DNA binding protein recognition method based on deep sparse representation network

Abstract

A DNA binding protein recognition method based on a deep sparse representation network comprises the following specific steps: acquiring a DNA binding protein sequence data set, and dividing the DNA binding protein sequence data set into a training set and a test set; calculating a specificity scoring matrix of all sequences in the DNA binding protein sequence data set by adopting PSL-BLAST software; respectively filling or cutting all the specificity scoring matrixes into new specificity scoring matrixes with the same size; and constructing and training a DNA binding protein recognition classifier model by adopting a deep sparse representation network, and inputting the new specificity scoring matrix into the DNA binding protein recognition classifier model to complete the recognition of the DNA binding protein sequence. The constructed and trained DNA binding protein recognition classifier model is an end-to-end network, the prediction precision can be obviously improved, errors are small, potential features of a specific scoring matrix can be learned robustly by adopting a convolutional auto-encoder, classification is performed through a sparse representation layer, and the generalization ability of the model is improved.

Description

technical field [0001] The invention relates to a DNA-binding protein identification method based on a deep sparse representation network. Background technique [0002] DNA-binding proteins are specialized proteins capable of binding and interacting with DNA. DNA-binding proteins are involved in many biological processes, such as: identification of specific nucleotides, regulation of transcription and regulation of gene expression. At the same time, DNA-binding proteins are important components of anticancer drugs, antibiotics and steroids, and play an important role in the research of anticancer drugs and the treatment of genetic diseases. Early DNA-binding protein identification methods are generally biological experiments, such as filter binding, genetic analysis, chromatin immunoprecipitation and X-ray crystallography. Biological experimental methods are time-consuming and labor-intensive, and cannot meet the needs of large-scale protein sequence detection. [0003] C...

AI Technical Summary

Problems solved by technology

[0004] The training goals of each step of this method are inconsistent, and there is a deviation from the macro goal, so that the trained model is difficult to achieve the optimal result; and each step has errors, and the error of the previous step will affect the training of the next step. As a result, the accumulation of errors eventually leads to poor final results

Images