General model for predicting performance of organic compound and prediction method

Abstract

The invention relates to a general model for predicting the performance of an organic compound and a prediction method. The general model can accurately predict the physical and chemical performance and ADME / Tox performance of an organic compound; and the physical and chemical properties of the organic compound and the ADME / Tox performance depend on the change of free energy related to the performance, so that a linear free energy relation model-LFER model for predicting the performance of the organic compound is deduced by using a chemical thermodynamic method, according to the linear free energy relation model (LFER model), a general formula for predicting the performance of the organic compound by using a fat-water partition coefficient is shown as a formula I, the accuracy of predicting the performance of the organic compound is high, and the problems that a QSPR model in the prior art is large in uncertainty, poor in prediction precision and poor in prediction capability and needs to be applied by professionals can be effectively solved.

Description

technical field [0001] The invention belongs to the technical field of computational chemistry, and in particular relates to a general model and a prediction method for predicting the properties of organic compounds. Background technique [0002] With the rapid development of the field of chemistry and chemical engineering and the rapid growth of new organic compounds, it is becoming more and more important to predict the physicochemical properties and pharmacokinetic properties of organic compounds (including drug candidates) by computational methods. For example, before synthesizing a drug candidate, accurate prediction of its absorption, distribution, metabolism, excretion and toxicity (ADME / Tox) can significantly reduce the cost and time of drug development and increase the success rate. [0003] At present, the commonly used method to predict the properties of organic compounds is to establish a structure-activity quantitative relationship model (QSPR) through mathemati...

AI Technical Summary

Problems solved by technology

[0004] However, prior art using QSPR S There are several problems in predicting the performance of compounds by the model, specifically: (1) the uncertainty of the model, because the molecular structure descriptor in the model is selected from thousands of molecular structure descriptors by mathematical and statistical methods , the resulting model is highly dependent on the compounds used in the training set, the quality of the data and the modeling approach employed

(2) The model can only predict compounds within its application domain (Applicability domain), but the application domain of the model is difficult to determine

QSPR S Models are usually only suitable for compounds within their application domain, and the prediction accuracy for other compounds is not high

(3) Overfitting, many models have good statistical results for the training set, but these models are not good at predicting the compounds used to test the model

Moreover, the better the statistical results, the worse the predictive ability. An important reason for this phenomenon is that some molecular descriptors in the model are only applicable to the compounds in the training set.

(4) It is difficult for non-professionals to practice and use QSPR S Models to predict compound performance require specialized training to build QSPR S Model and utilize QSPR S Model predicts compound performance

Images