Harnessing AI for Safer Drug Development: Predicting Adverse Drug Reactions

Adverse drug reactions (ADRs) remain a leading cause of hospital admissions and treatment discontinuation globally. Traditional methods often struggle to detect rare or delayed side effects associated with medications, highlighting the need for more advanced predictive tools. A pioneering study from the Medical University of Sofia introduces a deep learning model capable of forecasting potential ADRs based solely on a drug's chemical structure.

The model employs a neural network trained on extensive pharmacovigilance data. It analyzes SMILES codes—standardized representations of molecular structures—to predict the likelihood of six major ADRs: hepatotoxicity, nephrotoxicity, cardiotoxicity, neurotoxicity, hypertension, and photosensitivity. The researchers report that the model effectively identified known reactions with a low rate of false positives, demonstrating promising accuracy.

Testing on well-characterized drugs showed predictions aligned with their established side-effect profiles. For instance, erythromycin was predicted to have a 94.06% probability of causing hepatotoxicity, cisplatin showed an 88.44% chance of nephrotoxicity and a 75.8% probability of hypertension. The model also estimated a 22% chance of photosensitivity with cisplatin, whereas the experimental compound ezeprogind had a 64.8% likelihood of photosensitivity. For the novel molecule enadoline, the model indicated minimal ADR risks.

These findings suggest that AI-driven models could become invaluable tools in early drug discovery and ongoing safety assessments. They have the potential to streamline the identification of safety risks before clinical trials and regulatory approval. While current models focus on chemical structure, future improvements could incorporate dosing information and patient-specific factors, further enhancing predictive accuracy.

This innovative approach marks a significant step toward integrating artificial intelligence into pharmacovigilance, ultimately contributing to the development of safer medications and improved patient outcomes.

Source: https://medicalxpress.com/news/2025-08-drug-safety-ai-adverse-reaction.html