The Transformative Role of AI and Generative Chemistry in Modern Drug Discovery

Generative AI for de novo drug design has emerged as a promising paradigm for navigating the vast drug-like chemical space. In this workshop, we first provide a broad overview of the applications of AI throughout the drug discovery pipeline, from target identification to lead optimization. We then offer a critical and technical introduction to the main generative approaches, from SMILES-based chemical language models to 3D-native diffusion models, discussing their strengths, limitations, and practical applicability. We then argue that benchmark performance metrics, commonly used to evaluate these models, are insufficient proxies for real-world drug discovery impact, due to the fundamental gap between retrospective model validation and prospective generalization into unknown chemical subspaces. Finally, we present ITForge, an integrated, target-aware pipeline for generative fragment growing developed at ITALFARMACO, which combines a reinforcement learning-based generative module with a multi-stage scoring, filtering, and ranking workflow designed to guide generative models from the known into the unknown in a controlled manner.