Heterogeneity of APA regulation among cancer types.

(A) Hierarchical clustering using gene set-level APA estimates on 598 single-cells derived from 3 cancer types (280 tumor cells from 11 BC patients, 288 tumor cells from 3 GBM patients, and 30 tumor cells from 1 RCC patients). BC; Breast cancer, GBM; Glioblastoma, RCC; Renal cell carcinoma, T1,T2: Multiple regions in the brain, RLPS; Relapse. A total of 417 gene sets were used as variable components for hierarchical clustering and dimensional reduction (see Methods). (B) Unsupervised tSNE on the gene set-level APAs separating tumor cells along with cancer types. The dot colors represent patients which are the same as in Fig 4A. (C) APA map of top10 gene sets significantly (p<0.01) switched in each cancer type. The colors for cancer types and patients are the same as in Fig 4A.