A Machine Learning Framework to Improve Rat Clearance Predictions and Inform Physiologically Based Pharmacokinetic Modeling.

During drug discovery and development, achieving appropriate pharmacokinetics is key to establishment of the efficacy and safety of new drugs. Physiologically based pharmacokinetic (PBPK) models integrating in vitro-to-in vivo extrapolation have become an essential in silico tool to achieve this goal. In this context, the most important and probably most challenging pharmacokinetic parameter to estimate is the clearance. Recent work on high-throughput PBPK modeling during drug discovery has shown that a good estimate of the unbound intrinsic clearance (CLint,u,) is the key factor for useful PBPK application. In this work, three different machine learning-based strategies were explored to predict the rat CLint,u as the input into PBPK. Therefore, in vivo and in vitro data was collected for a total of 2639 proprietary compounds. The strategies were compared to the standard in vitro bottom-up approach. Using the well-stirred liver model to back-calculate in vivo CLint,u from in vivo rat clearance and then training a machine learning model on this CLint,u led to more accurate clearance predictions (absolute average fold error (AAFE) 3.1 in temporal cross-validation) than the bottom-up approach (AAFE 3.6-16, depending on the scaling method) and has the advantage that no experimental in vitro data is needed. However, building a machine learning model on the bias between the back-calculated in vivo CLint,u and the bottom-up scaled in vitro CLint,u also performed well. For example, using unbound hepatocyte scaling, adding the bias prediction improved the AAFE in the temporal cross-validation from 16 for bottom-up to 2.9 together with the bias prediction. Similarly, the log Pearson r2 improved from 0.1 to 0.29. Although it would still require in vitro measurement of CLint,u., using unbound scaling for the bottom-up approach, the need for correction of the fu,inc by fu,p data is circumvented. While the above-described ML models were built on all data points available per approach, it is discussed that evaluation comparison across all approaches could only be performed on a subset because ca. 75% of the molecules had missing or unquantifiable measurements of the fraction unbound in plasma or in vitro unbound intrinsic clearance, or they dropped out due to the blood-flow limitation assumed by the well-stirred model. Advantageously, by predicting CLint,u as the input into PBPK, existing workflows can be reused and the prediction of the in vivo clearance and other PK parameters can be improved.

Investigation of simplified physiologically-based pharmacokinetic models in rat and human.

Whole-body physiologically-based pharmacokinetic (PBPK) models have many applications in drug research and development. It is often necessary to inform these models with animal or clinical data, updating model parameters, and making the model more predictive for future applications. This provides an opportunity and a challenge given the large number of parameters of such models. The aim of this work was to propose new mechanistic model structures with reduced complexity allowing for parameter optimization. These models were evaluated for their ability to estimate realistic values for unbound tissue to plasma partition coefficients (Kpu) and simulate observed pharmacokinetic (PK) data. Two approaches are presented: using either established kinetic lumping methods based on tissue time constants (drug-dependent) or a novel clustering analysis to identify tissues sharing common Kpu values or Kpu scalars based on similarities of tissue composition (drug-independent). PBPK models derived from these approaches were assessed using PK data of diazepam in rats and humans. Although the clustering analysis produced minor differences in tissue grouping depending on the method used, two larger groups of tissues emerged. One including the kidneys, liver, spleen, gut, heart, and lungs, and another including bone, brain, muscle, and pancreas whereas adipose and skin were generally considered distinct. Overall, a subdivision into four tissue groups appeared most physiologically relevant in terms of tissue composition. Several models were found to have similar abilities to describe diazepam i.v. data as empirical models. Comparability of estimated Kpus to experimental Kpu values for diazepam was one criterion for selecting the appropriate PK model structure.

A "middle-out approach" for the prediction of human drug disposition from preclinical data using simplified physiologically based pharmacokinetic (PBPK) models.

Simplified physiologically based pharmacokinetic (PBPK) models using estimated tissue-to-unbound plasma partition coefficients (Kpus) were previously investigated by fitting them to in vivo pharmacokinetic (PK) data. After optimization with preclinical data, the performance of these models for extrapolation of distribution kinetics to human were evaluated to determine the best approach for the prediction of human drug disposition and volume of distribution (Vss) using PBPK modeling. Three lipophilic bases were tested (diazepam, midazolam, and basmisanil) for which intravenous PK data were available in rat, monkey, and human. The models with Kpu scalars using k-means clustering were generally the best for fitting data in the preclinical species and gave plausible Kpu values. Extrapolations of plasma concentrations for diazepam and midazolam using these models and parameters obtained were consistent with the observed clinical data. For diazepam and midazolam, the human predictions of Vss after optimization in rats and monkeys were better compared with the Vss estimated from the traditional PBPK modeling approach (varying from 1.1 to 3.1 vs. 3.7-fold error). For basmisanil, the sparse preclinical data available could have affected the model performance for fitting and the subsequent extrapolation to human. Overall, this work provides a rational strategy to predict human drug distribution using preclinical PK data within the PBPK modeling strategy.

Can We Predict Clinical Pharmacokinetics of Highly Lipophilic Compounds by Integration of Machine Learning or In Vitro Data into Physiologically Based Models? A Feasibility Study Based on 12 Development Compounds.

While high lipophilicity tends to improve potency, its effects on pharmacokinetics (PK) are complex and often unfavorable. To predict clinical PK in early drug discovery, we built human physiologically based PK (PBPK) models integrating either (i) machine learning (ML)-predicted properties or (ii) discovery stage in vitro data. Our test set was composed of 12 challenging development compounds with high lipophilicity (mean calculated log P 4.2), low plasma-free fraction (50% of compounds with fu,p < 1%), and low aqueous solubility. Predictions focused on key human PK parameters, including plasma clearance (CL), volume of distribution at steady state (Vss), and oral bioavailability (%F). For predictions of CL, the ML inputs showed acceptable accuracy and slight underprediction bias [an average absolute fold error (AAFE) of 3.55; an average fold error (AFE) of 0.95]. Surprisingly, use of measured data only slightly improved accuracy but introduced an overprediction bias (AAFE = 3.35; AFE = 2.63). Predictions of Vss were more successful, with both ML (AAFE = 2.21; AFE = 0.90) and in vitro (AAFE = 2.24; AFE = 1.72) inputs showing good accuracy and moderate bias. The %F was poorly predicted using ML inputs [average absolute prediction error (AAPE) of 45%], and use of measured data for solubility and permeability improved this to 34%. Sensitivity analysis showed that predictions of CL limited the overall accuracy of human PK predictions, partly due to high nonspecific binding of lipophilic compounds, leading to uncertainty of unbound clearance. For accurate predictions of %F, solubility was the key factor. Despite current limitations, this work encourages further development of ML models and integration of their results within PBPK models to enable human PK prediction at the drug design stage, even before compounds are synthesized. Further evaluation of this approach with more diverse chemical types is warranted.

LipMetE (Lipophilic Metabolism Efficiency) as a Simple Guide for Half-Life and Dosing Regimen Prediction of Oral Drugs.

The in vivo half-life is a key property of every drug molecule, as it determines dosing regimens, peak-to-trough ratios and often dose. However, half-life optimization can be challenging due to its multifactorial nature, with in vitro metabolic turnover, plasma protein binding and volume of distribution all impacting half-life. We here propose that the medicinal chemistry design parameter Lipophilic Metabolism Efficiency (LipMetE) can greatly simplify half-life optimization of neutral and basic compounds. Using mathematical transformations, examples from preclinical GABAA projects and clinical compounds with human pharmacokinetic data, we show that LipMetE is directly proportional to the logarithm of half-life. As the design parameter LipMetE can be swiftly calculated using the readily available parameters LogD, intrinsic clearance and fraction unbound in human liver microsomes or hepatocytes, this approach enables rational half-life optimization from the early stages of drug discovery projects.

Evaluation of the Success of High-Throughput Physiologically Based Pharmacokinetic (HT-PBPK) Modeling Predictions to Inform Early Drug Discovery.

Minimizing in vitro and in vivo testing in early drug discovery with the use of physiologically based pharmacokinetic (PBPK) modeling and machine learning (ML) approaches has the potential to reduce discovery cycle times and animal experimentation. However, the prediction success of such an approach has not been shown for a larger and diverse set of compounds representative of a lead optimization pipeline. In this study, the prediction success of the oral (PO) and intravenous (IV) pharmacokinetics (PK) parameters in rats was assessed using a "bottom-up" approach, combining in vitro and ML inputs with a PBPK model. More than 240 compounds for which all of the necessary inputs and PK data were available were used for this assessment. Different clearance scaling approaches were assessed, using hepatocyte intrinsic clearance and protein binding as inputs. In addition, a novel high-throughput PBPK (HT-PBPK) approach was evaluated to assess the scalability of PBPK predictions for a larger number of compounds in drug discovery. The results showed that bottom-up PBPK modeling was able to predict the rat IV and PO PK parameters for the majority of compounds within a 2- to 3-fold error range, using both direct scaling and dilution methods for clearance predictions. The use of only ML-predicted inputs from the structure did not perform well when using in vitro inputs, likely due to clearance miss predictions. The HT-PBPK approach produced comparable results to the full PBPK modeling approach but reduced the simulation time from hours to seconds. In conclusion, a bottom-up PBPK and HT-PBPK approach can successfully predict the PK parameters and guide early discovery by informing compound prioritization, provided that good in vitro assays are in place for key parameters such as clearance.

Translating pharmacology models effectively to predict therapeutic benefit.

Many in vitro and in vivo models are used in pharmacological research to evaluate the role of targeted proteins in a disease. Understanding the translational relevance and limitation of these models for analyzing a drug's disposition, pharmacokinetic/pharmacodynamic (PK/PD) profile, mechanism, and efficacy, is essential when selecting the most appropriate model of the disease of interest and predicting clinically efficacious doses of the investigational drug. Selected animal models used in ophthalmology, infectious diseases, oncology, autoimmune diseases, and neuroscience are reviewed here. Each area has specific challenges around translatability and determination of an efficacious dose: new patient-specific dosing methods may help overcome these limitations.

Pediatric Pharmacokinetics and Dose Predictions: A Report of a Satellite Meeting to the 10th Juvenile Toxicity Symposium.

On April 24, 2019, a symposium on Pediatric Pharmacokinetics and Dose Predictions was held as a satellite meeting to the 10th Juvenile Toxicity Symposium. This symposium brought together scientists from academia, industry, and clinical research organizations with the aim to update each other on the current knowledge on pediatric drug development. Through more knowledge on specific ontogeny profiles of drug metabolism and transporter proteins, integrated into physiologically-based pharmacokinetic (PBPK) models, we have gained a more integrated understanding of age-related differences in pharmacokinetics (PKs), Relevant examples were presented during the meeting. PBPK may be considered the gold standard for pediatric PK prediction, but still it is important to know that simpler methods, such as allometry, allometry combined with maturation function, functions based on the elimination pathway, or linear models, also perform well, depending on the age range or the mechanisms involved. Knowledge from different methods and information sources should be combined (e.g., microdosing can reveal early read-out of age-related differences in exposure), and such results can be a value to verify models. To further establish best practices for dose setting in pediatrics, more in vitro and in vivo research is needed on aspects such as age-related changes in the exposure-response relationship and the impact of disease on PK. New information coupled with the refining of model-based drug development approaches will allow faster targeting of intended age groups and allow more efficient design of pediatric clinical trials.

Correction to: Use of Physiologically Based Pharmacokinetic (PBPK) Modeling for Predicting Drug-Food Interactions: an Industry Perspective.

An Erratum to this paper has been published: https://doi.org/10.1208/s12248-020-00535-z.

Use of Physiologically Based Pharmacokinetic (PBPK) Modeling for Predicting Drug-Food Interactions: an Industry Perspective.

The effect of food on pharmacokinetic properties of drugs is a commonly observed occurrence affecting about 40% of orally administered drugs. Within the pharmaceutical industry, significant resources are invested to predict and characterize a clinically relevant food effect. Here, the predictive performance of physiologically based pharmacokinetic (PBPK) food effect models was assessed via de novo mechanistic absorption models for 30 compounds using controlled, pre-defined in vitro, and modeling methodology. Compounds for which absorption was known to be limited by intestinal transporters were excluded in this analysis. A decision tree for model verification and optimization was followed, leading to high, moderate, or low food effect prediction confidence. High (within 0.8- to 1.25-fold) to moderate confidence (within 0.5- to 2-fold) was achieved for most of the compounds (15 and 8, respectively). While for 7 compounds, prediction confidence was found to be low (> 2-fold). There was no clear difference in prediction success for positive or negative food effects and no clear relationship to the BCS category of tested drug molecules. However, an association could be demonstrated when the food effect was mainly related to changes in the gastrointestinal luminal fluids or physiology, including fluid volume, motility, pH, micellar entrapment, and bile salts. Considering these findings, it is recommended that appropriately verified mechanistic PBPK modeling can be leveraged with high to moderate confidence as a key approach to predicting potential food effect, especially related to mechanisms highlighted here.

Global Sensitivity Analysis of the Rodgers and Rowland Model for Prediction of Tissue: Plasma Partitioning Coefficients: Assessment of the Key Physiological and Physicochemical Factors That Determine Small-Molecule Tissue Distribution.

In physiologically based pharmacokinetic (PBPK) modelling, the large number of input parameters, limited amount of available data and the structural model complexity generally hinder simultaneous estimation of uncertain and/or unknown parameters. These parameters are generally subject to estimation. However, the approaches taken for parameter estimation vary widely. Global sensitivity analyses are proposed as a method to systematically determine the most influential parameters that can be subject to estimation. Herein, a global sensitivity analysis was conducted to identify the key drug and physiological parameters influencing drug disposition in PBPK models and to potentially reduce the PBPK model dimensionality. The impact of these parameters was evaluated on the tissue-to-unbound plasma partition coefficients (Kpus) predicted by the Rodgers and Rowland model using Latin hypercube sampling combined to partial rank correlation coefficients (PRCC). For most drug classes, PRCC showed that LogP and fraction unbound in plasma (fup) were generally the most influential parameters for Kpu predictions. For strong bases, blood:plasma partitioning was one of the most influential parameter. Uncertainty in tissue composition parameters had a large impact on Kpu and Vss predictions for all classes. Among tissue composition parameters, changes in Kpu outputs were especially attributed to changes in tissue acidic phospholipid concentrations and extracellular protein tissue:plasma ratio values. In conclusion, this work demonstrates that for parameter estimation involving PBPK models and dimensionality reduction purposes, less influential parameters might be assigned fixed values depending on the parameter space, while influential parameters could be subject to parameters estimation.

Dissolution and Translational Modeling Strategies Toward Establishing an In Vitro-In Vivo Link-a Workshop Summary Report.

This publication summarizes the proceedings of day 2 of a 3-day workshop on "Dissolution and Translational Modeling Strategies Enabling Patient-Centric Product Development." Patient-centric drug product development from a drug product quality perspective necessitates the establishment of clinically relevant drug product specifications via an in vitro-in vivo link. Modeling and simulation offer a path to establish this link; in this regard, physiologically based modeling has been implemented successfully to support regulatory decision-making and drug product labeling. In this manuscript, case studies of physiologically based biopharmaceutics modeling (PBBM) applied to drug product quality are presented and summarized. These case studies exemplify a possible path to achieve an in vitro-in vivo link and encompass (a) development of biopredictive dissolution methods to support biowaivers, (b) model-informed formulation selection, (c) predicting clinical formulation performance, and (d) defining a safe space for regulatory flexibility via virtual bioequivalence (BE). Workflows for the development and verification of absorption models/PBBM and for the establishment of a safe space using dissolution as an input are described with examples. Breakout session discussions on topics, such as current challenges and some best practices in model development and verification, are included as part of the Supplementary material.

Importance of target-mediated drug disposition for small molecules.

Target concentration is typically not considered in drug discovery. However, if targets are expressed at relatively high concentrations and compounds have high affinity, such that most of the drug is bound to its target, in vitro screens can give unreliable information on compound affinity. In vivo, a similar situation will generate pharmacokinetic (PK) profiles that deviate greatly from those normally expected, owing to target binding affecting drug distribution and clearance. Such target-mediated drug disposition (TMDD) effects on small molecules have received little attention and might only become apparent during clinical trials, with the potential for data misinterpretation. TMDD also confounds human microdosing approaches by providing therapeutically unrepresentative PK profiles. Being aware of these phenomena will improve the likelihood of successful drug discovery and development.

Development of a Novel Simplified PBPK Absorption Model to Explain the Higher Relative Bioavailability of the OROS® Formulation of Oxybutynin.

A new minimal Segmented Transit and Absorption model (mSAT) model has been recently proposed and combined with intrinsic intestinal effective permeability (P eff,int ) to predict the regional gastrointestinal (GI) absorption (f abs ) of several drugs. Herein, this model was extended and applied for the prediction of oral bioavailability and pharmacokinetics of oxybutynin and its enantiomers to provide a mechanistic explanation of the higher relative bioavailability observed for oxybutynin's modified-release OROS® formulation compared to its immediate-release (IR) counterpart. The expansion of the model involved the incorporation of mechanistic equations for the prediction of release, transit, dissolution, permeation and first-pass metabolism. The predicted pharmacokinetics of oxybutynin enantiomers after oral administration for both the IR and OROS® formulations were in close agreement with the observed data. The predicted absolute bioavailability for the IR formulation was within 5% of the observed value, and the model adequately predicted the higher relative bioavailability observed for the OROS® formulation vs. the IR counterpart. From the model predictions, it can be noticed that the higher bioavailability observed for the OROS® formulation was mainly attributable to differences in the intestinal availability (F G ) rather than due to a higher colonic f abs , thus confirming previous hypotheses. The predicted f abs was almost 70% lower for the OROS® formulation compared to the IR formulation, whereas the F G was almost eightfold higher than in the IR formulation. These results provide further support to the hypothesis of an increased F G as the main factor responsible for the higher bioavailability of oxybutynin's OROS® formulation vs. the IR.

Translating Human Effective Jejunal Intestinal Permeability to Surface-Dependent Intrinsic Permeability: a Pragmatic Method for a More Mechanistic Prediction of Regional Oral Drug Absorption.

Regional intestinal effective permeability (P(eff)) values are key for the understanding of drug absorption along the whole length of the human gastrointestinal (GI) tract. The distal regions of the GI tract (i.e. ileum, ascending-transverse colon) represent the main sites for GI absorption when there is incomplete absorption in the upper GI tract, e.g. for modified release formulations. In this work, a new and pragmatic method for the estimation of (passive) intestinal permeability in the different intestinal regions is being proposed, by translating the observed differences in the available mucosal surface area along the human GI tract into corrections of the historical determined jejunal P(eff) values. These new intestinal P(eff) values or "intrinsic" P(eff)(P(eff,int)) were subsequently employed for the prediction of the ileal absorption clearance (CL(abs,ileum)) for a set of structurally diverse compounds. Additionally, the method was combined with a semi-mechanistic absorption PBPK model for the prediction of the fraction absorbed (f(abs)). The results showed that P(eff,int) can successfully be employed for the prediction of the ileal CL(abs) and the f(abs). P(eff,int) also showed to be a robust predictor of the f(abs) when the colonic absorption was allowed in the PBPK model, reducing the overprediction of f(abs) observed for lowly permeable compounds when using the historical P(eff) values. Due to its simplicity, this approach provides a useful alternative for the bottom-up prediction of GI drug absorption, especially when the distal GI tract plays a crucial role for a drug's GI absorption.

Analysis of the impact of controlled release formulations on oral drug absorption, gut wall metabolism and relative bioavailability of CYP3A substrates using a physiologically-based pharmacokinetic model.

Controlled release (CR) formulations are usually designed to achieve similar exposure (AUC) levels as the marketed immediate release (IR) formulation. However, the AUC is often lower following CR compared to IR formulations. There are a few exceptions when the CR formulations have shown higher AUC. This study investigated the impact of CR formulations on oral drug absorption and CYP3A4-mediated gut wall metabolism. A review of the current literature on relative bioavailability (Frel) between CR and IR formulations of CYP3A substrates was conducted. This was followed by a systematic analysis to assess the impact of the release characteristics and the drug-specific factors (including metabolism and permeability) on oral bioavailability employing a physiologically-based pharmacokinetic (PBPK) modelling and simulation approach. From the literature review, only three CYP3A4 substrates showed higher Frel when formulated as CR. Several scenarios were investigated using the PBPK approach; in most of them, the oral absorption of CR formulations was lower as compared to the IR formulations. However, for highly permeable compounds that were CYP3A4 substrates the reduction in absorption was compensated by an increase in the fraction that escapes from first pass metabolism in the gut wall (FG), where the magnitude was dependent on CYP3A4 affinity. The systematic simulations of various interplays between different parameters demonstrated that BCS class 1 highly-cleared CYP3A4 substrates can display up to 220% higher relative bioavailability when formulated as CR compared to IR, in agreement with the observed data collected from the literature. The results and methodology of this study can be employed during the formulation development process in order to optimize drug absorption, especially for CYP3A4 substrates.

The use of ROC analysis for the qualitative prediction of human oral bioavailability from animal data.

PURPOSE: To develop and evaluate a tool for the qualitative prediction of human oral bioavailability (Fhuman) from animal oral bioavailability (Fanimal) data employing ROC analysis and to identify the optimal thresholds for such predictions. METHODS: A dataset of 184 compounds with known Fhuman and Fanimal in at least one species (mouse, rat, dog and non-human primates (NHP)) was employed. A binary classification model for Fhuman was built by setting a threshold for high/low Fhuman at 50%. The thresholds for high/low Fanimal were varied from 0 to 100 to generate the ROC curves. Optimal thresholds were derived from 'cost analysis' and the outcomes with respect to false negative and false positive predictions were analyzed against the BDDCS class distributions. RESULTS: We successfully built ROC curves for the combined dataset and per individual species. Optimal Fanimal thresholds were found to be 67% (mouse), 22% (rat), 58% (dog), 35% (NHP) and 47% (combined dataset). No significant trends were observed when sub-categorizing the outcomes by the BDDCS. CONCLUSIONS: Fanimal can predict high/low Fhuman with adequate sensitivity and specificity. This methodology and associated thresholds can be employed as part of decisions related to planning necessary studies during development of new drug candidates and lead selection.

Animal versus human oral drug bioavailability: do they correlate?

Oral bioavailability is a key consideration in development of drug products, and the use of preclinical species in predicting bioavailability in human has long been debated. In order to clarify whether any correlation between human and animal bioavailability exist, an extensive analysis of the published literature data was conducted. Due to the complex nature of bioavailability calculations inclusion criteria were applied to ensure integrity of the data. A database of 184 compounds was assembled. Linear regression for the reported compounds indicated no strong or predictive correlations to human data for all species, individually and combined. The lack of correlation in this extended dataset highlights that animal bioavailability is not quantitatively predictive of bioavailability in human. Although qualitative (high/low bioavailability) indications might be possible, models taking into account species-specific factors that may affect bioavailability are recommended for developing quantitative prediction.