Pipeline for analysing RNA-seq exon level data with lot of samples?

Question

[sarin.heikki.v]

Hi,

What sort of pipeline would you recommend on doing RNA-seq exon-level analysis for about 100 samples?

Our goal is study differential exon expression and splicing in our data. We have case and control groups with three time points and our focus is to study how does the exon expression and spicing differ between groups across all timepjoint when accounting for the within sample variability. For gene-level analysis we have used DESeq2 and LRT-testing.

I've looked into DEXseq but with a lot of sample is seems to be really computationally too demanding. Other option that we've considered have been diffSplice in limma and diffSplice in edgeR but not quite sure if those are the way to go.

Would really appreciate comments on what would be the best approach to the exon-level data. Is it possible to do LRT-testing with exon-level data with some of these methods?

Answer 1

[Satyajeet Khare]

I have used diffSplice in edgeR with glmQLFit as suggested in the manual. You can plot fold change in exon expression. Not tried DEXSeq yet, but I am thinking of moving back to Cuffdiff with SpliceR. Cuffdiff may not be ideal for all vs all comparison.

[sarin.heikki.v]

I think in my case the glmQLFit can't be applied since I'm using glmLRT test and not glmQLFTest. Correct me if I'm wrong.