1396-DiffPharma: 3D Molecular Generation

The paper introduce DiffPharma, a novel drug discovery framework that utilizes a conditional diffusion model to design three-dimensional molecules based on specific protein-ligand interactions. By incorporating an interaction particle framework, the system explicitly models essential contacts like hydrogen bonds and hydrophobic interactions to ensure generated molecules satisfy precise geometric constraints. Experimental results show that this method significantly outperforms existing models in interaction reproducibility, achieving high similarity to reference ligands across diverse protein targets. Beyond structural accuracy, the research validates the chemical feasibility of the designs through retrosynthetic analysis and molecular dynamics simulations, particularly in a case study involving the SARS-CoV-2 main protease. Ultimately, DiffPharma offers a sophisticated approach for exploring the vast chemical space while maintaining the pharmacological integrity required for effective drug design.References:Sako M, Yasuo N, Sekijima M. Interaction-constrained 3D molecular generation using a diffusion model enables structure-based pharmacophore modeling for drug design[J]. npj Drug Discovery, 2026, 3(1): 8.前往小宇宙评论区与主播互动