Figure 1. Schematic overview of DRW-GM.

The gene expression profiles are translated into pathway profiles (rows represent pathways and columns represent samples) by DRW-based pathway activity inference. The global gene–metabolite graph is constructed on KEGG metabolite pathways. The input W₀ of DRW is the initial weights of genes and metabolites, which are obtained from gene expression profiles and metabolomic profiles respectively. The output W_∞ is the probability weight vector of nodes when DRW reaches a steady state. W_∞ measures the topological importance of genes by incorporating both gene and metabolite information. The pathway activity is a expression value vector synthetized by topologically weighted differential gene expression vectors in the pathway (Formula (2)).