Differential isoform usage with replicates using Jensen-Shannon divergence

[Golsheed]

Hello,

I'm trying to do a differential isoform usage test for a target gene g across two conditions (i=1,2). Suppose that gene g has K isoforms. For each condition i, I have a set of J samples, where each sample is a vector of size K. The kth element of each such vector indicates the (MISO-estimated) proportion of isoform k among all the isoforms. Basically, I want to test whether the relative abundancy of the K isoforms differ across the two conditions.

I know that some packages exist that do this kind of analysis. But I want to get an overall idea of how the statistical test in this context works. Specifically, if one decides to use the Jensen-Shannon metric to test for the differential isoform usage, how would the results explained in the manual for CuffDiff2 generalize to this setting? Note that I'm assuming I have J replicates (samples) within each condition. I understand how the test works for a single-replicate problem, but I don't know how to generalize it to multiple replicates.

Any comments would be greatly appreciated.

Thanks,
Golsheed