Deep learning Residue2vec-based protein structure prediction method

Abstract

The invention discloses a deep learning Residue2vec-based protein structure prediction method. The method comprises the following steps of: giving input sequence information, regarding a known protein structure on a PDB website as a corpus to train, partitioning the proteins with known structures into residues with lengths of n, obtaining the expression of each residue in a vector space through a CBOW model and a Huffman code, and judging the similarities between the residues through calculating the distances between residue vectors, so as to obtain the front N fragment structures on each residue position of a query sequence and then form a fragment library of Residue2vec; carrying out random folding on the query sequence to form an initial conformation; randomly selecting a residue with the length of n, and carrying out dihedral angle replacement on the residue and fragments in the fragment library; and comparing the energy, if the energy is decreased, receiving the conformation, and if the energy is increased, receiving the conformation via a Metropolis criterion and finally obtaining a metastable-state conformation through continuous iteration. According to the method disclosed by the invention, the matching degree and prediction precision in the query sequence are relatively high.

Description

technical field [0001] The present invention relates to the fields of bioinformatics and computer applications, in particular to a protein structure prediction method based on deep learning Residue2vec. Background technique [0002] Protein molecules play a vital role in the process of biological and cellular chemical reactions. Their structural models and bioactive states have important implications for our understanding and cure of many diseases. Only when proteins are folded into a specific three-dimensional structure can they produce their unique biological functions. Therefore, to understand the function of a protein, it is necessary to obtain its three-dimensional structure. [0003] The ab initio prediction method of protein structure needs to solve two basic problems: (1) construct an appropriate energy model to calculate the interaction between different atoms in the protein; (2) develop an effective algorithm to find the global minimum of conformational space ene...

AI Technical Summary

Problems solved by technology

But so far there is no perfect method to predict the three-dimensional structure of proteins. Even if good prediction results are obtained, it is only for some proteins. At present, the main technical bottlenecks lie in two aspects. First, On the one hand, it lies in the sampling method. The existing technology is not strong in sampling the conformation space;

Images