MONTE: Methylation-based Observation Normalization and Tumorpurity Estimation

Bulk DNA methylation profiling is widely used to study cancer epigenomics in clinical settings. However, these measurements aggregate signals from malignant and non-malignant cells, introducing composition-dependent confounding that complicates tumor-intrinsic interpretation and cross-cohort analyses. While existing methods can estimate tumor purity and, in some cases, correct methylation measurements, they are often constrained by cancer-specific reference models, predefined probe sets, or limited adaptability to new datasets and purity definitions. We present MONTE (Methylation-based Observation Normalization and Tumor purity Estimation), a cancer label-free method for tumor purity inference and CpG-resolved methylation purification from bulk DNA methylation data. MONTE learns probe-wise relationships between observed methylation and tumor purity using a stabilized linear model with empirical Bayes variance moderation and infers purity in new samples via signal-to-noise weighted aggregation of probe-level effects, without requiring matched normal samples or predefined probe sets. MONTE achieves robust purity estimation across cancer types using a single pan-cancer model and can be efficiently recalibrated to specific contexts through Bayesian transfer learning. MONTE provides a flexible, scalable, and interpretable framework for both methylation-based tumor purity estimation and correction, yielding tumor-intrinsic profiles for downstream analysis.