relaxed lognormal clock, and three population models (Bayesian SkyGrid, Bayesian Skyline, and Constant Size). Before conducting the complete BEAST analysis, we evaluated each combination using a Markov Chain Monte Carlo (MCMC) analysis with a total chain length of 100 million and sampling every 10,000 iterations. We then summarized the results using NSLogAnalyser and determined the optimal model based on the marginal likelihood value for each combination. The results indicated that the model incorporating the Bayesian Skyline and the relaxed lognormal clock yielded the highest marginal likelihood value in our sample. Then, we proceeded to perform a timecalibrated Bayesian phylogenetic inference analysis for each lineage. The following settings were configured: the "GTR" substitution model, “4 gamma categories”, the "Relaxed Clock Log Normal" model, the "Coalescent Bayesian Skyline" tree prior, and an MCMC chain length of 100 million, with sampling every 10,000 iterations.

(4) Convergence was assessed using Tracer, with all parameter effective sampling sizes (ESS) exceeding 200. Maximum clade credibility trees were generated using TreeAnnotator. Finally, key divergence time points (with 95% credible intervals) were estimated, and the tree was visualized using FigTree. 

For the key lineages, L2b and L3b (carrying the resistome, posing antimicrobial resistance (AMR) risks, and exhibiting intercontinental transmission events), we have redrawn Figure 2 based on the updated BEAST analysis results (New Figure 2). For L1, L2a, and L3c, we have added supplementary figures to provide a more detailed visualization of their respective BEAST analysis outcomes (New Supplementary Figures 3-5). The revised BEAST analysis indicates that the origin of L3b in China can be traced back to as early as 1683 (95% CI: 1608 to 1839). In contrast, the earliest possible origin of L2b in China dates back to 1880 (95% CI: 1838 to 1902). This indicates that the previous manuscript's assumption that L2b is an older lineage compared to L3b may be inaccurate. 

Furthermore, In the revised manuscript, we specifically estimated the time points for the first intercontinental transmission events for the two major lineages, L2b and L3b. Our results indicate that L2b, likely underwent two major intercontinental transmission events. The first occurred around 1893 (95% CI: 1870 to 1918), with transmission from China to South America. The second major transmission event occurred in 1923 (95% CI: 1907 to 1940), involving the spread from South America to Europe. In contrast, the transmission pattern of L3b appears relatively more straightforward. Our findings show that L3b, an S. Gallinarum lineage originating in China, only underwent one intercontinental transmission event from China to Europe, likely occurring around 1790 (95% CI: 1661 to 1890) (New Supplementary Figure 7). Based on the more critical BEAST analysis for each lineage, we have revised the corresponding conclusions in the manuscript. We believe that the updated BEAST analysis, performed using a more accurate recombination removal approach, significantly enhances the rigor and credibility of our findings.

(3) Using a distance of 100 SNPs for a transmission is completely arbitrary. This would at least need to be justified in terms of the evolutionary rate and serial interval. 

Using single nucleotide polymorphism (SNP) distance to trace pathogen transmission is a common approach (J Infect Dis. 2015 Apr 1;211(7):1154-63) and in our previous studies (hLife 2024; 2(5):246-256. mLife 2024; 3(1):156-160.). When the SNP distance within a cluster falls below a set threshold, the strains in that cluster are considered to have a potential direct transmission link. It is generally accepted that the lower the threshold, the more stringent the screening process becomes. However, there is little agreement in the literature regarding what such a threshold should be, and the appropriate SNP cut-off for inferring transmission likely depends critically on the context (Mol Biol Evol. 2019 Mar 1;36(3):587-603).

In this study, we compared various thresholds (SNPs = 5, 10, 20, 25, 30, 35, 40, 50, 100) to ensure clustering in an appropriate manner. First, we summarized the tracing results under each threshold (Author response image 4), which demonstrated that, regardless of the threshold used, all strains associated with transmission events originated from the same location (New Figure 3a).

Author response image 4.

Clustering results of 45 newly isolated S. Gallinarum strains using different SNP thresholds of 5, 10, 15, 20, 25, 28, 30, 50, and 100 SNPs. The nine subplots represent the clustering results under each threshold. Each point corresponds to an individual strain, and lines connect strains with potential transmission relationships.

In response to your comments regarding the evolutionary rate, we estimated the overall evolutionary rate of the S. Gallinarum using BEAST. We applied the methodology described by Arthur W. Pightling et al. (Front Microbiol. 2022 Jun 16; 13:797997). The numbers of SNPs per year were determined by multiplying the evolutionary rates estimated with BEAST by the number of core SNP sites identified in the alignments. We hypothesize that a slower evolutionary rate in bacteria typically requires a lower SNP threshold when tracing transmission events using SNP distance analysis. Pightling et al.'s previous research found an average evolutionary rate of 1.97 SNPs per year (95% HPD, 0.48 to 4.61) across 22 different Salmonella serotypes. Our updated BEAST estimation for the evolutionary rate of S. Gallinarum suggests it is approximately 0.74 SNPs per year (95% HPD, 0.42 to 1.06). Based on these findings, and our previous experience with similar studies (mBio. 2023 Oct 31;14(5):e0133323.), we set a threshold of 5 SNPs in the revised manuscript.

Then, we adopted the newly established SNP distance threshold (n=5) to update Figure 3a and New Supplementary Figure 8. The heatmap on the far right of New Figure 3a illustrates the SNP distances among 45 newly isolated S. Gallinarum strains from two locations in Zhejiang Province (Taishun and Yueqing). New Supplementary Figure 8 simulates potential transmission events between the bvSP strains isolated from Zhejiang Province (n=95) and those from China with available provincial information (n=435). These analyses collectively demonstrate the localized transmission pattern of bvSP within China. Our analysis using the newly established SNP threshold indicates that the 45 strains isolated from Taishun and Yueqing exhibit a highly localized transmission pattern, with pairs of strains exhibiting potential transmission events below the set threshold occurring exclusively within a single location. Subsequently, we conducted the SNP distance-based tracing analysis for the 95 strains from Zhejiang Province and those from China with available provincial information (n=435) (New Supplementary Figure 8, New Supplementary Table S8). Under the SNP distance threshold (n=5), we identified a total of 91 potential transmission events, all of which occurred exclusively within Zhejiang Province. No inter-provincial transmission events were detected. Based on these findings, we revised the methods and conclusions in the manuscript accordingly. We believe that the updated version well addresses your concerns.

Nevertheless, the final revised and updated results do not change the conclusions presented in our original manuscript. Instead, applying a more stringent SNP distance threshold allows us to provide solid evidence supporting the localized transmission pattern of S. Gallinarum in China. 

(4) The HGT definition is non-standard, and phylogeny (vertical inheritance) is not controlled for.  

The cited method: 

'In this study, potentially recently transferred ARGs were defined as those with perfect identity (more than 99% nucleotide identity and 100% coverage) in distinct plasmids in distinct host bacteria using BLASTn (E-value {less than or equal to}10−5)' 

This clearly does not apply here, as the application of distinct hosts and plasmids cannot be used. Subsequent analysis using this method is likely invalid, and some of it (e.g. Figure 6c) is statistically very poor. 

Thank you for raising this important question. In our study, Horizontal Gene Transfer (HGT) is defined as the transfer of genetic information between different organisms, a process that facilitates the spread of antibiotic resistance genes (ARGs) among bacteria. This definition of HGT is consistent with that used in previous studies (Evol Med Public Health. 2015; 2015(1):193–194; ISME J. 2024 Jan 8;18(1):wrad032). In Salmonella, the transfer of antimicrobial resistance genes via HGT is not solely dependent on plasmids; other mobile genetic elements (MGEs), such as transposons, integrons, and prophages, also play significant roles. This has also  been documented in our previous work (mSystems. 2023 Dec 21;8(6):e0088323). Given the involvement of various MGEs in the horizontal transfer of ARGs, we propose that the criteria for evaluating horizontal transfer via plasmids can also be applied to ARGs mediated by other MGEs.

In this study, we adopted stricter criteria than those used by Xiaolong Wang et al. Specifically, we defined two ARGs as identical only if they exhibited 100% nucleotide identity and 100% coverage. To address concerns regarding the potential influence of vertical inheritance in our analysis, we have made the following improvements. In the revised manuscript, we provide a more detailed table that includes the co-localization analysis of each ARG with mobile genetic elements (New Supplementary Table 9). For prophages and plasmids, we required that ARGs be located directly within these elements. In contrast, for transposons and integrons, we considered ARGs to be associated if they were located within a 5 kb region upstream or downstream of these elements (Nucleic Acids Res. 2022 Jul 5;50(W1):W768-W773). 

In the revised manuscript, we first categorized a total of 621 ARGs carried by 436 bvSP isolates collected in China according to the aforementioned criteria and found that 415 ARGs were located on MGEs. After excluding the ARGs not associated with MGEs, we recalculated the overall HGT frequency of 10 types of ARGs in China, the horizontal ARGs transfer frequency in three key regions, and the horizontal ARGs transfer frequency within a single region (New Supplementary Table 7). Based on the results, we updated relevant sections of the manuscript and remade Figure 6. The updated manuscript describes the results of this section as follows:

“Horizontal transfer of resistome occurs widely in localized bvSP

Horizontal transfer of the resistome facilitates the acquisition of AMR among bacteria, which may record the distinct acquisition event in the bacterial genome. To compare these events in a geographic manner, we further investigated the HGT frequency of each ARG carried by bvSP isolated from China and explored the HGT frequency of resistome between three defined regions. Potentially horizontally transferred ARGs were defined as those with perfect identity (100% identity and 100% coverage) and were located on MGEs across different strains (Fig. 6a). We first categorized a total of 621 ARGs carried by 436 bvSP isolates collected in China and found that 415 ARGs were located on MGEs. After excluding the ARGs not associated with MGEs, our findings reveal that horizontal gene transfer of ARGs is widespread among Chinese bvSP isolates, with an overall transfer rate of 92%. Specifically, 50% of the ARGs exhibited an HGT frequency of 100%, indicating that these ARGs might underwent extensive frequent horizontal transfer events (Fig. 6b). It is noteworthy that certain resistance genes, such as tet(A), aph(3'')-Ib, and aph(6)-Id, appear to be less susceptible to horizontal transfer.

However, different regions generally exhibited a considerable difference in resistome HGT frequency. Overall, bvSP from the southern areas in China showed the highest HGT frequency (HGT frequency=95%). The HGT frequencies for bvSP within the eastern and northern regions of China are lower, at 92% and 91%, respectively (Fig. 6c). For specifical ARG type, we found tet(A) is more prone to horizontal transfer in the southern region, and this proportion was considerably lower in the eastern region. Interestingly, certain ARGs such as aph(6)-Id, undergo horizontal transfer only within the eastern and northern regions of China (Fig. 6d). Notably, as a localized transmission pathogen, resistome carried by bvSP exhibited a dynamic potential among inter-regional and local demographic transmission, especially from northern region to southern region (HGT frequency=93%) (Fig. 6e, Supplementary Table 7).”

We also modified the current version of the pipeline used to calculat the HGT frequency of resistance genes. In the revised pipeline, users are required to provide a file specifying the locations of mobilome on the genome before formally calculating the HGT frequency of the target ARGs. The specific code and data used in the calculation have been uploaded to https://github.com/tjiaa/Cal_HGT_Frequency.

However, we also acknowledge that the current in silico method has some limitations. This approach heavily relies heavily on prior information in existing resistome/mobilome databases. Additionally, the characteristics of second-generation sequencing data make it challenging to locate gene positions precisely. Using complete genome assemblies might be a crucial approach to address this issue effectively. In the revised manuscript, we have also provided a more detailed explanation of the implications of the current pipeline.

Regarding your second concern, "some of it (e.g., Figure 6c) is statistically very poor," the horizontal ARG transfer frequency calculation for each region was based on the proportion of horizontal transfer events of ARGs in that region to the total possible transfer events. As a result, we are unable to calculate the statistical significance between the two regions. Our aim with this approach is to provide a rough estimate of the extent of horizontal ARG transfer within the S. Gallinarum population in each region. In future studies, we will refine our conclusions by developing a broader range of evaluation methods to ensure more comprehensive assessment and validation.

(5) Associations between lineages, resistome, mobilome, etc do not control for the effect of genetic background/phylogeny. So e.g. the claim 'the resistome also demonstrated a lineage-preferential distribution' is not well-supported. 

Thank you for your comments. We acknowledge that the associations between lineages and the mobilome/resistome may be influenced by the genetic background or phylogeny of the strains. For instance, our conclusion regarding the lineage-preferential distribution of the resistome was primarily based on New Figure 4a, where L3 is clearly shown to carry the most ARGs. Furthermore, we observed that L3b tends to harbor bla_{_TEM-1B}, _sul2, and tet(A) more frequently than other lineages. However, we recognize that this evidence is insufficient to support a definitive conclusion of “demonstrated a lineage-preferential distribution”. Therefore, we have re-examined the current manuscript and described these findings as a potential association between the mobilome/resistome and lineages.

(6) The invasiveness index is not well described, and the difference in means is not biologically convincing as although it appears significant, it is very small. 

Thank you for pointing this out. For the invasiveness index mentioned in the manuscript, we used the method described in previous studies. (PLoS Genet. 2018 May 8;14(5), Nat Microbiol. 2021 Mar;6(3):327-338). Specifically, Salmonella’s ability to cause intestinal or extraintestinal infections in hosts is related to the degree of genome degradation. We evaluated the potential for extraintestinal infection by 45 newly isolated S. Gallinarum strains (L2b and L3b) using a model that quantitatively assesses genome degradation. We analyzed samples using the 196 top predictor genes, employing a machine-learning approach that utilizes a random forest classifier and delta-bitscore functional variant-calling. This method evaluated the invasiveness of S. Gallinarum towards the host, and the distribution of invasiveness index values for each region was statistically tested using unpaired t-test. The code used for calculating the invasiveness index is available at https://github.com/Gardner-BinfLab/invasive_salmonella. In the revised manuscript, we added a more detailed description of the invasiveness index calculation in the Methods section as follows:

Lines 592-603: “Specifically, Salmonella’s ability to cause intestinal or extraintestinal infections in hosts is related to the degree of genome degradation. We evaluated the potential for extraintestinal infection by 45 newly isolated S. Gallinarum strains (L2b and L3b) using a model that quantitatively assesses genome degradation. We analyzed each sample using the 196 top predictor genes for measuring the invasiveness of S. Gallinarum, employing a machine-learning approach that utilizes a random forest classifier and deltabitscore functional variant-calling. This method evaluated the invasiveness of S. Gallinarum towards the host, and the distribution of invasiveness index values for each region was statistically tested using unpaired t-test. The code used for calculating the invasiveness index is available at: https://github.com/Gardner-BinfLab/invasive_salmonella.”

Regarding the second question, 'the difference in means is not biologically convincing as although it appears significant, it is very small,' we believe that this difference is biologically meaningful. In our previous work, we infected chicken embryos with different lineages of S. Gallinarum (Natl Sci Rev. 2023 Sep 2;10(10):nwad228). The virulence of thirteen strains of Salmonella Gallinarum, comprising five from lineage L2b and eight from lineage L3b, was evaluated in 16-day-old SPF chicken embryos through inoculation into the allantoic cavity. Controls included embryos that inoculated with phosphate-buffered saline (PBS). The embryos were incubated in a thermostatic incubator maintained at 37.5°C with a relative humidity ranging from 50% to 60%. Prior to inoculation, the viability of the embryos was assessed by examining the integrity of their venous system and their movements; any dead embryos were excluded from the study. Overnight cultures resuspended in PBS at a concentration of 1000 CFU per 100 μL were administered to the embryos. Mortality was recorded daily for a period of five days, concluding upon the hatching of the chicks. 

It is generally accepted that strains with higher invasive capabilities are more likely to cause chicken embryo mortality. Our experimental results showed that the L2b, which exhibits higher invasiveness, with a slightly higher to cause chicken embryo death (Author response image 5). 

Author response image 5.

The survival curves of chicken embryos infected with bvSP isolates from S. Gallinarum L2b and S. Gallinarum L3b. Inoculation with Phosphate Buffer Saline (PBS) were considered controls. 

(7) 'In more detail, both the resistome and mobilome exhibited a steady decline until the 1980s, followed by a consistent increase from the 1980s to the 2010s. However, after the 2010s, a subsequent decrease was identified.' 

Where is the data/plot to support this? Is it a significant change? Is this due to sampling or phylogenetics? 

Thank you for highlighting these critical points. The description in this statement is based on New Supplementary Figure 11. On the right side of New Supplementary Figure 11, we presented the average number of Antimicrobial Resistance Genes (ARGs) and Mobile Genetic Elements (MGEs) carried by S. Gallinarum isolates from different years, and we described the overall trend across these years. However, we realized that this statement might overinterpret the data. Given that this sentence does not impact our emphasis on the overall increasing trends observed in the resistome and mobilome, as well as their potential association, we decided to remove it in the revised manuscript.

The revised paragraph would read as follows:

Lines 261-268: “Variations in regional antimicrobial use may result in uneven pressure for selecting AMR. The mobilome is considered the primary reservoir for spreading resistome, and a consistent trend between the resistome and the mobilome has been observed across different lineages, from L1-L3c. We observed an overall gradual rise in the resistome quantity carried by bvSP across various lineages, correlating with the total mobilome content (S11 Fig). Furthermore, we investigated the interplay between particular mobile elements and resistome types in bvSP.”

(8) It is not clear what the burden of disease this pathogen causes in the population, or how significant it is to agricultural policy. The article claims to 'provide valuable insights for targeted policy interventions.', but no such interventions are described. 

Thank you for your constructive suggestions. Salmonella Gallinarum is an avian-specific pathogen that induces fowl typhoid, a severe systemic disease characterized by high mortality rates in chickens, thereby posing a significant threat to the poultry industry, particularly in developing countries (Rev Sci Tech. 2000 Aug;19(2):40524). In our previous research, we conducted a comprehensive meta-analysis of 201 publications encompassing over 900 million samples to investigate the global impact of S. Gallinarum (Sci Data. 2022 Aug 13;9(1):495). Our findings estimated that the global prevalence of S. Gallinarum is 8.54% (with a 95% confidence interval of 8.43% to 8.65%), with notable regional variations in incidence rates.

Our previously analysis focused on the prevalence of S. Gallinarum (including biovars SP and SG) across six continents. The results revealed that all continents, except Oceania, exhibited positive prevalences of S. Gallinarum. Asia had the highest prevalence at 17.31%, closely followed by Europe at 16.03%. In Asia, the prevalence of biovar SP was higher than that of biovar SG, whereas in Europe, biovar SG was observed to be approximately two hundred times more prevalent than biovar SP. In South America, the prevalence of S. Gallinarum was higher than that of biovar SP, at 10.06% and 13.20% respectively. Conversely, the prevalence of S. Gallinarum was relatively lower in North America (4.45%) compared to Africa (1.10%) (Author response image 6).

Given the significant economic losses caused by S. Gallinarum to the poultry industry and the potential risk of escalating antimicrobial resistance, more targeted policy interventions are urgently needed. Further elaboration on this implication is provided in the revised “Discussion” section as follows:

Lines 401-416: “In summary, the findings of this study highlight that S. Gallinarum remains a significant concern in developing countries, particularly in China. Compared to other regions, S. Gallinarum in China poses a notably higher risk of AMR, necessitating the development of additional therapies, i.e. vaccine, probiotics, bacteriophage therapy in response to the government's policy aimed at reducing antimicrobial use ( J Infect Dev Ctries. 2014 Feb 13;8(2):129-36). Furthermore, given the dynamic nature of S. Gallinarum risks across different regions, it is crucial to prioritize continuous monitoring in key areas, particularly in China's southern regions where the extensive poultry farming is located. Lastly, from a One-Health perspective, controlling AMR in S. Gallinarum should not solely focus on local farming environments, with improved overall welfare on poultry and farming style. The breeding pyramid of industrialized poultry production should be targeted on the top, with enhanced and accurate detection techniques (mSphere. 2024 Jul 30;9(7):e0036224). More importantly, comprehensive efforts should be made to reduce antimicrobial usage overall and mitigate potential AMR transmission from environmental sources or other hosts (Vaccines (Basel). 2024 Sep 18;12(9):1067; Vaccines (Basel). 2023 Apr 18;11(4):865; Front Immunol. 2022 Aug 11:13:973224).”

Author response image 6.

A comparison of the global prevalence of S. gallinarum across continents.

(9) The abstract mentions stepwise evolution as a main aim, but no results refer to this. 

Thank you for raising this issue. In the revised manuscript, we have changed “stepwise evolution” to simply “evolution” to ensure a more accurate and precise description.

(10) The authors attribute changes in population dynamics to normalisation in China-EU relations and hen fever. However, even if the date is correct, this is not a strongly supported causal claim, as many other reasons are also possible (for example other industrial processes which may have changed during this period). 

Thank you for raising this critical issue. In the revised manuscript, we conducted a more stringent BEAST analysis for each lineage, as described earlier. This led to some changes in the inferred evolutionary timelines. Consequently, we have removed the corresponding statement from the “Results” section. Instead, we now only provide a discussion of historical events, supported by literature, that could have facilitated the intercontinental spread of L2b and L3b in the “Discussion” section. We believe these revisions have made the manuscript more rigorous and precise.

Lines 332-342: “_The biovar types of _S. Gallinarum have been well-defined as bvSP, bvSG, and bvSD historically ( J Vet Med B Infect Dis Vet Public Health. 2005 Jun;52(5):2148). Among these, bvSP can be further subdivided into five lineages (L1, L2a, L2b, L3b, and L3c) using hierarchical Bayesian analysis. Different sublineages exhibited preferential geographic distribution, with L2b and L3b of bvSP being predominant global lineage types with a high risk of AMR. The historical geographical transmission was verified using a spatiotemporal Bayesian framework. The result shows that L3b was initially spread from China to Europe in the 18^th-19^th century, which may be associated with the European hen fever event in the mid-19th century (Burnham GP. 1855. The history of the hen fever: a humorous record). L2b, on the other hand, appears to have spread to Europe via South America, potentially contributing to the prevalence of bvSP in the United States.”  

(11) No acknowledgment of potential undersampling outside of China is made, for example, 'Notably, all bvSP isolates from Asia were exclusively found in China, which can be manually divided into three distinct regions (southern, eastern, and northern).'.

Perhaps we just haven't looked in other places?

We appreciate the reviewer's observation regarding the sampling distribution of isolates in this study. We acknowledge that while the isolates were collected from 15 different countries with, a significant proportion originated from China (Author response image 1). This focus is due to several reasons:

(1) As once a globally prevalent pathogen across the 20th century, S. Gallinarum was listed by the World Organization for Animal Health (WOAH) due to its economic importance. After 30 years of implementation the National Poultry Improvement Plan in the US, it was almost eradicated in high-income countries, and interestingly, it became an endemic pathogen with sporadic outbreaks in most low- or middle-income countries like China and Brazil. Given the vast expanse of China's land area and the country's economic factors, implementing the same measures remains a challenging endeavour. 

(2) S. Gallinarum is an avian-specific pathogen, particularly affecting chickens, and its distribution is closely linked to chicken meat production in different countries. In some high chicken-producing developing countries, such as China and Brazil, there are more frequent reports of fowl typhoid. Data from the United States Department of Agriculture (USDA) on annual chicken meat production for 2023/2024 show that the global distribution of S. Gallinarum aligns closely with the overall chicken meat production of these countries (https://fas.usda.gov/data/production/commodity/0115000).  

(3) Our primary objective was to investigate the localized resistome adaptation of S. Gallinarum in regions. Being a region with significant disease burden, China has reported numerous outbreaks (Sci Data. 2022 Aug 13;9(1):495; Sci Data. 2024 Feb 27;11(1):244) and a high AMR prevalence of this serovar (Natl Sci Rev. 2023 Sep 2;10(10):nwad228; mSystems. 2023 Dec 21;8(6):e0088323), making it an excellent example for understanding localized resistance mechanisms. 

Nevertheless, a search of nearly a decade of literature on PubMed and a summary of the S. Gallinarum genome available on public databases indicate that the dataset used is the most complete. Furthermore, focusing on a specific region within China allowed us to conduct a detailed and thorough analysis. However, we highly agree that expanding the study to include more isolates from other countries would enhance the generalizability of our findings, and we are actively collecting additional S. Gallinarum genome data. In the revised manuscript, we modified this sentence to indicate that this phenomenon is only observed in the current dataset, thereby avoiding an overly absolute statement:

Lines 131-135: “For the bvSP strains from Asia included in our dataset, we found that all originated from China. To further investigate the distribution of bvSP across different regions in China, we categorized them into three distinct regions: southern, eastern, and northern (Supplementary Table 3)”.

(12) Many of the conclusions are highly speculative and not supported by the data. 

Thank you for your comment. We have carefully revised the manuscript to address your concerns. We hope that the changes made in the revised version meet your expectations and provide a clearer and more accurate interpretation of our findings.

(13) The figures are not always the best presentation of the data: 

a. Stacked bar plots in Figure 1 are hard to interpret, the total numbers need to be shown.

Panel C conveys little information. 

b. Figure 4B: stacked bars are hard to read and do not show totals. 

c. Figure 5 has no obvious interpretation or significance. 

Thank you for your comments. We have revised the figures to improve the clarity and presentation of the data.

In summary, the quality of analysis is poor and likely flawed (although there is not always enough information on methods present to confidently assess this or provide recommendations for how it might be improved). So, the stated conclusions are not supported. 

Thank you for your valuable feedback. We have carefully revised the manuscript to address your concerns. We hope that the updated figures and tables, and new data in the revised version meet your expectations and provide more appropriate interpretation of our findings.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors): 

This reviewer enjoyed reading this well-written manuscript. The authors are encouraged to address the following comments and revise the manuscript accordingly. 

(1) Title: The authors use avian-restrict Salmonella to refer to Salmonella Gallinarum. Please consider using Salmonella Gallinarum in the title. Also, your analysis relates to resistome and mobilome. Would it make sense to add mobilome in the manuscript? 

Thank you for your guidance. In the revised manuscript, we have changed the title to “Avian-specific Salmonella enterica Serovar Gallinarum transition to endemicity is accompanied by localized resistome and mobilome interaction”. We believe that this revised title more accurately reflects the content of our study.

(2) Abstract: This study uses 45 isolates from your labs. However, you failed to include these 45 isolates in the Abstract. Also, please clarify the sources of these isolates (from dead chickens, or dead chicken embryos? You wrote in two different ways in this manuscript). Also, I am not entirely convinced how the results from these 45 isolates will support the overall conclusion of this work. 

Thank you for your thorough review and constructive comments on the manuscript. In the revised version, we have added a description of 45 newly isolated S. Gallinarum strains in the Abstract to provide readers with a clearer understanding of the dataset used in this study.

Lines 36-41: “Using the most comprehensive whole-genome sequencing dataset of Salmonella enterica serovar Gallinarum (S. Gallinarum) collected from 16 countries, including 45 newly recovered samples from two related local regions, we established the relationship among avian-specific pathogen genetic profiles and localization patterns.”

Furthermore, the newly isolated S. Gallinarum strains were obtained from dead chicken embryos. We think your second concern may arise from the following description in the manuscript: “All 734 samples of dead chicken embryos were collected from Taishun and Yueqing in Zhejiang Province, China. After the thorough autopsy, the liver, intestines, and spleen were extracted and added separately into 2 mL centrifuge tubes containing 1 mL PBS. The organs were then homogenized by grinding.” In fact, all the collected dead chicken embryos were aged 19 to 20 days. At this developmental stage, collecting the liver, intestines, and spleen for isolation and cultivation of S. Gallinarum is possible. To avoid any confusion, we have included a more detailed description of the dead chicken embryos in the revised manuscript as follows:

Lines 447-451: “All 734 samples of dead chicken embryos aged 19 to 20 days were collected from Taishun and Yueqing in Zhejiang Province, China. After a thorough autopsy, the liver, intestines, and spleen were extracted and added separately into 2 mL centrifuge tubes containing 1 mL PBS. The organs were then homogenized by grinding.”

Regarding your concern about the statement, “I am not entirely convinced how the results from these 45 isolates will support the overall conclusion of this work,” we would like to clarify the significance of these new isolates. Our research first identified distinct characteristics in the 45 newly isolated S. Gallinarum strains from Taishun and Yueqing, Zhejiang Province. Specifically, we found that most of the strains from Yueqing belonged to sequence type ST92, whereas the majority from Taishun were ST3717. Additionally, there were significant differences between these geographically close strains in terms of SNP distance and predicted invasion capabilities. These findings suggest that S. Gallinarum may exhibit localized transmission patterns, which forms the basis of the scientific question and hypothesis we originally aimed to address. Furthermore, in our previous work, we collected 325 S. Gallinarum strains. By incorporating the newly isolated 45 strains, we aim to provide a more comprehensive view of the population diversity, transmission pattern and potential risk of S. Gallinarum. We will continue to endeavour to understand the global genomic and population diversity in this field.

Finally, we revised the sentences that could potentially raise concerns for readers: 

Lines 175-177: “To investigate the dissemination pattern of bvSP in China, we obtained forty-five newly isolated bvSP from 734 samples (6.1% overall isolation rate) collected from diseased chickens at two farms in Yueqing and Taishun, Zhejiang Province.”  >  “To investigate the dissemination pattern of bvSP, we obtained forty-five newly isolated bvSP from 734 samples (6.1% overall isolation rate) collected from diseased chickens at two farms in Yueqing and Taishun, Zhejiang Province.”

(3) The manuscript uses nomenclature and classification into different sublineages. Did the authors establish the approaches for defining these sublineages in this group or did you follow the accepted standards? 

Thank you very much for raising this important issue. The biovar types of Salmonella Gallinarum have historically been well-defined as S. Gallinarum biovar

Pullorum (bvSP), S. Gallinarum biovar Gallinarum (bvSG), and S. Gallinarum biovar Duisburg (bvSD) (J Vet Med B Infect Dis Vet Public Health. 2005 Jun;52(5):214-8). However, there seems to be no widespread consensus on the population nomenclature for the key biovar bvSP. In a previous study, Zhou et al. classified bvSP into six lineages:

L1, L2a, L2b, L3a, L3b, and L3c (Natl Sci Rev. 2023 Sep 2;10(10):nwad228). However, our more comprehensive analysis of S. Gallinarum using a larger dataset and hierarchical Bayesian clustering revealed that L3a, previously considered a distinct lineage, is actually a sublineage of L3c. Upon further review of our initial manuscript, we realized that the original submission did not strictly follow the lineage order proposed by Zhou et al. To avoid confusion in the typing system, we have adjusted the lineage nomenclature in the revised manuscript to reflect the corrected order (see Author response table 1).

(4) This reviewer is convinced with the analysis approaches and conclusion of this work.

In the meantime, the authors are encouraged to discuss the application of the conclusion of this study: a) can the data be somehow used in the prediction model? b) would the conclusion from S. Gallinarum have generalized application values for other pathogens. 

Thank you for your constructive comments on the manuscript. 

a) can the data be somehow used in the prediction model?

We believe that genomic data can be effectively used for constructing prediction models; however, the success of such models largely depends on the specific traits being predicted. In this study, we utilized a random forest prediction model based on 196 top genes (PLoS Genet. 2018 May 8;14(5)) to predict the invasiveness of 45 newly isolated strains. In relation to the antimicrobial resistance (AMR) issue discussed in this paper, we also conducted relevant analyses. For instance, we explored the use of image-based models to predict whether a genome is resistant to specific antibiotics (Comput Struct Biotechnol J. 2023 Dec 29:23:559-565). We are confident that the incorporation of newly generated data will facilitate the development of future predictive models, and we plan to pursue further research in this area.

b) would the conclusion from S. Gallinarum have generalized application values for other pathogens.

This might be explained from two perspectives. First, the key role of the mobilome in facilitating the spread of the resistome, as emphasized in this study, has also been confirmed in research on other pathogens (mBio. 2024 Oct 16;15(10):e0242824). Thus, we believe that the pipeline we developed to assess the horizontal transfer frequency of different resistance genes across regions applies to various pathogens. On the other hand, due to distinct evolutionary histories, different pathogens exhibit varying levels of adaptation to their environments. In this study, we found that S. Gallinarum tends to spread highly localized; however, this conclusion may not necessarily hold for other pathogens.

Reviewer #2 (Recommendations for the authors): 

The authors would need to: 

(1) Address my concerns about genomic analyses listed in the public review. 

Thank you for your valuable feedback. We have carefully reviewed your concerns and made the necessary revisions to address the points raised about genomic analyses in the public review. We sincerely hope that these modifications meet your expectations and provide more robust analysis. We appreciate your thoughtful input and remain open to further suggestions to improve the manuscript.

(2) Add more detail on the genomic methods and their outputs, as suggested above. 

We have added further details to clarify the methodologies and outputs as mentioned above. Specifically, we expanded the description of the data processing, and the bioinformatic tools used for analysis. To ensure clarity, we also included an expanded discussion of the key outputs, highlighting their implications. We hope these revisions meet your expectations.

(3) Critically rewrite their introduction to make it clear what problem they are trying to address. 

Thank you for your guidance. In the revised manuscript, we have made the necessary modifications to the Introduction section to more clearly articulate the problem we aim to address.

(4) Critically rewrite their conclusions so they are supported by the data they present, and make it clear when claims are more speculative. 

Thank you for your guidance. In the revised manuscript, we have made the recommended modifications to the relevant sections of the conclusion as outlined above.

More minor issues I identified: 

(1) Typo in the title 'avian-restrict'. 

Done.

Line 1: “Avian-specific Salmonella enterica Serovar Gallinarum transition to endemicity is accompanied by localized resistome and mobilome interaction.”

(2) 'By utilizing the pipeline we developed' -- a pipeline has not been introduced at this point. 

In the revised manuscript, we have removed this section from the 'Abstract'.

Lines 46-48: “Notably, the mobilome-resistome combination among distinct lineages exhibits a geographical-specific manner, further supporting a localized endemic mobilome-driven process.”

(3) 'has more than 90% serovars' -- doesn't make sense. 

Revised.

Lines 82-83: “Salmonella, a pathogen with distinct geographical characteristics, has more than 90% of its serovars frequently categorized as geo-serotypes.”

(4) 'horrific mortality rates that remain a disproportionate burden'. 

Revised.

Lines 83-87: “Among the thousands of geo-serotypes, Salmonella enterica Serovar Gallinarum (S. Gallinarum) is an avian-specific pathogen that causes severe mortality, with particularly detrimental effects on the poultry industry in low- and middle-income countries.”

(5) What is the rate, what is a comparison, how is it disproportionate? 

Thank you for your valuable feedback. It is challenging to accurately estimate the specific prevalence of S. Gallinarum, particularly due to the lack of comprehensive data in many countries. Numerous cases likely go unreported. However, S. Gallinarum is more commonly detected in low- and middle-income countries. Here, we provide three evidence supporting this observation. First, in our previous research, we conducted a comprehensive meta-analysis of 201 studies, involving over 900 million samples, to evaluate the global impact of S. Gallinarum (Sci Data. 2022 Aug 13;9(1):495). The estimated prevalence in 17 countries showed that Bangladesh had the highest rate (25.75%) of S. Gallinarum infections. However, for biovar Pullorum (bvSP), Argentina (20.69%) and China (18.18%) reported the highest prevalence rates. Second, previous studies have also reported that S. Gallinarum predominantly occurs in low- and middleincome countries (Vet Microbiol. 2019 Jan:228:165-172; BMC Microbiol. 2024 Oct 18;24(1):414). Finally, S. Gallinarum was once a globally prevalent pathogen in the 20th century. Following the implementation of eradication programs in most high-income countries, it was listed by the World Organization for Animal Health and subsequently became an endemic pathogen with sporadic outbreaks. However, similar eradication efforts are challenging to implement in low- and middle-income countries, leading to a disproportionately higher incidence of S. Gallinarum in these regions.

In the revised manuscript, we have rephrased this sentence to enhance its accuracy:

Lines 83-87: “Among the thousands of geo-serotypes, Salmonella enterica serovar Gallinarum (S. Gallinarum) is an avian-specific pathogen that causes severe mortality, with particularly detrimental effects on the poultry industry in low- and middle-income countries.”

(6) 'we collected the most comprehensive set of 580 S. Gallinarum isolates', -> 'we collected the most comprehensive set S. Gallinarum isolates, consisting of 580 genomes'. 

Revised.

Lines 97-100: “To fill the gaps in understanding the evolution of S. Gallinarum under regional-associated AMR pressures and its adaptation to endemicity, we collected the most comprehensive set S. Gallinarum isolates, consisting of 580 genomes, spanning the period from 1920 to 2023.” 

(7) Sequence reads are not available, and use a non-standard database. The eLife policy states: 'Sequence reads and assembly must be included for reference genomes, while novel short sequences, including epitopes, functional domains, genetic markers and haplotypes should be deposited, together with surrounding sequences, into Genbank, DNA Data Bank of Japan (DDBJ), or EMBL Nucleotide Sequence Database (ENA). DNA and RNA sequencing data should be deposited in NCBI Trace Archive or NCBI Sequence Read Archive (SRA).' So the sequences assemblies and reads should ideally be mirrored appropriately. 

Thank you for your valuable suggestion regarding submitting the genome data for the newly isolated 45 S. Gallinarum strains. The genome data have been deposited in the NCBI Sequence Read Archive (SRA) under two BioProjects. The “SRA Accession number” for each strain have been added to New Supplementary Table 1. We believe this will ensure that the data are more readily accessible to a broader audience of researchers for download and analysis. We have revised the corresponding paragraph in the manuscript as follows:

Lines 606-608: “For the newly isolated 45 strains of Salmonella Gallinarum, genome data have been deposited in NCBI Sequence Read Archive (SRA) database. The “SRA Accession” for each strain are listed in Supplementary Table 1.”

(8) You should state at the start of the results which data is public, and how much is newly sequenced. 

Revised.

Lines 109-112: “To understand the global geographic distribution and genetic relationships of S. Gallinarum, we assembled the most comprehensive S. Gallinarum WGS dataset (n=580), comprising 535 publicly available genomes and 45 newly sequenced genomes.”

変数の値の変化一覧表

プログラムの動きが見えるツール

使用方法

1.ここをクリック

2.Visualize Executionボタンをクリック

変数の値の変化一覧表

У меня в конце строки.

Пожалуй, зверь тут был намного лучше... Скот же собирательное, в единственном это ругательство. Механического монстра, механическое чудовище, робобыка?

変数の値の変化一覧表

プログラムの動きが見えるツール

1.ここをクリック

2.Visualize Executionボタンをクリック

DOI: 10.7554/eLife.98584

Resource: RRID:BDSC_23691

Curator: @maulamb

SciCrunch record: RRID:BDSC_23691

---

What is this?

DOI: 10.7554/eLife.98584

Resource: RRID:BDSC_11373

Curator: @maulamb

SciCrunch record: RRID:BDSC_11373

---

What is this?

DOI: 10.7554/eLife.98584

Resource: RRID:BDSC_32869

Curator: @maulamb

SciCrunch record: RRID:BDSC_32869

---

What is this?

DOI: 10.7554/eLife.98584

Resource: RRID:BDSC_7063

Curator: @maulamb

SciCrunch record: RRID:BDSC_7063

---

What is this?

DOI: 10.7554/eLife.98584

Resource: RRID:BDSC_4937

Curator: @maulamb

SciCrunch record: RRID:BDSC_4937

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_32713

Curator: @maulamb

SciCrunch record: RRID:BDSC_32713

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_5137

Curator: @maulamb

SciCrunch record: RRID:BDSC_5137

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_27818

Curator: @maulamb

SciCrunch record: RRID:BDSC_27818

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_60688

Curator: @maulamb

SciCrunch record: RRID:BDSC_60688

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_60720

Curator: @maulamb

SciCrunch record: RRID:BDSC_60720

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_24580

Curator: @maulamb

SciCrunch record: RRID:BDSC_24580

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_10046

Curator: @maulamb

SciCrunch record: RRID:BDSC_10046

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_31853

Curator: @maulamb

SciCrunch record: RRID:BDSC_31853

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_23473

Curator: @maulamb

SciCrunch record: RRID:BDSC_23473

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_33095

Curator: @maulamb

SciCrunch record: RRID:BDSC_33095

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_10975

Curator: @maulamb

SciCrunch record: RRID:BDSC_10975

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_33132

Curator: @maulamb

SciCrunch record: RRID:BDSC_33132

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_23638

Curator: @maulamb

SciCrunch record: RRID:BDSC_23638

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_23842

Curator: @maulamb

SciCrunch record: RRID:BDSC_23842

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_24590

Curator: @maulamb

SciCrunch record: RRID:BDSC_24590

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_41744

Curator: @maulamb

SciCrunch record: RRID:BDSC_41744

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_56583

Curator: @maulamb

SciCrunch record: RRID:BDSC_56583

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_7673

Curator: @maulamb

SciCrunch record: RRID:BDSC_7673

---

What is this?

DOI: 10.7554/eLife.101439

Resource: RRID:BDSC_6596

Curator: @maulamb

SciCrunch record: RRID:BDSC_6596

---

What is this?

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

DOI: 10.1371/journal.pbio.3002941

Resource: Bloomington Drosophila Stock Center (RRID:SCR_006457)

Curator: @maulamb

SciCrunch record: RRID:SCR_006457

---

What is this?

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 182, in init if 'link' in row['document']: TypeError: argument of type 'NoneType' is not iterable

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 182, in init if 'link' in row['document']: TypeError: argument of type 'NoneType' is not iterable

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 182, in init if 'link' in row['document']: TypeError: argument of type 'NoneType' is not iterable

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers

Traceback (most recent call last): File "/home/ubuntu/dashboard/py/create_release_tables.py", line 54, in format_anno_for_release parsedanno = HypothesisAnnotation(anno) File "/home/ubuntu/dashboard/py/hypothesis.py", line 231, in init self.links = row['document']['link'] TypeError: string indices must be integers