DIADpredictor: in silico prediction for drug-induced autoimmune diseases (DIAD) with machine learning
Abstract
The incidence and complexity of drug-induced autoimmune diseases have been on the rise in recent years, and the emergence of a large number of new drugs, including biologics, has accelerated this process. It is very important to reduce the failure caused by drug toxicity in the later stage of drug development or even in clinical trials, reduce the risk of disease and social medical burden, and identify potential toxicity problems as soon as possible. In this research, we focused on in silico toxicity prediction of chemical DIAD on humans based on structurally diverse organic chemicals. The model developed with the support vector machine (SVM) achieved a prediction accuracy of 77.08% on the test set and 75.98% on external validation.We hope this model can provide valuable reference for chemical DIAD evaluation.