Fig 1.
An overview of the analyses performed.
a) A workflow describing the data processing steps from protein structures in the PDB and cancer-related somatic mutations in COSMIC and ICGC to residue-level bi-partite protein interaction networks. b) The percentage of residues within surface, intermediate and core regions that harbor mutations for oncogenes (n = 56) and tumor suppressors (n = 47) with 3D structures. c) Focusing only on surface residues, the percentage of residues within interface and non-interface regions that harbor mutations for oncogenes and tumor suppressors with 3D structures.
Fig 2.
Characterizing the structural location of missense mutations in tumor suppressors (TS), oncogenes (OG) and other genes.
Fisher’s exact tests were performed separately for each set of genes. Shown are the odds ratios and 95% confidence intervals within each set of genes when making comparing the number of mutations located at a) surface versus core residues, b) surface interface versus surface non-interface residues.
Fig 3.
A bipartite protein-residue interaction network constructed for cancer genes and their interaction partners.
Edges involved in self-interaction are not shown. a) An example of a network describing how proteins, interface residues and mutations are represented in the bipartite network model. In a protein-protein interaction network, the nodes representing proteins A–D are directly connected to one another. In our bi-partite protein residue interaction network, interface residues are displayed between proteins. For example, residue 1 on protein A (A_1) is involved in the protein-protein interface between A and B along with A and C. Residues mutated in cancer are shown in the bi-partite protein mutated residue interaction network. For example, A_1 is mutated in at least one cancer in one patient whereas residue B_1 –present above but absent here—is not. b) A bipartite network showing cancer genes and their immediate interactors. c) A bipartite network displaying only residues that were mutated in one or more tumors. The circled interface residues interact with multiple proteins.
Fig 4.
Bipartite B2M interaction networks.
a) A bipartite network of the tumor suppressor B2M, its partners and the interface residues by which they interact. b) A bipartite network showing only the subset of residues that were observed to harbor missense mutations in cancer patients. The size here of the residue nodes represent the number of tumors in which the residue was mutated.
Fig 5.
Mutations affecting TP53 interaction interfaces and their impact on patient survival.
a) A bipartite network of the tumor suppressor TP53, its partners and the interface residues by which they interact. b) A bipartite network showing only the subset of residues observed to harbor missense mutations in cancer patients. The size here of the residue nodes represent the number of tumors in which the residue was mutated. c) A Kaplan Meier survival plot of patients from TCGA harboring mutations in TP53 at residues 175, 248 or 273. d) The TP53 mutations R175, R273 and R248 displayed on the crystal structure of TP53 as a homotetramer.
Fig 6.
Bipartite protein-residue interaction networks of 34 cancer genes known to form homo-oligomers displaying only those residues involved in oligomerization.
Edges are colored according to functional predictions made using VEST. Red lines indicate that mutations affecting that residue were predicted to be functional (VEST > 0.75), blue lines indicate a neutral prediction (VEST < 0.25), and dashed grey lines indicate mutations could not confidently be assigned a functional or neutral label.