LuxHMM

LuxHMM applies a hidden Markov model to bisulfite sequencing data to segment the genome and identify differentially methylated regions for DNA methylation and epigenetic analysis.

Key Features:

• Hidden Markov Model (HMM) segmentation: models spatial correlation among neighboring CpG sites to segment genomic sequences into distinct methylation regions.
• Bayesian regression with multiple covariates: estimates region-level methylation effects while adjusting for biological and technical covariates.
• Bisulfite sequencing experimental parameters: integrates biochemical details specific to bisulfite conversion into the statistical model.
• Inference methods: supports variational inference for fast genome-scale analyses and Hamiltonian Monte Carlo (HMC) for more rigorous posterior sampling.
• Differentially methylated region (DMR) detection: identifies DMRs at the regional level rather than focusing solely on individual cytosines.
• Benchmarking on real and simulated data: evaluated using real and simulated bisulfite sequencing data with competitive performance relative to other methods.

Scientific Applications:

• DMR identification: detection of differentially methylated regions for studying epigenetic regulation across the genome.
• Covariate-adjusted methylation analysis: modeling methylation in complex experimental designs to separate biological signals from technical variation.
• Epigenetic studies of disease and biological processes: region-level methylation analysis to investigate roles of epigenetic modifications in health and disease.
• Method evaluation and simulation studies: use with real and simulated bisulfite sequencing data for performance assessment and benchmarking.

Methodology:

Uses Hidden Markov Model (HMM) for genome segmentation; Bayesian regression with multiple covariates for region-level methylation inference; incorporates bisulfite sequencing experimental parameters; and supports variational inference and Hamiltonian Monte Carlo (HMC) for inference.