Best practices of comparing RNA-seq data from two species with few or no replicates

Question

[Saad Khan]

Hi,

What would comprise as best practices for comparing RNA-seq data from two species where you don't have many replicates. Can software like cuffdiff/edgeR be used when doing cross species differential expression analysis ( only taking the orthologus genes). Any suggestions would be appreciated.

Thanks

[bw.]

Mapping genes across species and doing differential expression are 2 separate problems. cuffdiff, edgeR only solves the 2nd one. What species are you dealing with?

[Saad Khan]

I have mapped the data. Now I just want to compare the expression. I want to know what are the better ways of comparing the data. I am dealing with Dog and Human.

[bw.]

I see, misunderstood your question..

Answer 1

[chefer]

Do you have different conditions?

If so, one way to do this for orthologs is to compare the ratio/differential expression status of a gene across species. For example, DogGene1 and HumanGene1 are orthologs. DogGene1 is differentially expressed between Condition1 and Condition2, but HumanGene1 show similar expression (not differentially expressed) between Condition1 and Condition2.

[Devon Ryan]

The downside to that method, of course, is that it leads the unwary to compare p-values, which is a bad idea.

[Saad Khan]

We can also use log fold change?

[Devon Ryan]

Depends on how accurate those are. The fewer samples you have the less meaningful fold-change estimates become.

[chefer]

You are correct, you need to have confidence in determining differential expression status. OP should make sure that he is able to determine differential expression within a species first (with confidence). It would be suspect to compare p-values across species. With this method you will end with a set of orthologs that show species co-expression (are DE in the same condition in both species), or show species specific differential expression (only DE in one of the species).