2013 Poster Sessions : Genomics Approach to Identifying Transcription Factor Functions and Complexes

Student Name : Harendra Guturu
Advisor : Gill Bejerano
Research Areas: Artificial Intelligence
Abstract:
ChIP-seq technology allows identification of all binding sites of a chosen transcription factor (TF) in a particular cellular context, but practical challenges of ChIP-seq – namely antibody, cell and condition availability – greatly limit de facto exploration. Here, we develop a complementary computational approach based on large scale TF function discovery directly from the genome. We make highly accurate binding site predictions in human and mouse representing all major DNA binding domains. Using the binding sites as our foundation, we integrate protein structural information to statistically learns TF complexes at a low false discovery rate of 11%. Next, we developed a novel rigorous framework, termed PRISM, to combine these predictions of binding sites for TFs and their complexes with millions of facts about the function of all human and mouse genes and obtain a set of function predictions in a large variety of contexts, at a false discovery rate of 15%. We show our TF function predictions are highly enriched for experimentally validated roles, making PRISM an appealing biological discovery tool that complements ChIP-seq and several other experimental methods that characterize TF binding specificity.

Bio:
I am a 5th year EE PhD student in the Bejerano lab. I am currently interested in studying how transcription factors control human gene regulation by binding to cis-regulatory elements. Using the identified sites of interest as a foundation, I would like to understand the mechanisms of regulation and the functional roles in which the factors are involved.