Inferring the Properties of Transcription Factor Regulation

The regulatory targets of transcription factors are costly to directly detect using existing technologies. Many computational models have thus been developed to infer the genes targeted by TFs using gene expression profiles, position weight matrices modeling TF protein binding, histone modifications, and other secondary datasets. We develop a framework for scoring the potential targets of various TFs using models that take the profile of motif hits on the proximity of transcription start sites as input, and describe methods to validate this framework using expression datasets. These models are then extended to include cis-regulatory regions inferred from epigenetic data. Thesis Supervisor: Manolis Kellis Title: Professor of Computer Science