Deep learning-based prediction of TFBSs in plants.

Abstract

In plants, transcription factor binding sites (TFBSs) are usually determined by in vivo chromatin immunoprecipitation sequencing (ChIP-seq) or in vitro methods i.e., DNA affinity purification sequencing (DAP-seq). By contrast, in human research, computational approaches have already been deployed i.e., machine learning (ML) and deep learning (DL), to mine experimentally known TFBSs data. Recently, a deep convolutional neural network (CNN) method was first deployed to predict the TFBSs in arabidopsis (Arabidopsis thaliana) using available DAP-seq datasets. In vivo experiments, however, are labor- and cost-consuming, resulting in the lack of available experimentally based TFBSs data in most plants. Therefore, it is urgent to develop reliable computational approaches for TFBSs prediction in plants using the available ChIP-seq datasets.