Bias in Genome Scale Functional Analysis of Transcription Factors using\r\nBinding Site Data

Jair Zhou, Haixia Rachel Li

Abstract

Genome scale functional analysis of a transcription factor is performed by mapping its genomic binding sites to genes using a nearness criterion followed by enrichment analysis for the pre-defined functional categories such as GO categories. It can result in biased assessment of functionality of transcription factors. In the view of the enormous work underwent using this procedure and its simplicity and effectiveness, it is important to understand the bias in this procedure and the factors influencing it. We show that the bias originates from widely varying gene lengths, intergenic regions and gene density. We also demonstrate that the bias depends on multiple factors such as the nearness criterion and the choice of the associated parameters and the distribution of the binding sites of the transcription factor. Furthermore, we propose a resampling based procedure called reFABS for unbiased functional analysis of binding sites.