A Risk Stratification Model for Lung Cancer Based on Gene Coexpression Network and Deep Learning

Figure 1

Gene coexpression network construction and survival-related modules identification. (a) A schematic diagram summarizing our risk stratification modeling strategy. Gene coexpression network was constructed from the training set. Gene network modules were extracted based on topological overlap. Survival-related modules were identified from the training set and validated in the two test sets. We selected representative genes from survival-related modules, and built network-based prognostic scoring system using deep learning. (b) Gene dendrogram and modules identified by weighted gene coexpression network analysis from the training set. Modules were labeled with different colors. (c) Univariate Cox regression analysis of module eigengene in the training set was performed. Module eigengene is a representative expression value of genes of each module calculated by the principal component analysis. The dotted line represents cutoff value ( value = 0.05) for significance, and five modules were identified as survival-related network modules. (d) Survival-related network modules were validated in the two test sets using Cox regression analysis. Three modules from test set 1 and two modules from test set 2 were significantly associated with overall survival.

(a)

(b)

(c)

(d)