Mutated Tumor Suppressors Follow Oncogenes Profile by the Gene Hypermethylation of Partners in the Protein Interaction Networks

Latest therapeutic technologies have shown promising results in identifying and targeting oncogenes and tumor suppressors along with their related pathways for novel drug development. In this study, we use proteins network, mutation and methylation data of fusions and devised a novel network-based parameter called ‘preferential attachment score’ to categorize genes into oncogenes or tumor suppressors for 21 cancer types with 3091 fusions. We explored various features of these cancer genes to be considered for training and focused on 12 network and mutational parameters, based on an ABC-MCMC machine learning method. We considered the RNAseq data from TCGA for 21 cancers and the gene expression profiles for normal as well as cancer genes. The number of communities are more in leukemia, lymphomas, melanomas, glioblastoma than in sarcomas and carcinomas. In carcinoma, the cell cycle, DNA replication, spliceosome, proteasomes, mismatch repair, p53 signaling, nucleotide excision repair and ten other pathways were up-regulated. From TCGA, we found that tumor suppressors had highest mutation frequency in most tumor types, have higher degrees of connectivity, betweenness centrality, and lower clustering coefficients as well as shortest-path distances. Finally, most mutated tumor suppressors integrate hypermethylated partners in the protein interaction networks follow the patterns of oncogenes.