Abstract

We discuss methods to identify DNA regulatory elements by exploiting the correlation between sequence data and gene expression. We start by reviewing the contribution of M. G. Tadesse et al. (Identification ofDNAregulatory motifs using Bayesian variable selection. Bioinformatics, 20 (2005), 2553–2561) in the use of Bayesian methods for variable selection for the identification of binding sites for regulatory factors. We then propose an extension of their model to include gene regulators. Although the modeling frameworks for variable selection has been extensively studied in the literature, their application in genomic studies for the identification of regulatory elements represents a novel contribution. We report performances of the methodologies on the well-studied regulatory systems of Saccharomyces cerevisiae under heat shock.

Introduction

The study of gene regulation plays an important role in understanding gene expression. A biological understanding of this complicated process motivates the techniques for analyzing and modeling gene expression and gene regulation. Fundamental principles of genetics are transcription (the process of encoding information in DNA as mRNA) and translation (the process of making proteins from mRNA). Microarrays measure the abundance of mRNA and, thus, describe gene expression at the transcription level. Understanding how transcription is regulated in the cell can provide insights into developing rich models for analyzing microarrays.

The mechanisms that control gene transcriptions consist of many different classes of proteins and classes of DNA sequences [10]. The proteins involved, known as trans-acting factors or transcription factors, interact with control points of DNA sequences known as cis-acting regulatory sequences. In eukaryotic systems, such as human cells and yeast, RNA polymerase II is solely responsible for transcribing DNA to mRNA. This polymerase requires multiple sets of cis-acting regulatory sequences.