Figure 1

Overview of DEVC-net on extracting discriminatively interpretable features of a gene network by combining gene expression, and expression variance/covariance. a The framework of conventional differential expression analysis (DEA). Only differential expression is considered in the conventional DEA, which can be estimated in a multiple-sample manner (e.g., P-value from statistic test) or in a single-sample manner (e.g., fold-change). b The framework of conventional differential expression network (DEN). In the conventional DEN, the information of differential expression variance has not been considered. c The framework of the proposed DEVC-net. Compared to the conventional network-based approaches, DEVC-net has two advantages: one is to use differential expression variance and the other is to design the measurements of differential expression variance/covariance in a single-sample case. Obviously, DEVC-net can be easily applied in a multiple-sample case. Note that, the gene is labeled in red if it has differential expression between case or control, and in green if has differential expression variance; The gene is labeled in black if there is no significant difference between case and control; The gene pair is labeled in red if the two genes have differential expression covariance, otherwise black; Besides, PPI means protein–protein interaction, and PCC means Pearson correlation coefficient.