Method and device for model training, protein feature extraction and function prediction

Abstract

The embodiment of the invention provides a model training, protein feature extraction and function prediction method and device.The model training method comprises the steps that based on a three-dimensional structure of initial protein, an initial topological graph G is constructed, the initial topological graph G adopts amino acid residues of the initial protein as nodes, and the initial topological graph G adopts the amino acid residues of the initial protein as the nodes; the initial topological graph G adopts adjacent amino acid residue pairs as edges; and inputting the initial topological graph G into a pre-training model, training the pre-training model so as to obtain the pre-training model for performing feature extraction on the protein, the pre-training model comprising a graph neural network with SE (3) and other degeneration. Therefore, the prediction precision of related prediction such as protein feature extraction can be improved, the working cost is reduced, and the prediction efficiency is improved.

Description

technical field [0001] The embodiment of the present application relates to the field of computer technology, in particular to a method and device for model training, protein feature extraction and function prediction. Background technique [0002] Proteins, composed of linear chains of amino acids, are among the most versatile molecules in living organisms. They perform vital functions in universal biological mechanisms. Proteins naturally fold into three-dimensional structures based on their amino acid sequences, and structure has a direct impact on their function. With the advent of sequencing technology, it has become relatively easier to obtain protein sequences, but the prediction of protein functions still requires a lot of time and resources. [0003] Most of the existing protein pre-training methods are based on the amino acid sequence of the protein or its related homologous sequence data, and often do not use the three-dimensional structure information of the pr...

AI Technical Summary

Problems solved by technology

With the advent of sequencing technology, it has become relatively easier to obtain protein sequences, but the prediction of protein functions still requires a lot of time and resources

[0003] Most of the existing protein pre-training methods are based on the amino acid sequence of the protein or its related homologous sequence data, and often do not use the three-dimensional structure information of the protein. have a more direct impact on downstream tasks such as

Images