Pallatom-Ligand: an All-Atom Diffusion Model for Designing Ligand-Binding Proteins

Haochen Wang, Qianyi Wang, Rui Ma, Jiawei Guan, Weikun.Wu, Haobo Wang, Jiayi Dou

Keywords: Diffusion, Protein Design, Ligand Binding

Abstract: Small-molecule ligands extend protein functionality beyond natural amino acids, enabling sophisticated processes like catalysis, signal transduction, and light harvesting. However, designing proteins with high affinity and selectivity for arbitrary ligands remains a major challenge. We present Pallatom-Ligand, a diffusion model that performs end-to-end generation of ligand-binding proteins at atomic resolution. By directly learning the joint distribution of all atoms in the protein–ligand complexes, Pallatom-Ligand delivers state-of-the-art performance, achieving the highest *in silico* success rates in a comprehensive benchmark. In addition, Pallatom-Ligand's novel conditioning framework enables programmable control over global protein fold and atomic-level ligand solvent accessibility. With these capabilities, Pallatom-Ligand opens new opportunities for exploring the protein function space, advancing both generative modeling and computational protein engineering.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 23631