Metal organic compound energy prediction method based on deep learning

By constructing molecular graphs and graph convolutional networks based on deep learning methods, a deep interactive fusion of energy prediction for organometallic compounds is achieved. This addresses the shortcomings of existing energy prediction technologies, improves prediction accuracy and reliability, provides a quantitative correlation between energy and molecular structure, and supports material design and screening.

CN122157857APending Publication Date: 2026-06-05HEFEI INTELLIGENT SPACE TECHNOLOGY CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202610323187.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-17
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing technologies for predicting the energy of organometallic compounds suffer from several problems, including insufficient characterization of long-range interactions or specific coordination environments, easy loss of structure-electron coupling information, and neglect of the intrinsic correlation of various physical properties. These problems result in high computational costs, long processing times, and inaccurate prediction results.

Method used

A deep learning-based approach is adopted to construct a molecular graph and extract geometric features by combining graph convolutional networks. A deep interactive fusion of geometric and electronic structure features is achieved through a cross-attention and gating fusion network. A multi-task prediction network is designed to simultaneously predict the total energy and molecular structure, and the energy prediction value and its confidence interval are obtained through multiple forward propagations.

Benefits of technology

It improves the quality of basic features and the accuracy of energy prediction, provides reliable prediction results and their confidence intervals, and solves the problems of inaccurate feature characterization and lack of reliability of prediction results in traditional methods, providing more valuable decision-making basis for material design and screening.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a metal organic compound energy prediction method based on deep learning, and belongs to the technical field of compound property prediction. The method comprises the following steps: acquiring molecular structure information of a metal organic compound, calculating electronic structure characteristics, constructing a molecular map, and encoding to generate a geometric characteristic vector; outputting global fusion features through the multi-modal feature fusion model; predicting a total energy value and a molecular structure through a multi-task prediction network; and an energy prediction value and a confidence interval are obtained through statistics of a mean value and a variance of the prediction total energy value. According to the method, the molecular graph is constructed, the geometric features are extracted, and meanwhile, deep interactive fusion of geometric and electronic structure features is realized by using the cross attention and gating fusion network, so that the problems of inaccurate characterization of metal organic compound features and poor multi-modal feature fusion effect of a traditional method are solved.
Need to check novelty before this filing date? Find Prior Art