A Novel Approach to Predicting P450 Mediated Drug Metabolism. CYP2D6 Catalyzed N-Dealkylation Reactions and Qualitative Metabolite Predictions Using a Combined Protein and Pharmacophore Model for CYP2D6

Abstract

A combined protein and pharmacophore model for cytochrome P450 2D6 (CYP2D6) has been extended with a second pharmacophore in order to explain CYP2D6 catalyzed N-dealkylation reactions. A group of 14 experimentally verified N-dealkylation reactions form the basis of this second pharmacophore. The combined model can now accommodate both the usual hydroxylation and O-demethylation reactions catalyzed by CYP2D6, as well as the less common N-dealkylation reactions. The combined model now contains 72 metabolic pathways catalyzed by CYP2D6 in 51 substrates. The model was then used to predict the involvement of CYP2D6 in the metabolism of a "test set" of seven compounds. Molecular orbital calculations were used to suggest energetically favorable sites of metabolism, which were then examined using modeling techniques. The combined model correctly predicted 6 of the 8 observed metabolites. For the well-established CYP2D6 metabolic routes, the predictive value of the current combined protein and pharmacophore model is good. Except for the highly unusual metabolism of procainamide and ritonavir, the known metabolites not included in the development of the model were all predicted by the current model. Two possible metabolites have been predicted by the current model, which have not been detected experimentally. In these cases, the model may be able to guide experiments. P450 models, like the one presented here, have wide applications in the drug design process which will contribute to the prediction and elimination of polymorphic metabolism and drug-drug interactions.

Computational Models for Cytochrome P450: A Predictive Electronic Model for Aromatic Oxidation and Hydrogen Atom Abstraction

“…4). However, qualitative models that approach these models have been described for CYP2D6 (de Groot et al, 1999). The model described here was generated by correcting the computational predictions with experimental results for microsomes and single expressed enzymes.…”

“…The required diversity is accomplished by families and subfamilies of enzymes with generally broad substrate specificities, a very reactive oxygenating species and a broad regioselectivity. Thus, for many reactions, the electronic features of the substrate are all that are required to predict regioselectivity (Grogan et al, 1992;Harris et al, 1992;de Groot et al, 1995de Groot et al, , 1999Yin et al, 1995).…”

The Molecular Basis of CYP2D6-Mediated N-Dealkylation: Balance between Metabolic Clearance Routes and Enzyme Inhibition

“…Since several substrates with approximately 10 Å between the nitrogen atom and the site of oxidation were also known, refined models were constructed including this intramolecular distance (Lewis et al, 1997). Even though the early pharmacophore models are rather crude, they have been successful in predicting the influence of CYP2D6 in the metabolism of xenobiotics as well as predicting possible sites of metabolism [e.g., 75% predictability (de Groot et al, 1999)]. …”

“…In drug discovery research it is of great interest to identify whether a compound is a substrate or not, and considerable effort has been put into the development of models to identify substrates and/or inhibitors of CYP2D6. Several pharmacophore models have been published identifying common features in CYP2D6 substrates and inhibitors (Koymans et al, 1992;de Groot et al, 1997de Groot et al, , 1999Lewis et al, 1997). The pharmacophore comprises basic nitrogen atom 5-7 or 10 Å from the site of oxidation, a flat hydrophobic region near the site of oxidation and a negative molecular electrostatic potential coplanar with this hydrophobic region (see examples in Fig.…”

Towards integrated ADME prediction: past, present and future directions for modelling metabolism by UDP-glucuronosyltransferases

“…This method is likely to prove most valuable when conjugative regioselectivity data, which has generally been characterised poorly in the past, is included. Advances in the prediction of CYP regioselectivity using quantum chemical descriptors [59,60] is in no small part due to the availability of training data derived from thorough characterisation of metabolite regioselectivity. Diligence in the characterisation of conjugative regioselectivity is required in order to realise similar advances with in silico UGT models.…”