MALAX

MALAX implements a generalized linear mixed model with a Laplace approximation to perform epigenome-wide association analysis of bisulfite sequencing methylation data.

Key Features:

• Handling Low and Uneven Sequencing Depth: Handles low and uneven coverage inherent to bisulfite sequencing to enable robust methylation analysis at single nucleotide resolution.
• Confounding Factor Management: Incorporates multiple variance components to model complex confounders, extending beyond the single-variance-component approach used by MACAU.
• Computational Efficiency: Uses a Laplace approximation instead of Markov Chain Monte Carlo (MCMC) to directly approximate model likelihoods, potentially reducing processing time by over 50% compared to state-of-the-art methods.
• Model Likelihood Approximation: Employs the Laplace approximation for direct and efficient estimation of model likelihoods to support accurate association testing in EWAS.

Scientific Applications:

• Epigenome-wide association studies (EWAS): Detects methylation–trait associations from bisulfite sequencing data to investigate gene regulation and disease-related molecular mechanisms.

Methodology:

MALAX fits a generalized linear mixed model with multiple variance components to bisulfite sequencing methylation data and uses a Laplace approximation to approximate model likelihoods, avoiding MCMC.