A method for constructing a Gibbs free energy prediction model based on multiple environmental parameters and molecular fingerprints

CN118248235BActive Publication Date: 2024-11-05TIANJIN UNIV
View PDF -1 Cites 0 Cited by

Patent Information

Application Number
CN202410079062.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-19
Publication Date
2024-11-05
Estimated Expiration
2044-01-19

AI Technical Summary

Technical Problem

The existing Gibbs free energy prediction method does not consider environmental factors and lacks appropriate feature screening mechanisms, resulting in low prediction accuracy.

Method used

Using data encoding methods of multi-environment parameters and molecular fingerprints, one-dimensional molecular fingerprint features are converted into two-dimensional square matrix, and feature fusion and regression prediction are performed through a convolutional neural network architecture with multiple parallel inputs, including feature preprocessing and feature Fusion and regression prediction module.

Benefits of technology

It significantly improves the prediction accuracy of Gibbs free energy in biochemical reactions, reduces prediction errors, and performs more accurately than a single fingerprint feature method.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118248235B_ABST
    Figure CN118248235B_ABST
Patent Text Reader

Abstract

A Gibbs free energy prediction model based on multiple environmental parameters and molecular fingerprints: The present invention relates to a Gibbs free energy prediction model and a neural network technology, which solves the problem that the existing methods have low prediction accuracy. The technical solution adopted by the present invention is: a Gibbs free energy prediction model based on multiple environmental parameters and molecular fingerprints. First, a data encoding method is proposed to convert a one-dimensional feature vector into a two-dimensional square matrix. Second, a convolutional neural network architecture is proposed, which uses a multi-channel parallel input method to input different reaction codes based on different types of molecular fingerprints, and divides the substrates and products of the biochemical reaction into two channels and inputs them into the network. Feature preprocessing and feature fusion are performed through multiple convolutional layers, pooling layers and fully connected layers, and finally regression prediction of Gibbs free energy is performed.
Need to check novelty before this filing date? Find Prior Art