Fig. 4

Systems-level properties of cancer genes. a Percentage of genes with ≥ 1 gene duplicate covering ≥ 60% of the protein sequence. b Proportion of genes originating in pre-metazoan species. c, d Number of human tissues in which genes (c) and proteins (d) are expressed. In panel c, tissue types were matched between GTEx and Protein Atlas wherever possible, giving 43 unique tissues. In tissues represented in both datasets, genes were defined as expressed if they had ≥ 1 TPM in both datasets. Only genes present in both sources were compared (Additional file 2: Table S1). e Percentage of genes essential in ≥ 1 cell line and distribution of cell lines in which each gene is essential. Only genes with concordant annotation between OGEE and PICKLES were compared (Additional file 2: Table S1). f Percentage of proteins involved in ≥ 1 protein complex. g Median values of betweenness (centrality), clustering coefficient (clustering), and degree (connectivity) of human proteins in the protein-protein interaction network. h Median values of betweenness and degree of the target genes in the miRNA-target interaction network. The clustering coefficient is zero for all nodes, because interactions occur between miRNAs and target genes. Known, candidate, and all cancer genes were compared to the rest of human genes, while TSGs were compared to OGs. Significance was calculated using a two-sided Fisher test (a, b, e, f) or Wilcoxon test (c, d, g, h). *p < 0.05, **p < 0.01, ***p < 0.001. Enrichment and depletion of cancer genes in representative functional categories taken from level 1 of Reactome (i) and level 2 of KEGG (j). Significance was calculated comparing each group of cancer genes to the rest of human genes using a two-sided Fisher test. False discovery rates were calculated in each gene set separately. Only pathways showing enrichment or depletion are shown. The full list of pathways is provided in Additional file 2: Table S3