Recent breakthroughs in bioinformatics tools for quantification (e.g. Cufflinks/Kallisto/Salmon etc.) and tools which can identify differential transcript usage (DTU) (e.g. DRIMSeq, Cufflinks etc.) mean that from RNA-seq data we can now relatively easy obtain a genome wide analysis of transcripts that are differentially used betwen conditions.
What can you use these results for? What systematic analysis does this transcript level data enable?
I will take the liberty of giving one possible answers to my own question – but I’m very interested in other answers.
One analysis type that such data enables is the analysis of transcript switches with predicted potential consequences.
I myself have recently developed such a tool called IsoformSwitchAnalyzeR. IsoformSwitchAnalyzeR enables statistical identification (via DRIMSeq) of isoform switches with predicted functional consequences. The consequences analyzed can be chosen from a long list which includes gain/loss of protein domains, signal peptides changes in NMD sensitivity etc. The R package also enables easy visualization of isoform switches along with their consequences and it directly supports the output of Cufflinks/Cuffdiff, RSEM, Salmon and Kallisto.
Apart from enabling identification of interesting examples switch identification also enables systematic analysis of what genes are affected. For inspiration I will recommend one of my own articles (describing results obtained with IsoformSwitchAnalyzeR) as well Hector et al’s recent paper (which is not using IsoformSwitchAnalyzeR). Of particular interest and finesse is Hector’s analysis of how isoform switches can disrupt protein-protein interactions.
Looking forward to hear more ideas.
I am not sure if this has been done, is common, or are research lines, but here is what I think can be done with transcripts' differences (aside from comparing the change of the expression of each transcript):
Having the different transcripts, enables us to study also the relation between transcripts, which are co-expressed together or differently co-expressed. Maybe at the individual level the switch between transcripts of a gene is not related to a condition, but together with other transcripts it is.
Transcripts usage could be also linked to structural or genomic changes. Either short/small changes or big. At genomic level SNP, inversions or methylation..., or structural changes such as compression of a region, different distances between chromosomes/regions...
Additional insights can frequently be made in research by studying things at a lower level. By peeling back the higher layers, effects can be seen that are hidden (or reversed) when looking at aggregate data. This is generally referred to as Simpson's Paradox.
We know that RNA is functional at an isoform/transcript level, and in some cases the expression of different transcripts from the same gene can have different biologically-relevant outcomes. By looking at transcript-level information, even if the transcripts for a gene average out to be a similar level between different treatment conditions, variation at an individual transcript level can still be discovered. This is important when looking at differential expression of populations of cells, because it can clue researchers into finding out that a particular cell sub-population might be of interest; single cells typically will only produce single isoforms for each gene at a specific point in time.
I understand that this post is rather old but there has been some development in the field and this 2019 paper might be a good resource for interested individuals.
