question about exon usage across conditions?

[caswater]

Hi,

I want to quantify exon inclusion/skipping rates across a number of different conditions. So for each condition, I obtained FPKM values for each exon, and also FPKMs for each gene locus encompassing all the eons. Does it make sense if I use the ratio of EXON-FPKM/Gene-FPKM in each condition to quantify the exon usage in the condition?

Thank you

[Brian Bushnell]

Yes, it seems logical to me.

[dpryan]

So one notable problem with this approach is if you get a bit of a different mapping distribution between the groups (i.e., more/less 5' bias) then the results will be off. If you wanted to look at straight differential exon usage, you could just use DEXseq. If you wanted to specifically look at differential exon skipping, a good method would be (in this case, looking at skipping of exon #2):

1. Count the number of reads mapping from exon 1 to exon 2
2. Count the number of reads mapping from exon 1 to exon 3
3. Fit these counts with a beta-binomial model (this is the same as is increasingly used to look at bisulfite sequencing and is the most relevant model for the type of data you have).

There are a couple packages to do that with R, though I can't guarantee that they're pretty limited in terms of model complexity. If you need something that can handle more than 2 groups, just let me know, since I'm writing a package to do that (for methylation data, but the statistics are the same).