Method for predicting compound carcinogenic toxicity based on complex sampling and improved decision forest algorithm

Abstract

The invention relates to a method for predicting compound carcinogenic toxicity based on complex sampling and an improved decision forest algorithm. The method is suitable for calculating carcinogenic toxicity evaluation and virtual selection on a compound according to an organic small molecular structure, and comprises the following steps: firstly, adopting a relative force field to molecules with the molecular structure to carry out optimization and charge calculation, carrying out complex sampling to the compound with centralized original training to be used for generating a training subset, and fixing various relative descriptors in composition calculation molecules of descriptors according to a complex sampling algorithm result; secondly, optimizing a descriptor pool by using a method based on relative matrix analysis and factor analysis; and finally, carrying out data mining on carcinogenic toxicity data and corresponding chemical characteristics thereof of training set molecules by using the improved decision forest method to obtain a classified prediction confidence interval, a carcinogenic toxicity prediction mold and a judgment rule. The method has favorable application prospect in high throughput virtual selection and calculating the carcinogenic toxicity evaluation.

Description

technical field [0001] The invention relates to a compound carcinogenic toxicity prediction calculation method based on complex sampling and improved decision forest algorithm, which is suitable for virtual carcinogenic toxicity evaluation and screening of the compound according to the molecular structure information of the organic compound. Background technique [0002] Toxicity issues are an important factor in the failure of late-stage drug development. The carcinogenic toxicity of a compound refers to a long-term effect of the compound inducing the growth of malignant or benign tumors in the human body. Rodent biological testing is the main carcinogenic toxicity test method currently used. However, this approach suffers from the following problems: (1) high cost of testing (average test cost exceeds two million dollars); (2) time-consuming (3 to 5 years); (3) ethical considerations and public pressure - in R&D and Reduce or eliminate the use of animals in testing. Due...

AI Technical Summary

Problems solved by technology

However, this approach suffers from the following problems: (1) high cost of testing (average test cost over two million dollars); (2) time-consuming (3 to 5 years); (3) ethical considerations and public pressure - in R&D and Reduce or eliminate the use of animals in testing

[0007] In addition to performance limitations, current computational toxicity prediction methods often encounter the problem of random prediction, that is, changing the composition of even a small training data set may make diametrically opposite prediction results

Images