Pocket Contrastive Learning and Conformation Fusion for Robust Structure-Conditioned 3D Molecular Generation under Receptor Conformational Variation.

Pocket contrastive learning and conformation fusion enhance 3D molecular generation robustness under receptor conformational variation, improving success rates by significant margins.

• Why it matters: Receptor conformational changes cause distribution shifts that hinder accurate molecular design, especially during cross-docking, highlighting the need for methods that handle receptor flexibility effectively.
• What they did: A structure-conditioned 3D molecular generation framework was developed, integrating pocket-level contrastive learning with reference-conformation feature fusion across a stepwise generative process, trained on multiple datasets.
• The result: The approach achieved higher success, structural validity, and docking performance, particularly under cross-dock conditions, demonstrating improved robustness in structure-conditioned molecular generation.