Protein-ligand affinity prediction method based on interaction energy term and machine learning

Abstract

The present invention relates to a protein-ligand affinity prediction method based on interaction energy items and machine learning. The method distributes various interaction energy items between ligands and protein pockets on the main and side chains of 20 amino acid residues, And use the machine learning method to train the known active protein-ligand interaction energy information in the PDB library to obtain a model, and use the obtained model to score the affinity between the ligand and the protein. The present invention can fully consider the influence of main and side chains of different amino acid residues on the affinity by dispersing the interaction energy items; using machine learning to perform nonlinear fitting is beneficial to deal with the correlation or coupling between various interaction energies, thereby targeting It can significantly reduce the errors caused by different amino acid structures in the process of affinity calculation. The invention is more conducive to the prediction of the affinity of active molecules, so as to achieve the purpose of improving the prediction accuracy.

Description

technical field [0001] The invention relates to the technical field of scoring functions, in particular to a method for predicting protein-ligand affinity based on interaction energy between ligand and protein binding pocket residues and machine learning. Background technique [0002] In the process of drug design, scoring functions are usually used to predict the binding affinity between protein targets and their ligands, thereby improving the success rate of drug design and reducing the cost of drug screening. In recent years, it has been paid more and more attention by relevant scientific research institutions and pharmaceutical companies. Common scoring function methods can be divided into force field-based scoring functions, experience-based scoring functions and knowledge-based scoring functions. Commonly used scoring functions include PLP, ChemScore, X-Score, and GlideScore, etc. Among them, the scoring function based on experience is the most widely used. Empirica...

AI Technical Summary

Problems solved by technology

However, the current scoring function based on experience still has related problems such as poor correlation between predicted value and experimental value, large target dependence and poor sensitivity to congeners

There are many reasons for these problems. For example, too few interaction energy items cause the differentiation to be ignored to a certain extent. The protein-ligand complex data set is rarely enough to cause insufficient diversity. Linear regression ignores the coupling effect between each interaction energy item.

Images