Cuffcompare or Cuffmerge for tissue specificity?

Question

[colin.kern]

I have RNA-seq libraries from 8 tissues (2 biological replicates for each). I want to determine which transcripts have tissue-specific expression. After I build cufflinks annotations for each of the individual tissues, would it be better to combine them into a single annotation using cuffmerge or cuffcompare? I'm concerned that if only one of the tissues has an assembled transcript at a specific loci, cuffmerge may consider it to not have enough evidence to include. But that's exactly the type of thing I'm interested in finding. Am I misunderstanding how cuffmerge works? It seems like cuffcompare might be better in this situation as it just does a straightforward union of all fragments.

Answer 1

[vivekbhr]

You can safely use cuffmerge for merging. It causes no loss of detected transcripts.

Also you will find this post useful..

[colin.kern]

Thanks for the answer. I tried both and I get 33886 XLOCs with cuffmerge and 51144 with cuffcmp. Also almost twice the number of TCONS. It says in the cuffmerge manual that it "automatically filters a number of transfrags that are probably artfifacts." Could this be what's causing the difference?

[vivekbhr]

Did you use the reference annotation (ensembl gtf) with cuffcompare?