TileHMM

TileHMM applies Hidden Markov Models to analyze tiling array ChIP-chip data for genome-wide inference of transcriptional activity and protein–DNA interactions, including histone modification signals.

Key Features:

• Hidden Markov Model Framework: Employs HMMs to model the sequential nature of tiling array genomic data and capture underlying biological states.
• Robust Parameter Estimation: Implements robust maximum likelihood estimation techniques for accurate parameter inference.
• Baum-Welch and Viterbi Training: Uses the Baum-Welch algorithm and Viterbi training to derive and refine model parameters.
• t Emission Distributions: Incorporates t emission distributions to increase resilience against outliers in the data.
• Efficient Parameter Estimation Procedure: Combines two distinct approaches to parameter estimation into a single efficient procedure.
• Comparative Performance with TileMap: Demonstrates superior performance relative to TileMap in analyses focused on histone modifications.

Scientific Applications:

• ChIP-chip tiling array analysis: Analysis of genome-wide ChIP-chip tiling array data to detect regions of protein–DNA interaction.
• Chromatin structure and transcriptional regulation: Elucidation of chromatin features and transcriptional regulation at high genomic resolution.
• Histone modification studies: Detection and analysis of histone modification signals across the genome.

Methodology:

Uses an HMM tailored to ChIP-chip tiling array data with robust maximum likelihood estimation, Baum-Welch algorithm and Viterbi training for parameter fitting, incorporation of t emission distributions, and a combined two-approach parameter estimation procedure.