CYP Enzyme Kinetics

The Michaelis constant (K_m), maximum metabolic rate (V_max), and intrinsic clearance (CL_int) are important kinetic parameters for characterizing the activities of cytochrome P450 enzymes (CYPs) which constitute a superfamily of hemoproteins involved in the drug metabolism in the human body.

Predictor's Metabolism Module provides a SOM-specific human P450 enzyme kinetic models package that includes K_m, V_max, and CL_int models for rCYPs 1A2, 2C9, 2C19, 2D6, 3A4, and for CYP3A4 assayed in human liver microsomes (HLM). There is also a separate model for overall CYP clearance in HLM. The models are ANNEs incorporating rapidly calculated atomic and molecular descriptors representing charge, reactivity, steric effects, and local environment of the metabolic site. K_m, V_max, and CL_int models were developed independently from separate data sets.

Experimental K_m, V_max, and CL_int data were compiled from the literature with careful examination of the original articles. The data set contains 77-165 metabolic reactions with known pathways and regiochemical positions for each enzyme (101 for 1A2, 83 for 2C9, 77 for 2C19, 115 for 2D6, and 165 for 3A4), with kinetic parameters measured from in vitro metabolic studies on cloned virus-infected cells expressing human enzyme-specific microsomes. In addition, the data set includes 138 human liver microsomal CYP3A4-mediated metabolic reactions (Bu 2006). Note that experimental results obtained with recombinant enzyme can differ substantially from results obtained with liver microsomes due to differences in P450 reductase: CYP ratios and cytochrome b5 content (Rodrigues 1997).

For each model, 15-20% of the data set was set aside for use as an external test set.

Kinetic parameter units

All kinetics models were built in log space, and the displayed value is the antilog of the model output. Because K_m, V_max, and CL_int predictions are used in many different kinds of physiological, pharmacokinetic, and pharmacodynamic simulations to support risk assessment and decision-making, likely users differ in terms of their needs. The output units reflect that diversity of end use:

• K_m is reported in μM for all models

• V_max is reported in nmol/min/mg HLM protein for the CYP3A4_HLM_Vmax model and in nmol/min/nmol P450 for the recombinant CYP V_max models.

• Intrinsic clearances (CL_int) are expressed in μL/min/mg HLM protein for all models, values that have been scaled from the rCYP data based on the average expression level of each CYP in HLM. The scaling factors used are taken from Inoue (2006) and are included in the .mtb file provided for each model in your App Data folder. The files are user-editable, should you find it necessary or desirable to use other expression levels.

Reconciling kinetic parameter predictions

The models for each of three kinetic parameters estimated by ADMET Predictor are built independently, so they have different complexities and generally use different descriptors. Nonetheless, the endpoints they predict are related algebraically:

CL_int = V_max/K_m by definition

V_max = CL_int x K_m

K_m = V_max/CL_int

It is crucial to understand, though, that these constants are calculated at the atomic level, per each CYP site in a given molecule.

In 2013 we have implemented “reconciliation” strategy to ensure that the CL_int = V_max / K_m relationship holds for all atomic and molecular predictions. First, three independent predictive models dCLint, dVmax, and dKm (“d” stands for “direct”) were developed using their respective data sets. Naturally, the expected constraint was not satisfied for these predictions. To rectify it the following mathematical manipulations were made per atom:

Where, the reconciled rCL_int and rV_max were respective geometric averages of the direct predictions. Atomic rKm = dKm, i.e., it remained unchanged.

Molecular level Michaelis-Menten constants mCL_int and mV_max are then obtained by summing reconciled atomic contributions:

The molecular level mK_m, on the other hand, cannot be obtained by simple summation. Instead, it was defined as the ratio of reconciled molecular quantities:

The above procedure automatically ensured the strict adherence to the CL_int definition for all predictions at both atomic and molecular levels.

Performance statistics

Predicted atomic and molecular attributes are accessible via the Atomic Properties Display, the ADMET Predictor spreadsheet and through MedChem Designer. Molecular V_max and CL_int values are summed across the respective atomic predictions and an effective molecular K_m is calculated from their ratio.

• For K_m models, R² is 0.626-0.864 and test set RMSE is 0.422-0.554 log units.

• For V_max models, R² is 0.470-0.830 and test set RMSE is 0.465-0.566 log units.

• For CL_int models, R² is 0.647-0.861 and test set RMSE is 0.430-0.649 log units.

Performance plots for each set of CYP kinetic models are shown below. Note that CYP3A4_HLM_Km and CYP3A4_HLM_CLint are corrected for microsomal binding, whereas the other (rCYP) models are not.