Gene isoform expression and somatic mutation profiles provide new molecular taxonomy beyond tissue-histology classifications.

(A) TCGA multi-omics data employed in the study include matched RNA-seq for isoform expression (upper) and whole-exome-wide DNA-seq data for gene somatic mutation (lower) across 33 cancer types. (B) Cluster residence heatmap shows the number of samples from a given cancer type that resides within each of the 29 (C0 – C28) annotated Isoform clusters. (C) The UMAP visualization of 10,464 TCGA cancer samples based on expression of 73,599 isoforms. Each dot represents a TCGA sample and is colored/marked by the 29 Isoform clusters. (D-E) Samples on the same Isoform UMAP map are colored by TCGA cancer types (D) and tumor status (E). (F) Cluster residence heatmap shows the percent of each Isoform cluster that overlaps with each Mutation cluster (left) and TCGA iCluster (right). (G) Variation of information analysis of clustering schemes derived from various TCGA data types. The cluster membership of Aneuploidy, mRNA, miRNA, RPPA, DNA methylation (DNAmeth) and iCluster were derived from (Hoadley et al., 2018). The membership of Isoform was determined in this study. See also S1 Table and S1 Fig.