MRHMMs

MRHMMs models multivariate, regime-dependent relationships in genomic and genetic data using multivariate regression Hidden Markov Models to analyze differential gene regulation and combinatorial transcription factor patterns.

Key Features:

• Regime-aware multivariate modeling: Accommodates multiple distinct regimes where relationships among multivariate features vary across genomic regions or conditions.
• Flexible emission probability structures: Supports emission models including mixtures of multivariate normal distributions and logistic regression models.
• Extension of HMM framework: Extends traditional Hidden Markov Models with multivariate regression structures to capture dependencies among features.
• Computational efficiency (C implementation): Implemented in C to provide speed suitable for genome-wide datasets.
• Customizability: Allows implementation of alternative models tailored to specific research needs.

Scientific Applications:

• Genomics and genetics: Facilitates analysis of differential gene regulation influenced by gene functions and experimental conditions.
• Transcription factor combinatorics: Enables characterization of combinatorial patterns of transcription factors across conditions.
• General HMM-applicable studies: Applies to other biological contexts where hidden Markov models and multivariate feature regimes are relevant.

Methodology:

Uses multivariate regression Hidden Markov Models with emission probability structures including mixtures of multivariate normal distributions and logistic regression models; implemented in C.