DeepTFBS: A Hybrid Model Using Deep Learning Methods for Transcription Factor Binding Sites Prediction

Abstract

The formation, transmission and regulation of genetic data at the molecular level are complex combinatorial processes that are difficult to understand. Transcription factors, which form the basis of these processes, play a critical role in determining the properties and functions of cells by copying genetic information from DNA to RNA. Transcription factors, which control complex structures such as the nervous system, play a vital role in determining conditions such as disease and health by regulating gene expression. The binding sites of proteins on DNA determine the critical points of gene expression and contribute to the adaptation of cells to various conditions. Various methods have been developed in the literature for the prediction of transcription factor binding sites, which is an important step for the diagnosis and treatment of genetic diseases. Several studies have been developed with successful results obtained by using DNA sequence and shape features together. In this study, a hybrid method is proposed by combining different deep learning technologies to identify transcription factor interactions based on DNA sequences and shapes. 165 validated CHIP-Seq datasets were used in the study.