Random effect with LMM in EWAS

Question

[obidobi]

Dear all,

Let’s consider including an epigenetic kinship matrix into Linear Mixed Models (LMM) which could account for environmentally induced population structure in epigenome-wide association studies (EWAS). If a shared environment effect leads to genome-wide changes in DNA methylation, then controlling for epigenetic kinship via DMRs would remove the fixed effect since you basically remove the association between a specific DMR and the environment, right? So, using epigenetic kinship matrix as a fixed effect in EWAS is not a good idea in general, or how to avoid removing the fixed effect in case of possible DMR & environment interactions?

Thanks!