The expectations of in silico fragment-based drug design and future challenges

Fragment-based drug discovery (FBDD) employs the Grow-Merge-Link (GML) model to identify therapeutic compounds through a combination of experimental and computational techniques. Generative models facilitate compound design, predict interactions, and enhance chemical diversity.

This perspective highlights recent FBDD developments, especially in silico methods where AI-ML accelerates discovery. Reinforcement learning optimizes properties, saving resources for targets like kinases and GPCRs. Generative chemistry enables de novo design, boosting diversity and IP, with pocket-aware design ensuring relevance and ADMET evaluation. Advances in VAEs and reinforcement learning speed up analogue creation and SAR by catalogue.

AI is transforming FBDD by automating compound design, predicting fragment-protein interactions, and expanding chemical diversity through deep learning, generative models, and reinforcement learning. These tools accelerate hit-to-lead processes, improve drug properties, and support multi-objective optimisation. AI enables fragment generation, pocket-specific design, and large-scale virtual screening, aiding the targeting of challenging proteins and modalities such as PROTACs and molecular glues. Larger fragment libraries enhance model training; although experimental validation remains key, AI reduces development time, improves accuracy, and broadens FBDD's scope.