Computational Approaches That Predict Metabolic Intermediate Complex Formation with CYP3A4 (+b5)

Some mechanism-based inhibitors cause irreversible inhibition by forming a metabolic intermediate complex (MIC) with cytochrome P450. In the present study, 54 molecules (substrates of CYP3A and amine-containing compounds that are not known substrates of CYP3A) were spectrophotometrically assessed for their propensity to cause MIC formation with recombinant CYP3A4 (+b5). Comparisons of common physicochemical properties showed that mean (±S.D.) mol. wt. of MIC-forming compounds was significantly greater than mean mol. wt. of non-MIC-forming compounds, 472 (±173) versus 307 (±137), respectively. Computational pharmacophores, logistic regression, and recursive partitioning (RP) approaches were applied to predict MIC formation from molecular structure and to generate a quantitative structure activity relationship. A pharmacophore built with SKF-525A (2-diethylaminoethyl 2:2-diphenylvalerate hydrochloride), erythromycin, amprenavir, and norverapamil indicated that four hydrophobic features and a hydrogen bond acceptor were important for these MIC-forming compounds. Two different RP methods using either simple descriptors or 2D augmented atom descriptors indicated that hydro-phobic and hydrogen bond acceptor features were required for MIC formation. Both of these RP methods correctly predicted the MIC formation status with CYP3A4 for 10 of 12 literature molecules in an independent test set. Logistic multiple regression and a third classification tree model predicted 11 of 12 molecules correctly. Both models possessed a hydrogen bond acceptor and represent an approach for predicting CYP3A4 MIC formation that can be improved using more data and molecular descriptors. The preliminary pharmacophores provide structural insights that complement those for CYP3A4 inhibitors and substrates.