A practical comparison of methods for detecting transcription factor binding sites in ChIP-seq experiments

Abstract

Background: Chromatin immunoprecipitation coupled with massively parallel sequencing (ChIPseq)<br />is increasingly being applied to study transcriptional regulation on a genome-wide scale. While<br />numerous algorithms have recently been proposed for analysing the large ChIP-seq datasets, their<br />relative merits and potential limitations remain unclear in practical applications.<br />Results: The present study compares the state-of-the-art algorithms for detecting transcription<br />factor binding sites in four diverse ChIP-seq datasets under a variety of practical research settings.<br />First, we demonstrate how the biological conclusions may change dramatically when the different<br />algorithms are applied. The reproducibility across biological replicates is then investigated as an<br />internal validation of the detections. Finally, the predicted binding sites with each method are<br />compared to high-scoring binding motifs as well as binding regions confirmed in independent qPCR<br />experiments.<br />Conclusions: In general, our results indicate that the optimal choice of the computational<br />approach depends heavily on the dataset under analysis. In addition to revealing valuable<br />information to the users of this technology about the characteristics of the binding site detection<br />approaches, the systematic evaluation framework provides also a useful reference to the<br />developers of improved algorithms for ChIP-seq data.