ChIP-GSM

ChIP-GSM infers transcription factor (TF) modules and predicts active regulatory elements from ChIP-seq data using a Bayesian Gibbs sampler and logistic regression.

Key Features:

• Bayesian Gibbs sampling: Employs a Bayesian framework with a Gibbs Sampler to iteratively sample ChIP-seq read counts and estimate binding potentials and module abundances across genomic regions.
• Integration of multiple TFs: Integrates ChIP-seq profiles from multiple transcription factors, including master factors and mediators, to model complex TF interactions.
• Probabilistic binding estimation: Infers module-region probabilistic bindings to detect weak binding events and characterize TF-module interactions at cis-regulatory regions.
• Predictive modeling: Applies logistic regression on inferred probabilistic bindings to predict active regulatory elements, validated across multiple independent datasets.

Scientific Applications:

• Regulatory element prediction: Improves identification of functional regulatory regions by leveraging inferred TF modules and their binding dynamics.
• Gene expression studies: Provides insights into how active TF modules influence gene expression over time, as demonstrated in analyses of K562 cells.
• Functional cellular processes: Facilitates exploration of cellular processes mediated by TF modules and their roles in biological functions and disease states.

Methodology:

Collect and integrate ChIP-seq data from multiple TFs; use a Gibbs Sampler within a Bayesian framework to iteratively sample read counts and estimate module-region probabilistic bindings and module abundances across genomic regions; apply logistic regression on inferred probabilistic bindings to predict active regulatory elements.