Ligand Design Using Unique Conformations to Preferentially Dock a Specific Site on Collagen-Bound MMP1

Precise site-specific ligand design remains a major challenge in structure-based drug discovery. Most existing approaches screen ligands against binding pockets identified from static protein structures obtained by X-ray crystallography, NMR spectroscopy, cryo-electron microscopy, or AlphaFold predictions. However, protein function is governed by a structure–dynamics–function relationship, and ligand screening that does not account for binding competition across the protein surface or the receptor’s dynamic, substrate-dependent conformational states remains incomplete. Substrate-specific conformations are underexplored and may offer new opportunities for selective ligand design, although systematic workflows to identify and exploit such states remain limited. Previously, we showed that collagen alters matrix metalloprotease-1 (MMP1) dynamics and that R405 is a collagen-specific allosteric residue exhibiting strong dynamic correlations with the catalytic site. Here, we present a computational framework for substrate-specific allosteric ligand design using collagen-bound MMP1 as a model system. We characterized the conformational dynamics of free and collagen-bound MMP1 by all-atom molecular dynamics simulations, clustered the resulting conformational ensembles, and identified conformations unique to the collagen-bound state. These conformations were used as structural templates for machine-learning-based generation of approximately 150,000 candidate ligands, which were subsequently docked against both the R405-centered region and all detectable binding pockets on the MMP1 surface. Several candidate ligands were predicted to dock preferentially at the R405 region by at least 0.3 kcal/mol compared with competing surface pockets. Together, these results establish a generalizable computational workflow for identifying candidate ligands predicted to preferentially dock to substrate-specific allosteric conformations and provide a foundation for future experimental validation of selective allosteric modulation.