RNAseq Analysis of Samples from Multiple Ancestries

Question

[dhibar]

I am wondering if it is possible or if there is a common practice for differential gene expression comparisons when the samples come from different ethnic/ancestral backgrounds?

In GWAS it is common to test for association within a single ancestral group to avoid population stratification, but is this also required for RNAseq differential expression comparisons?

Has any of this been explicitly tested in the literature? My search so far has come up null.

Many thanks (from a first-time poster!)

Answer 1

[DG]

I certainly haven't seen any. I think it would depend a bit on what exactly you are studying. My guess would be that for the majority of cases either there is no mixed sources (rare disease studies typically occur within families) or that for the experiment it is expected that the actual biology would overwhelm any stratification issues (tumour expression, etc). Without having any data to back it up my assumption would be that gene expression data, for the most part, at a gross level, is far less susceptible to population stratification. There are some obvious exceptions to do with some phenotypic features such as skin pigmentation, but on the whole I think it isn't that much of a concern, especially when you are doing your comparisons with well defined clinical or cellular phenotypes.

I could be wrong, I haven't seen data on this one way or the other, but that would be my gut assumption based on biology in general. We expect lots of stratification on the pure genetic level, but most of those have minimal impacts, if any, on gene expression. With exceptions of course.