On 2019-07-09 19:52:17, user TeresitaMPorter wrote:
GitHub repos supporting this ms have been updated:
SCVUC bioinformatic pipeline:<br /> https://github.com/Hajibaba...
Code for figures, infiles, supporting files:<br /> https://github.com/terrimpo...
On 2019-07-09 14:21:02, user Rémi Allio wrote:
MitoFinder can be downloaded from:<br /> https://github.com/RemiAlli...
On 2019-07-09 13:54:28, user Sebastian Wicha wrote:
This is an important article, and differentiating between efficacy and potency-based pharmacodynamic interactions is highly important.
Unfortunately, the authors of the MuSyC paper have missed to include the General Pharmacodynamic Interaction (GPDI) model, published in Nature Comms in 2017, into their comparison. The GPDI model can also differentiate between efficacy and potency-based interactions and is compatible with Bliss Independence, Loewe Additivity and other additivity criteria.
Link to the GPDI paper: https://www.nature.com/arti...
On 2019-07-09 10:43:04, user Siri Leknes wrote:
Hi, interesting and important work! <br /> BUT: morphine is not the main active metabolite of heroin, 6AM is! See work by Boix, Kvello and others
Also I miss discussion of how to determine “equipotent” opioid doses
On 2019-07-09 10:31:05, user Waldeyr Mendes Cordeiro da Sil wrote:
It seems promising! I going to test it on my data soon! Thanks for this contribution.
On 2019-07-09 10:18:30, user Peter Manning wrote:
Hi Noemie- you're probably aware of it but we took a similar approach, but no SEM, in an old Ecotron experiment. Details here: https://www.sciencedirect.c...
On 2019-07-09 09:29:52, user Martin Steppan wrote:
A very fascinating study and topic with interesting implications! If I understand the results and design correctly, the genetic instruments quantifying the risk for early menarche also grasp a considerable amount of socioeconomic information (by being associated with educational attainment). Would it not be straight forward to control outcome variables (like age at first birth) for socioeconomic differences on the observational level first, before conducting Mendelian Randomization? MR is powerful, but if you have real confounding variables, is there a rationale not use them?
On 2019-07-08 23:53:42, user Bailey Ing wrote:
An insightful analysis in the equilibrium of Sparrowhawk and prey species. Migration and population change of a species is multi-factorial and limitations of each study including that of Newson et al, concerning interpolated annual frequency surfaces on predator abundance must be considered. This paper highlights the need for further sampling from data sets further afield for a comprehensive picture on the ecological impact and interaction of bird species.
On 2019-07-08 20:37:19, user Igor Schneider wrote:
Now in press at https://www.pnas.org/conten...
On 2019-07-08 19:34:01, user Ben B wrote:
A version of this manuscript has been accepted for publication and is available here: <br /> https://link.springer.com/a...
On 2019-07-08 17:22:54, user Anna Schwabe wrote:
In response to George Weiblen
George Weiblen wrote: We suggest that readers of this article might also consider reasons to question the central claim that NIDA cannabis is genetically closer to hemp than to marijuana.
We respond: In our sampling, NIDA was found to be genetically closer to the hemp-type samples than the marijuana-type samples. This was seen in multiple analyses. We do not claim that NIDA is supplying hemp for marijuana research, but we are confident that our analyses show that the research grade marijuana supplied by NIDA is genetically different from the retail marijuana samples analyzed in this study.
“Our results clearly demonstrate that NIDA Cannabis samples are substantially different from most commercially available drug-type strains, sharing a genetic affinity with hemp samples in most analyses.”
“…this study highlights the genetic difference between research grade marijuana provided by NIDA and commercial Cannabis available to medical and recreational users.”
Given both genetic and chemotypic investigations have concluded that NIDA is supplying product that does not align with what is available for consumers, our hope is that NIH, NIDA, and the University of Mississippi take this into consideration. Medical practitioners, researchers and patients deserve access to marijuana products that reflect the products available on the legal market.
George Weiblen wrote: First, the authors do not report the cannabinoid profiles of their samples, so it is unclear whether the NIDA samples are marijuana-type, hemp type, or intermediate, nor did they specify the batch number. The NIDA Drug Supply Program has materials available from all major varieties of cannabis.
We respond: The lack of inclusion of sample batch numbers was an oversight on our part. The “research grade marijuana” plant material samples were labeled as: <br /> 1. High THC: RTI log number 13494-22, reference number SAF 027355.<br /> 2. High THC/CBD: RTI log number 13784-1114-18-6, reference number SAF 027355.
One of the aims of the study was to determine where the NIDA samples fell on the genetic spectrum of Cannabis types. The phytochemical content was not considered in this study because it is widely known that phytochemical constituents change due to environmental conditions, which include age, storage conditions, and storage temperature.
Furthermore, the samples from NIDA were ordered and are labeled as “research grade marijuana”, which should need no further investigation into whether the samples received were indeed marijuana-type, hemp type, or intermediate.
George Weiblen wrote: Second, there is inconsistency between the individual-based metrics and population-based metrics. Statistics for population subdivision (Fst) and genetic distance (Nei's D) in Table 1 do not agree with Figures 1-4 in supporting the central claim. For example, the Fst values of the NIDA samples are more differentiated from hemp than they are from the three drug-type subclasses. According to Nei's D, the NIDA samples are more similar to "hybrid" and "indica" drug-types than they are to hemp. The authors point out that the small sample size (N = 2) of NIDA varieties in their study is not sufficient to accurately estimate population-level parameters so they emphasize the individual-based results instead. This represents a bias on the part of the authors, who could request more samples from the NIDA Program to improve their sample size.
We reply: The “populations” are not true populations per se, but rather are commonly referenced usage groups. Given the high degree of hybridization among these groups, we do not necessarily consider the six groups as unique and separate populations. Hemp and drug-type cannabis groups have consistently been found by several studies to be genetically separate, and we feel these may be considered populations, but the rise in cannabidiol popularity has led to the development of several lines that are hybrids between the two types.
There were only two types of Cannabis from NIDA because that is what we had access to through one of our co-authors. We are not opposed to incorporating more NIDA samples into our analyses if NIDA would like to provide them. However, we feel that the two samples we examined have an interesting genetic profile given this is what was supplied to researchers conducting marijuana research and will possibly inspire further investigation of additional material supplied by NIDA.
George Weiblen wrote: The authors place more rhetorical weight on the individual-based approach by devoting four figures to it. It also possible, however, that the greater number of similar individuals in the drug-type samples could exaggerate their separation from much smaller numbers of NIDA and hemp samples of individuals across all four individual-based metrics.
We respond: The drug-type and hemp-type samples are grouped as such because that is how they were presented. However, given hybridization levels and wide variation in THC/CBD levels, as well as over-reporting of these levels, we feel that, even though we grouped them as such, examining genetic relationships at the individual level rather than population level was appropriate for this investigative study.
In some cases, drug types fell out with the hemp type samples, and is likely an ancestral artifact given these analyses are among individuals within species. The opposite is not true of the hemp group- no sample designated as hemp had substantial genetic signal associated with the drug-types (<15%).
The individuals in the drug-type group are not all that similar in description, reported THC content, or genetically. We sampled a wide range of available strains and feel this appropriately represents the groups. We have 9 hemp samples (including ruderalis), 11 sativa, 14 hybrid, 10 indica, which is a good representation of each of these groups. The 3 CBD samples we expected to be hybrids of hemp and drug-types, which they were, and we feel although this group is small, we again reiterate the groups are artificial.
George Weiblen wrote: An even stronger potential artifact has to do with the microsatellite genotypes themselves as presented in the supplementary table. The hemp samples all have considerable missing data whereas no data is missing from the drug-type samples and the two NIDA samples have a large number of private alleles. It appears that most of the signal assigning the NIDA samples to hemp are due to alleles at only three of the ten loci. Complete microsatellite panels and preferably more NIDA samples are needed to evaluate the preferred interpretation.
We respond: We did not assume to assign the samples from NIDA as hemp, but rather made the observation and conclusion that the plant material supplied from NIDA labeled as “research grade marijuana” does not align genetically with marijuana available on the retail market. In fact, it is quite different, as indicated by the presence of private alleles. We are aware that there are three loci are contributing to the majority of the divergence between NIDA and drug-type samples. Considering that 3 loci represent 30% of our marker regions, this divergence is substantial. Private alleles are commonly used in population genetic studies to identify divergent groups. Although the inclusion of additional NIDA samples would be beneficial, additional sampling would in no way change the genotypes of the samples we have included in this study.
Regarding the missing data, we are attempting another round of reruns to fill in some of the missing data, some of which we have retrieved. We will include this data prior to publication.
Anna Schwabe and Mitchell McGlaughlin, University of Northern Colorado
On 2019-06-12 18:17:51, user George Weiblen wrote:
We suggest that readers of this article might also consider reasons to question the central claim that NIDA cannabis is genetically closer to hemp than to marijuana.
First, the authors do not report the cannabinoid profiles of their samples, so it is unclear whether the NIDA samples are marijuana-type, hemp type, or intermediate, nor did they specify the batch number. The NIDA Drug Supply Program has materials available from all major varieties of cannabis.
Second, there is inconsistency between the individual-based metrics and population-based metrics. Statistics for population subdivision (Fst) and genetic distance (Nei's D) in Table 1 do not agree with Figures 1-4 in supporting the central claim. For example, the Fst values of the NIDA samples are more differentiated from hemp than they are from the three drug-type subclasses. According to Nei's D, the NIDA samples are more similar to "hybrid" and "indica" drug-types than they are to hemp. The authors point out that the small sample size (N = 2) of NIDA varieties in their study is not sufficient to accurately estimate population-level parameters so they emphasize the individual-based results instead. This represents a bias on the part of the authors, who could request more samples from the NIDA Program to improve their sample size.
The authors place more rhetorical weight on the individual-based approach by devoting four figures to it. It also possible, however, that the greater number of similar individuals in the drug-type samples could exaggerate their separation from much smaller numbers of NIDA and hemp samples of individuals across all four individual-based metrics.
An even stronger potential artifact has to do with the microsatellite genotypes themselves as presented in the supplementary table. The hemp samples all have considerable missing data whereas no data is missing from the drug-type samples and the two NIDA samples have a large number of private alleles. It appears that most of the signal assigning the NIDA samples to hemp are due to alleles at only three of the ten loci. Complete microsatellite panels and preferably more NIDA samples are needed to evaluate the preferred interpretation.
George D. Weiblen and Jonathan Wenger, University of Minnesota<br /> Mahmoud ElSohly, University of Mississippi
On 2019-07-08 16:34:50, user Muhao Chen wrote:
Now online in Bioinformatics: http://dx.doi.org/10.1093/bioinformatics/btz328
On 2019-07-08 15:25:20, user Zhuang Wei wrote:
Good model. May be the best is to FIT the model curve to real measured living cell intensity changing of Plk4, sas6 and STIL. It may help to verify the model.
On 2019-07-08 15:15:30, user Marco wrote:
sounds this is not yet published or? but Slide-seq is out
On 2019-07-08 12:45:18, user Dave Lahr wrote:
Hello - I'm wondering if you have / can provide the data (or location of the data) that indicates for each cell line whether or not it is MSI?
On 2019-07-08 04:08:39, user Fraser Lab wrote:
This work describes the implementation of a data processing pipeline for acquiring high-resolution maps of microtubules (MTs) from cryo-electron microscopy (cryoEM) data using the RELION software. As in other pipelines for processing microtubule EM data, this implementation requires extensive custom processing because of the pseudosymmetric nature of most MTs assembled in vitro (also observed in vivo): a "seam" down the length of the assembly disrupts the otherwise helical symmetry. The broken symmetry means that existing methods for processing purely helical particles equate nonequivalent positions and produce low-quality reconstructions. The authors implement a treatment of these particles that accounts for the seam and produces high-resolution structures of the MT α and β asymmetric units. It builds on implementations of similar pipelines for the same purpose using other software, with the key advantage of conducting all steps in a single program that most cryoEM users are already familiar with. The pipeline consists of a set of scripts and a series of steps the user should complete in the RELION graphical user interface (gui) in order to obtain the asymmetric unit reconstructions. The authors test their pipeline on three example datasets with different decorators on the α or β subunits that aid in initial alignment and discrimination between the two, and note that they have successfully (with minor modifications) applied the pipeline to a more challenging dataset with both subunits decorated.
The major success of this paper is the clear and thorough description of the steps necessary to produce high-resolution α and β subunit reconstructions, complete with clear justification for each step and descriptions of expected results so that an advanced user can intervene when intermediate results deviate from expectations. This tool meets an immediate need in the structural biology community for analysis of MTs, which are inadequately reconstructed from cryoEM data by existing helical or strictly single-particle methods, and which play an important role in the cell interacting with a variety of other molecular machines. Ideally, it would be benchmarked against the other pipelines mentioned in the paper (e.g. https://github.com/nogalesl... from: https://www.ncbi.nlm.nih.go... but as one of us (JSF) knows from personal experience that it can be tricky to set up the necessary EMAN and frealign environments correctly to do such benchmarking properly. Here, the ability to complete this analysis without exporting steps to other programs could be a major boost in accessibility. Moreover, as the authors have built upon a popular cryoEM image processing program that has a gui highly accessible to novice and intermediate users as well as command-line tools that expert users may use for more advanced customizations and interventions, we anticipate this pipeline will be enthusiastically adopted by many users.
We also applaud the authors' choice to make the scripts open-source and publicly available on github, which will facilitate the active conversation between users and developers (and sometimes software developers who did not author the original work) that lead to major breakthroughs and advancements in later versions. However, we cannot comment on the scripts themselves yet as a full path to the source code is not provided in the manuscript and a search for “MiRP relion github” didn’t yield anything informative. We would like to request the authors include it in the revised manuscript and also provide it to us during the review process so we may evaluate this important component of the work. We recommend using Zenodo (https://zenodo.org/) to generate a DOI for a snapshot of the repository, which will also produce a timestamp and facilitate formal versioning.
We identify a few major weaknesses in the manuscript in its current form, all of which we hope the authors can address in a revision. First, the final, high-resolution reconstruction is the αβ dimer, not the C1 reconstruction of a full helical turn, which may not serve the goals of all users. The authors identify the final averaging step that disrupts the density for all but the αβ dimer directly opposite the seam and describe alternative approaches, including one implemented by the Carter group while the manuscript was in preparation. We would strongly encourage them to implement one of these approaches so that biological questions that require examination of the whole MT can also be addressed. We are also unclear on how the present implementation would preocess both the microtubule and fiducial protein for datasets with dynein or EB3 bound, and would like to see this explicitly discussed (or better, tested if EMPIAR datasets are available).
Second, at times the authors describe what they expect from data they have not processed, for example on page 14 lines 1-4. Given that they have the necessary tools in-hand and this work describes the method, we would press them to test this type of claim and describe the supporting evidence. They have also described processing a dataset with fiducial markers on both α and β monomers but not described the modifications they had to make to the pipeline for this dataset, and they have not yet (to our knowledge) used the method to process undecorated MTs. As they cite successful processing of undecorated MTs by the Nogales group, proof-of-concept processing of undecorated MTs would be an important component of making this pipeline at least as useful as existing methods. The other case where we would strongly prefer to see the authors test their claims is on page 12 lines 46-48, where they speculate on the effects of excluding some MTs from further analysis, and although the authors do not make any claims about performance using automated MT picking, we would be very interested to see this tested (even if the result is that it is discouraged).
Third, while we agree with the statement: “strikingly, although application of MiRP compared to standard helical processing has a negligible effect on the reported reconstruction resolution by FSC (Fig. 6c), the structural details are clearly superior in quality,” based on the snapshots of density shown at a specific contour in Fig 6 and Supp Fig 3, it is possible to use tubulin model refinement and other quantitative evaluations of the map to validate this statement. For example, compared to the standard helical map we would expect a higher quality reconstruction to have a smaller convergence rmsd when multiple independent models are rebuilt into the map (as in: https://www.ncbi.nlm.nih.go... a better EMRinger score (https://www.ncbi.nlm.nih.go... when evaluated with the same starting model rigid body fitted into the map; and lower B-factors/better model geometry when an atomic model is refined.
Fourth, reiterating that this is a methods paper, we find it critical that the raw data be made available on EMDB for all datasets described in the manuscript, not just the C1 reconstructions and symmetrized asymmetric units. This is important for reproducibility, open science, and the development of exciting new methods like this one using publicly available test data.
In sum, we find this an important piece of work that will immediately improve the ability of groups working on MTs to recover high-resolution structures, pending our several major reservations that we hope the authors will resolve in a revision. We also identify several minor points that could be improved, mostly regarding readability, and a few suggestions for alternative implementations of some steps in this or a future version of the pipeline. As the line numbers do not appear to be spaced the same as the text, for these points, we have indicated the closest line number to the line when printed.
Some figures or tables referenced in the main text are not referenced correctly, such as on page 9 line 31, Fig. 6aii (referencing a panel that does not exist).
Table 1 should include accession codes in the EMDB.
The authors might comment on the biological relevance of MTs with seams, given that these are more often encountered in vitro and much less often in vivo, preferably in the introduction.
We suggest a figure that visually highlights the symptoms of misalignment of the seam and/or helical averaging of MTs with seams. Including correlation coefficients with this figure could help illustrate the challenge this pipeline overcomes. This could also illustrate the signal boost of the superaverage and the symptoms of out-of-register units. The figure could be referenced at several points later in the text to explain why certain steps are necessary.
At the end of the first paragraph on page 6, the authors describe "structural constraints of MT polymers" but apply restraints in orientational and translational searches. It would be helpful to expand on the rigidity of these restraints and whether it varies with distance to further neighbors, if applicable. Ideally (or possibly in a future version), the authors could consider restraining each α or β monomer relative to its immediate neighbors and using this approach in combination with variable restraint rigidity to aid in reconstructions of monomers at the seam and in distorted regions.
Several points regarding resolution starting at the end of page 6 and continuing in the first paragraph on page 7 describe increments of resolution or changing pixel size by binning, which have no meaning in isolation (e.g. a difference of 0.2 Å or binning x 4). The starting or ending absolute quantities should be included (e.g. improvement of the resolution to 3.2 Å or a final pixel size of 3 Å). This is repeated on page 11 lines 11-12.
The authors describe that "there is a clear bias towards a certain range of Rot angles" on page 8 line 16, but as this is the expected behavior and not a ground truth, it should be described as such.
Similarly, on page 9 line 19, the authors intermix behavior on their test data ("As expected") with description of the method, and should more clearly separate these.
On page 8 lines 18-22, the authors describe their approach for using the bias toward one Rot angle to select the correct seam location. We recommend testing the alternative method of a grid search over correlation coefficients, or describing how this is effectively accomplished during the global search step.
On page 8 lines 21-22, the result of the Rot search is described as an approximation. It would be helpful to clarify whether this result is precise but sometimes inaccurate, or accurate but known to be imprecise.
On page 8 in the section on X-Y shift smoothing, the authors describe a remedy for out-of-register asymmetric units involving resetting excessively large shifts to zero and re-refining. We propose an alternative method by analysis of the distribution of X-Y shifts that identifies the out-of-register shift vector and adjusts excessively large shifts by modulo arithmetic. This would reduce the error in the reset shifts.
As part of the same description on page 8 lines 47-48, the authors describe enforcing all X/Y shifts in a MT to follow a single slope and intercept, and should clarify whether this is constrained or restrained.
The authors could expand on the process of 'segment average' image generation on page 4 line 29 and the source of known helical parameters on page 4 line 33.
The sample preparation for cryoEM section starting on page 3 could include greater detail, e.g. Vitrobot parameters during blotting and freezing.
The authors could clarify the difference between defects and switches in PF number on page 7, lines 27-28.
The description of a 'clean' seam on page 10 line 48 is confusing. Describing this as a MT with no seam might be clearer, if that is the correct interpretation.
There are a couple creative uses of the word "allocation" — on page 9 line 6 we suggest substituting with "positioning" and on page 13 line 22 we suggest substituting with "assignment".
The wording on page 5 line 52 seems to imply the RELION nomenclature preceded the Euler angle nomenclature used in many other applications, so we recommend dropping the modifier "former".
The wording on page 7 line 5 implies the previously implemented approaches are lacking in some way, and we recommend dropping the modifier "albeit". This is repeated on page 12 line 42.
The authors could describe which of the operations through the gui could alternatively be run on the command line on page 5, lines 59-60.
On page 13 line 21, the authors claim their implementation is "the only way to avoid introducing artefacts." This may be an overly bold claim.
We are unsure what the authors mean to do with the "41 Å shifted positions" on page 9 line 7.
Some of the word choices could be made more accessible to all readers, for example by using the more common and equivalent "while" instead of "whilst" in several instances.
The sentence "Initial Tilt ... picking coordinates" on page 6 lines 54-56 is unwieldy and could be rephrased.
We find the sentence beginning "In other words" on page 8 lines 2-4 redundant and unnecessary.
The qualifier "data collection parameters" should not accompany ice thickness on page 9 line 42.
We would prefer sticking to one set of units on page 10 line 1 and substituting "sub-10 Å" for "subnanometer".
The abbreviation "(DQE)" on page 9 line 43 is not used again and may be omitted.
There are several spacing errors throughout the text. Between a numeral and a unit, there should be no space, except where the next character is º or %.
On page 8 line 4 and in several other instances, where "however" is an interjection, it should also be preceded by a comma, e.g. "this register is, however, very error prone".
There is a typo on page 3 line 52 (1 mg/ml --> 1 mg/mL), an incorrect abbreviation on page 4 line 6 (sec --> s), an overly dense abbreviation on page 4 line 59 (4xbin), a typo on page 4 line 53 (smoothened --> smoothed), a typo on page 5 line 14 (psi/tilt/ranges --> Psi/Tilt ranges), use of redundant "around" and "~" modifiers on the same quantities on page 6 line 53, incorrect pluralisation on page 7 lines 33-34 (confidence --> confidences), an unnecessary word "score" on page 8 line 7, a missing word "good" in "good signal to noise and good angular distribution" on page 9 line 33, an unnecessary hyphen in "ice-thickness" on page 9 line 42, and an unnecessary comma in "reconstructions, remains" on page 9 line 56.
We review non-anonymously, Iris Young and James Fraser (UCSF).
On 2019-07-07 18:59:07, user Ben wrote:
Nice structures, it should be noted that absolutely all viruses are dynamic to be able to function as molecular machines, otherwise they would not be able to assemble/disassemble and interact with their hosts. One should not describe a system as 'highly dynamic' based on few distinct conformations seen in high-res structures, be it EM or crystallography.
On 2019-07-07 18:08:28, user anon wrote:
Lovely structures, but are they really dynamic? Dynamic implies that your structures are actually moving (i.e. there is some time-resolved aspect to your analysis), whereas your cryoEM merely shows that they are flexible.
On 2019-07-07 17:04:45, user Jiarui wrote:
Nice work! Thank you for the tremendous efforts of comparing all these methods! However, I think that different algorithms accept different inputs. For example, scvis uses principal components instead of raw-counts as inputs, otherwise, the error models and the outputs do not make any sense. Typical t-SNE implementations also either explicitly or implicitly do PCA first, and use the top PCs, e.g., 30 PCs as inputs.
On 2019-07-07 15:07:26, user Robert Gourdie wrote:
This manuscript was accepted for publication pending minor revision in the Journal of the American Heart Association July 3, 2019.
On 2019-07-07 02:53:42, user jeff ellis wrote:
the N termini of flax L6 and M proteins have been shown to direct to specific sub cellular locations. this may be the function of all plant NLR N termini
On 2019-07-07 00:13:13, user Sonya Aldret wrote:
Good paper. How long time of CDNF perfusion? CDNF induces some ECG electrical distubance during baseline?
On 2019-07-07 00:03:23, user Buert Sohrrem wrote:
The amounts of CDNF released from ischemic/ reperfused hearts are sufficient to generate protection in another heart?<br /> Ischemic preconditioning is able to increase CDNF secretion? Does the activation of the KDEL receptor by the CDNF have any genomic effects?
On 2019-07-02 18:41:34, user Wyllian Schartchter wrote:
I was impressed with the effect of CDNF on the transient calcium. Could CDNF have any effect on Ryanodine receptors?<br /> Could the activation of the KDEL receptor act on the Ryanodine receptor?<br /> Great job.Your research tells a story from beginning to end.<br /> WS
On 2019-07-02 15:20:12, user Wyllian Shatlecher wrote:
I was impressed with the effect of CDNF on the calcium transient. Could CDNF have any effect on Rianodine receptors?<br /> Could the activation of the KDEL receptor act on the Reanodine receptor?<br /> Great job. Your research tells a story from the beginning to end.
On 2019-06-27 18:30:25, user Morrisonw Ferreds wrote:
The authors attempted to co-immunoprecipitate the binding between CDNF with the KDEL receptor?
On 2019-06-27 18:24:53, user Downey wrote:
Still not entirely clear whether the activated receptor is the KDEL receptor. If this is really the receptor, it will be a noncanonical function of the receiver. Congratulations.
On 2019-07-06 16:46:46, user Valentina Carreno wrote:
Hi, <br /> I'm analyzing the stage 1 process of the binding score prediction, I'm confused after the first convolution filter how they got the 96x30 matrix. I understand theres 96 filters but where is the 30 from? I'd greatly appreciate if anyone could help me.
On 2019-07-06 06:07:10, user Andre wrote:
A potential gamechanger!
On 2019-07-05 22:41:36, user Charles Warden wrote:
Figure 2 in your paper reminds me of Figure 2 in Yizak et al. 2019, except I think that other paper's strategy of showing the range of absolute counts has a more intuitive interpretation than showing residuals (even though I see that you are trying to correct for multiple factors, and you want to show that age affects the differences among the skin-exposed samples)
Was there any communication between the labs? At this point, it might be good to cite that other paper. However, I think that paper could have possibly benefited from waiting a little longer and possibly being in a lower impact journal (and/or being split into separate papers):
http://cdwscience.blogspot....
In general, having similar conclusions from different studies should help provide confidence (although you are both using the same dataset, so it isn't really independent validation). However, I think there can also be value if you take more time to do things carefully (and there can be other factors that delay paper submission). So, I think this is something interesting to think / talk about either way.
On 2019-07-05 11:35:18, user Carolyn Lawrence-Dill wrote:
It's submitted and under review! Comments here would also help, so please do give us some feedback if you have ideas.
On 2019-07-05 06:18:41, user David Lawrie wrote:
Very cool paper. :) There are some previous human population genetics studies that should be cited in the Discussion as they are germane to your argument about selection on synonymous sites: https://www.genetics.org/content/188/4/931.full and https://genome.cshlp.org/content/28/10/1442.short. The following paper was submitted to biorxiv around the same time as yours, but likewise could be included in future revisions: https://www.biorxiv.org/content/10.1101/441337v1.full. Note that none of these papers relate selection to codon optimality as yours does - in fact, the second paper is looking specifically at splicing. However, these papers do find evidence for selection on synonymous sites in humans using variant data and would thus be good to cite.
On 2019-07-04 16:20:49, user Ilia Kats wrote:
Hi,
awesome paper. I noticed two mistakes however:
1) The R_G loss function (page 20) should read $R_{\mathcal{G}} = <br /> \sum_{n=1}^N\left( -Y_n\log(\mathcal{G}) - (1 - Y_n)\log(1 - <br /> \mathcal{G})\right)$ . You are passing the output of a sigmoid <br /> activation (i.e. 1/(1 + exp(-G))to Keras's binary_crossentropy loss, <br /> which converts<br /> it back into logits (i.e. recovering G) before passing it to <br /> TensorFlow's. sigmoid_cross_entropy_with_logits
2) The simplification of KL-Divergence (page 16) should read $2 + \sum_n p_n \log p_n$
On 2019-07-04 11:27:21, user Rahmi Lale wrote:
If you have any comments, questions or opinions please drop us a line.<br /> Rahmi Lale
On 2019-07-04 07:11:31, user Takanori Eguchi wrote:
This work was finally published in "Cancers (Basel) 2019, 11(6), 792.<br /> https://doi.org/10.3390/can...<br /> https://www.mdpi.com/2072-6...
On 2019-07-03 18:31:06, user Charles Warden wrote:
1) A peer-reviewed version of this article is now available:
https://bmcgenomics.biomedc...
2) Since BMC Genomics doesn't have a Disqus comment system, I apologize for posting this here (although I hope this changes in the future).
However, I noticed "HPV" was defined as an abbreviation in the peer-reviewed version, without being used in that version of the paper.
In the pre-print, I do see some analysis of "72 TCGA HNSC tumor samples with valid Human Papillomavirus (HPV)," but it looks like they changed the comparison to tumor-versus-normal (while accidentally keeping the abbreviation in their draft).
On 2019-07-03 10:38:50, user Kimon Frousios wrote:
First impression: The abstract reads like "Tools based on different models capture our simulated truth less well than the analysis method we created specifically using the same model as the simulation." This would be true even for a blatantly wrong model, so it is not a good way to introduce your work.
On 2019-07-03 08:01:16, user i came here from twitter wrote:
great paper! Just wanted to let you know that you have left a TODO note in the Star Methods: "(add picture in supplementals of courtship arenas build by Daniel)"
On 2019-07-03 03:16:52, user nkb wrote:
Great paper, Dr. Gunes! Very small comment: I believe there is a typo in the right panel of supplemental figure 1B. The legend indicates there is a significant correlation, but the stat doesn't correspond.
On 2019-07-03 00:37:40, user Charles Warden wrote:
Thank you for putting together this pre-print.
The methodology is different, but I think there are also some differences in terms of the overall conclusions with a study that I did in yeast (using EvoFold and RNAz to predict functional RNAs in yeast).
There are some things that I probably would have done differently in retrospect (or at least would have done in an ideal situation), so there is actually the paper as well as a fairly long comment:
https://journals.plos.org/p...
https://github.com/cwarden4...
In terms of overall conclusions, these would be my questions:
1) How are you defining "widespread" (versus "substantial," "significant," etc.)?
For example, could you do something like Figure 1 with your strategy? This would show there are a non-trivial number of functional secondary structures within coding regions (at synonymous sites), but the coding percentage is higher in yeast versus other Eukaryotes.
In that PLOS ONE paper we say "We found that EvoFold had a greater propensity to predict coding fRNAs in vertebrates than in yeast. A previous study for conserved fRNAs in the human genome using the EvoFold program found that 23% of the predicted fRNAs were found within coding regions [3]. For comparison, only 18% of the comparative data set used in this study was coding regions (as measured by the proportion of phastCons elements found within coding regions) [13]. In contrast, we found only 33% of fRNAs in coding regions of yeast, which contain 86.1% of the phastCons elements [13]. Another way to understand this comparison is to note that 303 coding fRNAs were found in 65,348 phastCons CDS blocks in yeast while 12736 coding fRNAs were found in 23,580 phastCons CDS blocks in vertebrates [3], [13]. Thus, in terms of the proportion of coding fRNAs to phastCons CDS blocks, coding fRNAs are about 10 times more likely to be found in vertebrates than yeast.". If this is true, I would be concerned that "widespread" is not the best choice of words for this result.
2) The main functional enrichment result from that paper is for ribosomal proteins (and that is perhaps made even more clear with Comment Figure C1, at least in terms with the association with CAI), but that seems different than your results.
While I am not exactly sure what to recommend checking with your study, perhaps you can check associations with CAI (Codon Adaptation Index, for codon usage)? The ribosomal proteins tend to have high CAI - at least on a log-scale that makes something that is relatively easy to visualize.
3) It's not really as related to the Coding fRNA paper, but perhaps the individual points and/or density distributions can be added to Figure 1B and Figure 1C?
Currently, the p-value seems low, but the difference is not highly predictive. If it were possible to show some sort of functional consequence, that might help showing there is a biological difference (and not just a statistically significant difference). However, the experimental validation is also something that I wasn't able to include in that earlier paper (although I do think it would be best if that was planned for future projects).
On 2019-07-02 22:50:21, user Charles Warden wrote:
I thought this result was interesting in terms of emphasizing the value of replicates, even with scRNA-Seq (in Figure 2B, for Acne / GA / Psoriasis).
I apologize that I had some difficulty figuring this out with quick read-through of the paper, but are these biological replicates (different patients and/or collection sites) or technical replicates (such as varying batches or total sorted cell number)?
On 2019-07-02 20:45:49, user Yichao Li wrote:
Hello,
Is there any reason that you did, "After alignment, fragments that were longer than 120bp were filtered away". I know the original CUT&RUN paper treated 120bp and 150bp separately. Still, reads above 120bp could mean something, right?
Can you also compare your peak caller with this: Improved CUT&RUN chromatin profiling and analysis tools? They claimed that MACS2 could have many false positive on cut&run data.
Thanks,<br /> Yichao
On 2019-07-02 16:52:42, user Yeh Lab wrote:
Beautiful work, unexpected mechanism, and important finding! Thanks so much for sharing on preprint with the community!
On 2019-07-02 14:48:53, user DVB at LCSB wrote:
The article is now available online at: https://www.nature.com/arti...
On 2019-07-02 10:02:34, user Yo Yehudi wrote:
Hey all - really nice to see the use of HumanMine GO enrichment in your <br /> paper! I'm one of the development team for InterMine, who runs <br /> HumanMine. I was wondering if you'd be willing to help us out by citing <br /> us where you're quoting HumanMine? We have citation guidelines here: <br /> https://intermineorg.wordpr...
On 2019-07-01 23:23:49, user Charles Warden wrote:
I guess it doesn't really need to be a comment (since Tweets are listed above), but I did have an additional question:
https://twitter.com/cecilej...
Even though the y-axis is missing, am I correct that the x-axis is supposed to be the p-value? If so, then why don't all the distributions overlap when the p-value is equal to 1.00 (on the right side)?
On 2019-07-01 19:48:25, user Julius Adler wrote:
July 2, 2019: some changes to April 18, 2019
Drosophila Mutants that Are Motile but Respond Poorly to All Stimuli Tested Mutants in RNA splicing and RNA Helices, Mutants in The Boss
Lar L. Vang and Julius Adler
The following idea was presented in 2011 in “My Life with Nature” by Julius Adler, p. 60:
“Recently I conceived a new idea. “The Boss is the thing inside every organism – humans, other animals, plants, microorganisms – that is in charge of the organism. I don’t mean this in any mystical or spiritual or religious sense, but rather I mean it in terms of chemistry and physics. You may think that The Boss is a wild idea, and certainly the evidence for it is poor, but I think it’s true, and at least it’s a hypothesis to be tested.”
Now we have tested this idea:
Adler and Vang (2016) and Vang and Adler p. 13, 2018) reported Drosophila mutants that lack all responses to external and internal stimuli at 34 degrees but at room temperature these mutants are not deficient. This means that activity by the Boss can be eliminated at 34 degrees but the activity is still present at room temperature.
And they (Vang and Adler, 2016) reported a Drosophila mutant that lacks responses to all stimuli tested at both 34 degrees and room temperature. That indicates that this mutant lacks behavioral action by The Boss.
(It must be admitted that the defects in these mutants were caused by defects in The Boss.)
What is The Boss? It is a mechanism that acts as described in Figure 10 of Adler, 2016:
https://uploads.disquscdn.c... https://uploads.disquscdn.c...
Fig. 10 of Adler, 2016
The idea that each organism has something in control of the organism is novel. Before this, it was believed that each organism has properties that are largely independent of each other. Now it is suggested that all the properties are controlled by a single factor, The Boss, which directs both the interior and the outside of the organism. The Boss is to be found in humans, other animals, plants, and microorganisms. The evidence for this idea is incomplete.
Adler J (2011) My life with nature. Ann Rev Biochem 80 42-70.
Adler J (2016). A search for The Boss: The thing inside each organism that is in charge. Anat Physiol Biochem Int J Vol.1, 2016.
Adler J, Vang LL (2016) Decision making by Drosophila flies. bioRxiv March 24, 2016.
Vang LL, Adler J (2018) Drosophila mutants that are motile but respond poorly to all stimuli tested: Mutants in RNA splicing and RNA helices, mutants in The Boss. bioRxiv October 1, 2018.
On 2019-06-26 15:30:51, user Julius Adler wrote:
June 27, 2019: some changes to April 18, 2019
Drosophila Mutants that Are Motile but Respond Poorly to All Stimuli Tested Mutants in RNA splicing and RNA Helices, Mutants in The Boss
Lar L. Vang and Julius Adler
It is proposed here that every organism has something in charge of it. This is called “The Boss”.
The idea derives from the discovery that mutants of The Boss fail in response to all stimuli tested, even though the mutants are motile (Julius Adler and Lar Vang, 2016; Vang and Adler, p. 13, 2018).
What is The Boss? It is a mechanism that acts as described in Figure 10 from Adler, 2016:
https://uploads.disquscdn.c... <br /> Fig. 10 of Adler, 2016
The idea that each organism has something in control of the organism is novel. Before this, it was believed that each organism has properties that are largely independent of each other. Now it is suggested that all the properties are controlled by a single factor, The Boss, which directs both the interior and the outside of the organism. The Boss is to be found in humans, other animals, plants, and microorganisms. The evidence for this idea is incomplete or poor.
Adler J (2016). A search for The Boss: The thing inside each organism that is in charge. Anat Physiol Biochem Int J Vol.1, 2016.
Adler J, Vang LL (2016) Decision making by Drosophila flies. bioRxiv March 24, 2016.
Vang LL, Adler J (2018) Drosophila mutants that are motile but respond poorly to all stimuli tested: Mutants in RNA splicing and RNA helices, mutants in The Boss. bioRxiv October 1, 2018.
On 2019-06-11 19:25:06, user Julius Adler wrote:
June 11, 2019: here are some changes of April 18, 2019
Drosophila Mutants that Are Motile but Respond Poorly to All Stimuli Tested<br /> Mutants in RNA Splicing and RNA Helicase, Mutants in The Boss
Lar L. Vang and Julius Adler
Decision making in Figure 1 has here been replaced by action selection because action selection is more widely used in the literature than decision making:<br /> https://uploads.disquscdn.c... <br /> Figure 1. The mechanism of behavior. Action selection (also called decision making) occurs when several sensory stimuli are encountered together. In the absence of sensory stimuli, the organism moves in straight flights alternating with rapid turns called saccades, see Fig. 2 of Lance Tammero and Michael Dickinson (2002). Central processing produces the response. The Boss controls this.
Tony Prescott (2008) says, “Action selection describes the task of choosing ‘what to do next’. The task is to decide what action to perform. Selecting between alternative actions has been a focus of research in ethology, psychology, neurobiology, computational neuroscience, artificial intelligence, and robotics." “Effective decision-making, one of the most crucial functions of the brain, entails the analysis of sensory information and the selection of appropriate behavior in response to stimuli; here, we consider the current state of knowledge on the mechanisms of decision-making and action selection in the insect brain,” say Andrew Barron et al. (2015).
Summary: Does The Boss exist?
Adler and Vang (2016) reported Drosophila mutants that lack all responses to external and internal stimuli at 34 degrees but at room temperature these mutants are not deficient. This could mean that activity by The Boss can be eliminated at 34 degrees but it is still present at room temperature.
Vang and Adler (2018) reported a Drosophila mutant that lacks responses to all stimuli at both 34 degrees and room temperature. That indicates that this mutant lacks action by The Boss.
Where does The Boss act?
One possibility for the action of The Boss could be at action selection of Fig. 1 above. How it would act there is not yet known. Further work is needed.
Another possibility for the action of The Boss could be at saccades described in the legend of Fig. 1 above. Again, how it would act there is not yet known. Further work is needed.
Relation between The Boss and consciousness
Many views exist about consciousness, as reported in Wikipedia by Marc Bekoff with seven others: Richard Frackowiak together with seven other neuroscientists (2004) felt that this was still too soon for a definition of consciousness, "We have no idea how consciousness emerges from the physical activity of the brain…At this point the reader will expect to find a careful and precise definition of consciousness. You will be disappointed. Consciousness has not yet become a scientific term that can be defined in this way. Currently we all use the term consciousness in many different and often ambiguous ways. Precise definitions of different aspects of consciousness will emerge…but to make precise definitions at this stage is premature…It has been defined somewhat vaguely as: subjectivity, awareness, sentience, the ability to experience or to feel, wakefulness, having a sense of selfhood, and the executive control of the mind.” Max Velmans and Susan Schneider (2005) wrote, “Anything that we are aware of at a given moment forms part of our consciousness, making conscious experience at once the most familiar and most mysterious aspect of our lives.” “Despite the difficulty in definition, many philosophers believe that there is a broadly shared underlying intuition about what consciousness is", according to John Searle (2005). See also “Animal Minds, Beyond Cognition and Consciousness, in Favor of Animal Consciousness" by Donald Griffin (2001). Wikipedia says, “Over the last 20 years, many scholars have begun to move toward a science of consciousness. Antonio Damasio (1999) and Gerald Edelman (2003) are two neuroscientists who have led the move to neural correlates of the self and of consciousness.” In summary, the term consciousness is widely used but it has not been experimentally defined.
In contrast to consciousness, The Boss is based on experimental results (see above, Summary: Does The Boss exist?). The Boss is the leader of the organism. But more is needed to establish this.
Adler J, Vang LL (2016) Decision making by Drosophila flies. bioRxiv March 24, 2016.
Barron AB, Gurney KN, Meah LFS, Vasilaki E, Marshall JAR (2015) Decision-making and action selection in insects: inspiration from vertebrate-based theories. Front Behav Neurosci 9:216 doi:10.3389.
Damasio A (1999) The Feeling of What Happens: Body, Emotion and the Making of Consciousness, Harcourt Brace.
Edelman GM (2003) Naturalizing consciousness: A theoretical framework. Proc Natl Acad Sci USA 100:5520-5524.
Frackowiak R and seven other neuroscientists (2004) 269 -301 chapter 16, The neural correlates of consciousness.
Frighetto G, Zordan MA, Castiello U, Megighian A (2018) Mechanisms of selection for the control of action in Drosophila melanogaster. bioRxiv April 17, 2018.
Griffin D ((2001) Animal minds: beyond cognition to consciousness. U Chicago Pres.
Heisenberg M (2019) Outcome learning, outcome expectations, and intentionality in Drosophila. Cold Spring Harb Lab Press 22:294-298.
Prescott TJ (2008) Action selection. Scholarpedia 3:2705, 1-14.
Searle J (2005) Consciousness in Honderich T (ed). The Oxford companion to philosophy. Oxford U Press.
Tammero LF, Dickinson MH (2002) The influence of visual landscape on the free flight behavior of the fruit fly Drosophila melanogaster. J Exper Biol 205:327-343.
On 2019-06-04 18:42:14, user Julius Adler wrote:
June 4, 2019: here are some changes of April 18, 2019
Drosophila Mutants that Are Motile but Respond Poorly to All Stimuli Tested<br /> Mutants in RNA Splicing and RNA Helicase, Mutants in The Boss
Lar L. Vang and Julius Adler
Decision making in Figure 1 has here been replaced by action selection because action selection is more widely used in the literature than decision making:
https://uploads.disquscdn.c...
Figure 1 modified. The mechanism of behavior. This applies to all organisms: microorganisms, plants, animals including humans. Action selection (also called decision making) occurs when several sensory stimuli are encountered together. Central processing produces the response. The Boss controls this.
Tony Prescott (2008) says, “Action selection describes the task of choosing ‘what to do next’. The task is to decide what action to perform. Selecting between alternative actions has been a focus of research in ethology, psychology, neurobiology, computational neuroscience, artificial intelligence, and robotics." “Effective decision-making, one of the most crucial functions of the brain, entails the analysis of sensory information and the selection of appropriate behavior in response to stimuli; here, we consider the current state of knowledge on the mechanisms of decision-making and action selection in the insect brain,” say Andrew Barron et al. (2015).
Summary: Does The Boss exist?
Adler and Vang (2016) reported Drosophila mutants that lack all responses to external and internal stimuli at 34 degrees but at room temperature these mutants are not deficient. This could mean that activity by The Boss can be eliminated at 34 degrees but it is still present at room temperature.
Vang and Adler (2018) reported a Drosophila mutant that lacks responses to all stimuli at both 34 degrees and room temperature. That indicates that this mutant lacks action by The Boss.
Relation between The Boss and consciousness
Many views exist about consciousness, as reported in Wikipedia by Marc Bekoff with seven others: Richard Frackowiak together with seven other neuroscientists (2004) feel that this was still too soon for a definition of consciousness, "We have no idea how consciousness emerges from the physical activity of the brain…At this point the reader will expect to find a careful and precise definition of consciousness. You will be disappointed. Consciousness has not yet become a scientific term that can be defined in this way. Currently we all use the term consciousness in many different and often ambiguous ways. Precise definitions of different aspects of consciousness will emerge…but to make precise definitions at this stage is premature…It has been defined somewhat vaguely as: subjectivity, awareness, sentience, the ability to experience or to feel, wakefulness, having a sense of selfhood, and the executive control of the mind.” Max Velmans and Susan Schneider (2005) wrote, “Anything that we are aware of at a given moment forms part of our consciousness , making conscious experience at once the most familiar and most mysterious aspect of our lives.” “Despite the difficulty in definition, many philosophers believe that there is a broadly shared underlying intuition about what consciousness is", according to John Searle (2005). See also “Animal Minds, Beyond Cognition and Consciousness, in Favor of Animal Consciousness" by Donald Griffin (2001). Wikipedia says, “Over the last 20 years, many scholars have begun to move toward a science of consciousness. Antonio Damasio (1999) and Gerald Edelman (2003) are two neuroscientists who have led the move to neural correlates of the self and of consciousness.” In summary, the term consciousness is widely used but it has not been experimentally defined.
In contrast to consciousness, The Boss is based on experimental results (see above, Summary: Does The Boss exist?). The Boss is the leader of the organism. But more is needed to establish this.
Adler J, Vang LL (2016) Decision making by Drosophila flies. bioRxiv March 24, 2016.
Barron AB, Gurney KN, Meah LFS, Vasilaki E, Marshall JAR (2015) Decision-making and action selection in insects: inspiration from vertebrate-based theories. Front Behav Neurosci 9:216 doi:10.3389.
Damasio A (1999) The Feeling of What Happens: Body, Emotion and the Making of Consciousness, Harcourt Brace.
Edelman GM (2003) Naturalizing consciousness: A theoretical framework. Proc Natl Acad Sci USA 100:5520-5524.
Frackowiak R and seven other neuroscientists (2004) 269 -301 chapter 16, The neural correlates of consciousness.
Frighetto G, Zordan MA, Castiello U, Megighian A (2018) Mechanisms of selection for the control of action in Drosophila melanogaster. bioRxiv April 17, 2018.
Griffin D ((2001) Animal minds: beyond cognition to consciousness. U Chicago Pres.
Prescott TJ (2008) Action selection. Scholarpedia 3:2705, 1-14.
Searle J (2005) Consciousness in Honderich T (ed). The Oxford companion to philosophy. Oxford U Press.
Vang LL, Adler J (2018) Drosophila mutants that are motile but respond poorly to all stimuli tested: Mutants in RNA splicing and RNA helicase, mutants in The Boss. bioRxiv October 1, 2018.
On 2019-07-01 07:59:51, user Paul Macklin wrote:
Now published!
On 2019-07-01 05:25:07, user Eric Hwang wrote:
Our new manuscript on the role of Ran in non-centrosomal microtubule nucleation.
On 2019-07-01 05:17:30, user Gal Haimovich wrote:
This is very interesting, and a cool method, but the authors should verify that CYP3A4 protein is not secreted as well in their mice, e.g. as suggested here: https://bpspubs.onlinelibra...<br /> Also, in fig 1B, there are very faint bands of CYP3A4 protein in ScA and VsA which the authors ignore. This could explain the 5EU-RNA in adipose tissue.<br /> There is also clear HD5EU staining in adipose tissue at level similar to kidney(fig 2C). Perhaps a quantitative analysis of these images (instead of showing a single image example) will show clear differences.<br /> I am not convinced that the 5EU-RNA from adipose tissue originated from Liver, or that the differences seen are a result of the ccfRNA and not the change in diet that changes transcriptional or RNA decay programs.
On 2019-06-30 13:58:35, user S2 wrote:
I like this paper. I have two questions. If the authors would like, please answer them.
Which do you think epithelial arroy is mainly depend on composition of placozoan epithelial sheet (very thin cells, AJ only junctions, etc) or specific features of placozoan molecules ?
Do not Individual placozoas bond after they fission ?
Thank you.
On 2019-06-30 05:33:50, user Richard Anthony wrote:
This is an excellent piece of very careful work and makes a lot of sense. <br /> So POA tags panD for degradation by clpC. <br /> I guess this would also explain the need for stress (a low pH) as clpC needs to be sufficiently induced? Is that correct?
On 2019-06-29 05:41:04, user Jay wrote:
Interesting effect on bovine endometrial stromal cells. Were cells isolated as described in author lab https://cronfa.swan.ac.uk/R... ?
On 2019-06-28 18:50:38, user Jorge Velásquez wrote:
Can't wait to read. We've been arguing for sometime that good discrimination statistics do not necessarily relate to model performance and hence, supplementary statistics like expert ratings may help in better selecting models, even if subjectively. https://journals.plos.org/p...
On 2019-06-28 18:36:39, user JAZLYN MOONEY wrote:
I was wondering if any of the authors could comment on their reasoning behind using the unsupervised model of ADMIXTURE and then comparing with RFMix which requires haplotype data? It seems like you would want to use the supervised model of ADMIXTURE with the same haplotypes you used in RFMix to make the comparison fair.
On 2019-06-28 18:15:59, user Fraser Lab wrote:
This manuscript by Rocchio and co-workers investigates the structural basis for the interaction between the molecular chaperone Spy and the client protein Im7. Although the Spy-Im7 complex crystallizes, only the Spy portion is readily modelable and there is only residual electron density for the Im7 molecule. This attribute of the cyrstal system means that investigating the structural basis for the interaction is challenging for traditional X-ray crystallography methods. In a previous publication (Horowitz et al., 2016), an approach was designed to overcome this challenge; 4-iodophenylalanine (pI-Phe) was introduced into the Im7 peptide at one of eight positions, and anomalous data was collected for the Spy-Im7 complexes. The Spy-Im7 interaction was mapped using a combination of the anomalous signals from pI-Phe, the residual electron density from Im7, with plausible Im7 conformations generated using molecular dynamics simulations. The original work was the subject of criticism (Wang, 2018), which focused mainly on whether or not the interpretation of the weak anomalous signal was valid. The authors have acknowledged aspects of the criticism (Horowitz et al., 2018), while emphasizing the cross-validation build into their approach, and that other biophysical methods (e.g. NMR) corroborate their proposed model for the Spy-Im7 interaction. The major successes of the present manuscript are that: i) it increases the sensitivity of pI-Phe assignment, validating parts of the previous paper, and ii) it contributes to the development of anomalous scattering as a tool for interrogating protein structure and function.
In the present work, Rocchio and co-workers focus on optimizing data collection and analysis for anomalous signals from pI-Phe. They placed pI-Phe at one of three positions (only one of which cleanly overlaps with the previous publication) in the Im7 peptide and collected diffraction data around the iodine L absorption edge (λ = 2.38 Å and 2.76 Å). Because the iodine scattering factor is 3.9-fold lower at the longer wavelength, peaks in the anomalous density map that are present at 3.28 Å but absent at 2.76 Å, were assigned to iodine. To test signal reproducibility, they adjusted the goniometer κ angle, and collected additional datasets from the same crystals, and also collected repeat datasets from different crystals. The reproducibility was moderate; three signals were observed in all the datasets, one signal was observed only in one of the datasets, and two signals were observed in three out of four datasets. The authors used a cut-off of 6 σ for assigning peaks. Why was this cut-off chosen? Were peaks present at site I3 (e.g. Fig. 1A/C) at lower electron density thresholds? We suggest that showing the anomalous difference maps at various thresholds is a good idea for a supplemental figure to appreciate the sparsity of the signal at high sigma values and the choice of the cut-off at 6 σ .
The manuscript could be strengthened if the authors addressed the lack of reproducibility for some of the iodine signals in more detail. They state that it is “…clear that the ability to detect low intensity signals is highly dependent on the crystal and the resolution.” Could it be that different crystallization conditions have shifted the binding mode of Im7? I noticed that the concentration of zinc acetate was listed as between 70-270 mM – could this have contributed to the difference? The authors also mention that three out of the four binding sites were identified in the previous work. To make a fair comparison, it might be worth showing all of the proposed binding sites from the previous paper (Horowitz et al., 2016) (e.g. on Fig. 5). On a related note, how confident are the authors in the refined occupancies for the iodine atoms? Is it possible to perform refinement with different initial occupancies to see if they converge on a particular value?
The major weakness of the manuscript lies in its failure to extend the increased sensitivity of iodine assignment to an improved understanding of the Spy-Im7 interaction. The authors begin the manuscript by introducing the READ algorithm – why hasn’t the more sensitive assignment of pI-Phe been fed back into READ? It would be interesting to see if whether the ensembles generated by the two anomalous data collection strategies agree with each other.
Overall this is a well written article that will be of interest to a wide section of the structural biology community. The improved capability of modern beamlines to collect anomalous data at long wavelengths, as described in this paper by Rocchio et al., may help to interrogate the structure and function of macromolecules that were previously intractable to traditional approaches.
Minor points
Consider revising the manuscript title – it is possible for a residue to be partially occupied and not dynamic - “conformationally heterogeneous” is probably more accurate. Or leave it as is, but clearly define the differences and the ability of crystallography to inform between dynamics and heterogeneity somewhere in the manuscript
It might be helpful to add a couple of sentences in the introduction to describe what is known about the Spy-Im7 interaction from orthogonal methods (see point 6).
Please include a section in the methods describing how the ITC experiment was performed.
Page 4 line 31: Typo - “we reasoned that we should be able to specifically distinguish iodine anomalous signals from other with elements”
Page 2 line 14: Shouldn’t it be Residual Electron and Anomalous Density not Residual Anomalous and Electron Density?
Please either label Fig 1D with the location of the “crook of Spy’s cradle” or include a new figure. It might be helpful to have a “cartoon” type schematic in the introduction to illustrate what is known about the Spy-Im7 interaction.
Fig S4-6: Please label the iodine peaks with the labels used in Fig. 1 (e.g. I1, I2, I3, I4).
PDB 6OWX: Is imidazole 214 in chain A modelled correctly?
It’s notable that they have co-crystalized Spy with 10 peptides (8 in the previous publication, 1 overlapping in both, and 2 novel ones in this publication) - have soaks been attempted? This could also add an interesting experimental control where the anomalous signal should be displaced by excess unlabeled client peptide but maintained (or enhanced) by labeled peptide.
Peak height is a good measure and relatively unbiased, but the anomalous maps have the potential to inform (in a relatively model unbiased manner) on the occupancy and B-factor directly. Showing multiple concentric contours or plotting density as a distance from peak center along a pseudo atomic radius will help to clarify the differences in the profiles. For example, in table 4, the peak heights for the 4th and 5th rows are the same, but the occupancy is 2-fold different.
11 - Figure 4 is pretty difficult to follow, even for us. We could do with a bit more annotation of where the disordered peptide backbone is predicted to trace. It also seems like the aromatic group attached to the I is a fairly strong modeling constraint that could help guide the eye in this figure.
12 - These data are especially valuable for methods developers and given the issues raised by Wang on the previous paper, it is especially important to put as raw data as possible out into the public. Papers like this should be held to a higher data deposition standard (mtzs are not enough!): the authors should deposit their raw diffraction data at SBGrid Data Bank (https://data.sbgrid.org/) or an equivalent resource to enable future re-use and validation.
We review non-anonymously, James Fraser and Galen Correy (UCSF), and will have posted this review as a public comment on the preprint: https://www.biorxiv.org/con...
References<br /> Horowitz, S., Salmon, L., Koldewey, P., Ahlstrom, L., Martin, R., Quan, S., Afoine, P., van den Bedem, H., Wang, L., Xu, Q., Trievel, R., Brooks, C. & Bardwell, J. (2018). Nat. Struct. Mol. Biol. 25, 989–991.<br /> Horowitz, S., Salmon, L., Koldewey, P., Ahlstrom, L. S., Martin, R., Quan, S., Afonine, P. V., Van Den Bedem, H., Wang, L., Xu, Q., Trievel, R. C., Brooks, C. L. & Bardwell, J. C. A. (2016). Nat. Struct. Mol. Biol. 23, 691–697.<br /> Wang, J. (2018). Nat. Struct. Mol. Biol. 25, 989–991.
On 2019-06-28 16:24:55, user Mete Civelek wrote:
Interesting results. It looks like in Figure 5, many well known causal regulators of CAD and T2D are not captured. For example, TG with CAD, and other lipid measurements with T2D.
On 2019-06-27 20:29:47, user George Davey Smith wrote:
This is a really interesting paper, with lots of good things, including a clear discussion of the latent causal variable approach and its relationship with that. A few comments:<br /> (1) A new term, "correlated pleiotropy" is introduced (see figure 1), which is covered by the widely used term in Mendelian randomization of "vertical pleiotropy", but it is vertical pleiotropy when you have mis-specified the primary phenotype (in this case, unknown). see https://www.ncbi.nlm.nih.go... for vertical pleiotropy with mis-specified primary phenotype <br /> (2) This issue is generally addressed in MR by bi-directional MR applying Steiger filtering between M and Y in fig 1 (see https://journals.plos.org/p.... It would be very interesting to know how this approach and CAUSE compare<br /> (3) The approach to HDL cholesterol would be through multivariable MR (which gets the right answer) - currently this is not implementable in CAUSE. It would be interesting to know if in principle if MVMR could be implemented in CAUSE<br /> (4) The SBP to BMI association in figure 5 is interesting. It says conventional MR shows SBP lowers BMI . The SBP GWAS used is adjusted for BMI, so collider bias is introduced and you should get this effect. This is the right answer to the wrong question. CAUSE gets the wrong answer to the wrong question
On 2019-06-27 18:34:47, user Viraj Deshpande wrote:
Nice work! Have you tried running this on samples with larger circles that we observe in cancer? You might be interested in our method for detecting circles in cancer genomes along with validated datasets (see page 5 paragraph 2): https://www.nature.com/arti...
On 2019-06-27 17:29:31, user Jon Moulton wrote:
The more severe phenotype of mbnl2 morphants is consistent with expected outcomes of genetic compensation. You propose off-target effects as the source of the difference, which is also possible. Injecting an mbnl2 Morpholino into an mbnl2 mutant would differentiate between these mechanisms; an off-target effect would appear as a different phenotype in the mutant, while the Morpholino in the compensated background would produce no additional effect by interaction with mbnl2RNA. In the absence of additional phenotype in the Morpholino-injected mutant, the more severe effect of the Morpholino in the wild-type embryos shows the effect of an acute (non-compensated) knockdown.
On 2019-06-27 07:02:36, user Joe Li wrote:
This paper has been published in Environment DNA https://onlinelibrary.wiley...
On 2019-06-26 23:18:50, user AC wrote:
A privilege to be able to read advance copies of research like this. I will sit down this evening to have a proper read. A couple of citations are incorrectly dated e.g Robinson 2008 and I think one of the Danser papers (just wrote a project paper about Nepenthes so I read some of these this week) but probably this will be corrected in review. Congrats!
On 2019-06-26 21:33:46, user Charles Warden wrote:
It looks like there is a peer-reviewed version of this paper now available: https://journals.plos.org/p...
On 2019-06-26 15:12:50, user David Bernstein wrote:
A revised version of this manuscript has been published under the title "Metabolic network percolation quantifies biosynthetic capabilities across the human oral microbiome". https://doi.org/10.7554/eLi...
On 2019-06-26 13:04:53, user Serbulent Unsal wrote:
Although paper mentioned it takes 32 minutes to calculate sequence and protein embeddings for human proteome with an 1080 card it takes about 5 hours for me with and RTX 5000 gpu with 20 core server and only for human protein embeddings.
On 2019-06-26 11:36:46, user David Escors wrote:
Pending the proper link to the peer-reviewed version by BioRxiv staff, here is the final published version:
On 2019-06-26 10:58:45, user Paul Kersey wrote:
This is cool, but I don't quite understand the use case - if you can get 1-3 contig assemblies using short reads only, why do you need to use long reads at all? If there's a sweet spot for this, isn't it for genomes that are too large/repetitive to assemble with short reads alone?
On 2019-06-26 08:20:51, user Richard Börner wrote:
Nice work! And thanks for continuing this exciting journey to find the best possible solution for the state identification and model selection of noisy multi-state trajectories. Can we implement DISC in our software package MASH-FRET - https://rna-fretools.github... ?
On 2019-06-26 07:33:37, user jeff ellis wrote:
Can the authors please compare the C terminal repeat structures in the LRR region to repeats previously described in the L and M flax rust resistance genes where 1, 2, 3 and 4 larger order repeats each consisting of several LRR repeat units have been described
On 2019-06-26 04:18:40, user Tom Schneider wrote:
In Figure 5E the authors called the images "Sequence consensus"s. However, they are just sequence logos. Notice that a logo contains ALL possible sequences in the stack of letters at every position. So any sequence could be read from a logo, including the worst possible binding sequence. The "strict" consensus would be the most frequent base at every position (the top letter of a standard logo). Also, a sequence with the most frequent base at every position is a rare sequence that is not representative of the roughly Gaussian distribution of binding sites, see figure 2 of: https://doi.org/10.1006/jtb... and https://alum.mit.edu/www/to... So it is better to avoid the word "consensus" unless you really mean the most strong sequence because the consensus is usually not a naturally occurring sequence. It is never appropriate to call a sequence logo a "consensus". You could call them a "motif".
On 2019-06-26 02:23:03, user John F. Allen wrote:
Remarkable paper, explaining much. I've long wondered why plastid DNA encodes NAD(P)H dehydrogenase subunits only in land plants.
On 2019-06-25 20:05:56, user Alejandro Heuck wrote:
Nice work. I have a few experimental questions.<br /> 1) Does EHEC has constitutive secretion of T3SS effectors? How is Type 3 secretion regulated?<br /> 2) Have you purified OMV using flotation assays with ultra-centrifugation in sucrose cushion and show that any protein aggregate or filament/s present in the supernatant do not contribute to the processes?<br /> 3) Are the proteins inside/outside OMVs? A proteinase K treatment will answer this question.
Thanks
On 2019-06-25 20:05:15, user Shiya Song wrote:
The supplemental material only contains supplemental figures but not tables.
On 2019-06-25 16:45:15, user Thomas N. Seyfried wrote:
In their recent study, Sperry et al conclude that the U87-MG model of glioblastoma can utilize fatty acids and ketone bodies for growth. However, their data presented in Figures 1C & 1D argue against this conclusion. It is clear that neither supplementation with fatty acids (Fig. 1C) nor B-OHB (Fig. 1D) could replace glucose as a fuel for maintaining U87 proliferation under lower glucose conditions. If fatty acids and B-OHB could be utilized for growth, then FA & B-OHB supplementation should be able to replace glucose for U87 growth under low glucose conditions. Their data do not show this.
It is also important to mention that human GBM express abnormalities in mitochondrial number, structure, and function all of which will compromise energy production through OxPhos (DOI: 10.1177/1759091418818261). These abnormalities will force the tumor cell to rely more heavily on fermentation then on OxPhos for growth. Neither fatty acids nor ketone bodies are fermentable fuels and cannot replace either glucose or glutamine, which are fermentable through substrate level phosphorylation in the cytoplasm and mitochondria, respectively (DOI: 10.1177/1759091418818261). The data presented in Figures 1C and 1D support this view in showing that neither fatty acids nor B-OHB can replace glucose for maintaining U87 growth rate. Hence, their data do not support their main conclusion that U87 can utilize fatty acids and ketone bodies for growth.<br /> Thomas N. Seyfried
On 2019-06-24 01:13:01, user Harley K wrote:
We thank the P. Mukherjee for the comments. While we certainly do not argue that our study proves the futility of the ketogenic diet, our results strongly support the notion that GBM are able to utilize fatty acids and ketones for growth. Our data are consistent with recent observations but do disagree with some prior observations. Clearly, the metabolism of the host mouse strain is important. However, in this case, P. Mukherjee is in error, as NSG mice, despite the name, do not have diabetes, as their immunodeficiency prevents the development of diabetes.
On 2019-06-21 20:07:00, user Purna Mukherjee wrote:
Sperry et al., have provided another interesting study showing how a ketogenic diet fails to manage glioblastoma growth in the U87-MG xenograft mouse model. The results are consistent with the previous findings of Dang et al (PLoS One. 2015 Jul 20;10(7):e0133633, doi: 10.1371/journal.pone.0133633), and Kalaany and Sabatini (Nature. 2009 Apr 9;458(7239):725-31. doi:10.1038/nature07782.) showing that neither caloric restriction nor ketogenic diet have any therapeutic effects on brain tumor growth when the tumors are grown in the brains of Non-Obese Diabetic/Severely Compromised Immuno Deficiency mice (NOD/SCID). It is important to mention that NOD/SCID mice not only have a compromised innate and/or adaptive immune system but also express characteristics of both Type-1 and Type-2 diabetes (Chaparro et al, PNAS, 2006; DOI:10.1073/pnas.0604317103). These findings are inconsistent with other studies showing that caloric restriction and restricted ketogenic diets can reduce U87-MG growth when the tumors are grown in mice that do not have characteristics of Type 1 or Type 2 diabetes (DOI 10.1007/s11060-013-1154-y; DOI:10.1158/1078-0432.CCR-04-0308; doi:10.1186/1743-7075-4-5). Although Sperry et al were careful in their in vitro experiments to maintain normal glucose physiology, they chose a mouse host for their in vivo studies that has no relevance to either normal human or mouse physiology. It remains unclear whether glucose and ketone levels would be linked to tumor growth in this mouse host. The sensitivity of some tumors to metabolic therapy is dependent on host energy metabolism and microenvironment, which are abnormal in NOD/SCID mice. Hence, their conclusions that ketogenic diet fails to manage glioblastoma growth must be viewed with caution.
On 2019-06-25 11:19:45, user Vera Kozjak-Pavlovic wrote:
So, it is an IMS protein after all :)
On 2019-06-25 06:47:33, user JJ wrote:
It would be better to have the tested versions of these tools.
On 2019-06-24 21:29:29, user ThePatrickWatsonLab wrote:
Hello! Very nice paper-- we are also interested in post-translational modification of SR proteins. I am wondering if you could provide more clarity regarding your phos-tag gels. They way they are currently labeled, they are difficult to interpret. Is it possible to include size markers?
On 2019-06-24 13:28:47, user Jomar Fajardo Rabajante wrote:
The order of the authors should be JFRabajante, ELAnzia, CSGokhale. Sorry for the mistake.
On 2019-06-24 13:23:27, user Edgar Gonzalez-Kozlova wrote:
Congrats!, the paper present evidence and discusses a lot of sources that influence patient survival in a comprehensive analysis of prognostic scores.
On 2019-06-24 08:18:07, user Omer Markovitch wrote:
This is interesting. Have you checked the most recent Lipid World literature [J. R. Soc. Interface, 0159 (2018)] ? This simulations seems to be directly related to the present experiments [Physical Biology 8, 066001 (2011)] ; And these experiments are relevant too: [<br /> 2014 Oct 7;107(7):1582-90]
On 2019-06-24 08:16:54, user @julio_r_banga wrote:
An island approach for parameter fitting in systems biology was presented back in 2012:<br /> https://bmcsystbiol.biomedc...<br /> and refined later:<br /> https://bmcbioinformatics.b...
On 2019-06-22 19:10:11, user Andre wrote:
An important detailed study demonstrating mutation effects on retrograde axonal transport and neuronal and neuromuscular disease.
On 2019-06-22 14:11:57, user Mariana Mateos wrote:
Peer-reviewed version now available at https://doi.org/10.1038/s41....<br /> Mateos, M., N. O. Silva, P. Ramirez, V. M. Higareda-Alvear, R. Aramayo, and J. W. Erickson. 2019. Effect of heritable symbionts on maternally-derived embryo transcripts. Scientific Reports 9: 8847
On 2019-06-21 12:47:15, user Ian Corfe wrote:
Very cool comparison of the methods. I have a question - since much of the discussion is based around the monophyly or otherwise of Allotheria, did you try this kind of comparison with one of the datasets from Jin Meng's group that both suggest Allotherian monophyly and include all the known euharamiyids? That would seem to be the appropriate comparison to make rather than the Krause phylogeny that doesn't include much of the recently discovered material - as you say, only one euharamiyid...
On 2019-06-21 09:37:39, user Wouter De Coster wrote:
Dear authors,
Thank you for the very interesting work. It is perhaps a semantic detail, but I would suggest reconsidering the following sentence in the Background section: "TRs can be further divided into two types based on the length of the repeat unit; repeats with less than six base pair repeat units are classified as microsatellites or short tandem repeats (STRs) and those with more than 6 base pair repeat units are known as minisatellites [4]." This definition excludes TRs with exactly 6 bases. I would argue that STRs are 2-6 bp (inclusive), or 1-6 if you wish to include homopolymers here. There is some inconsistency in this terminology in the literature, as some use VNTR as a synonym for minisatellites, while others use it like you to include both microsatellites and minisatellites.
Furthermore, while constantly having to update tools is cumbersome, it would be interesting to know to which extent your results for nanopore targeted capture sequencing and WGS would be improved by using a more recent Guppy basecaller.
Best,<br /> Wouter De Coster
On 2019-06-20 18:03:04, user Max Yin wrote:
Cool story!) We also have a recent story about survival, era and stringent response. The paper is under review at the moment.
On 2019-06-20 17:27:15, user BGS wrote:
The original posting was missing Table 1, sorry about that. Fixed now.
On 2019-06-20 17:16:54, user Taj Azarian wrote:
Great manuscript. I have been trying to replicate and the one piece of information I found missing was the gDNA concentrations in the controls and the depleted samples. Before attempting qPCR, I was using this to determine how successful the extraction and depletions were. Thanks!
On 2019-06-11 01:47:01, user John Osei wrote:
Excellent, but I have some questions, although I am planning to adopt this procedure for my microbiome analysis: <br /> 1. Can the same approach be used for meta-transcriptomics of the microbiome i.e., after using the saponin treatment, can total RNA be isolated without being degraded or affected?<br /> 2. Can this be extended to the gut and other microbiome besides the lower respiratory microbiome?
On 2019-06-20 12:38:38, user Chris Fromme wrote:
We are uploading a revised version with updated MTF/NPS/DQE curves. We've learned that the knife-edge method overestimates DQE at high dose-rates. The revision does not change the overall conclusions or any of the single-particle experiments.
On 2019-06-20 09:17:16, user John Aplin wrote:
How confident are you that Epi9 is a ciliated cell? It is distant from other ciliated groups in the SNE. It is hard to see in your colouring scheme in Fig 2 how it maps in the pseudo time analysis -- is it at the most distant (left hand) end of the secretory cell repertoire? Do you think that ciliated cells are also secretory?
On 2019-06-20 06:33:44, user Jean-Claude Dujardin wrote:
No impact on phylogenomic studies (maybe representation bias), but huge impact for studies<br /> aiming to link parasite genome variation and clinical phenotype (drug<br /> resistance, virulence). For these, forget analyses of cultivated parasites. Warning for anyone working with cultivated pathogens
On 2019-06-20 06:33:04, user Jean-Claude Dujardin wrote:
Development and validation of a genome capture method to sequence #Leishmania directly in host tissues. Sensitive, excellent performances for calling of SNPs, ploidy, CNVs.....but parasite genomes very different from those of derived cultured isolates
On 2019-06-19 14:00:05, user Adam Adler wrote:
In the past few years there has been a significant advancement in the field of full-length, single cell, paired B cell repertoire sequencing and functional validation, all of which has been published in peer-reviewed journals before the posting of this preprint. The last paper cited in this preprint highlighting paired B cell repertoire analysis is from 2016. By our count, there have been at least seven papers since then that are similar to and/or go beyond the scope of this presented manuscript, none of which were cited.
Specifically, I am an author and directly supervised five different studies performed at GigaGen that utilized our single cell microfludic platform and yeast display system to deep sequence and screen millions of single human, mouse, and humanized mouse B cells, followed by extensive functional validation in four of the studies: Adler et al., mAbs, 2017a (PMID:28846502); Adler et al., mAbs, 2017b (PMID:28846506); Adler et al., mAbs, 2018 (PMID:29376776); Medina-Cucurella et al., Antibodies, 2019 (DOI:antib8010017); Asensio et al., mAbs, 2019 (PMID:30898066).
In addition, MedImmune (Rajan et al., Commun Biol, 2018; PMID:30271892) and George Georgiou's lab (Wang et al., Nat Biotech, 2018; PMID:29309060) also published similar work after our 2017 mAbs studies were published.
At minimum I would have expected several of these recent studies to have been cited in this preprint, and further it would be useful to describe how the work presented here goes beyond what has been described in these already published articles.
-- Adam Adler, Chief Scientific Officer, GigaGen
On 2019-06-18 15:44:59, user Robert Flight wrote:
The presence of multiplicative error is a large part of the reason -omics data are often log or some other transform applied before doing *almost* anything with it, if the person working with the data is aware of these issues.
Given the difficulties in working with the multiplicative error data (as evidenced by some of the very hairy derivations and equations by the authors), I would be very curious to know how the additive correction applied to log-transformed data behaves compared to the full additive + multiplicative correction.
On 2019-06-17 12:03:27, user 141xgc wrote:
Effective July 1, 2019 we will be using Scopus and IPDD instead of Web of Science and Derwent for publications and patenst respectively. The ETL scripts for the latter two data sources are available under the MIT license.
On 2019-06-17 02:09:22, user Mau Mauleon wrote:
Dear all: The final print version at Gigascience journal contains updates and edits, so please cite the one at doi: 10.1093/gigascience/giz028
On 2019-06-15 21:21:43, user ebarthelmess wrote:
HI folks - we'd love any feedback on this preprint!
On 2019-06-15 15:51:16, user Mitchell Thompson wrote:
We are currently getting an Open Access License for the code described in our preprint as required by DOE. As soon as the license has been approved we will provide the repository link.
On 2019-06-14 19:35:43, user No name wrote:
Is the undigested PCR product by Cas9 itself not knocked-in for the cells?<br /> Also, in case of N-terminal knock-in, there should be some cells that are knocked-in and expressing GFP not following the expression of gene-of-interest because of a frame-shift after GFP sequence.<br /> How do you think these problems?
On 2019-06-14 16:47:55, user Harry Harris wrote:
Hi, please put CRU TS paper in the refs. Also, it's not a reanalysis dataset (see the paper!).
On 2019-06-14 16:15:01, user Roo Vandegrift wrote:
Thanks for the heads-up, I hadn't seen that one!
On 2019-06-11 21:02:27, user AS wrote:
Interesting study. Have you seen this:
On 2019-06-14 12:33:25, user Kevin Folta wrote:
The first author on this paper should be Tautvydas Shuipys. He did the vast majority of the work, and it will appear that way in publication (and on the preprint). During the submission process something got shuffled and I'm really sad about that. Tautvydas did very nice work here, lots of replicates and has a very nice story. It is the first molecule invented from randomness that has a discrete effect on plants. How cool is that?
On 2019-06-13 18:29:10, user Laura Sanchez wrote:
Dear Rappez et al, this preprint was discussed in a lab meeting and we would like to offer the following for review. Thank you for posting this very interesting manuscript. Best, The Sanchez Lab
The manuscript by Rappez et al details a very clever adaptation of MALDI based imaging mass spectrometry to examine the lipidomics of single cells. This proof of principle paper thoroughly described the experimental, computational and statistical details that contributed towards the multimodal imaging of mass spectrometry-based data with microscopy images. It is noteworthy that the authors took great detail to explain the technicality of the manuscript and provide excellent statistical analyses. The crux of the experiment relies on the researchers utilizing the ablation marks left by the AP-SMALDI source being co-registered with microscopy images to correlate the different types of data (cell type or state with lipid profiles). While the premise of the paper and execution of the technical aspect was well done, the biological context, limitations of the method, limitations of the existing methodology, and the metabolite identifications could be improved to enhance the manuscript. Below please find a list of our major and minor critiques for this manuscript.
Major: <br /> The title of manuscript is somewhat misleading. The large majority of the ions that were annotated and validated were lipids. While it seemed both positive and negative mode data were collected the majority of the manuscript was centered on lipids. This however may be amended if the authors were to provide a list of the 88 metabolites used to discriminate cells, but based on the tables offered in SI Dataset 1, this was primarily focused on lipids.
Along these lines, the authors may wish to acknowledge in the introduction what the current limitations of single cell analyses really are. It is clear they have overcome the hurdle of the physical measurement as it relates to data collection and co-registering images but another clear limitation which was overlooked was the copy number (abundance, number of moles, etc) of any given metabolite. We suspect that lipids ended up being a primary focus since they comprise a large proportion of the cell wall and are thus in high copy number (in addition to their efficient ionization, particularly in negative mode).
The figure legends and text tended to be highly redundant and technical in certain places which did not aid in the overall understanding of certain aspects of the manuscript. For example Figure 3B,F,G description verbatim with line 163-165 regarding the richness and relevance types of sentences. This occurs throughout. In general it would be helpful if the figure captions were not just recapitulations of the technical body of the text but instead a highlight of the take-away message for the figure.
As a take-home message from this paper, it would be helpful if the authors could posit how this may be extended to other single cell applications, biomedical applications, etc which may also tie into a short discussion of the limitations and advantages of this exciting approach compared to others that currently exist.
The number of replicates conducted for each experiment is concerning. As far as we were able to discern, it appears that in the HeLa-NIH3T3 cell experiment there is only one biological replicate with two technical measurements being utilized to draw conclusions about metabolomic intermixing. Is there a reason that so few replicates were used? A consideration that should be weighed with any single cell type of experiment in the number of single cells measured in order to identify significance and meaning in the data. How many measurements across biological and technical replicates should be considered? Is one control for biological (or human) error enough to use the collected data to deduce significance? And largely, each cell was only sampled with one ablation mark. It is hard to therefore reconcile certain decision points, for instance the description of poor quality cells (Figure S11 and text). How does one determine if a cell is poor quality? Would it be more accurate to say the ablation of a specific cell lead to a poor quality spectrum? A better justification that relies on the data instead of metabolites ID-ed (such as signal to noise or intensity) may be a more reproducible workflow want to exclude data from the resulting analyses.
A brief justification of the cells chosen would be beneficial especially for those not familiar with mammalian cell types. It was interesting that the fibroblasts didn't overwhelm the HeLa cells, but this may also be due to the culturing techniques mentioned which were not explained in detail and made this difficult to gauge whether the cells were cultured correctly in the first place.
Annotations/validations in supplemental data S1.
No consistency regarding the reported masses (nominal masses vs. sometimes all 5 decimals) - this makes it seem like data doesn’t match.
The large difference in spectra between the standards/library matches to the sample does not confirm these annotations, esp for cis-9-octadecenoic acid, which also doesn’t appear to be retention time matched as agradient was used rather than an isocratic LC method.
This also applies to TG(50:1), TG(52:2), and linoleic acid. Pointing to a minor fragment that does not match the mirrored fragment in the standard spectrum or is very close to the signal to noise in the baseline is not convincing.
One structure was checked for fragmentation assignment (AMP), the proposed fragments appear to be generated via the ChemDraw fragmentation tool which creates homolytic cleavage, it is unlikely that this is the case for every fragment. The authors should carefully consider fragmentation mechanisms before assigning peaks in a spectrum. We recommend reference : DOI: 10.1039/c5np00073d
Largely, with the other annotations (88 metabolites not listed or described in the manuscript), it would be more appropriate to use the word putative annotation as no other methods for validation was used. We suggest that the authors use levels described in PMCID: PMC3772505.
Metabolic intermixing. Lipid profiles are the main basis. Since the list of 88 metabolites never provided, were there other things in list? How do the cells start to have the same lipids, is one cell endocytosing another? Are all the cells actually alive? How do you know if the cell walls are intact? Can cell wall associated vs extracellular vs intracellular be differentiated with this method?
Replicates with sample size, what was omitted from 4C? Not enough sample size to support some of these claims, more biological replicates could demonstrate how stable the method and measurement is, esp regarding metabolic intermixing.
Minor
Many of the graphs and charts utilized red-green schemes, this would not be color blind friendly, additionally, when printed in greyscale the cell differentiations and color maps are not distinguishable. Could figure 3F&G be changed to a single color gradient to help with printing in greyscale?
Scale bars for the color gradients should be labelled, are they always 0 to 100?
Many of the figure panels had an excessive amount of data which didn’t necessarily help convey the central message of the data, could some of these be moved to the SI? <br /> Is it possible to include a plain bright field image unaltered or overlaid in figures 3 and 7? At times it seemed that the cells were not necessarily in a monolayer and this would help clear that up.
Along these lines, the cartoons used in Figure 4H/E seem like they don’t necessarily portray a monolayer? This lead to some confusion as to what was actually being measured versus discarded from the analysis.
The word novel was used in excess, some elements of this work has been done in slightly different aspects, this detracts from the actual novelty of experimental data alignment.
Regarding the HeLa-NIH3T3, if the cells were cultured at a 50:50 ratio, why was the region selected for analysis not also 50:50? This seemed like a disconnect.
Figure 7. The take home message with this figure is diluted. Is the take home message that every cell is unique? How would we reconcile this with reality? ie if every cell is metabolically unique therapeutic treatments likely would be ineffective (seeing as they work on the majority of a population in theory). Or if every cell is unique, have you not sampled a large enough population to extend these results? Again how many single cells should be measured experimentally in order to draw inferences, and is an in depth analysis on 5 cells enough? We suspect that by simplifying the figure would remove these questions.
Are raster spots are bigger than the diameter of the cell? Figure 2 is misleading because the laser is shown as smaller than the cells but in reality the cells are much smaller based on reported diameters in Figure 4 legend. Methods: beam diameter, line 683. 15-43 microns; Perhaps an inclusion of a burn mark or adding a real image of the ablation marks on the cell using a microscope image would help clear this up? Expand Figure S9 to have the cells for a visual aid.
On 2019-06-13 11:34:53, user paul Brear wrote:
A similar concept was discussed in the paper below using CHES as the probe
On 2019-06-13 02:02:57, user Harvard2TheBigHouse wrote:
Suggested opening lines (apologies in advance for any errors or false assumptions):
Mapping the origin and dispersal of the hominids who would become modern homo sapiens has traditionally relied on applying the neutral theory to modern mtDNA, specifically only the 7% of mtDNA comprising the D-loop. This approach bore the Out of Africa theory, which has persisted in the literature despite several glaring and growing flaws. [[Recent studies, however, have established… ]]
[[…largely a result of parallel mutations]] in coding regions of slow and fast-evolving protein coding sequences of the autosome [[rather than recent common ancestry and admixture as commonly assumed.]]
[[…by frequency and hence there were many private or low frequency SNPs in our set.]] <br /> - What exactly does “set” refer to here? One of the earlier three mentioned? All three combined?
[[That the observed sharing (24%) in fast SNPs was lower than the 85% for common SNPs was because we did not filter the SNPs by frequency and hence there were many private or low frequency SNPs in our set.]] <br /> -Shouldn’t a high amount of sharing between common SNPs be the expected result? The point you are looking to make here isn’t clear to me.
[[we next studied only rare SNPs with the same allele frequency in a racial group for either fast or slow evolving variants]] <br /> - Do you mean “either or” here, or “both fast and the slow evolving”?
[[For each major racial group, we selected a set of SNPs that have the alternative allele appearing only twice in the group (either two heterozygous or one homozygous)]] <br /> - What I think you mean is “looking within each racial group separately” you identified the set of SNPs, but I’m not sure.
[[Thus, these results invalidated the present notion of a very recent common ancestor (<100 ky) and large amount of admixture as suggested by the Out of Africa model and support a more distant common ancestor and limited admixture as advocated by the multiregional model.]] <br /> - The invalidation you are claiming isn’t obvious on its face here, I’m not sure how you draw this conclusion based on the data you went over earlier. It would be helpful to spell out how this invalidates OOA exactly.
[[by calculating the average pairwise genetic distance (PGD) per group in different types of SNPs, including the slow set and the random set as defined above]] <br /> - At this point it would be helpful to remind us what exactly the random set is, and what its defining characteristics are.
[[also the hom PGD resulting from hom mismatches that should better represent neutral diversity]] <br /> - What is a “hom mismatch” exactly? Do you mean the allele is mismatched at the same loci in a different racial group?
[[thereby deeming the Out of Africa model untenable as it is based on the now disapproved assumption of neutrality for a random set of genome wide SNPs.]] <br /> - Out of Africa was never based on genome wide SNPs at all, just mtDNA, so this is very confusing.
[[To confirm if we have made the appropriate cut-off in selecting the slow SNPs as our phylogeny-informative set of neutral SNPs, we verified that the next set of just slightly less conserved nonsyn SNPs]] - The word “neutral” seems contradictory here, because I thought the slow SNPs were being selected for because they’re in protein-coding regions conserved between monkeys and humans.
I will have more comments once these questions are resolved, figuring them out will make it a lot easier to move forward.
On 2019-06-12 20:12:25, user Kayla R Nygaard wrote:
New version published in Autism Research: https://onlinelibrary.wiley.com/doi/full/10.1002/aur.2154
On 2019-06-12 13:13:39, user Guest wrote:
It's impossible that Modern Sicilians would have 43% North African <br /> admixture. Every other study estimates ~5%. You obviously did something <br /> wrong.
On 2019-06-12 07:36:54, user Bowen Tang wrote:
where is the etable?
On 2019-06-11 21:47:44, user Aaron Lukaszewski wrote:
Also using the EVS (Study 3), we find complementary evidence: https://journals.sagepub.co...
On 2019-06-11 16:30:50, user Janice Valencia wrote:
Great work!<br /> I am curious about the ycf2 gene, it looks like you do not have it in the IR of monocots or eudicots...
On 2019-06-11 11:06:41, user Cistanthe longiscapa wrote:
I have commented on this article in a manuscript. Not singling out this work....just convenient. Comments apply to the whole paradigm.
On 2019-06-11 08:51:06, user Zhou Xu wrote:
During the revision process, we found that a subset of Chlamydomonas reinhardtii telomeres have blunt ends by hairpin ligation assay, which is an important finding with the regards to the evolution of such structures in plants. Because this assay is very sensitive to the quality of extracted genomic DNA, we were able to detect blunt ends only using a strain that does not have a rigid cell wall and gives higher quality DNA. The results in this preprint have now been refined and updated in the version published in Life Science Alliance (https://www.life-science-al....
On 2019-06-10 15:31:44, user Chun wrote:
Remarkable to see the disruption of GPCR dimer with a single point mutation.
On 2019-06-10 09:07:55, user Jolinda Pollock wrote:
Please note that the 16S rRNA gene sequence data will be made publicly available when the manuscript finds a home in a journal.
I look forward to hearing feedback on our work. Thank you!
On 2019-06-10 04:15:26, user Paul Macklin wrote:
Now fully published in Molecular Systems Design and Engineering, available as open access at:
On 2019-06-08 05:20:39, user Jinyong Wang wrote:
Welcome independent peers from anywhere coming to my lab to repeat our protocol with my guide. I will cover all the traveling and accomodation expenses for at least five individuals from five independent labs. I am also happy to share our key materials with peers who are interested in this method for their own research purposes, potentially. Tweet me @JinyongWang3 or email me (wang_jinyong@gibh.ac.cn) for your potential interests. <br /> Jinyong from GIBH, Chinese academy of sciences.
On 2019-06-07 20:30:44, user J Wallace wrote:
Is anyone else concerned that Equation 2 puts the PCAs _after_ the SNP? (I checked the R code; it does it too, then appears to take simple Type I SS for p-values.) That seems like a basic methodological error. SNP effects should always be fit _after_ population structure. The github repository was missing the PCA file so I couldn't check it on the supplied code, but I bet fixing that would resolve most/all of the "problems" mentioned in this paper. (If the issue persists, I'd like to see if/how it relates to using kinship matrices, which are just as common as PCs.)
On 2019-06-07 14:25:21, user Giorgio Scita wrote:
A revised version of this manuscript has been accepted for publication in Nature Materials.
On 2019-06-07 14:24:07, user Olivier Gandrillon wrote:
Dear authors
We read with interest your BioRxiv preprint and would like to communicate the following comments:
We find that in general, you tend to make very strong statements that tend to contradict the existing litterature. For example, the bursting model you derive is exactly the one that is tacitly considered in many existing biology papers. It is interesting to state it more explicitly than usually done in biological communities, but still, the very same bursting model was considered in [Shahrezaei & Swain, 2008] and the negative binomial distribution was also derived in that paper.
As you note in your preprint, the negative binomial distribution is nothing but an alternative parameterization of the Poisson-gamma distribution, which itself is a rigorous first order approximation of the Poisson-beta distribution derived from the mechanistic two-state promoter model. This approximation corresponds to the so-called bursty regime, which is biologically relevant and well accepted.
Also, the Poisson layer becomes negligible when mRNA quantities span a high range, which happens in practice (e.g., figure 5.A of [Albayrak et al., 2016]). The counterpart of your bursting model is then a piecewise-deterministic Markov process that is well-established [Friedman et al., 2006]. Related to that, you really should check and cite existing literature, for example about rigorous convergence results [Crudu et al., 2012] or the application to single-cell expression data [Herbach et al., 2017].
Ulysse Herbach, Olivier Gandrillon
Refs:
[Shahrezaei & Swain, Analytical distributions for stochastic gene expression, PNAS 2008]
[Albayrak et al., Digital Quantification of Proteins and mRNA in Single Mammalian Cells, Molecular Cell 2016]
[Friedman et al., Linking stochastic dynamics to population distribution: an analytical framework of gene expression, Phys Rev Lett 2006]
[Crudu et al., Convergence of stochastic gene networks to hybrid piecewise deterministic processes, The Annals of Applied Probability 2012]
[Herbach et al., Inferring gene regulatory networks from single-cell data: a mechanistic approach, BMC Systems Biology 2017]
On 2019-06-07 11:21:21, user Lauri Törmä wrote:
Happens to me in flies all the time. I've never had a correct knock-in.
On 2019-06-06 18:27:31, user Amrit Singh wrote:
Hi Y-h. Taguchi, this is interesting. The datasets in the mixOmics R-library are a small subset of the original data used in the manuscript (https://www-ncbi-nlm-nih-go.... Please see https://github.com/singha53... for the entire breast cancer data (train and test). I wonder if there is a sparse version of HOSVD, in order to perform variable selection which is already implemented in DIABLO using soft-thresholding.<br /> Best,<br /> Amrit
On 2019-06-06 15:58:13, user Andrey Yurkov wrote:
Dear Primrose,
I suggest to consider the following publication<br /> https://link.springer.com/a...
It has been demonstrated that the occurrence of Saccharomyces paradoxus shows a clear seasonal pattern. Using a conventional isolation plating technique, authors isolated Saccharomyces paradoxus in a substantial proportion.
On 2019-06-06 07:13:16, user Moshe Olshansky wrote:
Supplementary Material section says that the authors reimplemented the HiCCUPS algorithm in https://github.com/mirnylab.... Indeed, I was able to clone that folder and there is dotfinder.py file there which is supposed to do the job. However, there isn't even a hint about how this file can be used.<br /> Is it possible to write some instructions?
On 2019-06-06 01:36:12, user Joe Flood wrote:
Very irritating the way these authors insist on using ISOGG nomenclature for Y haplogroups, but can't be bothered keeping abreast of changes. 'R1b1a2' used for the Catacomb samples has at various times in the past been Z2103, M269, V88 and is now BY15383. My guess is they probably mean the Z2103 'Yamna' subclade'
On 2019-06-05 17:58:30, user Ryan Kelly wrote:
Note that the eDNA index we discuss in this paper is also equivalent to the classical ecology "Wisconsin double-standardization" as implemented in vegan with appropriate margins specified.
On 2019-06-05 07:20:57, user Connor Lynch wrote:
Thanks for the fascinating paper! It was the subject of much interesting discussion at our lab, and some questions came up that I was wondering if you could resolve: <br /> - Was a sham surgery used for the controls during the cervical ligation, and if not, was there a reason for this? <br /> - Are there negative controls for the IHC images? Some of the stains overlap almost perfectly, whether this is because of colocalization or fluorescence bleed-through is not entirely clear (e.g. figure 9)
On 2019-06-04 23:09:40, user Charles Warden wrote:
With my own samples, HIBAG was something that I also used for SNP chip data (in addition to SNP2HLA), but I thought the best sense of robustness was from checking for concordance between SNP chip and DNA-Seq data (using bwakit and HLAminer):
https://github.com/cwarden4...
So, I initially thought you might have only used SNP chip data (since I saw SNP2HLA), but I think some of those other programs used Illumina sequencing data (unless I am misunderstanding something). So, perhaps it is at least worth adding bwakit for comparison? I would have expected that as a common program for HLA typing for Illumina reads.
Thank you very much for posting this pre-print!
On 2019-06-04 22:39:55, user German Matias Traglia wrote:
I have a question: have you use the sequences of SGI-3 reported by Petrovska 2016? or have you use the sequence of SGI-3 reported by moreno 2012? Because,the two SGIs are two different islands. SGI-3 of Petrovska is SGI-4 (in the introduction appointments is like SGI-3) [SGI-4, note addendum for nomenclature change from SGI-3 (Petrovska et al., 2016)] while SGI-3 is the one reported by Moreno. What SGI did you use to this paper?
On 2019-06-04 22:05:53, user Charles Warden wrote:
I overlooked this link the first time that I saw this paper, so I hope this is useful for other people:
https://www.nygenome.org/bi...
I think having raw data from dbGaP is also important, so I hope that will be available soon.
Thank you very much for posting this very interesting paper!
On 2019-06-04 19:24:12, user Pedro Luna wrote:
Nice study, I would suggest to measure nestedness and modularity using frequency based algorithms. Also It would interesting to remove alien species from the networks and then recalculate the network descriptors to actually see what is the effect of alien species in the network structure. As you present your work you are only comparing network a with network b, your data has huge potential to make more interesting insights.
On 2019-06-04 15:29:55, user Timothy Hla wrote:
This work has just appeared in print in the Journal of Experimental Medicine - here is the link to the paper - <br /> http://jem.rupress.org/cont...
On 2019-06-04 14:45:30, user Florian Wennmans wrote:
Interesting paper overall. I like the completness of the study with the use of tools that are usually not included in these comparison studies. However, I was a bit surprised by some results here, in particular regarding snooper. We tested this tool in my lab few years ago and obtained reasonable performances compared to other callers. If I recall, snooper needs known germline variants (such as the ones you provided to Mutect) to filter polymorphisms accurately, which seems not to have been the case here. That could maybe explain the FPs that ''often coincident with germline variants''. Finally, unless authors updated their tool, snooper's output needed to be filtered on class probabilities, not only on ''PASS'' flag, which again seems not to have been the case here either. If you don't play with class probabilities, given that it reports any mutation with a class probability over 0.5, you will end up with a lot of errors.
On 2019-05-31 22:52:18, user Charles Warden wrote:
I thought this was a very interesting paper.
I like that you showed joint variant calls for GATK HaplotypeCaller and MuTect2. My impression is that the joint HaplotypeCaller strategy may be currently underutilized, but I realize that having a mix of somatic and germline calls may make things tricky. I noticed the methods say "fi?ltered out the variants that were not homozygous for the reference allele in the healthy sample." What if you 1) filter any variant not called jointly in both tumor and normal samples and/or 2) provide a metric for the combined variants (perhaps for missense variants in cancer-related genes). I realize this may not be completely fair, but sometimes the cancer is due to a mix of somatic and germline variants in the same gene. So, if one metric is too low and one metric is too high (but both are provided), perhaps that is a fair overall representation of what can be done with joint HaplotypeCaller variants?
Plus, I find the earlier RNA-Seq "Best Practice" filters to be helpful. While things are already a little crowded, are you testing the effect of using GATK HaplotypeCaller with parameters like -dontUseSoftClippedBases and GATK VariantFiltration with parameters like -window 35 -cluster 3 -filterName QC -filter "QD < 2.0" -filterName FS -filter "FS > 30.0" (or equivalents in latest version of GATK)? I wonder if this could improve precision.
Also, my understanding is that this is all simulated data. I think this other study shows lower concordance among lower frequency variants. Could you possibly make use of that data for additional benchmarks (and report empirical recovery of variants with different Illumina sequencers)?
On 2019-06-03 14:42:33, user Mick Watson wrote:
Cool paper! Suggest citing https://genomebiology.biome... ;-)
On 2019-06-03 11:47:35, user aquape wrote:
Thanks a lot, very interesting findings, but the interpretation seems to assume that all these fossils were fossil relatives of humans, but not of the African apes. This is an unproven assumption. Asian apes have lots of fossils relatives, it's believed, but for some reason (anthropocentric bias?) it's traditionally assumed that the African apes have virtually no fossil relatives, whereas humans are thought to have innumerable fossil relatives or even ancestors. This is statistically impossible, of course: if orangutans have so many fossil relatives, why nhy not chimps, bonobos or gorilla?<br /> The solution is not so difficult IMO: most so-called human traits of australopithecines (e.g. vertical spine, thick enamel, low pelvis, full plantigrady etc.) are not uniquely-human-derived, but are hominid- or even hominoid-primitive, and were probably already present in most Miocene hominoids (and sometimes lost in African ape evolutions). All apes have centrally-placed spines (vs. dorsally-placed spines in monkeys & most mammals), this suggests that ape ancestors were already "vertical", not for running over open plains, but for climbing vertically (arms overhead) and/or wading bipedally (google e.g. "bonobo wading") and/or hanging from branches (suspensory).<br /> It is likely IMO that the East-African and the South-African australopithecines were no close relatives of each other, but that both branches evolved in parallel (allopatric parallel evolution) from more gracile to more robust (e.g. afarensis->boisei // africanus->robustus), google e.g. "ape and human evolution 2018 Verhaegen". This would help explain the different "stages" the paper describes in australopithecine limb bone evolution, with early australopiths Lucy & Little Foot resembling the ancestral condition, and Pan and Gorilla apes evolving in parallel longer upper limbs, but humans evolving longer lower limbs.
On 2019-06-03 11:14:14, user Felix Jonas wrote:
Dear Junai,
thanks for sharing your protocol with the wider community it looks really interesting. I was wondering if you tried washing away the proteinase K instead of deactivating. As some protocols recommend to not heat after Tn5 to avoid dissociation of the dsDNA before the first 72C step. Also do you usually heat activate the HiFI enzyme before using it?
Thanks,<br /> Felix
On 2019-06-03 09:29:22, user aquape wrote:
Thanks for this stimulating paper, but fossil & genetic evidence suggests that archaic Homo, probably already 1.8 Ma, had acquired most of our anatomy & control of the oral cavity, pharynx & larynx, e.g. MYH16 inactivation of the masticatory musculature, hyoidal descent, incisiform canine teeth within a closed, parabolic tooth-row etc. The most parsimonious solution for these & other innovations (versus apes, habilis & australopithecines) is the intercontinental dispersal of early-Pleistocene Homo along African & Eurasian coasts & rivers, where part of their diet consisted of littoral foods, which are extremely rich in brain-specific nutrients (DHA, taurine, iodine etc.) and best explain the dramatic brain enlargement seen in archaic Homo ("seafood is brainfood"), google e.g. "coastal dispersal of Pleistocene Homo 2018 biology vs anthropocentrism".
On 2019-06-03 07:47:43, user Ashwin wrote:
How to access supporting information for this article. ?
On 2019-06-03 07:20:46, user Kacper Rogala wrote:
Good effort!<br /> Please allow me to make a few small suggestions before you submit this manuscript for peer review.<br /> First, consult with a serious centriole structural biologist -- they might help you tie up some loose ends.<br /> Second, figures 5C and S7 can be considered nonsensical. Unfortunately, those things are non trivial, and plugging in a sequence to I-TASSER and making up a theory around a largely disordered protein will not take you very far. Also, the equivalent of S38 in human STIL was previously shown to interact with G-BOX of SAS-4/CPAP -- see figure S3 in Hatzopoulos 2013 (PRxxPxP motif is only a small section of the binding interface). Phosphorylation of S38 probably just makes Ana2 stick stronger to Sas4, and any discussion regarding conformational changes is flimsy. You might want to consider removing / trimming those sections completely, and adjusting your Fig. 6.<br /> Third, your interpretation of Ana2(FL) binding Sas4 stronger than the Ana2(1-60) fragment is not quite right. Instead of extra binding sites, think about the avidity effect provided by the super-strong oligomeric structure of SAS-5/Ana2//STIL.<br /> Finally, it would be nice if you guys showed some love for the worm folks, and cited them properly for a range of SAS-5 discoveries!<br /> Anyway, it's great to see that someone is pushing on the procentriole front. Good luck in peer review!
On 2019-06-03 02:25:23, user Wenming Xiao wrote:
We replaced this article with the pre-peer review version. Peer reviewed version will be published on a journal soon. Apologize for confusion!
On 2019-06-01 17:32:23, user Tyler A. Elliott wrote:
A really interesting paper, and methodology to help determine what TEs in a genome might be active. Would love to see this applied to other genomes and see what it does.
I'm wondering if the authors noticed any effect of age on the autonomous or pervasive transcription of L1? Mainly I'm thinking about the GTEx data and the bias towards samples collected from older individuals. I know there is some evidence of this, not sure if it's been found in humans: https://link.springer.com/a...
On 2019-06-01 17:04:26, user Anders Sejr Hansen wrote:
We recently found out that although I thought we uploaded the supplementary materials, they do not seem to be available for download. Sorry about that. Here is a link where the Supplementary Materials and Figures can be found: https://www.dropbox.com/s/u...
On 2019-06-01 14:57:38, user Titus Brown wrote:
It seems to me like the methods section is missing for this preprint, or am I missing something myself? :)
On 2019-06-01 10:32:22, user Dr. Rajesh Kumar wrote:
This is one of the cool paper where Reichert´s lab from HHU, Düsseldorf first time shows the that Mitochondrial cristae and crista junctions (CJs) both undergo frequent cycles of fusion and fission on a second-time scale. This is how the organelle itself undergoes frequent cycles of fusion and fission from the surface/outer membrane. <br /> The discovery of the dynamics/plasticity of the mitochondrial inner membrane (MIM) under physiological conditions was possible due to the STED super-resolution microscope. This is an important cellular process that may have a direct impact on the kinetics of chemical reactions, the structure of the OXPHOS system and cellular metabolism. I guess this is exciting to further zoom in the process that might open a new door where cristae dynamics may a promising therapeutic target to modulate metabolic dysfunction/mitochondrial disease.<br /> Congrats to the whole team !!
On 2019-05-30 21:23:07, user Aathmaja Anandhi Rangarajan wrote:
Great work on Mitochondrial dynamics. Beautifully designed experiments with great microscopy.
On 2019-05-30 20:29:23, user Arun Kumar Kondadi wrote:
This article changes our view of mitochondrial cristae plasticity. We used live-STED and other techniques to unravel a model of cristae and CJs dynamics .....
On 2019-05-31 23:25:09, user Charles Warden wrote:
Very interesting paper:
What if you plot precision (for breast cancer), with pathogenic variants for BRCA1 versus BRCA2 (versus current PRS predictions)?
On 2019-05-31 22:56:24, user Charles Warden wrote:
I'm sorry, but I noticed the following typo:
The PDF abstract is correct. However, the abstract on this page has a typo for the GitHub Link:
https://github.com/lbcbsci/ra --> should be https://github.com/lbcb-sci/ra
On 2019-05-31 14:58:19, user Lutz Fischer wrote:
Hi,
you state in the supplement that the FDR is calculated as
FDR = Ndecoy/(3*Ntarget)
This to me appears to be underestimating the FDR. It is true that the random space for false positives cross-linked peptides can be 3 times larger - but only for the case that both peptides of an identification are wrong and henceforth random. <br /> For high-scoring matches, the far more likely case is, that one peptide is correctly identified and one is incorrect. In these cases only one peptide is random - resulting in a linear space for random matches. For these you would then underestimate the FDR threefold.
Another question: Did you split the data into inter and intra cross-links for FDR-analysis?
Lutz
On 2019-05-31 12:40:48, user Alex Leitao wrote:
Very interesting article, as usual.
Just to let you know that figure 6D is missing the wt/NimB5 labels.
On 2019-05-30 19:19:55, user John Rinn wrote:
During the revision of our manuscript a study was published van Heesch et al. Cell 2019 demonstrating the Tug1 locus can encode a peptide. We noticed that some data in this study is the same data contributed by the authors in our pre-print (with different representative images). These data will be removed from our revised manuscript to uphold data quality standards (specifically Lewandowski Supplemental Figure 6C-D -- that is similar to -- van Heesch et al Figure 4H-I)
On 2019-05-30 08:21:59, user Martin Modrák wrote:
Liked the preprint and great that all the code is available. Good job! Should my main takeaway be that amplicon sequencing of the V4 (or other regions) in 16S is not the best investment of time and money and we should change to other marker genes? Or would you caution against that interpretation of your data?
A few nitpicky suggestions follow:
I would use heatmaps or contour plots in Figure 1 instead of the 3D surface which (to me) was very hard to decipher.
I feel that when you combine recall and precision as in Figure 3b, it might be better to use F1 than their sum (not that would likely change your conclusions much) - if that was a conscious decision, might be worth discussing in the paper.
One thing that is not clear is whether you used the whole 16S rRNA as a marker gene or just the region used in common amplicon sequencing protocols (from Figure 3c I would guess you used whole length).
On 2019-05-29 23:23:37, user Liang Huang wrote:
A heavily-revised version of this paper has been accepted to the Proceedings of ISMB 2019 and will be published by the journal Bioinformatics in July 2019 as:
Liang Huang, He Zhang, Dezhong Deng, Kai Zhao, Kaibo Liu, David Hendrix, and David Mathews (2019). LinearFold: Linear-Time Approximate RNA Folding by 5'-to-3' Dynamic Programming and Beam Search. Bioinformatics, Vol. 35, July 2019, Special Issue of ISMB 2019 Proceedings.
https://www.iscb.org/cms_ad...
Stay tuned -- the final PDF will be online on the publisher's website (Oxford University Press) in July 2019.
In the meantime, please try out our LinearFold server and code:
On 2019-05-29 20:33:16, user Leonardo Pereyra wrote:
Hi, I enjoyed to read your manuscript. I saw that you measured the the expression level of NRC-silenced tomato leaves by Semi-quantitative RT-PCR. Perhaps by measuring the silencing percentage by real-time PCR you can do a better correlation between the observed phenotypes of tomato silenced plants and the expression level of each silenced NRC gene, and to estimate independently if the NRC2 a/b and NRC3 are redundant helper proteins, if there is a relatively preference by one of them. Of course as you mentioned you will answer the question by knock-out with CRISP-CAS9.
Greetings from Mexico!!!
Leonardo Pereyra <br /> PhD. student.
On 2019-05-29 14:39:32, user Victoria wrote:
I appreciate that the authors do acknowledge that the recombinant and native enzymes may possess different properties. This assumption is often underestimated, especially by biochemists with chemical background. Yet I would like to pinpoint that the statement of the manuscript that "The biochemical and enzymatic properties of the native KADHC in mammalian tissues have not been characterized", is not true. In 2017, a paper was published (PMID: 28601082), where it was shown that a rather high levels of OGDH activity in the brain homogenates are devoid of the OADH activity. That said, one cannot exclude formation of the hybrid complexes in the cell lysates as an artefact of the detergent extraction and/or further purification procedures. Regarding the OADH activity of the purified OGDH complexes, one could add other couple of references to the discussion: PMID: 10848975; PMID: 8495733
On 2019-05-29 12:04:23, user Enora Dupas wrote:
Some revisions have been made since this preprinted post. To read the full version of the article, please visit : https://doi.org/10.1016/j.m...
On 2019-05-28 21:23:35, user Kristof Theys wrote:
We intend to make all code used for this study available to anyone that can benefit from it. We plan to use the repository https://github.com/ktheyss to share the framework.
On 2019-05-28 21:15:48, user Kristof Theys wrote:
All R code will be shortly posted on https://github.com/ktheyss/..., to be used in future studies or to be improved.
On 2019-05-28 16:04:37, user Mia Shin wrote:
Members of the Lander Lab at Scripps Research in La Jolla, California discussed this manuscript at a Journal Club and would like to share our thoughts with the authors as well as the broader scientific community.
In this manuscript, Rubinstein et al introduce “Shake-it-off,” a cryo-EM specimen preparation device assembled by the authors using parts from an ultrasonic humidifier, homemade self-wicking EM grids, 3D printed parts, and a Raspberry Pi single board computer. Notably, all components can be manufactured using open-source files shared on the internet or readily purchased. Rubinstein stated via Twitter (@RubinsteinJohn) that the SIO device was constructed for ~$1000 Canadian dollars (approximately $740 US dollars). He added, “This is #frugalscience.” Indeed, SIO is remarkably less expensive than the cryo-EM specimen preparation devices that are commonly used by the cryo-EM community, indicating that this device could readily be adopted in-house by any lab interested in pursuing cryo-EM.
The SIO device attempts to address several substantial limitations that the community currently faces during cryo-EM specimen preparation using traditional blot-plunge devices: 1) more than 99.9% of the (often precious) sample applied to the EM grid is blotted away and trashed, 2) the plunge-freezing process occurs on the timescale of several seconds to minutes, which can lead to problematic air-water interface interactions for macromolecules (preferred orientation, complex disassembly, denaturation/aggregation, etc.), and 3) questionable reproducibility of ice thickness from grid to grid. According to the authors of the manuscript, only 1 uL of sample is required for applying sample to the ultrasonic humidifier, a 3-4-fold reduction in wasted starting material. The authors report plunge-freezing by SIO is on the timescale of 100 ms, which should substantially reduce hydrophic effects from the air-water interface compared to traditional blotting. Additionally, the resulting grids appear to consistently have a large “mountain” of frozen sample with a ring of optimal ice for data acquisition at its periphery, thereby reproducibly delineating where to image on each grid.
The authors used SIO to prepare samples of equine apoferritin that resolved to better than 3 Å resolution, demonstrating the device’s ability to vitrify robust samples for high resolution cryo-EM analyses. As SIO appears to address long-standing specimen preparation problems faced by the cryo-EM community for a fraction (a very small fraction!) of the price of the sophisticated Spotiton/Chameleon devices, we are excited about this promising design and its potential to revolutionize specimen preparation for cryo-EM labs worldwide.
However, we have several points we’d like the authors to address prior to this manuscript’s publication in a peer-reviewed journal:
Major Points:<br /> 1. Figure 4 shows two atlases of apoferritin grids prepared using SIO. Both grids have a circular “mountain” of ice that occupies ~25% of the grid area circumvented by a narrow region of suitable ice for high-resolution imaging. We are curious about the reproducibility of these types of grids. Although these two grids appear nearly identical, do grids containing different protein sample with different buffer conditions (e.g. salts and detergents) produce the same “mountain” of ice phenotype? Additionally, we invite the authors to speculate as to the origins of this region of thick ice, and whether or not it can be correlated to a region of the ultrasonic spray that may emit larger droplets than other regions. Presumably the drops released from the piezo are smaller than this mountain, suggesting that the drops puddle together during the plunging? Could this happen within 100 ms? Perhaps if this is the case, the ultrasonic spray can be re-positioned or improved such that a larger region of the grid may be amenable for high-resolution imaging.<br /> 2. While the authors were able to resolve samples of equine apoferritin prepared using SIO to sub-3A resolution, we would really like to know whether its purported plunge-freezing speed is able to overcome preferred specimen orientation at the air-water interface, as has been reported for Spotiton/Chameleon. In addition to the benefits of using less sample and reproducibility, minimizing air-water interactions would likely be one of the primary motivations for other groups to build their own SIO plunge-freezing device. We request that the authors include images of hemagglutinin, a sample that exhibits pathological preferred orientation (Tan et al., Nat. Methods 2017), to test SIO’s ability to overcome preferred specimen orientation. Additionally, the inclusion of tilt-series to assess the percentage of proteins associated with the air-water interface (a la Noble et al., eLife 2018) would be greatly informative.
Minor Points:<br /> 1. Is the micrograph shown in Figure 4C representative for the dataset or is the best micrograph. Near the bottom left quadrant, there is a circular area characteristic of protein denaturation. While we see this routinely in our micrographs from a variety of samples and datasets, particularly in areas where the ice is very thin, it would be troubling if this type of pathology was present in many or most micrographs (or worse).<br /> 2. How is the piezo device is cleaned after each sample is prepared? We are able to see in the GUI that there is a button for cleaning the piezo device, but no description of the mechanism was found.<br /> 3. The 1 uL droplet that is applied to the piezo device is obviously much smaller than the surface area of the ultrasonic humidifier. Does one achieve different results depending on location of sample placement on the piezo device? If this is the case, perhaps there can be modifications to the design of the piezo device to identify the optimal location for sample placement for spraying.<br /> 4. Lastly, is there a potential health hazard associated with the ultrasonic spray emitting sample. Do you recommend the user to use a mask or placing SIO into a covered chamber to avoid any aerosol contamination being emitted from the device, particularly in the case of BSL-2 level samples (i.e. prions)?
On 2019-05-28 15:21:10, user Darren Thomson wrote:
We published this Spk phenomenon in Candida albicans a few years ago, if you want to take a look.
On 2019-05-28 12:47:57, user Jacob Pollier wrote:
The metabolite identification in this paper seems seriously flawed. It is not because a compound has the same (accurate) mass as bayogenin 3-O-cellobioside that it is bayogenin 3-O-cellobioside. To properly identify a metabolite, you need to compare the mass, retention time and fragmentation spectra of the peak in the rice extract with those of an authentic standard. To claim a dicot-specific metabolite to be present in a monocot without any proper standard is a blatant overstatement and makes this study totally unreliable. The same is valid for the other metabolites discussed in this paper, their IDs are absolutely unreliable. Also the chromatographic method is not well-described. Column? Solvents? Gradient? Ionization? Not to mention the gene identification...
On 2019-05-28 09:14:29, user Mikko Rautiainen wrote:
Please make it clear in the text that the GraphAligner in your comparison is not actually GraphAligner, but just the bit-parallel DP extension algorithm. I recommend using terms like "bit-parallel" or "Rautiainen et al." or something else that won't get confused with GraphAligner.
I ran GraphAligner (version 1.0.7, bioconda, "/usr/bin/time -v GraphAligner -g MHC1.vg -f M3.fastq -a alns.gam -t 4") on your MHC1 graph and M3 reads. It aligned all reads in 2 minutes cpu-time (30 sec wall-time) on my laptop. This might be an another software for your seed-and-extend heuristic comparison. Since the alignments are outputted in .gam format, you can even reuse your pipeline for comparing to vg.
On 2019-05-28 08:11:24, user Andres Romanowski wrote:
Sounds interesting! I will give it a thorough read!
On 2019-05-28 02:20:02, user Huiwang Ai wrote:
There is a relevant paper published on Nature Methods two years ago: https://www.nature.com/arti...
On 2019-05-27 14:41:04, user Florian Freimoser wrote:
This work has now been published in Molecular Microbiology:<br /> https://onlinelibrary.wiley.com/doi/full/10.1111/mmi.14272
On 2019-05-27 14:19:11, user bli wrote:
Dear authors, maybe I haven't searched thoroughly enough, but I haven't found indications regarding the availability of tRNAscan-SE 2.0. What is the official site hosting the source code and usage documentation ?
Thanks in advance.
On 2019-05-27 13:15:29, user Zhihua Han wrote:
Wonderful, I want use this on human intervertebral disc
On 2019-05-26 22:16:50, user DeboraMarks wrote:
Dear Authors
You might want to consider comparing your approach for variant prediction to results from following two paper: plausibly the state-of-art for variant prediction from sequence:
2.( Unsupervised) Variational autoencoder, Riesselman, Ingraham, Marks "Deep generative models of genetic variation capture the effects of mutations" . Nature Methods 2018 <br /> https://www.nature.com/articles/s41592-018-0138-4<br /> Compared to 40 deep mutational scans
On 2019-05-26 18:29:32, user Thomas Munro wrote:
I think the suggestion that ORCID profiles should include a contact link is excellent, and deserves to be mentioned and emphasized in the abstract. It has several advantages over institutional email forwarding and webmail addresses:
*any published email address attracts spam. Academic addresses can currently be scraped from article landing pages, and even exported in bulk from pubmed. <br /> The resulting spam is harder to filter than the non-academic variety, since the scammers know the recipient's full name and research topic. This problem is a gift to deceptive ('predatory') journals and conferences, and a nuisance for authors. The difficulty of filtering also increases the risk that authentic messages will be missed.<br /> *as noted in the previous comment, free webmail accounts increase the difficulty of confirming the true identity and affiliation of the recipient, and were used in recent fake peer review scams.<br /> *institutions have little incentive to maintain forwarding addresses for former employees and students. It would impose ongoing costs with little benefit to the institution.
All of these problems could be overcome if ORCID introduced a web contact link, forwarding to an unpublished email address, with a captcha to prevent spam. Links like this have been used by Nature journals for years.
*This would allow authors to unilaterally stop publishing their e-mail addresses, solving the spam-bait problem.<br /> *This method would also offer greater persistence, creating one standard point of contact for all researchers who want to remain contactable.<br /> *The point of contact would be verifiably linked to a particular researcher.<br /> *Finally, if the option of posting the message publicly were added, the incentive to respond would be stronger. A failure or refusal to provide data on request would be public knowledge. This would also spare authors the need to reply to the same question or request repeatedly, providing a public space to share links or answer questions.
On 2019-05-26 16:32:30, user Robert Kingsley wrote:
A peer-reviewed and modified version of this preprint is now available at https://www.frontiersin.org...
On 2019-05-25 05:07:40, user Alex Crits-Christoph wrote:
Thank you for sharing this work!
(1) To try to clear up some confusion in the literature - Dormibacteraeota are referred to as "ANG-CHLX" in Diamond et al 2019 ( https://www.nature.com/arti... ) due to some overlap in time frame when that work was done and the candidate phylum name was proposed and a slow reaction time. It is great to see more analysis of this ubiquitous phylum.
(2) It has been previously reported that sequences of candidate phylum Rokubacteria have been misannotated as Nitrospirae (a sister phylum) in genomic databases ( https://www.frontiersin.org... ) - due to the abundance of Rokubacteria in soils at the eel river czo site I wonder if some of the Nitrospirae data shown here could actually be sequences from Rokubacteria.
On 2019-05-25 00:06:05, user Anders wrote:
I just noticed that the supplementary materials are not available, even though I thought we uploaded them. Sorry about that. If anyone is interested, they can be downloaded here: https://www.dropbox.com/s/p...
On 2019-05-24 21:29:22, user Ana wrote:
Can the authors discuss how the algorithms were selected? Just googling dimension reduction single cell, 2 of the top 3 papers were not compared and even not cited. The other one is cited but not compared in the manuscript. <br /> Townes, F. William, et al. "Feature Selection and Dimension Reduction for Single Cell RNA-Seq based on a Multinomial Model." bioRxiv (2019): 574574.<br /> Becht, Etienne, et al. "Dimensionality reduction for visualizing single-cell data using UMAP." Nature biotechnology 37.1 (2019): 38.<br /> Ding, J., Condon, A., & Shah, S. P. (2018). Interpretable dimensionality reduction of single cell transcriptome data with deep generative models. Nature communications, 9(1), 2002.
On 2019-05-24 18:41:01, user Edgar Gonzalez-Kozlova wrote:
Very interesting!, I wonder if the definition of species for bacteria needs to be readjusted ?
On 2019-05-24 15:13:57, user Peter Ellis wrote:
Transcription and translation are ongoing throughout spermatogenesis. This model has no basis in known cell biology.
On 2019-05-24 03:10:13, user Alex Combes wrote:
This work has been published as follows:<br /> Kidney organoid scRNA-seq and a comparison to human fetal kidney are presented in: Combes & Zappia et al., Genome Medicine 2019 doi: 10.1186/s13073-019-0615-0 <br /> Analysis of the variability in human kidney organoids including some single cell data: Phipson, Er, Combes et al., Nature Methods 2019 doi: 10.1038/s41592-018-0253-2 <br /> Analysis of the developing mouse kidney cell types and signalling between them: Combes & Phipson et al., Development 2019 doi: 10.1242/dev.178673
On 2019-05-23 20:10:53, user Fredy Barneche wrote:
This study is now published in Genome Biology: doi.org/10.1186/s13059-019-...
On 2019-05-23 13:30:52, user MohanBalasubramanian wrote:
This work will soon be published in Int J Mol Sciences in a special issue edited by Kensaku Sakamoto on "Expanding and Reprogramming the Genetic Code", which carries ~15 articles on this subject.
On 2019-05-23 00:14:09, user Charles Warden wrote:
I apologize that I won't be able to look into the raw data in the immediate future (and was the data deposited into a public database?), there were a couple things that seemed strange about the report:
1) Table 1 seems too high for high-coverage variants (not to mention low-coverage variants). The variant counts also seem low. Is there something special about those regions? Do they tend to be at higher copy number, or of some other reason that makes them easier to detect?
For higher coverage data, I think you had to look within regions that were easier to call (which is not representative for a person's overall set of variant calls, even within CDS regions), to predictive statistics in the 99+ percent range: http://cdwscience.blogspot....
2) For the HLA genes, I was expecting more of a difference for the HLA-D genes than the HLA-A/B/C genes (Figure 1), but maybe that is because I need to check more samples processed with different technologies. For example, you can see my HLA calls with various strategies in the "HLA Analysis Results" section here: https://github.com/cwarden4...
On 2019-05-22 23:31:32, user Charles Warden wrote:
I thought this part was interesting:
"How much does it actually cost to run a workshop like EDAMAME? The first year, we ran the workshop for less than $14,000; students paid their own expenses of room and board; and no workshop fees were charged."
On 2019-05-22 20:23:29, user Salvatore Pappalardo wrote:
The manuscript was finally published on PLOS ONE, on May 1st 2019. <br /> It is possible to access and download the full paper: https://journals.plos.org/p...<br /> Salvatore E Pappalardo, corresponding author.
On 2019-05-22 14:18:52, user A. S. wrote:
When you write "Mitochondria evolved from archaebacteria", you are probably confusing alphaproteobacteria (which are Eubacteria) with archaebacteria.
On 2019-05-22 12:14:15, user John Novembre wrote:
This looks very interesting and I hate to be the person to complain about overlooked work but... I’d love to see comparisons to prior work on an effective number of codons to deal with varying GC content (Nc’, from “Accounting for Background Nucleotide Composition When Measuring Codon Usage Bias” MBE 2002). I think your approaches likely have advantages, but a comparison, or at least a few more words of discussion, would be helpful. Happy to chat via email too if it’s helpful. John Novembre
On 2019-05-21 21:39:25, user Charles Warden wrote:
I see that you mention "precisionFDA" in the discussion, but I think having continual comparisons overtime (from individual companies or labs) has benefits.
For example, these are my experiences (with Veritas WGS data and Genos Exome data):
http://cdwscience.blogspot....
Also, while you can specify the regions to compare (like I did for CDS regions), there are also other apps where you can get more details reports.
For example, Vcfeval + Hap.py Comparison was suggested to me, and it helped me troubleshoot error messages to get the main variant comparison function working.
On 2019-05-21 19:32:06, user Emma wrote:
I'm not allowed to post the accepted manuscript here, but there were some important corrections. There's a copy at: http://eeg.github.io/lab/_s...
On 2019-05-21 19:05:51, user Michele Busby wrote:
Nice paper! One thing I think I didn't do very well in my application was how I handled library complexity. I think I modeled it as if you could sequence forever and keep getting more reads. Did you look at how library size affects the saturation?
On 2019-05-19 23:04:36, user Charles Warden wrote:
Thank you for putting together this pre-print.
I am sure that there are some situations where higher read coverage can be beneficial. Admittedly, I think other applications like mutation calling would have a relatively greater need for more reads (and that would depend upon the evenness of coverage for your library type, and possibly what genes you have the greatest need to check mutations in), but I think it is perfectly reasonable to focus on the differential expression part for one paper.
That said, when I saw the tweet mentioning "We find > 70% published studies would have benefitted from increasing number of reads sequenced", I was a little worried about the influence it could have on readers for the following reasons:
1) If somebody is considering purchasing a Desktop sequencer for RNA-Seq analysis, I think 2-6 Million reads to cover genes with above average expression may be a better option than using targeted gene panels. For example, if you do re-analysis (with unique read counts and updated differential expression methods, like DESeq2, limma-voom, etc.), I think the MiSeq data from the cuffdiff2 paper shows reasonable results (for treatments with clear gene expression changes).
2) In most cases, I am more concerned about people having replicates than needing more reads (at least for gene expression).
I apologize that I think it may be a little while before I can focus more on point #1, but I tried to take a quick look at this paper.
I think that it is great that you performed benchmarks with DESeq2, edgeR, and limma-voom (although maybe you want to change “limma” to “limma-voom” in the abstract?). I apologize for not being able to find this on the superSeq page (although I did find the reminder for the previous biocLite() command for dependencies to be helpful), but are tables of pre-processed counts (and their gene lists with all 3 methods) readily available for the 1,021 contrasts?
I am also glad that you are looking at differentially expressed gene counts (and not just unique read sequences) for your rarefaction plot, since I think that is a more relevant measure for whether you get functionally relevant results. However, in terms of the vignette example, I think the difference between 1338.968 “Estimated number of discoveries” at read depth of 1 and 1888.286 “Estimated number of discoveries” at read depth of 3x is within the range that could be achieved from changing the p-value method and/or changing the FDR cutoff (from 0.05 to 0.25 or 0.50, for example).
Similarly, I am concerned about some of the maximum gene counts in the pre-print, which look like pretty much the entire genome in Figures 2 (and are already above 2000-4000 genes in the theoretical example in Figure 1). I think the best balance for functional enrichment is often around 1000-2000 total genes (~5-10% of genes). So, I would be interesting in knowing if your framework can answer a question like “What is the range of reads needed to identify 1000 or 2000 differentially expressed genes?.”
While some treatments have greater effects, I think 10-20 million reads for a human polyA library is probably usually OK (and perhaps double that for a ribosome-depleted library, with a lower exonic percentage). I think that is pretty much what Figure 1A shows (although that looks like close to 30 million reads), but I am wondering if there is a figure derived from your ~1000 comparisons (and/or a parameter that can be added to plot pre-computed values in the R package).
Also, am I correctly understanding that you downloaded pre-processed counts? Did you look at some of the most extreme differences and test reprocessing the samples to see if that helped the differentially expressed gene counts become more similar? There are situations where I would prefer to start from FASTQ files and process all samples the same way.
For example, 60,000 in Figure 5 seems like it probably includes transcripts – is it possible to only look at unique gene-level counts (that is admittedly what I would be interested in checking)? Or, are there outliers that can be excluded if you only look at human and mouse experiments (trying to control for annotation effects)? Also, I’m not hugely concerned about the annotation in model organisms like yeast or fly, but the total number of genes in the genome is going to have some effect (both in terms of the effect on the differential expression models, as well as having very different genome sizes and maximum gene counts).
Finally, going back to my original point #2, I would expect replicates should help reduce false positives. With large enough sample sizes, I would expect to pick up more subtle effects. However, with 1-3 replicates, I think fewer genes to narrow down candidates may be beneficial (rather than increasing the number of genes identified). For example, at an estimated FDR of 0.05, how many genes are identified between biological replicates for the same group (to see if increased sensitivity may actually be affecting the accuracy of the estimation to allow more false positives, which seems likely if you are identifying >20% of the genome, in my opinion).
Or, it is a slightly different point, but I think 6 replicates are used in Figure 3A. If 6 replicates exist for an experiment, what is the effect of having 3 replicates at the current coverage versus 6 replicates at halved coverage? Sometimes, getting people to even do comparisons with triplicates can be a challenge.
I apologize that this is kind of a long comment, but that is because I think this is an important topic. When I get the point of being able to post some pre-prints, I realize that answering questions from long commenters can take time, but I think that is very important for the scientific community (in terms of helping put together the best possible paper for peer-review).
On 2019-05-21 12:17:17, user Brendan Barrett wrote:
now in press: https://www.nature.com/arti...
On 2019-05-21 10:33:38, user Annkatrin Bressin wrote:
Hi, I'm really interested in the data. Is the data publicly available? Thank you in advance.
On 2019-05-20 18:59:04, user Aaron Gitler wrote:
Comments on Bolognesi et al.
By Aaron Gitler and Steven Boeynaems (Stanford University)
Many human neurodegenerative diseases are associated with protein aggregation. In Alzheimer disease, Amyloid-beta (A) and tau form aggregates, in Parkinson disease, synuclein forms aggregates, and in Huntington disease, a form of huntingtin with an expanded polyglutamine tract forms aggregates. Whether or not these aggregates cause disease is an area of intense study, with important therapeutic implications. For example, if the aggregates are what is causing disease, then strategies to dissolve or degrade the aggregates would be warranted. But if the aggregates were potentially protective, then therapeutic strategies to promote aggregate formation of toxic entities would be warranted. For the neurodegenerative diseases ALS and FTD, the RNA-binding protein TDP-43 is mislocalized to the cytoplasm and forms aggregates in the degenerating neurons of patients with these diseases.
TDP-43 harbors a prion-like domain, which is able to drive liquid-liquid phase separation. The liquid phases formed by TDP-43 are thought to lead to the formation of solid insoluble aggregates. Several mutations in the gene encoding TDP-43 have been identified in familial and sporadic ALS patients. Almost all of these mutations are located in the prion-like domain. Some reports have indicated that these pathogenic mutations can increase the aggregation-propensity of TDP-43. Whether the loss of TDP-43 from the nucleus (where it regulates a diverse array of mRNA targets) or rather its accumulation in the cytoplasm in an aggregated form (perhaps causing dysregulation of other RNAs and RNA-binding proteins) causes ALS (i.e., loss of function vs. gain of function) is still unresolved and an area of intense interest.
In this manuscript by Bolognesi and colleagues, the authors performed a systematic analysis of the prion-like domain of TDP-43 – engineering over 50,000 mutations – assessing effects on aggregation and toxicity in a yeast model. They made several key findings:
1) Mutations that change hydrophobicity and aggregation are predictive of changes in toxicity<br /> 2) Surprisingly, the mutations they introduced that increased hydrophobicity and aggregation actually decreased toxicity! <br /> 3) The mutations that increase the liquid-like properties of TDP-43 increased toxicity<br /> 4) Using a clever analysis of double mutant combinations they were able to infer the existence of specific structures within TDP-43’s prion-like domain.
Overall, this is a very exciting new story that provides a comprehensive and deep analysis of TDP-43 prion-like domain with surprising findings that mutations that increase aggregation actually decrease toxicity and the ones that increase the liquid phases of TDP-43 are the more toxic ones. This study has important implications for thinking about how TDP-43 contributes to ALS and FTD and also provides a framework and new method to study other neurodegenerative disease-associated aggregating proteins.
We have several comments and suggestions for the authors to consider.
1) RNA-binding seems to be required for TDP-43 toxicity in yeast (PMID: 18434538 and PMID: 20740007). The authors focused their mutational scan on the prion-like domain of TDP-43, which does not include the two upstream RNA-recognition motifs (RRMs). The specific mutations they introduced to the prion-like domain could potentially influence the ability of TDP-43 to bind RNA. It would be interesting for the authors to test several of their mutations for impact on RNA-binding. The binding site for TDP-43 (UGUGUGUG) has been characterized and several of its targets are known. It is possible that the mutations that decrease toxicity do so by decreasing TDP-43’s ability to bind RNA. This result would still be interesting and important and would provide insight into potential communication between the prion-like domain and the RRMs. The authors could also introduce additional mutations in the RRMs (e.g., FW) in the context of their most toxic mutants in order to define if the enhanced toxicity of these new mutations depends on RNA-binding or not.
2) Kaganovich, Kopito, and Frydman have described distinct subcellular quality control compartments, to which distinct misfolded proteins are routed (PMID: 18756251). These include a juxtanuclear compartment (JUNQ) and a peripheral perivacuolar compartment named insoluble protein deposit (IPOD). Proteins in the JUNQ are more mobile than those in the IPOD. Toxicity of a model misfolded protein can be reduced by directing it away the JUNQ and towards the IPOD (PMID: 22967507). The authors should consider analyzing whether their toxic and non-toxic mutants accumulate in yeast JUNQs or IPODs, using the established markers for these compartments.
3) Expression level of TDP-43 strongly correlates with toxicity in yeast (and other systems). Surprisingly, the authors see (in Extended Fig. 5b), that the toxic mutants are expressed at much lower levels than the non-toxic ones. This result is unexpected and it would be helpful to extend the analysis to additional mutants, and potentially longer time of protein induction (e.g., >8 hours).
4) While recurrent ALS mutations in TDP-43 were in general moderately more toxic when expressed in yeast, they do not necessarily associate with increased charge or decreased hydrophobicity. Rather, they seem enriched for G/A/N to S/T mutations (5/12). While such mutations could increase hydrophilicity substantially via phosphorylation, these human disease mutations do not replicate the pattern observed for the very toxic yeast mutations. An explanation why “severe” mutants are not observed in humans is that such mutations are potentially not tolerated during development, leaving only the possibility for moderate mutations to cause age-related disease. Hence, while the signature of increased hydrophilicity/decreased hydrophobicity is clearly associated with toxicity in yeast, this finding does not necessarily illuminate the biophysics behind TDP-43 toxicity in humans.
5) Some of the well-established TDP-32 ALS mutants have been previously studied in vitro, where authors have shown that these mutants perturb liquid phase separation of TDP-43 (PMID: 27545621). It would be great if the authors could specifically image these mutants in their yeast system. Additionally, evaluation of key mutants, both ALS-relevant ones and the ones with strong yeast phenotypes, in a human cellular context would be crucial for interpreting these results. It is known that condensates may present with different material properties between yeast and human cells (PMID: 26238190), and yeast cells potentially lack the (non-)canonical chaperone systems that may alter TDP-43 phase behavior (e.g. PMID: 29677512, PMID: 30100264).
6) Currently, the round, dynamic, perinuclear TDP-43 condensates are described as being liquid(-like). However, this claim is only weakly supported. Micronscale dynamic liquid behavior should be observed before such claims can be made (i.e., plastic deformation, wetting, fusion/fission). As mentioned above, investigating the relationship of these condensates to IPOD/JUNQ will be important.
7) This paper once more provides evidence for the fact that aggregates may be beneficial by sequestering toxic protein species (e.g. oligomers, liquid droplets). Interestingly, a recent study identified stress-induced liquid-like wildtype TDP-43 condensates that were toxic to human cells (PMID: 30853299). However, it remains unclear what the effect of ALS-associated mutations is on the latter ones, and how they relate to the TDP-43 pathology in human patients. While these two studies suggest that indeed “liquid droplet toxicity” can indeed happen in yeast and human cell culture, it remains to be tested whether this relates to the disease. Especially, in the context of yeast: Overexpression of TDP-43 out of its endogenous context induces spontaneous cytoplasmic condensation and toxicity, which is generally not observed in human systems. This is likely caused by TDP-43 perturbing endogenous yeast RNA metabolism and proteostasis pathways. Any molecular perturbation (i.e. mutations) that make TDP-43 inert for yeast clients (i.e. aggregation in this study, RNA-binding mutants in PMID: 18434538 and PMID: 20740007), are expected to reduce toxicity. While loss of RNA binding is indeed beneficial in yeast, in the human system RNA binding protects against toxicity (PMID: 30826182). The latter observation suggests that toxicity in yeast –and the associated biophysical features of TDP-43– may not always be correlated with toxicity in human system
On 2019-05-20 08:22:53, user Jiri Hulcr wrote:
Hello Kirk et al.<br /> I am glad to see that you all are still interested in ambrosia beetles. With respect to the hypothesis about cycloheximide tolerance in some Ophiostomatales as related to their symbiosis with ambrosia beetles, I think it is important to consider that the tolerance is widespread in that fungal family, pre-dating their engagement with the beetles. I don't know the literature that well, but I would guess that cycloheximide tolerance is probably present in many Ophios that are not relevant to the beetles, or even compete with them. <br /> Have you measured whether the compound occurs in the galleries in quantities comparable to those in your media? That might get us close to a test whether it is related to higher fitness of the fungus, which would support your hypothesis. <br /> Thank you!<br /> Jiri Hulcr
On 2019-05-19 13:54:30, user Ana Pedro wrote:
All the original data mentioned in this pre-print can be found by consulting the original data from Tucker and Pedro, F1000 Research (2018)