Mixture Regression Modeling of Next-Generation Sequencing Data

• Activity Number: 132
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Naim Rashid*+
• Companies: The University of North Carolina at Chapel Hill
• Keywords: Next Generation Sequencing ; Mixture Regression Modeling ; Model-based Clustering ; Chip-Seq

Abstract

Next Generation Sequencing platforms offer greater dynamic range, resolution, and genomic coverage in measuring local enrichment of signal compared to microarrays. In this talk we propose a three component mixture regression model consisting of enriched, background, or zero-inflated components to classify local signal enrichment. Because it has been shown that factors such as GC content and mappability can artificially influence enrichment, we include these in our model as covariates to adjust the component means for their effects. We show that our method performs better than existing ones, in addition to being applicable to both high and low signal to noise ration datasets and performing well with and without input control. We demonstrate the model's performance through simulation and both ChIP-Seq and FAIRE-seq datasets. A software implementation of the method is discussed.