Linking CoMFA and protein homology models of enzyme-inhibitor interactions: an application to non-steroidal aromatase inhibitors

An approach to compare quantitatively a ligand-based (CoMFA) model and an enzyme active site model was investigated. The active site of the cytochrome P450 human aromatase was constructed by homology modeling techniques and two structurally different non-steroidal aromatase inhibitors were docked into it. A CoMFA model was then developed on a related series of non-steroidal inhibitors by correlating their inhibitory activity (expressed as -log IC50 values) versus only 11 steric descriptors (i.e. C-sp3-ligand steric interaction energies). The resulting 3D-QSAR coefficients (II) and the steric field values of the aromatase active site calculated at the same points of the CoMFA lattice (i.e. eleven C-sp3-protein steric interaction energies) were pair-wise compared. Specifically, when a positive coefficient was associated with a negative or low (< 5 kcal/mol) value of the protein steric field or, alternatively, a negative coefficient was associated with a large positive value of the protein steric field we recorded as many matches. When a 3D-QSAR coefficient did not correspond to the protein steric potential in the sense described above we considered that point as a mis-matching point. In our view, in spite of several limitations, such a comparison represents a valuable criterion to evaluate quantitatively how convergent are the results from a 3D-QSAR CoMFA model and a homology-built protein 3D structure. (C) 2000 Elsevier Science Ltd. All rights reserved.