Do Multi-Reads Matter in ChIP-Seq Data Analysis?

• Activity Number: 148
• Type: Invited
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Sunduz Keles*+
• Companies: University of Wisconsin at Madison
• Keywords: High throughput sequencing ; ChIP-Seq ; Mixture models ; Mappability ; Transcription factors

Abstract

The introduction of next generation sequencing enabled a myriad of creative ways to answer genome-wide questions. In particular, chromatin immunoprecipitation coupled with sequencing (ChIP-Seq) has become a powerful technique for large scale profiling of transcription factor binding and chromatin modifications. A ChIP-Seq experiment generates millions of short reads. The first step of data analysis is to map reads to reference genome and retain reads that map to a single location in the genome (uni-reads). Restraining the analysis to unireads leads to reduced sequencing depth and further poses a significant challenge for identifying binding locations that reside in regions of genome that have been duplicated over evolutionary time.

We investigate the effects of discarding multi-reads in ChIP-Seq data analysis and illustrate that their incorporation can increase sequencing depth up to 20%. We develop a model-based method for taking into account mapping uncertainty in ChIP-Seq data analysis. We quantify and characterize gains from multi-reads in case studies from the ENCODE project, and support our conclusions with both computational and experimental validations.