Method for predicting properties of chemical molecules based on quantum graph neural network

Abstract

The invention relates to a method for predicting the properties of chemical molecules based on a quantum graph neural network. Firstly, chemical molecules are expressed in the form of an undirected graph, and in the undirected graph, features of a starting node and a terminating node of each edge are spliced to obtain a spliced matrix. And after the splicing matrix corresponding to each edge is coded into quantum data, preset unitary transformation is performed, and information of the nodes is updated according to classical data converted from a unitary transformation result.

Description

technical field [0001] The invention mainly relates to the field of chemical molecule prediction, more specifically, to a method for solving chemical molecule property prediction based on a quantum graph neural network. Background technique [0002] As a tool for chemistry learning and research, computers can not only help with text and image processing, but are also irreplaceable in chemistry learning-related data processing, map simulation analysis, molecular mechanics, and quantitative calculations. The prediction of chemical molecular properties is a sub-problem of computer-aided chemical molecule research and development, such as the prediction of pharmacological analysis and toxicological analysis, which belongs to the field of biomedicine. The problem of predicting the properties of chemical molecules is usually to predict the properties of the whole or functional groups given a known molecule. It is required that the predicted results can provide meaningful suggesti...

AI Technical Summary

Problems solved by technology

[0006] It is worth noting that classical graph neural network algorithms cannot be processed on quantum devices in the same way that they are processed on electronic chips

For the prediction of chemical molecular properties, some excellent performances have been achieved based on the current classical graph neural network model, but in the message passing stage between nodes in the classical graph neural network algorithm, its operation has a high degree of parallelism and consumes a lot of computing resources , so that the current computer computing power is stretched

Images