Cytochrome P450 (P450) 3A4 is the most abundant human P450 enzyme and has broad selectivity for substrates. The enzyme can show marked catalytic regioselectivity and unusual patterns of homotropic and heterotropic cooperativity, for which several models have been proposed. Spectral titration studies indicated one binding site for the drug indinavir (M(r) 614), a known substrate and inhibitor. Several C-terminal aminated peptides, including the model morphiceptin (YPFP-NH(2)), bind with spectral changes indicative of Fe-NH(2) bonding. The binding of the YPFP-NH(2) N-terminal amine and the influence of C-terminal modification on binding argue that the entire molecule (M(r) 521) fits within P450 3A4. YPFP-NH(2) was not oxidized by P450 3A4 but blocked binding of the substrates testosterone and midazolam, with K(i) values similar to the spectral binding constant (K(s)) for YPFP-NH(2). YPFP-NH(2) inhibited the oxidations of several typical P450 substrates with K(i) values 10-fold greater than the K(s) for binding YPFP-NH(2) and its K(i) for inhibiting substrate binding. The n values for cooperativity of these oxidations were not altered by YPFP-NH(2). YPFP-NH(2) inhibited the oxidations of midazolam at two different positions (1'- and 4-) with 20-fold different K(i) values. The differences in the K(i) values for blocking the binding to ferric P450 3A4 and the oxidation of several substrates may be attributed to weaker binding of YPFP-NH(2) to ferrous P450 3A4 than to the ferric form. The ferrous protein can be considered a distinct form of the enzyme in binding and catalysis because many substrates (but not YPFP-NH(2)) facilitate reduction of the ferric to ferrous enzyme. Our results with these peptides are considered in the context of several proposed models. A P450 3A4 model based on these peptide studies contains at least two and probably three distinct ligand sites, with testosterone and alpha-naphthoflavone occupying distinct sites. Midazolam appears to be able to bind to P450 3A4 in two modes, one corresponding to the testosterone binding mode and one postulated to reflect binding in a third site, distinct from both testosterone and alpha-naphthoflavone. The work with indinavir and YPFP-NH(2) also argues that room should be present in P450 3A4 to bind more than one smaller ligand in the "testosterone" site, although no direct evidence for such binding exists. Although this work with peptides provides evidence for the existence of multiple ligand binding sites, the results cannot be used to indicate their juxtaposition, which may vary through the catalytic cycle.

No abstract

The in vitro–in vivo extrapolation (IVIVE) approach for predicting total plasma clearance (CLtot) has been widely used to rank order compounds early in discovery. More recently, a computational machine learning approach utilizing physicochemical descriptors and fingerprints calculated from chemical structure information has emerged, enabling virtual predictions even earlier in discovery. Previously, this approach focused more on in vitro intrinsic clearance (CLint) prediction. Herein, we directly compare these two approaches for predicting CLtot in rats. A structurally diverse set of 1114 compounds with known in vivo CLtot, in vitro CLint, and plasma protein binding was used as the basis for this evaluation. The machine learning models were assessed by validation approaches using the time- and cluster-split training and test sets, and five-fold cross validation. Assessed by five-fold validation, the random forest regression (RF) and radial basis function (RBF) models demonstrated better prediction performance in eight attempted machine learning models. The CLtot values predicted by the RF and RBF models were within two-fold of the observed values for 67.7 and 71.9% of cluster-split test set compounds, respectively, while the predictivity was worse in the time-split dataset. The predictivity of both models tended to be improved by incorporating in vitro parameters, unbound fraction in plasma (f u,p), and CLint. CLtot prediction utilizing in vitro CLint and the well-stirred model, correcting for the fraction unbound in blood, was substantially worse compared to machine learning approaches for the same cluster-split test set. The reason that CLtot is underestimated by IVIVE is not fully explained by considering the calculated microsomal unbound fraction (cfu,mic), extended clearance classification system (ECCS), and omitting high clearance compounds in excess of hepatic blood flow. The analysis suggests that in silico machine learning models may have the power to reduce reliance on or replace in vitro and in vivo studies for chemical structure optimization in early drug discovery.

Accurate prediction of oral pharmacokinetics remains challenging. This study investigated quantitative approaches for the prediction of the area under the plasma concentration−time curve after oral administration (AUC p,oral ) to rats using the in vitro−in vivo extrapolation (IVIVE), in silico model using machine learning approaches and the combination of the in silico model and in vitro data. A set of 595 structurally diverse compounds with determined AUC p,oral at 1 mg/kg, in vitro intrinsic clearance (CL int ), an unbound fraction in plasma (f u,p ) in rats, and kinetic solubility at pH 6.8 was used for this assessment. Prediction models developed by two different types of machine learning techniques (i.e., random forest regression and Gaussian processes) were evaluated using three validation methods implementing the time and cluster-split training and test set and fivefold cross-validation. The developed machine learning models have a square of correlation coefficient (R 2 ) in the range of 0.381−0.685 with 33−45% of the compounds being predicted within 2fold of the observed AUC p,oral value. The predictivity was improved by incorporating CL int , f u,p , and solubility as explanatory variables with R 2 = 0.554−0.743. In cases where extraction by the liver is the main elimination pathway and intestinal extraction is negligible, AUC p,oral can be expressed by dose, CL int , and f u,p based on a well-stirred model. By using this conventional IVIVE approach, only 1.7−5.0% of compounds were predicted within the 2-fold error with R 2 = 0.354−0.487. Two empirical scaling factors (ESFs) determined by linear regression analysis and machine learning approaches improved the predictivity of AUC p,oral with 33−44% predicted within twofold variability. The IVIVE using ESF predicted by random forest regression showed better predictivity of AUC p,oral with R 2 = 0.471−0.618, while it still showed lower predictivity than machine learning approaches applied directly to AUC p,oral prediction. This study demonstrated that the combination of in silico and in vitro parameters is useful to improve the predictivity of the machine learning model for rat AUC p,oral and supports consideration for predicting AUC p,oral for human and other non-clinical species in a similar manner. KEYWORDS: machine learning, quantitative structure−activity relationship (QSAR), oral clearance prediction, in silico, plasma protein binding, in vitro−in vivo extrapolation (IVIVE), bottom-up approach, well-stirred model

Human pharmacokinetic (PK) predictions play a critical role in assessing the quality of potential clinical candidates where the accurate estimation of clearance, volume of distribution, bioavailability, and the plasma-concentration-time profiles are the desired end points. While many methods for conducting predictions utilize in vivo data, predictions can be conducted successfully from in vitro or in silico data, applying modeling and simulation techniques. This approach can be facilitated using commercially available prediction software such as GastroPlus which has been reported to accurately predict the oral PK profile of small drug-like molecules. Herein, case studies are described where GastroPlus modeling and simulation was employed using in silico or in vitro data to predict PK profiles in early discovery. The results obtained demonstrate the feasibility of adequately predicting plasma-concentration-time profiles with in silico derived as well as in vitro measured parameters and hence predicting PK profiles with minimal data. The applicability of this approach can provide key information enabling decisions on either dose selection, chemistry strategy to improve compounds, or clinical protocol design, thus demonstrating the value of modeling and simulation in both early discovery and exploratory development for predicting absorption and disposition profiles.