Comparative study of protein X-ray and NMR structures: molecular docking-based virtual screening

Molecular docking-based virtual ligand screening is a powerful computational approach for identifying potential binders from large chemical libraries. Protein structures used in docking screens are commonly derived from X-ray crystallography or NMR spectroscopy, yet their impact on screening performance remains unclear. To address this, we conducted virtual screening using Glide against both apo and holo X-ray and NMR structures of 18 proteins. While no statistically significant difference in screening performance was observed for apo structures overall, X-ray apo structures tended to perform better in cases where better ligand enrichment than random selection was achieved. Similarly, single holo X-ray and NMR structures did not exhibited statistically significant difference in screening performance either. However, when multiple holo X-ray structures and NMR conformers (from one PDB ensemble) per protein were used, X-ray structures outperformed NMR conformers in most cases. In addition, for consensus enrichment which leverages multiple structures/conformers per protein to optimise ligand ranking, X-ray holo structures exhibited better performance than NMR holo conformers, suggesting that X-ray structures with chemically diverse co-crystalized ligands may introduce more relevant binding-site configurations than the NMR conformers with higher structural diversity but the same bound ligand. Overall, the better performance by X-ray holo structures could be partially attributed to the higher numbers of hydrogen bonds and hydrophobic contacts, formed between proteins and docked ligands.