Fig. 3
Micro-scale pan-cancer subgroups demonstrate distinct subgroup-specific patterns. a Patient distribution of 12 cancer types (COAD and READ are treated as one type) in pan-cancer subgroups with k = 9. Subtypes of three cancers were used. For example, BRCA-Her2 stands for the Her2+ breast cancer subtype, while BRCA stands for the breast cancer samples with no subtype information. For a patient set of each cancer type or subtype in the subgroups, the significance of enrichment was evaluated using a Chi-squared test (see Methods). P-values lower than 1 × 10−10 were set as 1 × 10−10 for convenient visualization. The number denotes the size of the corresponding patient cohort of each cancer type or subtype in this subgroup. b The number of overlaps between each pair of the 9 sets of significantly differentially influenced genes in each subgroup. c The number of overlaps of the biological functional terms derived from the corresponding significantly differentially influenced genes sets of each subgroup using DAVID. Genes and functional terms were selected with false discovery rate (FDR) q-value smaller than 0.05. d Selected GO terms (biological processes) from the functional analysis using the 9 sets of significantly differentially influenced genes for each subgroup. Bars represent the significance with -log10(FDR) (green) and the number of enriched genes (red) of the corresponding GO term