“…In general, these methods either (1) model the RNA-Seq counts directly using generalized linear mixed models (GLMMs) [ 8 – 10 ], or (2) transform the counts into continuous measures that can then be analyzed using linear mixed models (LMMs) assuming a normal distribution [ 11 , 12 ]. The former approach has the benefit of modeling the RNA-Seq data directly, but model convergence and type 1 error rate control can be problematic at the smaller samples sizes common in RNA-Seq studies, depending on the GLMM estimation approach [ 8 , 13 ]. The alternative of using transformed counts is appealing since LMMs have been extensively studied and are generally faster to fit.…”