To explore the reproducibility of module membership for individual genes, we directly compared this quantity between networks for conserved modules (Fig. 3 and Supplementary Fig. 3 online). Module membership was markedly consistent across networks, indicating that a gene’s network position is highly reproducible. Notably, despite the use of different microarrays and samples from different individuals, module membership was highly correlated between CTX and CTX_95 (M9: rho = 0.70, P = 9.0 × 10−271; M15: rho = 0.81, P < 1.0 × 10−300; M16: rho = 0.63, P = 3.4 × 10−203; Fig. 3). Similar results were obtained for comparisons of CTX with CTX_ILMN (M9: rho = 0.65, P = 1.6 × 10−191; M15: rho = 0.73, P = 5.0 × 10−264; M16: rho = 0.61, P = 3.0 × 10−162; Supplementary Network Analysis). These results validate the reproducibility of module membership values for individual genes and indicate that conserved gene coexpression modules reflect consistent underlying sources of variation in microarray data generated from human brain.
