LuxUS

LuxUS models differential DNA methylation using a generalized linear mixed model that incorporates spatial correlation among neighboring cytosines to detect differentially methylated cytosines and regions from bisulfite sequencing data.

Key Features:

• Spatial Correlation Structure: Incorporates cytosine random effect correlations to model spatial dependence between adjacent cytosines.
• Generalized Linear Mixed Model (GLMM): Uses a GLMM that accommodates binary and continuous covariates and includes replicate and cytosine random effects.
• Probabilistic Programming: Fits model parameters using Stan for Bayesian inference and efficient computation.
• Statistical Testing: Performs hypothesis testing using Savage-Dickey Bayes factor estimates for covariates of interest.
• Accounting for Experimental Variation: Accounts for experimental variation including bisulfite conversion efficiency within the likelihood model.

Scientific Applications:

• Epigenetic research: Analyzing DNA methylation patterns and their role in gene regulation.
• DMR and DMC detection: Identifying differentially methylated cytosines and differentially methylated regions (DMRs) from bisulfite sequencing data.
• Disease association studies: Detecting methylation changes associated with disease mechanisms and phenotypes.
• Complex experimental designs: Analyzing methylation data with binary and continuous covariates and replicate structures.

Methodology:

Implements a likelihood model tailored for bisulfite sequencing data, incorporates spatial correlation via cytosine random effects within a GLMM, fits parameters using Stan, and uses Savage-Dickey Bayes factor estimates for statistical testing.