RJMCMCNucleosomes

RJMCMCNucleosomes infers genome-wide nucleosome numbers and positions using reversible jump Markov chain Monte Carlo within a Bayesian Multinomial-Dirichlet t-mixture framework.

Key Features:

• Multinomial-Dirichlet prior: Employs a Multinomial-Dirichlet prior/classification to model multinomial read distributions.
• t-mixture model: Implements a t-mixture model to capture overdispersion in sequencing-derived signals.
• Reversible jump MCMC: Uses reversible jump Markov chain Monte Carlo to estimate both the number of nucleosome components and their genomic positions.
• Hierarchical mixture distributions: Incorporates hierarchical mixture distributions for robust modeling of component heterogeneity.
• Dynamic model selection: Allows the number of nucleosome components to vary during inference for flexible model selection.
• Benchmarking and validation: Validated on simulated datasets and MNase-Seq data from Saccharomyces cerevisiae and compared against PING and NOrMAL.
• Alternative to EM/BIC: Provides a Bayesian alternative to EM with BIC and heuristic strategies for estimating nucleosome numbers and positions.
• Relevance to sequencing technologies: Targets nucleosome mapping relevant to data produced by ChIP-chip, ChIP-Seq, and MNase-Seq technologies.

Scientific Applications:

• Genome-wide nucleosome profiling: Maps nucleosome positions across genomes to study chromatin organization and gene regulation.
• MNase-Seq analysis: Identifies nucleosome locations from MNase-Seq data, with demonstrations on Saccharomyces cerevisiae.
• Method comparison and benchmarking: Enables quantitative comparisons to PING and NOrMAL using simulated and experimental datasets.
• Robust estimation in complex landscapes: Improves estimation of nucleosome number and position where EM/BIC or heuristic methods may be insufficient.

Methodology:

Performs Multinomial-Dirichlet classification within a t-mixture and hierarchical mixture framework, using reversible jump MCMC to estimate the number of components and their genomic positions.