A Detection Method for Identifying the Shared Targets of Multiple DNA-Binding Proteins from ChIP-chip Tiling Array Data

• Activity Number: 379
• Type: Topic Contributed
• Date/Time: Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Nancy N. Wang*+
• Companies: University of California, Berkeley
• Keywords: ChIP-chip ; tiling arrays ; multivariate HMM ; pseudo-ROC

Abstract

Gene regulation often involves the chromosomal binding of multiple proteins. Individually, each protein has a distinct set of binding targets. When assembled into a complex, these proteins share essentially the same targets. Biologists often perform ChIP-chip experiments to study DNA binding proteins. The sequential data contain measurements tiled along the chromosomes. Peaks generally occur at the binding sites. Existing methods can only analyze the ChIP-chip data of one protein at a time. We propose to analyze multiple proteins simultaneously using a multivariate hidden Markov model. To establish a benchmark for optimizing our method, we created sets of pseudo-positive and pseudo-negative common peaks, based on the individual outputs of the two best existing methods. Our method achieves pseudo-sensitivities of over 90%, when the pseudo-specificities are controlled above 80%.