PHYSICS-BASED VS. AI APPROACHES IN DRUG DESIGN: COMPLEMENTARY TOOLS OR COMPETING PARADIGMS?

Recent advances in artificial intelligence (AI), particularly in protein structure prediction, have significantly impacted the field of computational drug design. More recently, emerging AI autonomous agents are enabling automated hypothesis generation, experiment design, and dynamic system exploration, bringing AI to the forefront of structural biology and medicinal chemistry. While AI excels at navigating the multidimensional nature of drug discovery - encompassing structure, function, selectivity, and pharmacokinetics - physics-based methods such as docking and molecular dynamics simulations remain essential for tasks that require mechanistic insight and interpretability. This talk will present a user's perspective on the integration of AI and physics-based tools within drug design workflows, highlighting practical applications of both approaches in modelling protein structures, predicting biological targets, investigating ligand–protein interactions, and evaluating ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties. Emphasis will be placed on how these methodologies can complement one another, rather than compete, in building robust and efficient in silico pipelines. Challenges related to accessibility, data quality, and methodological integration will also be discussed, with reflections on how research groups can benefit from the strengths of both paradigms to address the multifactorial complexity of drug discovery.