iSeq

iSeq identifies immunoprecipitation-enriched genomic regions from ChIP-seq and ChIP-chip data using Bayesian hidden Ising models to model spatial correlation and high-order interactions.

Key Features:

• Bayesian Hidden Ising Models: Employs a Bayesian hierarchical hidden Ising model to infer IP-enriched regions from chromatin immunoprecipitation data.
• Spatial Correlation Handling: Explicitly models the spatial correlation of probe intensities inherent in ChIP-chip microarray data due to hybridization to neighboring probes.
• High-Order Interactions: Incorporates high-order interactions within the Ising framework to capture complex dependencies and intrinsic spatial structures in genomic signals.
• Multi-platform Compatibility: Applicable to ChIP-chip data from Affymetrix tiling arrays and Agilent promoter arrays and accommodates differing genomic resolutions.
• Parameter Estimation via Gibbs Sampling: Uses the Gibbs sampler for model parameter estimation.
• Dataset Types Supported: Accommodates controlled and uncontrolled datasets, with or without replicates.
• Comparative Performance: Demonstrates comparable performance to a Bayesian hierarchical model, hierarchical gamma mixture model, and TileMap hidden Markov model on Affymetrix tiling arrays and superior sensitivity and false discovery rates on Agilent promoter arrays.

Scientific Applications:

• Protein-DNA Interactions: Identifies genomic regions of protein binding to support studies of transcription factor binding and regulatory element mapping.
• Histone Modifications: Detects regions enriched for histone modification marks to inform chromatin state and gene regulation analyses.
• DNA Methylation Patterns: Analyzes methylation-associated enrichment signals to study epigenetic regulation and genomic stability.

Methodology:

Implements a Bayesian hierarchical hidden Ising model with high-order interactions to model spatial correlation in ChIP-chip data, with parameter estimation performed via the Gibbs sampler.