PRMeth

PRMeth estimates cell-type proportions and infers methylation profiles from bulk DNA methylation data when reference profiles are only partially available, enabling deconvolution of tumor and blood tissue mixtures.

Key Features:

• Iteratively optimized non-negative matrix factorization (NMF) framework: PRMeth uses an iterative NMF procedure to jointly estimate cell-type proportions and methylation profiles for unknown cell types in tissue mixtures, applicable to blood and solid tumors.
• Handling partial reference data: Operates when methylation profiles are available for only a subset of known cell types, enabling composition inference without a complete reference panel.
• Benchmark performance: Across three benchmark datasets, PRMeth outperformed five alternative methods in accurately recovering cell-type proportions and reconstructing methylation profiles of unknown components.
• Biological consistency and significance: Applications to four TCGA tumor types produced estimated immune cell proportions largely consistent with prior studies, supporting biological plausibility of the inferred compositions.
• Facilitating immunotherapy research: By reducing dependence on complete methylation reference data, PRMeth enables broader investigation of immune infiltration and tumor microenvironment composition relevant to cancer immunotherapy.

Scientific Applications:

• Tumor heterogeneity analysis: Deconvolution of mixed tumor tissues using DNA methylation data to characterize cellular composition and heterogeneity.
• Immunotherapy response prediction: Estimation of immune cell fractions within tumors to support association studies and response stratification analyses.

Methodology:

Inputs are DNA methylation profiles for a subset of known cell types observed in tissue mixtures; an iteratively optimized NMF procedure jointly estimates the proportions of all constituent cell types and methylation profiles for unknown cell types present in the mixture.