Classification for cancer phenotypes and different cancer types. (A) Heat maps of five differential edges in P-SSN for a tumor sample and all normal samples in LIHC, UCEC, LUAD and BRCA. Colors of sample labels corresponded to various clusters, i.e. red label corresponded to the tumor sample, and black labels corresponded to the normal samples. The accuracy of classification for a tumor sample and total normal samples in LIHC, UCEC, LUAD and BRCA is all 100% (Supplementary File 1). (B) The clustering dendrogram of four types of cancer (the green bar, the red bar, the blue bar and the purple bar correspond to the UCEC, the BRCA, the LUAD and the LIHC, respectively) was generated by hierarchical clustering based on network distance among P-SSNs. For visualization, we randomly selected 30 samples from four types of cancer for classification. The accuracy of the classification of four types of cancer is 100% (Supplementary Figure S2).