Hello, I'm trying to perform eQTL analysis using an interaction term. (maybe with MatrixEQTL) My purpose is to get eQTLs having different effects according to the gender. In my case, it's impossible to perform analyses separately for both gender due to the sample size.
In general eQTL analysis, It is said that covariates like below should be adjusted during analysis. 1) Known covariates from metadata 2) Hidden covariates from PCA or PEER results for gene expression data 3) Genetic covariates from PCA for genotype data
But I'm not sure if it's right to adjust covariates in the same way. I think hidden covariates will include effects of gender. So my worry is on removal of effects from gender during eQTL analysis. Can I put the 'gender' variable into 'known covariates' to be adjusted? Can I adjust top PCAs or PEERs even though the 'gender' variable explains the most part of variance in gene expression data?
I've heard that Combat and SVA function from SVA R package can correct for known & hidden covariates, preserving an effect from a variable of interest (='gender' for my case). Then, is it a good idea to use gene expression data already corrected with combat/sva as the input of eQTL analysis? (using only genotype PCs as input covariates in eQTL analysis)
I'm so confused about how to adjust gene expression data. Please give me advice. Thank you in advance.
