Copy number-aware deconvolution of tumor-normal DNA methylation profiles
SUMMARYAberrant DNA methylation is a hallmark of cancer development. Bisulfite sequencing of tumor samples can be used to study changes in the cancer methylome, but data interpretation is confounded by admixed normal cells and copy number changes. Here, we introduce CAMDAC, a novel method for Copy number-Aware Methylation Deconvolution Analysis of Cancers, enabling tumor purity, allele-specific copy number and deconvolved tumor methylation rate profiling from bulk tumor samples. We apply CAMDAC to 122 multi-region samples from 38 non-small cell lung cancers profiled by reduced representation bisulfite sequencing as part of TRACERx study. CAMDAC-derived purity and copy number profiles confirm those derived from matched whole-genome and whole-exome sequencing. Purified tumor methylation rates greatly enhance the accuracy of tumor-normal and tumor-tumor differential methylation calling and, in contrast to bulk signals, capture phylogenetic relationships between tumor clones. CAMDAC directly links genetic mutations and epigenetic changes, yields insights into allele-specific methylation and reveals (epi)genetic heterogeneity in solid tumors.