Hi All,

I am attempting to apply the Jensen-Shannon divergence strategy for detecting differential isoform usage, which is implemented by cuffdiff2 (detailed in the supplement of Trapnell et al.), to estimates calculated by eXpress. Before pressing ahead I wanted to make sure that I am not making an egregious error; that the output of eXpress is compatible with the JS procedure.

Specifically, do `eff_counts`

and `eff_length`

(generated by eXpress) correspond directly to the conditional fragment abundances and adjusted transcript lengths (referred to as gamma and l(t) in Trapnell et al.), respectively? And is it appropriate to back calculate Var[gamma] directly from the FPKM CIs given by eXpress?

Hi sah9b!

I am currently attempting to do the same. In my case I would like to try to apply the JSD on the transcripts reconstructed with Ballgown (and then try to apply the same logic to the exons, but I am not sure if this is really correct). Did you find the solution? How did it work? Would you recommend it?

Thanks in advance!