Integrative Statistical Models for High-Throughput Genomic Data

• Activity Number: 532
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Caucus for Women in Statistics
• Author(s): Sunduz Keles*+
• Companies: University of Wisconsin-Madison
• Keywords: Next generation Sequencing ; ChIP-seq ; Matrix factorization ; Clustering

Abstract

Integrative statistical models for high throughput genomic data

Recent efforts from consortium projects such as ENCODE, mouse ENCODE and the RoadMap EpiGenomics have mapped the transcriptional and epigenomic state of multiple cell lines and primary tissues at unprecedented detail. However, tools that can statistically phenotype a genomic region based on its epigenomic state are significantly lacking, severely limiting the use of these valuable datasets by researchers outside of these consortia. We develop general statistical models that integrate multiple regulatory genomic datasets such as transcription factor occupancies, histone marks, and DNA methylation, that together capture the epigenomic states of cells to characterize genomic loci of varying lengths and predict regulatory potential of a given set of loci by integrating a diverse set of regulatory genomic data.