OpenTox Summer School 2024

An Integrated Approach for toxicophore identification and characterization

Adverse drug reactions (ADRs) linked to toxicophores (structural features associated with toxicity) are a major cause of drug withdrawals. Traditional methods for toxicophore identification are expensive, time-consuming, and often rely on animal testing. In contrast, computational tools offer a faster, cheaper, and more ethical alternative for drug discovery. We propose an integrated in silico framework for toxicophore identification within the drug discovery pipeline. The framework utilizes molecular docking and advanced AI-based QSAR modeling to identify protein-drug interactions responsible for ADRs. As a proof-of-concept, the framework is applied to the drug pair Troglitazone and Rosiglitazone, withdrawn due to hepatotoxicity and suspected cardiotoxicity. Building on previous studies, 15 relevant pathways were identified across three domains (RNA-seq analysis, reporter assay, and disease-based). An additional domain captured shared pathways between both drugs. Proteins from these pathways were retrieved from databases, prepared, and docked to each drug using inverse molecular docking. Nearly 80% of the proteins in each pathway achieved a favorable minimum score (<-6), suggesting a potential interaction relevant to toxicity and therapeutic effect. This initial analysis provides a foundation for further investigation. The framework will leverage insights from this docking study using biological, chemical, and pharmacokinetic data. This combined data will be used develop silico models for the reliable prediction of toxicophores, ultimately improving drug safety and efficacy.