I would like to do a per gene ancestry covariate. I believe this could be important because ancestry for portions of chromosome can be different in the same individual so there is no way to correct across all genes with a simple covariate.
Is there any way to do this with DESeq2? I am already using CQN to do gene length and GC content normalizations.
There is not a way to provide different covariates for certain genes in DESeq2. The only way I can think to hack it would be to break the analysis into groups of genes with the same covariate. You could use the same size factors (or normalization factors) for all genes, but just estimate dispersions and coefficients separately for all groups of genes that share the same covariate.