Drug-similar compound toxicity predicating method based on deep learning

Abstract

The invention relates to the field of drug-similar compound toxicity predicating technology, and provides a drug-similar compound toxicity predicating method based on deep learning. The method comprises the steps of 1, performing character extraction, and generating a fingerprint sequence of a to-be-detected drug-similar compound through module fingerprint generating software; 2, performing noisereduction preprocessing on the characteristic of the module fingerprint sequence; 3, performing characteristic dimension reduction on the module fingerprint sequence after characteristic preprocessingby means of a DX algorithm, and updating the module fingerprint sequence; and 4, performing toxicity prediction on the module fingerprint sequence after dimension reduction screening by means of a stack type self-coding neural network model. According to the method, the module fingerprint sequence is used as the module characteristic for describing a large amount of redundancy of fingerprint characteristic items of the drug-similar compound, thereby designing and realizing the characteristic dimension reduction method. The fingerprint characteristic which is screened again after dimension reduction evaluation is used as input of cascade hidden layer learning, thereby realizing high-efficiency predication to the toxicity of the drug-similar compound.

Description

technical field [0001] The invention relates to the technical field of toxicity prediction of drug-like compounds, and is a method for predicting the toxicity of drug-like compounds based on deep learning. Background technique [0002] Compound toxicity is one of the important properties of pharmacokinetics and one of the main reasons for the failure of drug development. Putting drug safety evaluation in the early stage of new drug development can help shorten the development cycle and reduce development costs. The toxicity of compounds is also an important starting point for human safety in daily life. People are exposed to a large number of chemicals every day. While improving the quality of life, they also pose potential hazards to human health and the environment. It can be seen that it is very necessary to evaluate the safety of compounds. [0003] Different from traditional in vivo and in vitro toxicity assessment experiments. With the development and application of ...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a method for predicting the toxicity of drug-like compounds based on deep learning, which overcomes the deficiencies of the above-mentioned prior art, and can effectively solve the technical obstacles in the high-throughput screening and identification of drug leads in the prior art, resulting in The high cost of safety assessment of large batches of drug-like compounds

Images