Systematic evaluation of cell-type deconvolution pipelines for sequencing-based bulk DNA methylomes.

Area of Science:

• Epigenetics and Genomics
• Computational Biology

Background:

• DNA methylation sequencing offers single-molecule resolution for methylome analysis.
• Cell-type heterogeneity analysis is a key application, but deconvolution methods require systematic evaluation.

Purpose of the Study:

• To benchmark six sequencing-based deconvolution methods against two array-based methods.
• To assess the performance of deconvolution methods in various synthetic bulk sample scenarios.
• To propose a best-practice strategy for DNA methylation deconvolution using sequencing data.

Main Methods:

• Benchmarking of six sequencing-based methods (Bayesian epiallele detection, DXM, PRISM, csmFinder+coMethy, ClubCpG, MethylPurify) and two array-based methods (MeDeCom, Houseman).
• Individual assessment of informative region selection and cell-type composition estimation steps.
• Evaluation using synthetic bulk DNA methylation samples.

Main Results:

• Performance of cell-type deconvolution is influenced by the number of cell types in the mixture.
• Array-based methods generally outperformed sequencing-based methods.
• Current sequencing-based methods struggle with identifying cell-type-specific signals and confounding methylation patterns.

Conclusions:

• A best-practice deconvolution strategy for sequencing data is proposed.
• Limitations in current sequencing-based deconvolution methods are identified.
• Future research should focus on improving the ability of sequencing-based methods to resolve cell-type-specific methylation signals.