AI-Guided Fragment-Based Drug Design for Virtual Library Screening and Hit Optimization

Review Article

Abstract

Recent developments in artificial intelligence have transformed the fragment-based drug design (FBDD), changing traditional approaches to drug discovery. Machine learning and deep learning algorithms now enable rapid exploration of vast chemical spaces, precise prediction of fragment properties, and optimization of binding interactions. The use of AI-driven methods with FBDD has enhanced virtual library screening efficiency, improved hit identification accuracy, and accelerated the fragment-to-lead optimization process. Deep generative models and physics-informed neural networks have shown remarkable capabilities in designing vast fragment libraries and predicting their physicochemical properties. Graph neural networks and reinforcement learning algorithms have proven particularly effective in binding affinity prediction and fragment elaboration methods. The combination of AI technologies with experimental methods, including X-ray crystallography, NMR spectroscopy, and surface plasmon resonance, has established new paradigms in structure-based drug design. Success stories in developing kinase inhibitors and targeting protein-protein interactions highlight the practical impact of AI-guided FBDD. These AI-enabled virtual library screening helps in reducing drug discovery timelines and improve success rates in lead optimization