Fig. 2: Subtypes stratification and survival analysis on different numbers of clusters.

a According to the cophenetic parameters of NMF, TCGA-LUAD samples were divided into four subtypes. The line x (solid line) was drawn according to the nmf function. The line y (dotted line) was set according to the nmfEstimateRank function. These results were plotted to choose the optimal rank value. And, the x-axis value is the possible factorization rank (2–6), and the y-axis represents the cophenetic value. b Consensus map of NMF clustering results in TCGA-LUAD cohort. c, d and e The OS of patients with the 2, 3 and 4 subtypes were determined by Kaplan–Meier survival analysis, respectively. The P-values were calculated by the Log-rank test.