iChip

iChip identifies enriched genomic regions in Chromatin Immunoprecipitation microarray (ChIP-chip) experiments by modeling spatial dependencies with hidden ferromagnetic Ising models.

Key Features:

• Hidden Ferromagnetic Ising Model: Employs a hidden ferromagnetic Ising model to capture spatial dependencies in ChIP-chip data and identify enriched genomic regions.
• Bayesian Hierarchical Framework: Implements a Bayesian hierarchical model to enable robust statistical inference across genomic resolutions and sample sizes.
• Metropolis within Gibbs Sampling Algorithm: Uses a Metropolis-within-Gibbs sampling algorithm to simulate posterior distributions and estimate model parameters.
• Compatibility with Multiple Platforms: Processes microarray data from Affymetrix, Agilent, and NimbleGen platforms and supports single and multiple replicates.
• Performance and Efficiency: Shows comparable performance to TileMap HMM and BAC on high-resolution Affymetrix data, significantly outperforms them on low-resolution Agilent data, and is computationally more efficient than the BAC method which uses Markov Chain Monte Carlo (MCMC).

Scientific Applications:

• Transcription factor binding sites: Detects transcription factor binding sites from ChIP-chip enrichment signals.
• DNA methylation patterns: Investigates DNA methylation patterns measurable by microarray-based enrichment assays.
• Histone modifications: Profiles histone modifications using ChIP-chip enrichment data.

Methodology:

Analysis combines a hidden ferromagnetic Ising model within a Bayesian hierarchical framework with parameter inference via Metropolis-within-Gibbs sampling to simulate posterior distributions.