Extended Data Fig. 10: Deconvolution of previously published 450K DNA methylation array data.

(A) Deconvolution of pancreatic islet methylomes. Methylation arrays from 53 male and 34 female non-diabetic donors⁴⁸ were analysed atlas methylomes, revealing detailed cellular composition including previously uncharacterized alpha and delta cells. No statisticaly significant sex differences in cellular composition were observed. (B) Analysis of 865 pulmonary methylomes from TCGA⁴⁹. WGBS-based markers for lung alveolar epithelium and lung bronchial epithelium cells reveal differential cell populations in 443 LUAD, 11 SCLC, 337 LUSC, 32 normal adjacent (LUAD), and 42 normal adjacent (LUSC) lung methylomes. Note that only alveolar cell DNA is identified in lung adenocarcinomas, while small cell lung cancer and squamous cell carcinomas contain also bronchial DNA, consistent with the presumed cellular origins of each type of lung cancer. Note that epithelial cells are a minority in both lung adenocarcinoma and normal lungs. This is probably due to the abundance of stromal cells in bulk preparations of either normal lungs or lung cancers. (C) DNA methylation from 721 cancerous and 97 normal breast biopsies from TCGA. WGBS-based markers for breast luminal and basal epithelial cells were used to study the cellular composition in TCGA⁵⁰, which were classified into five subtypes using PAM50, a 50-gene expression-based classification⁶⁶. Different cell composition is observed for normal-like, basal-like, luminal A, luminal B, and Her2-enriched PAM50 subtypes, compared to healthy breast biopsies. The low fraction of breast basal cells in breast cancer is likely to result from the abundance of non-epithelial cells in both the normal breast and breast cancer. Box plots mark median and interquartile range (IQR), with 1.5*IQR whiskers.