A protein structure design method and device based on deep learning

Abstract

The present invention relates to a protein structure design method and device based on deep learning, the method comprising: determining the gene sequence or molecular crystal structure information of the targeted protein according to biomarkers; inputting the gene sequence or molecular crystal structure information of the targeted protein into the geometric graph neural network model; use the trained geometric graph neural network model to generate an amino acid sequence; construct a protein skeleton model based on the generated amino acid sequence and homologous proteins; optimize the protein skeleton model according to proteomics and molecular dynamics. The present invention binds protein data to corresponding DNA sequences and mRNA sequences, on the one hand, improves the interpretability and effectiveness of amino acid sequences, and on the other hand reduces the time and effort of screening, repeated adsorption, elution, and amplification for protein design or verification. In the process, the calculation amount of the model is reduced through the geometric graph neural network.

Description

technical field [0001] The present invention relates to the fields of biological information and deep learning, in particular to a method and device for designing protein structures based on deep learning. Background technique [0002] As an important part of living organisms, proteins participate in most biological functions of living organisms. In particular, important physiological activities in the human body are completed by protein, which is an important material basis for life activities. As a class of macromolecules with the widest distribution and the most complex functions in organisms, the study of proteins has always been an important part of biology. [0003] The molecular structure of a protein includes four levels, in which the arrangement order of the amino acid sequence is called the primary structure of the protein; the polypeptide chain of the protein is regulated by hydrogen bonds between oxygen groups, regular and periodic The stable structure of the p...

AI Technical Summary

Problems solved by technology

[0009] In order to solve the problems of high repetition, low validity and poor interpretability of the amino acid sequence generated in the existing model-based protein design, the present invention provides a protein structure design method based on deep learning in the first aspect of the present invention, including the following Steps: Determine the gene sequence or molecular crystal structure information of the targeted protein according to the biomarker; input the gene sequence or molecular crystal structure information of the targeted protein into the geometric graph neural network model; use the trained geometric graph neural network model to generate Amino acid sequence; construct a protein skeleton model according to the generated amino acid sequence and its homologous protein; optimize the protein skeleton model according to proteomics and molecular dynamics

Images