Computational prediction of transcription factor binding sites based on an integrative approach incorporating genomic and epigenomic features

초록

Transcription factor binding sites (TFBSs) are often predicted by sequence-based methods that use a position weight matrix or consensus sequences. However, the degeneracy of TFBSs makes the prediction of them very challenging. Recent completion of the encyclopedia of DNA elements project provides many useful additional information enabling researchers to tackle these difficulties from a noble point of view. In this paper we developed an integrative TFBS prediction method incorporating genomic as well as epigenomic features, such as DNA methylation, histone modification and chromatin accessibility. We found that (i) an integration of various features facilitates more accurate TFBS prediction, (ii) the proximity range of -500 to 500 nt relative to the transcription start site of a gene resulted in slightly more accurate prediction than the other ranges, and (iii) the proximity of an epigenomic feature contributes more than the other properties of epigenomic or genomic features to the accurate prediction of TFBSs. This study demonstrates that epigenomic features play a critical role in an integrative approach for predicting TFBSs.

키워드