cross species RNAseq

[sidderb]

Hi All,

I have paired-end RNAseq data from the same tissue in two species (human and rat), which I want to compare. It seems to me this is fraught with issues:

1. The annotation quality for the two species is different and so the reads determined to align to orthologous features will likely be different. It is often the UTRs which are not well annotated (it seems), so perhaps excluding reads that bind to them is a way forward?
   
2. You are wholly reliant on the quality of the orthologue mapping.
   
3. Orthologous *genes* are mapped, not transcripts, meaning that a representative value for the gene needs to be derived from (possibly) multiple expressed variants. One could simply sum read counts from all the exons (with htseq-count) for this?
   
4. Count based tests such as DEseq (after mapping orthologous gene pairs) seems inappropriate given the differences in gene/transcript length between species etc. However, if I calculate RPKM in each species what would an appropriate test for differential expression be?
   
5. I suspect there are others, but this seems enough for now!

I wonder if anyone else has wrestled with these and has any useful advice?

Cheers.