DBMol: Design of High-Affinity, Target-Specific Small Molecules through Structure Prediction Models
DBMol uses structure-prediction models as the optimization signal for generating new small-molecule ligands.
The paper describes an alternating process: gradient optimization improves predicted pocket interactions and binding affinity, then a flow-matching model projects the result back into chemically valid molecules. Its experiments report stronger Boltz-2 affinity proxy scores, pocket-specific binding, and improved pocket coverage while preserving diversity. The authors also test against held-out metrics, including AlphaFold-3-based evaluation, to limit self-confirmation from optimizing against Boltz-2. ArXiv · AI/CL/LG's note
The paper describes an alternating process: gradient optimization improves predicted pocket interactions and binding affinity, then a flow-matching model projects the result back into chemically valid molecules. Its experiments report stronger Boltz-2 affinity proxy scores, pocket-specific binding, and improved pocket coverage while preserving diversity. The authors also test against held-out metrics, including AlphaFold-3-based evaluation, to limit self-confirmation from optimizing against Boltz-2. ArXiv · AI/CL/LG's note
score 5