Predicting enhancer regions and transcription factor binding sites in D. melanogaster

Abstract

Identifying regions in the genome that have regulatory function is important to the fundamental biological problem of understanding the mechanisms through which a regulatory sequence drives specific spatial and temporal patterns of gene expression in early development. The modENCODE project aims to comprehensively identify functional elements in the C. elegans and D. melanogaster genomes. The genome- wide binding locations of all known transcription factors as well as of other DNA- binding proteins are currently being mapped within the context of this project [8]. The large quantity of new data that is becoming available through the modENCODE project and other experimental efforts offers the potential for gaining insight into the mechanisms of gene regulation. Developing improved approaches to identify functional regions and understand their architecture based on available experimental data represents a critical part of the modENCODE effort. Towards this goal, I use a machine learning approach to study the predictive power of experimental and sequence-based combinations of features for predicting enhancers and transcription factor binding sites.