Figure 1
Schematic view of the methodology. The inputs include (a) a DNA methylation profile measured at genomic loci, which we use to compute (b) methylation value at the gene level, and (c) a gene expression profile. (d) We construct an integrative network, in which nodes represent genes and edges model the association between individual gene pairs based on both expression and methylation data (Eq. (1) in “Methods”). (e) For each module, we compute eigengenes as weighted averages of the expression and DNA methylation level of all genes in that module (Eqs. (2)–(4) in “Methods”). (f) We employ the eigengenes as robust biological signatures (i.e., biomarkers) for survival analysis. (g) While not implemented in this study, the eigengenes could also be utilized for other downstream data mining analyses.
