A novel strategy is applied to obtain quantitative insights on factors influencing biological affinity in protein-ligand complexes. This approach is based on the detection of ligand binding by (15)N and (1)H amide chemical shift differences in two-dimensional (15)N-heteronuclear single-quantum correlation spectra. Essential structural features linked to affinity can be extracted using statistical analysis of (15)N and (1)H amide chemical shift differences in congeneric series relative to uncomplexed protein spectra, as demonstrated for 20 MMP-3 inhibitors in complex with human matrix metalloproteinase stromelysin (MMP-3). The statistical analysis using PLS led to a significant model, while its chemical interpretation, highlighting the importance of particular residues for affinity, are in agreement to an X-ray structure of one key compound in the homologue MMP-8 binding site.