Transcription Factor Binding Sites Prediction Based on Histone Acetylations using Neural Networks

Histone acetylation is one of the most important epigenetic processes that regulate gene expression. In other words, chromatin exposes DNA to transcription factors and gene regulators by histone tail acetylation in nucleosomes. There are some studies to show the relation between gene regulation and histone acetylation. In this paper, our main goal is to propose a computational method for transcription factor binding site prediction based on a pattern of 18 types of histone acetylations. In this regard, we analyze 18 types of histone acetylations near SP1 binding sites on Chromosome 1 in human CD4+T cells. The results show that 12 out of 18 marks are strongly correlated with transcription factor binding sites. Then, we implement a multilayer perceptron neural network with supervised training. This network is trained using binding sites of various transcription factors of SP1 in chromosome 1 and 18 types of histone acetylations near them. Finally, this network is applied for predicting binding sites of various transcription factors on chromosomes 1 and 2.