A deep learning approach for DNA-binding protein identification and functional annotation based on self-attention mechanism

Abstract

The invention discloses a deep learning method for DNA binding protein identification and functional annotation based on a self-attention mechanism. A data set is selected from a protein database, and the constructed deep learning model is trained and tested by using the selected data set, and then the trained deep learning model is used for training and testing. The deep learning model can predict whether protein can bind to DNA; wherein, the deep learning model includes coding layer, embedding layer, long short-term memory neural network layer (LSTM), convolutional neural network layer (CNN) and self-attention Layer (Self‑Attention). Through the deep learning model based on the self-attention mechanism, the present invention can predict whether it has the function of binding with DNA according to the primary sequence information of the protein, and find out the region with the binding function.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a deep learning method for DNA binding protein identification and function annotation based on a self-attention mechanism. Background technique [0002] DNA-binding proteins (DBPs) are proteins that contain functional domains that can bind to DNA, and are widely present in various organisms. It plays a vital role in biological regulation mechanisms, including DNA replication, DNA transcription, DNA repair, DNA recombination and other biological functions. Therefore, the effective determination of the functional region of DNA and protein binding is not only crucial for protein functional analysis, but also plays a crucial role in the development of new drugs. In a protein sequence that can bind to DNA, there is one or several DNA-binding domains, which are composed of several amino acid residues and can interact with DNA. This region is generally called DNA-binding domain. ...

AI Technical Summary

Problems solved by technology

[0005] At present, deep learning methods can achieve higher prediction accuracy than machine learning methods, but both types of methods cannot complete functional annotation and cannot obtain functional domains that combine DNA and proteins

Images