MethylResolver-a method for deconvoluting bulk DNA methylation profiles into known and unknown cell contents

Area of Science:

• Epigenetics
• Computational Biology
• Cancer Research

Background:

• Bulk tissue DNA methylation profiling is crucial for identifying disease biomarkers.
• Cellular heterogeneity in tissues complicates the interpretation of methylation data.
• In silico deconvolution offers an efficient alternative to experimental fraction measurement.

Purpose of the Study:

• To develop and validate MethylResolver, a novel computational method for deconvoluting immune cell fractions from DNA methylation profiles.
• To assess MethylResolver's accuracy and performance compared to existing methods, especially in complex tumor mixtures.

Main Methods:

• Developed MethylResolver, a method based on Least Trimmed Squares regression.
• Applied MethylResolver to infer leukocyte subset fractions from methylation profiles of tumor samples.
• Benchmarked MethylResolver against other deconvolution approaches using simulated and real-world data.

Main Results:

• MethylResolver demonstrates superior accuracy in deconvolution, particularly with increasing unknown cellular content.
• The method successfully resolves tumor purity-scaled immune cell-type fractions without requiring cancer-specific signatures.
• Pan-cancer analysis of TCGA data revealed eosinophil fraction as a predictor of improved cervical carcinoma survival and B cell fraction as a predictor of poor survival in papillary renal cell carcinoma.

Conclusions:

• MethylResolver provides a robust and accurate tool for in silico deconvolution of immune cell fractions from DNA methylation data.
• The method enhances the utility of bulk tissue methylation profiling for cancer research and biomarker discovery.
• Identified novel immune cell-type associations with patient survival in specific cancer types, highlighting potential therapeutic targets.