Transcription factor binding site prediction method fusing with DNA shape features

Abstract

The invention relates to a transcription factor binding site prediction method fusing DNA shape features, belongs to the field of bioinformatics, and provides a new model for predicting transcription factor binding sites by using CNN in combination with DNA sequence and shape feature information by combining knowledge of structural biology, genomics and a deep learning neural network. Meanwhile, a special data set containing DNA shape features and DNA sequence information is constructed, and corresponding DNA shape information is added on the basis of a traditional transcription factor prediction data set. Therefore, the prediction accuracy of the DNA transcription factor binding site is improved.

Description

Technical field [0001] The present invention is in the field of bioinformatics, structural biology and combined knowledge of genomics design and implement a new method wherein a shape of the fusion DNA transcription factor binding site prediction. Background technique [0002] Transcription factor (TF) expression of many genes may be coordinated by binding to regulate transcription of the genomic region. With these primary cellular mechanism to regulate critical regulator of cellular processes and adapt to environmental stimuli. In fact, TF sequence or the number of changes may be hereditary diseases, complex diseases, autoimmune deficiencies and the main reason for cancer. How TF regulatory sequence specific DNA (referred to as TF binding sites, or simply of TFBS, such as promoters, enhancers) binding to coordinately regulated gene transcription and protein synthesis is a very important process, which in many biological processes from a key role. Over the past decade, have gener...

AI Technical Summary

Problems solved by technology

Although an improvement over similar models that only consider motif information, this approach was trained and validated using only normalized DNase-seq data from 15 cell types and information on 5 specific core histone modifications, which may will cause the trained model to predict well only the TF binding preference in that specific cellular context

In summary, there are few attempts to combine the three-dimensional structural features of DNA into the predictive modeling of TFBS

Images