Abstract:Traditional toxicity testing reliant on animal models is costly and low throughput, posing a significant challenge with the increasing numbers of chemicals that humans are exposed to in the environment. The purpose of this investigation was to build optimal prediction models for various human in vivo/organ-level toxicity end points (extracted from ChemIDPlus) using chemical structure and Tox21 in vitro quantitative high-throughput screening (qHTS) bioactivity assay data. Several supervised machine learning algorithms were applied to model 14 human toxicity end points pertaining to vascular, kidney, ureter and bladder, and liver organ systems. Three metrics were used to evaluate model performance: area under the receiver operating characteristic curve (AUC-ROC), balanced accuracy (BA), and Matthews correlation coefficient (MCC). The top four models, with AUC-ROC values >0.8, were derived for endocrine (0.90 ± 0.00), musculoskeletal (0.88 ± 0.02), peripheral nerve and sensation (0.85 ± 0.01), and brain and coverings (0.83 ± 0.02) toxicities, whereas the best model AUC-ROC values were >0.7 for the remaining 10 toxicities. Model performance was found to be dependent on the specific data set, model type, and feature selection method used. In addition, chemical structure and assay data showed different levels of contribution to the prediction of different toxicity end points. Although in vitro assay data, when combined with chemical structure, slightly improved the predictive accuracy for most end points (11 out of 14), a noteworthy finding was the near equal success of the structure-only models, which do not require Tox21 qHTS screening data, and the relatively poor performance of assay-only models. Thus, the top-performing structure-only models from this study could be applied for hazard screening of large sets of chemicals for potential human toxicity, whereas the largest assay contributions to models (i.e., cellular targets) could be used, along with the top-contributing structural features, to provide insight into toxicity mechanisms.

Research Progress in Novel Predictive Method of Drug Metabolic Toxicity in Vitro

Prediction of Metabolic Drug Interaction in vivo by Using in vitro Drug Metabolism Data

Research progress in drug toxicology based on systems biology

Applications of Machine Learning Methods in Drug Toxicity Prediction.

Development of novel tools for the in vitro investigation of drug-induced liver injury

Progress in Researches on in Vitro Alternative Model for Nephrotoxicity Evaluation

Review of machine learning and deep learning models for toxicity prediction

[Advances and Related Issues in the Use of in Vitro Methods to Predict Metabolic Clearance Rate of New Drugs].

Three-dimensional Culture of Hepatocytes for Prediction of Drug-Induced Hepatotoxicity

Human Hepatobiliary Organoids: Recent Advances in Drug Toxicity Verification and Drug Screening

Predictive Models for Human Organ Toxicity Based on in Vitro Bioactivity Data and Chemical Structure

In Silico Prediction Of Chemical Toxicity For Drug Design Using Machine Learning Methods And Structural Alerts

Predicting Organ Toxicity Using &Itin Vitro&It Bioactivity Data And Chemical Structure

Predicting Organ Toxicity Using in Vitro Bioactivity Data and Chemical Structure.

Advantages,disadvantages and applications of in vitro hepatic metabolic models

In Silico Prediction Of Adverse Drug Reactions And Toxicities Based On Structural, Biological And Clinical Data

Theoretical Consideration on the Prediction of in Vivo Toxicity from in Vitro Toxicity: Effect of Bio-Uptake Equilibrium, Kinetics and Mode of Action

Recent advances in the development of in vitro liver models for hepatotoxicity testing

In Silico Prediction of Chemical Reproductive Toxicity Using Machine Learning

In silico prediction of potential drug-induced nephrotoxicity with machine learning methods

Simulation and Prediction of in Vivo Drug Metabolism in Human Population Based on in Vitro-in Vivo Extrapolations