MeDeCom

MeDeCom decomposes complex DNA methylomes to uncover and quantify latent methylation components and their proportions in heterogeneous, multi-cellular samples.

Key Features:

• Reference-free framework: Operates without predefined reference data to enable decomposition of heterogeneous methylation profiles.
• Constrained Non-Negative Matrix Factorization (NMF): Uses constrained NMF to factorize methylation matrices into non-negative latent components and sample-specific proportions.
• Biologically motivated regularization: Incorporates a regularization function within the NMF framework to bias solutions toward biologically plausible methylation patterns.
• Unsupervised exploratory analysis: Performs reference-free, unsupervised decomposition that does not require labeled training data.
• Scalability for large datasets: Designed to handle large-scale, high-throughput epigenomic DNA methylation datasets.

Scientific Applications:

• Cell-type deconvolution: Identify and quantify cell-type-specific latent methylation components and their proportions from complex methylomes.
• Confounding correction for cell composition: Isolate cell composition as a major source of methylation variation to improve interpretation of epigenomic studies.
• Biomarker and target discovery: Recover biologically relevant methylation components to support identification of methylation-based biomarkers or therapeutic targets.

Methodology:

Applies constrained non-negative matrix factorization (NMF) with a biologically motivated regularization in a reference-free, unsupervised decomposition to derive latent methylation components and sample-specific proportions from DNA methylomes.