Fig. 3: Characteristics of multi-omics periphery and core (MOPC).

a Functional profiles. The excess overlap between the peripheral and core region was calculated with ten biological functional datasets. The items with index excess overlap >1 (excess white line) mark the functional profiles of the peripheral and core regions. The error bar corresponds to the fluctuation of the results across 15 cancers. The results of peripheral regions (excess overlap >1, green bar) are slightly weaker than those of the core regions (grey bar), showing that their functionality should not be ignored. b The distribution of the number of cancer MOPCs enriched in KEGG pathways. c Enriched pathways shared or specific among cancer MOPCs. The ten representative pathways are focused on and their enrichment results are displayed (−log₁₀(p-value), hypergeometric test). d–f Regulatory relationship analysis in eQTL, based on cis-eQTL and trans-eQTL in the PancanQTL database (see ‘Methods’). d We counted the number of core genes directly regulated by peripheral genes. The z-score represents the statistical significance compared with that of 1000 random experiments. e We counted the number of core genes regulated by eQTL. The excess overlap values between the cores and the ground truth egene (see ‘Methods’) are given. The CNV (O4) core has the most excess overlap with egene (mean, 1.5). f We counted the number of peripheral genes, which regulate core genes, and showed the proportion of peripheral genes that are responsible. The periphery of CNV (O4) accounts for the largest proportion (average, 0.52). g The omnigenic neighbourhood portrays cancer similarity. We used simAB to calculate the relationship between cancers (‘Methods’), where the grey image represents the result of the cancer similarity analysis based on core genes, and the green image represents the result of the cancer similarity analysis based on the omnigenic neighbourhood. The points represent the similarities between cancers. We verified the results by comparison with DO similarity, symptom similarity, and comorbidity RR. The fitted line is the Pearson correlation coefficient between the predicted and known similarity between cancers. The shading indicates the 95% confidence interval. In the three similarity verification experiments, based on the omnigenic neighbourhood, the Pearson correlation coefficient increased by 6.12-fold, 1.26-fold, and 2.31-fold, respectively. h Common mechanism of rectum adenocarcinoma (READ) and colon adenocarcinoma (COAD) in the periphery. The Venn diagram shows the number of overlapping peripheral genes between COAD and READ (2101, Jaccard coefficient = 0.54). The heat map is also used to show the enrichment results (−log₁₀(p-value), hypergeometric test) on the shared peripheral genes in ten classic cancer signalling pathways⁴² that frequently undergo gene variations. The error bars indicate the 95% confidence intervals.