Measuring Co-Binding Among Transcription Factors

• Activity Number: 649
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Yingying Wei*+ and Hongkai Ji
• Companies: JHSPH and Johns Hopkins University
• Keywords: Genomics ; Mixture model ; Survival Copula ; EM algorithm ; ChIP-seq

Abstract

How transcription factors (TFs) cooperate together is a key scientific problem in genomics. However, because certain regions of the genome are more accessible due to chromatin structure, traditional tests of independence, such as Pearson's chi-squared test, are always likely to reject the null hypothesis that the binding of two TFs along the genome is independent. To answer the important questions that how strong the association, in terms of co-binding, between two TFs is and how the associations vary among pairs of TFs, in this work we propose a unified approach to measure the co-binding association among TFs from ChIP-seq experiments. We generalize the idea of irreproducible discovery rate (IDR) to incorporate the missing data problem raised by the non-shared peaks from different types of TFs. Like IDR, our method creates curves, instead of the usual scalar measures of associations, to quantitatively characterize how the co-binding between two TFs changes from the most significant peaks to the least ones. Our curves are fitted by a survival copula mixture model using EM algorithm. We illustrate the effectiveness of our approach using simulations and real data examples.