Differential co-expression analysis of protein interaction network for human hepatocellular carcinoma (HCC) progression. Four arrows in this figure indicate major analysis flow. Arrow 1: for each protein interaction pair (edge), an expression correlation value was calculated for each of five HCC stages, and a differential correlation value (dC) was calculated for each of four stage transitions. Edges with the highest dC values were used as seeds in a search of a differential co-expression subnetwork. Four subnetworks were retrieved from the PIN for the four HCC stage transitions, respectively. Arrow 2: the four transition-wise subnetworks were combined into a union set, in which each edge was associated with four dC values. Arrow 3: the edges in the union subnetwork were clustered based on similarity in those four-dC data vectors. Arrow 4: six clusters of differential co-expression protein-interaction modules were determined, each characterized with a distinct, coherent co-expression change pattern over the whole HCC process.