“…Classical statistical tests are based on an experimental design having microarray/RNA-seq samples as subjects, where each subject has the same set of (GE) measurements [ 6 , 10 , 24 ]. In the usual supervised setting, the sampling model consists of M independent realizations (for M subjects) of ( X 1 , y 1 ), ( X 2 , y 2 ), …, ( X s , y s ), …, ( X M , y M ), where, X s represents the N -dimensional vector ( N : total number of genes) of the GE levels for s-th subject and y s is the corresponding class label (e.g., case: +1 vs. control: −1), s = 1, 2, …, M .…”