DE exons to DE transcripts

[waterV]

Hi everyone,
I aligned my reads using tophat and loaded the results in R to do the DE analysis using DESeq. I feed it an exon annotation and get DE exons. So far so good. I would now like to identify DE mRNAs from these DE exons. I can think of several ways of doing this but would like to know if anyone else has done this before and how it is "usually" done...
For example: identify the transcripts for which "most" (threshold) of the present exons (count > threshold2) have a FC > 2 (significant) and call them as DE...

Or maybe I am missing something...
Thanks a lot!

[NKAkers]

I tend to see people assemble transcripts first, then look for DE in transcripts. How about using cufflinks on your tophat output?

http://cufflinks.cbcb.umd.edu/

Here's a recent paper that shows some DE in isoforms:
http://onlinelibrary.wiley.com/doi/10.1002/em.21801/pdf

Hope that helps!

[waterV]

So people usually feed whole transcripts as features to count the reads? But then, aren't you throwing a lot of information since transcripts will likely overlap much more than exons and then you are discarding a lot of reads...?

[Thomas Doktor]

I would recommend trying DEXSeq for your analysis. It is based on DESeq and can relate alternative exons to isoforms.