Ensemble learning method for predicting DNA protein binding site

Abstract

The invention relates to an ensemble learning method for predicting a DNA protein binding site. The method comprises the following steps of obtaining protein sequence data of the DNA protein binding site; preprocessing the protein sequence data of the DNA protein binding site; establishing input data in a one-hot coding mode; combining extracted features, establishing features of amino acid on each protein sequence, and taking the features as the input data; carrying out oversampling on positive sample data through utilization of an SMOTE algorithm; dividing negative sample data into a plurality of parts according to the positive sample quantity, and combining each part of negative samples with the positive samples to form a new data subset, thereby obtaining N data subsets; training eachdata subset through utilization of convolutional neural networks; and integrating results of N convolutional neural networks through utilization of a majority voting method, thereby obtaining a predicting result. According to the method, the DNA protein binding site predicting problem under the imbalance data condition is solved, and the predicting accuracy is improved.

Description

Technical field [0001] The invention relates to the field of bioinformatics, in particular to an integrated learning method for predicting DNA protein binding sites. Background technique [0002] At present, DNA protein binding site detection mainly uses chromatin immunoprecipitation (Chromatin Immunoprecipitation, ChIP) and protein binding microarray (protein binding microarray, PBM). The ChIP-Seq technology, which combines ChIP with second-generation sequencing technology, can efficiently detect DNA segments that interact with histones and transcription factors in the whole genome. The principle of ChIP-Seq is: firstly, the DNA fragments bound by the target protein are specifically enriched by chromatin immunoprecipitation (ChIP), and then purified and library construction is performed; then the enriched DNA fragments are subjected to high pass Quantitative sequencing. Researchers accurately locate the millions of sequence tags obtained on the genome to obtain information abo...

AI Technical Summary

Problems solved by technology

[0005] However, most of these methods do not take into account the unbalanced nature of the DNA protein binding site data itself, that is, the data of the binding site is much less than the data of the non-binding site, so when building a prediction model, there is a relatively large gap. deviation

Images