MultiEndpointTox 2.0: A Multi-Endpoint Toxicity Prediction Platform for Drug Safety Assessment with OpenTox-Aligned Workflows

Abstract

MultiEndpointTox 2.0 is an open-source computational toxicology platform for multi-endpoint drug safety prediction from molecular SMILES input. The system integrates machine-learning models including XGBoost, LightGBM, and random forests to predict multiple toxicity endpoints including hepatotoxicity, nephrotoxicity, Ames mutagenicity, cytotoxicity, skin sensitization, and cardiotoxicity. The platform implements scaffold-based validation, SHAP explainability, structural alert detection, metabolite screening via SyGMa, pharmacogenetic sensitivity modeling, and molecular docking integration. MultiEndpointTox is deployed as a Dockerized FastAPI service with OpenAPI documentation, enabling integration into cheminformatics pipelines and automated drug discovery workflows. This archive contains the full source code, trained models, documentation, and reproducible environment configuration for the MultiEndpointTox platform.