histoneHMM

histoneHMM performs differential analysis of ChIP-seq data to detect broad histone modification domains (e.g., H3K27me3, H3K9me3) between experimental and reference samples.

Key Features:

• Bivariate Hidden Markov Model (HMM): Employs a bivariate HMM to model and classify broad histone modification domains across two samples.
• Aggregation and bivariate counts: Aggregates ChIP-seq short-read data over larger genomic regions and transforms them into bivariate read counts for downstream analysis.
• Unsupervised classification: Performs unsupervised classification of genomic regions without requiring manual tuning parameters.
• Probabilistic outputs: Produces probabilistic classifications of regions as modified in both samples, unmodified in both, or differentially modified.
• Performance and validation: Validated on broad repressive marks H3K27me3 and H3K9me3 and corroborated by qPCR and RNA-seq, showing improved detection of functionally relevant differentially modified regions versus existing methods.
• R and C++ implementation: Implemented in C++ and compiled as an R package compatible with Bioconductor for integration into R-based workflows.

Scientific Applications:

• Detection of differential histone modifications: Identifies genomic regions with differential enrichment of broad histone marks between conditions or treatments.
• Epigenetic regulation studies: Supports investigation of epigenetic regulation and its impact on gene expression by mapping changes in broad histone domains.
• Development and disease research: Facilitates analysis of chromatin state alterations relevant to development and disease contexts involving H3K27me3 and H3K9me3.

Methodology:

Aggregates ChIP-seq short-read data into larger genomic regions to produce bivariate read counts and applies an unsupervised bivariate HMM to classify regions as modified, unmodified, or differentially modified.