Universal compound structure-property correlation prediction method based on neural network

A compound structure and prediction method technology, applied in chemical property prediction, neural learning methods, biological neural network models, etc., can solve unrealistic problems and achieve high practicality and ingenious design effects

CN111798935APending Publication Date: 2020-10-20PHARMABLOCK SCIENCES (NANJING) INC
0 Cites 6 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2020-10-20

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a universal compound structure-property correlation prediction method based on a neural network, and the method comprises the following steps: 1, converting a molecular descriptor into a form of a feature vector, and constructing a data set; 2, dividing the data set into a training set and a test set, sending the training set and the test set to a full-connection neural network for training, and determining parameters of a full-convolution model; and 3, converting a to-be-predicted molecular descriptor into a form of a feature vector, and performing prediction accordingto the full convolution model. The prediction method has relatively high accuracy in prediction of solubility.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to a method for constructing a prediction model, in particular to a neural network-based method for predicting the structure-property correlation of universal compounds. Background technique

[0002] Solubility is a basic property of compounds, especially small molecular compounds. Generally speaking, for different compounds, due to the structure and spatial arrangement of the compounds themselves, different compounds will have different solubility in the same solution under the same conditions. The determination of solubility plays an important role in the chemical industry's chemical process, process preparation, medicine and the migration of chemical substances in the environment.

[0003] However, due to the wide variety of compounds in reality, and different solutions require different storage and determination conditions, it is impractical to measure the solubility of all compounds through practical methods. Establishing an...

Examples

Embodiment Construction

[0030] 1. Molecular descriptor

[0031] Molecular descriptors are the basis of QSPR, which refer to molecular properties and measurements that can be numerically expressed in one or more aspects of molecules, and can be understood and analyzed by computers. A molecular descriptor can be a direct numerical representation of the physical and chemical properties of a molecule, or it can calculate a variety of data indicators according to a specific algorithm. The former includes physical and chemical indicators such as the boiling point and melting point of molecular compounds, while the latter involves more about the energy of the outer layer of molecules, the distribution of outer electron charges between bonds, etc.

[0032] There are various calculation methods for molecular descriptors. Different software and software packages can calculate nearly 6,000 physical and chemical parameters covering the characteristic properties and structural characteristics of compounds at this...