Deep learning method for predicting binding site on antibody through sequence

Abstract

The invention relates to the field of antibody binding site prediction, and particularly provides a deep learning method for predicting a binding site on an antibody through a sequence. The method comprises the following steps: acquiring a plurality of hypervariable regions on the antibody, and connecting the plurality of hypervariable regions in series to form a hypervariable region sequence, adding an unknown type of amino acid among different hypervariable region sequences as a distinguishing identifier, wherein the characteristics of each amino acid in the hypervariable region sequence comprise word embedding characteristics and additional characteristics; combining the word embedding features and the additional features to obtain a final feature matrix, and inputting the feature matrix into a neural network model, wherein the neural network model adopts a bidirectional long-short-term memory network and a transformer encoder to learn sequence information of high-variable regions and interaction information between different high-variable regions, and predicts antibody binding sites. According to the invention, the antibody binding site can be accurately predicted by learning the interaction information between different hypervariable regions.

Description

technical field [0001] The present invention relates to the technical field of antibody binding site prediction, and more specifically, relates to a deep learning method for predicting the binding site on an antibody through sequence. Background technique [0002] An antibody is a protein that plays a vital role in the immune system. The structure of an antibody is related to the process of antibody-antigen binding such as figure 1 shown. The main goal of antibody binding site prediction is to accurately predict which amino acids on the hypervariable region of the antibody responsible for binding to the antigen during the immune process will contact the antigen, and explore the binding mode of the antibody, which is helpful for immune mechanism research, disease diagnosis and antibody design . [0003] The relevant research methods for antibody binding site prediction are mainly divided into two types: traditional methods and machine learning methods, which will be introdu...

AI Technical Summary

Problems solved by technology

[0008] In order to overcome the problem that the above existing methods do not take into account the interaction information between different hypervariable regions during the antibody antigen binding process, the present invention provides a deep learning method for predicting the binding site on the antibody based on the sequence, which can learn Interaction information between different hypervariable regions to accurately predict antibody binding sites

Images