Identifying transcription factor–DNA interactions using machine learning

Abstract Machine learning approaches have been applied to identify transcription factor (TF)–DNA interaction important for gene regulation and expression. However, due the enormous search space of genome, it is challenging build models capable surveying entire reference genomes, especially in species where were not trained. In this study, we surveyed a variety methods classification epigenomics data an attempt improve detection 12 members auxin response (ARF)-binding DNAs from maize soybean as assessed by DNA Affinity Purification sequencing (DAP-seq). We used prediction minimizing genome only unmethylated regions (UMRs). For identification DAP-seq-binding events within UMRs, achieved 78.72 % accuracy rate across ARFs on average encoding with count vectorization k-mer logistic regression classifier up-sampling feature selection. Importantly, selection helps uncover known potentially novel ARF-binding motifs. This demonstrates independent method TF-binding sites. Finally, tested model built DAP-seq directly found high false-negative rates, which accounted more than 40 ARF TFs tested. The findings study suggest potential use various predict TF–DNA interactions between varying degrees success.