A Statistical Framework for the Analysis of ChIP-Seq Data

• Activity Number: 157
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Pei Fen Kuan*+
• Companies: The University of North Carolina at Chapel Hill
• Keywords: ChIP-Seq ; Mixture Model ; Zero-inflated Negative Binomial Regression

Abstract

Chromatin immunoprecipitation followed by direct sequencing (ChIP-Seq) has revolutionized the experiments in profiling DNA-protein interactions and chromatin remodeling patterns. However, it is not free of sequencing and other sources of biases. Despite this, most of the existing tools do not consider such biases. We carefully study sources of bias in the underlying data generating process of ChIP-Seq technology by utilizing sequenced naked DNA and develop a model that captures the background signal in the ChIP-Seq data. We then proposed mixture models called MOSAiCS for analyzing ChIP-Seq data. Our modeling framework incorporates the variability in both the mappability and GC-content of regions on the genome and sequencing depths of the samples. We show that our model fits very well on real data and provides a fast model-based approach for ChIP-Seq data analysis.