Closing the loop: human-augmented, mechanistically enhanced AI for proactive management of drug-drug interactions

Drug-drug interactions (DDIs) represent a major challenge in pharmaceutical research for ensuring safe and effective medication use in clinical practice. Pharmacological DDI assays generate data that underpin clinical guidelines, drug interaction checkers, and decision support systems. Although these approaches remain indispensable, contemporary clinical practice is far more complex, shaped by polypharmacy, multimorbidity, diverse phenotypes, and social determinants of health. Artificial intelligence (AI) offers opportunities to integrate molecular, pharmacokinetic, and pharmacodynamic knowledge with real-world observations, enabling more proactive and patient-centered approaches to DDI risk assessment. This perspective proposes a conceptual framework for transitioning from static, rule-based DDI tools toward human-augmented AI systems in which clinician feedback is embedded as an integral component of model learning and interpretation, rather than serving solely as post hoc validation. Through structured incorporation of clinical expertise, experimental pharmacology is continuously contextualized against real-world decision-making. The manuscript outlines an AI ecosystem that is ethically grounded, mechanistically informed, and enriched through pharmacovigilance data and systematic clinician input. By operationalizing human-in-the-loop learning as a core design principle, this framework establishes human-augmented AI as a foundational paradigm for future DDI research, drug development, and personalized medication safety.