different pathway databases and relies on the robust GSEA algorithm for a single-omics enrichment analysis. In a final step, those scores will be combined to create a robust composite multi-omics pathway enrichment measure.
robust = you get a choice of multiple methods
multiGSEA package
bioconductor
he active participation of RBPs in regulated gene expressionmay be pertinent to an emerging concept of phase separationduring gene expression
test
Orb has a clear preference for canonical TTTT(A)1–3T
Orb and Orb2 Bind to Linear CPE Motifs.
We identified 3,128 targets for Orb and 2,154 for Orb2
thresholds might be transferred wrongly. To prevent this, flowLearn selects prototypes that define sample groups
flowLearn: fast and precise identification and quality checking of cell populations in flow cytometry
Reading this article might spark hypotheses as to how intron retention might have regulatory function....
I find this a great systematic review of how introns appear in the different phases of transcription and how they are found (or speculated) to be involved in amy aspects of :
We then devised a measure of co-transcriptional splicing for each exon by calculating the difference of intronic read coverage at the 3′ and 5′ ends of the flanking introns
Nascent transcripts at different stages of formation throughout the intron generate a gradient. The fact that this slope spans across individual introns, rather than across the entire transcript, can be explained by co-transcriptional splicing.
Raises question: by this argument, might lack of such a gradient be taken as intron retention as opposed to co-transcriptional splicing? How to distinguish?
Intronic RNAs were abundant in brain tissue, particularly for genes involved in axonal growth and synaptic transmission
we see similarly in our analysis of genes having > 25% intron retention in wildtype+naive
Finally, the fact that many IRTs are subject to NMD adds an additional layer of complexity, as the gene may be observed as expressed at a lower level due to the specific degradation of IRTs
I do not expect there is a bioinformatics approach to account for this. The article does not suggest a bioinformatics approach for normalizing for rate of NMD, nor have I seen one elsewhere.
s many of these harbour independently transcribed small RNAs, such as snoRNAs or microRNAs (Fig. 3C). If host transcripts are expressed at high levels, and a coverage cut-off is not implemented or used, it might be erroneously concluded that IR affects the coding gene.
Did we try to protect against this possibility with sample prep?
implement a coverage cut-off,
we have no cut-off
only ∼10–25% of sequencing reads in a typical paired-end library span one or more splice junctions
note: our percentage is > 25% in all samples, and in most closer to 33%
Pitfalls of splice-junction-only approaches to analyse IR from short read sequencing data
we do not take this approach
Hence, any bioinformatic tool for IR identification and differential analysis which does not include an approach to deal with multi-mapping reads and repetitive sequences in the genome is likely to miss substantial, functionally relevant biological complexity.
I have not a position on whether Bai is correct or not, but note this review article presents the issue as lacking consensus and does not propose a best practice.
filters in place to discard reads that map to more loci than a specific threshold
we use STAR configured to discard multi-mapping reads.
Long read sequencing captures full intron length but is limited by sequencing depth and accuracy
interesting, but, i don't think the reference really support this statement.
GSEA
this is the approach I have taken.
based on the orthology relationship or conservation of their flanking exons
we will implicitly be taking this approach, based on our use of the 27 species conservation analysis performed by UCSC.
Recently, enrichment of particular RNA binding sites has been observed in sequences of frequently retained introns as well as their flanking exons in human
did Teddy specifically scan for RNA binding sites to see if we can recapitulate this result? Similarly, did we evaluate whether Teddy's DREME analysis identified any known fly regulatory signals? We should!
benchmarking studies of best practices for IR detection, normalisation and quantitation similar to those available for differential gene expression analysis [50,51] need to be carried out.
send me that review/bake-off when you see it!
Both strategies also require a filtering step to segregate IR in differentially expressed genes from IR in non-differentially expressed genes
disagree
followed by a robust assessment of whether the read depth is adequate to sample intronic regions. This can be done by subsampling reads and examining splice junction or intron coverage statistics
here is that word robust again. The advice of "examining" is also not specific. We can do this, but it may still be heuristic/judgement call as to whether we can dial down the sequencing depth by 75%, as proposed.
validation rates for estimates in the levels of IR remain low. For example, one study reported a correlation coefficient (r) of only 0.63 between IR fold-changes determined using RNA-seq and qRT-PCR [19]
I would be very interested in our performing sufficient qRT-PCR to allow seeing if we "beat" this rate. I bet we do (at least for pure intron retention events).
IR detection can be confounded by antisense transcription
good point. Our protocol currently does not do so. I recommend we consider altering our protocol accordingly.
passing any coverage, splice junction boundary balance or other cut-off.
we have no such cut-offs. We should not need them. We depend upon analysis of variance to properly adjust for low/high expression levels in performing the contrast.
or uniquely mappable region of the intron
we do not use "uniquely mapple region". We arguable shoujld implement a variation of this and it might reveal new levels of changes in intron retention . However, we had extraordinarily high proportion (~97%) of uniquely mappable, reads, suggesting that such changes would likely be minimal.
coverage-based approaches
we use such a method
Depending on the statistical tests implemented by each tool, this may lead to biased and unreliable analyses.
This statement is too vague to be useful.
improvement in alternative splicing assessment when read depth is increased to 200–300 million reads per sample [19,43].
i can not find this in the references cited
recommend at least 70 million reads per sample
Alas, this recommendation must be dependent on the size of the "junction-ome", which they do not quote here.
Vast was studied by me during development of our method.
I re-read reference [19] and note their observation:
"The resulting PIR calls were robust with respect to sequencing depth, which was tested by randomly sampling between 1.25% and 80% of the reads in the original sample and recalculating PIR as described above (data not shown)."
This approach to testing is exactly what I suggested we could do to explore the effect of reducing read depth on our method. Note that they report "robust" without quantification. This is presumably because they do not know ground truth, so such comparisons are not really subject to quantification.
FPKM of constitutive and alternative splicing event were calculated by sum of the corresponding isoforms (constitutive or alternative splicing event may have multiple isoforms). Then, fisher’s exact test was applied to above FPKM to analyze differential alternative splicing between well-watered and salt-stress treatments
machine learning to derive a computational model that takes as input DNA sequences and applies general rules to predict splicing in human tissues
HTML, CSV, text, and Markdown
I'd love to see RSS added to this list. Want to get a feed updated daily with activity on page in the last day. Would use to focus shared reading / literature discussion groups. Otherwise, would love to see a plugin for microsoft teams (sort of like slack, except I use it where I work). Associate a H group with a Teams "channel". How can I help (test? spec?).
H has chance to be very disruptive in academic circles with additions such as this.
When viewing a page, press Shift+?
This is one of the first shortcuts you should learn.
(Unix-alikes.) system(cmd, input = ) now uses ‘shell-execution-environment’ redirection, which will be more natural if cmd is not a single command (but requires a POSIX-compliant shell). (Wish of PR#15508)
x