From Code to Cure: How Generative AI Is Reinventing Drug Discovery

The traditional drug development process has become too time-consuming, expensive, and ineffective to the point where it stretches sometimes beyond a decade and incurs hefty investments in billions, with exaggerated clinical failure rates. This paper comprehensively outlines how Generative AI (GenAI) is reconfiguring the established way of thinking by enabling intelligent, data-driven design cycles. We evaluate state-of-the-art models such as MolBART, Chemformer, AlphaFold2, and DiffDock across all aspects of drug development: from target identification to molecule generation, structural simulation, and synthesis planning. We propose a four-phase GenAI paradigm building on these innovations: Target-to-Design Translation (T2D), Simulation-Guided Optimization Loop (SGOL), Digital Synthesis Feasibility (DSF), and Clinical Feedback Integration (CFI). The closed-loop design incorporates explainability, synthetic access, and continuous feedback, keeping molecule design tightly coupled to clinical reality. We validated this framework via real-world case studies, augmented by a fully simulated EGFR inhibitor pipeline, and explored challenges of implementing this framework in areas of regulation, data governance, and infrastructure integration. As GenAI achieves maturity, the framework does more than provide a technical scaffold—it charts a scalable pathway toward AI-native pharmaceutical ecosystems that are faster, safer, and more adaptive.