PolyaPeak

PolyaPeak models ChIP-seq peak shapes with a Bayesian hierarchical framework to improve transcription factor binding site (TFBS) detection.

Key Features:

• Incorporation of peak shape information: Models the characteristic offset forward- and reverse-strand peak pair in ChIP-seq data and positions the TFBS between them.
• Multivariate Pólya mixture modeling: Represents peak shapes using a mixture of multivariate Pólya distributions to capture complex read-distribution patterns.
• Minorization-Maximization (MM) learning: Employs the MM algorithm to automatically learn peak-shape parameters from the data.
• Hierarchical integration of read counts: Integrates read count information with learned peak-shape models within a hierarchical Bayesian framework to distinguish signal from background.
• Comparative performance: Reported analyses on real datasets indicate improved TFBS detection relative to MACS, CisGenome, and PICS.

Scientific Applications:

• TFBS mapping: High-resolution identification of transcription factor binding sites from ChIP-seq experiments.
• Gene regulation studies: Improved detection of protein-DNA interactions to inform regulatory network and promoter/enhancer analyses.
• Epigenetics and chromatin research: Enhanced peak characterization for studies of chromatin state and epigenetic regulation.

Methodology:

Inputs aligned ChIP-seq reads on forward and reverse strands; models peak shapes with a mixture of multivariate Pólya distributions; learns shape parameters via the minorization-maximization (MM) algorithm; integrates read count information through a hierarchical Bayesian model to detect TFBS.