Molecule discovery method based on graph Bayesian optimization

Abstract

The invention relates to a molecule discovery method based on graph Bayesian optimization. The method comprises the steps of acquiring a candidate molecule set; randomly selecting a plurality of molecules from the candidate molecule set for property evaluation to obtain a molecule-property pair set; training an agent model according to the molecule-property pair set to obtain a trained agent model; performing property prediction on molecules in the candidate molecule set according to the trained agent model, and selecting desired molecules from the candidate set to perform property evaluationto obtain desired molecule properties; and finding molecules having desired properties according to the desired molecular properties. According to the molecule discovery method based on graph Bayesianoptimization, the molecules in the candidate set are predicted, then the molecules are selected according to the prediction result for evaluation to obtain the actual properties of the molecules, themolecules are selected accordingly for evaluation, the evaluation frequency of the molecules is reduced, and thus the analysis and evaluation cost is reduced.

Description

technical field [0001] The invention relates to the field of molecular discovery, in particular to a molecular discovery method based on graph Bayesian optimization. Background technique [0002] Molecular discovery has always been an important problem in computational chemistry, materials design and other fields. Its main purpose is to find molecules with certain desired properties from a large number of molecular spaces. In the process of molecular discovery, it has the following characteristics: (1) The molecular space is usually very huge. For example: the number of potential drug-like molecules is about 10 23 -10 60 . (2) Molecular space is discrete. (3) Evaluation of a certain property of a molecule (such as: drug-like properties, effectiveness, etc.) usually requires a high cost. For example: the application of density functional theory (Density functional theory) to estimate molecular properties, high-precision estimates are obtained based on high calculations,...

AI Technical Summary

Problems solved by technology

(3) Evaluation of a certain property of a molecule (such as: drug-like properties, effectiveness, etc.) usually requires a high cost

However, even if this method considers the cost when searching the vector space, it still needs to spend a lot of extra pre-training time to ensure the accuracy of encoding-decoding when training encoding-decoding tools (such as variational autoencoders), and finally Still can't lower the price of molecular discovery

Images