3D-QSAR of Angiotensin-Converting Enzyme and Thermolysin Inhibitors: A Comparison of CoMFA Models Based on Deduced and Experimentally Determined Active Site Geometries

The ability of comparative molecular field analysis (CoMFA), a three-dimensional, quantitative structure-activity relationship (3-D QSAR) paradigm, to predict the activity of inhibitors of angiotensin-converting enzyme (ACE) and thermolysin was examined. Correlations derived from computationally and experimentally determined alignment rules were compared. The correlations derived for the ACE series using alignment rules determined from a systematic conformational search (Mayer, D.; Naylor, C. B.; Motoc, I.; Marshall, G. R. J. Comput.-Aided Molec. Des. 1987, 1, 3–16) were comparable to those derived for the thermolysin inhibitors using alignment rules defined by crystallographic data. Models derived from potential fields alone, however, were insufficient for accurately quantifying and predicting the nature of enzyme-inhibitor interactions. The predictive ability of the ACE model for a series of molecules not included in the training set was improved by the addition of a zinc indicator variable which explicitly defined the nature of the zinc-ligand interaction, an effect not observed within the thermolysin series. The effects of additional parameters, such as torsional degrees of freedom and the change in conformational enthalpy, ΔH_conform = H_aligned − H_min, were also examined. Experimentally derived alignment rules based on known structures of three-dimensional complexes produced predictive correlations for thermolysin inhibitors comparable, but not superior, to the correlations for ACE inhibitors based on alignment rules which were computationally deduced. The use of the active analog approach to determine active site geometries in the absence of structural data on the receptor is strongly supported by these results. Additionally, the correlations indicate that 3-D QSARs based on alignment rules derived from structure-activity data alone can produce statistically significant predictive correlations for quite diverse, noncongeneric compounds.