MCMSeq

MCMSeq models RNA-Seq count data using a Bayesian hierarchical negative binomial generalized linear mixed model to analyze complex experimental designs with clustered and repeated measures.

Key Features:

• Bayesian hierarchical modeling: Implements a negative binomial generalized linear mixed model framework for RNA-Seq count data.
• Repeated-measures support: Handles an arbitrary number of repeated observations and correlated/clustered study designs.
• Covariate integration: Incorporates various covariates directly into the hierarchical model.
• Error-rate control via posterior inference: Maintains nominal false positive and false discovery rates during posterior inference.
• Statistical power demonstrated: Simulation studies report high sensitivity (recall) while controlling error rates, including at smaller sample sizes.
• R package implementation: Distributed as the mcmseq R package for computational analysis.

Scientific Applications:

• Longitudinal gene expression analysis: Quantifies transcriptional changes over time within individuals in longitudinal studies.
• Paired and correlated designs: Analyzes paired, repeated, and other correlated RNA-Seq experimental designs.
• Biomarker discovery in infectious disease: Applied to longitudinal RNA-Seq cohort data to identify genes associated with tuberculosis infection.

Methodology:

Models RNA-Seq counts directly with a Bayesian hierarchical negative binomial generalized linear mixed model, performs posterior inference, and incorporates covariates and arbitrary repeated measures.