Molecular attribute determination method and device, electronic equipment and storage medium

Abstract

The invention discloses a molecular attribute determination method and device, electronic equipment and a storage medium, belonging to the technical field of molecules. The method comprise the following steps: S1, representing molecular attribute data of each labeled molecule and molecular structure data of each unlabeled molecule by using molecular topological structure diagrams; S2, inputting the molecular topological structure diagrams of all the labeled and unlabeled molecules into a pre-constructed teacher model, and training all the molecular topological structure diagrams by utilizing semi-supervised learning to obtain parameters of a teacher model; S3, migrating the parameters of the teacher model to a pre-constructed student model; S4, adjusting the student model by using labeledmolecules; S5, using the adjusted student model to assign labels to all the unlabeled molecules to obtain a labeled molecule set; S6, feeding back the labeled molecular set to the teacher model; and S7, repeatedly executing the steps S2 to S6 until the teacher model and the student model are both converged, and predicting molecular attributes by utilizing the converged student model. According tothe invention, the attributes of unknown molecules can be accurately predicted.

Description

technical field [0001] The present application relates to the technical field of molecular determination, and in particular to a method, device, electronic equipment and storage medium for determining molecular properties. Background technique [0002] The discovery of new substances and materials can often drive technological development and social progress. From drug research and development to material development, molecular discovery is inseparable. In many fields, a candidate set of possible molecules is needed in order to find molecules with specific properties. For each molecule in the ensemble, its properties need to be predicted and, if the requirements are met, further studied. There are two main methods for predicting molecular properties. One is to use traditional physical methods such as density functional analysis, but this method is too time-consuming to handle large amounts of data; the other is data-driven methods. The relevant knowledge of molecular prope...

AI Technical Summary

Problems solved by technology

There are two main methods for predicting molecular properties. One is to use traditional physical methods such as density functional analysis, but this method is too time-consuming to handle large amounts of data; the other is data-driven methods. Learn the relevant knowledge of molecular properties from existing data and use it to predict properties of unknown properties

However, data-driven methods rely heavily on molecular data with known properties (called labeled molecules), and the number of labeled molecules is very small. At the same time, a large number of molecular data with unknown molecular properties (called unlabeled molecules) not being used

Images