I have RNAseq assemblies for the same organism from multiple different experiments. Let's say I have experiments A, B, and C each with conditions 1 and 2.

Under the assumption that exon usage doesn't vary between experiments, can I use cuffmerge on all the assemblies to get a better overall model of transcripts before only comparing a subset of assemblies with cuffdiff? Or will this throw off the statistics in the subsequence cuffdiff steps?

So can I run cuffmerge on A1,A2, B1, B2, C1, C2, followed by cuffdiff on only C1 & C2?

Thank you in advance for your input.