Perspectives on the use of machine learning for ADME prediction at AstraZeneca.

A drug's pharmacokinetic (PK) profile will determine its dose and the frequency of administration as well as the likelihood of observing any adverse drug reactions.It is important to understand these PK properties as early as possible in the drug discovery process, ideally, to accurately predict these prior to synthesising the molecule leading to significant improvements in efficiency.In this paper, we describe the approaches used within AstraZeneca to improve our ability of predicting the preclinical and human pharmacokinetic profiles of novel molecules using machine learning and artificial intelligence.We will show how combining chemical structure-based approaches with experimentally derived properties enables improved predictions of in vivo pharmacokinetics and can be extended to molecules that go beyond the classical Lipinski's rule-of-five space.We will also discuss how combining these in vitro and in vivo predictive models could ultimately improve our ability to predict the human outcome at the point of chemical design.

Extracellular vesicles as the next-generation modulators of pharmacokinetics and pharmacodynamics of medications and their potential as adjuvant therapeutics.

BACKGROUND AND MAIN BODY: Pharmacokinetics (PK) and pharmacodynamics (PD) are central concepts to guide the dosage and administration of drug therapies and are essential to consider for both healthcare professionals and researchers in therapeutic planning and drug discovery. PK/PD properties of a drug significantly influence variability in response to treatment, including therapeutic failure or excessive medication-related harm. Furthermore, suboptimal PK properties constitute a significant barrier to further development for some candidate treatments in drug discovery. This article describes how extracellular vesicles (EVs) affect different aspects of PK and PD of medications and their potential to modulate PK and PD properties to address problematic PK/PD profiles of drugs. We reviewed EVs' intrinsic effects on cell behaviours and medication responses. We also described how surface and cargo modifications can enhance EV functionalities and enable them as adjuvants to optimise the PK/PD profile of conventional medications. Furthermore, we demonstrated that various bioengineering strategies can be used to modify the properties of EVs, hence enhancing their potential to modulate PK and PD profile of medications. CONCLUSION: This review uncovers the critical role of EVs in PK and PD modulation and motivates further research and the development of assays to unfold EVs' full potential in solving PK and PD-related problems. However, while we have shown that EVs play a vital role in modulating PK and PD properties of medications, we postulated that it is essential to define the context of use when designing and utilising EVs in pharmaceutical and medical applications. HIGHLIGHTS: Existing solutions for pharmacokinetics and pharmacodynamics modulation are limited. Extracellular vesicles can optimise pharmacokinetics as a drug delivery vehicle. Biogenesis and administration of extracellular vesicles can signal cell response. The pharmaceutical potential of extracellular vesicles can be enhanced by surface and cargo bioengineering. When using extracellular vesicles as modulators of pharmacokinetics and pharmacodynamics, the 'context of use' must be considered.

After 50 Years of Hepatic Clearance Models, Where Should We Go from Here? Improvements and Implications for Physiologically Based Pharmacokinetic Modeling.

There is overwhelming preference for application of the unphysiologic, well-stirred model (WSM) over the parallel tube model (PTM) and dispersion model (DM) to predict hepatic drug clearance, CLH , despite that liver blood flow is dispersive and closer to the DM in nature. The reasoning is the ease in computation relating the hepatic intrinsic clearance ( CLint ), hepatic blood flow ( QH ), unbound fraction in blood ( fub ) and the transmembrane clearances ( CLin and CLef ) to CLH for the WSM. However, the WSM, being the least efficient liver model, predicts a lower EH that is associated with the in vitro CLint ( Vmax / Km ), therefore requiring scale-up to predict CLH in vivo. By contrast, the miniPTM, a three-subcompartment tank-in-series model of uniform enzymes, closely mimics the DM and yielded similar patterns for CLint versus EH , substrate concentration [S] , and KL / B , the tissue to outflow blood concentration ratio. We placed these liver models nested within physiologically based pharmacokinetic models to describe the kinetics of the flow-limited, phenolic substrate, harmol, using the WSM (single compartment) and the miniPTM and zonal liver models (ZLMs) of evenly and unevenly distributed glucuronidation and sulfation activities, respectively, to predict CLH For the same, given CLint ( Vmax and Km ), the WSM again furnished the lowest extraction ratio ( EH,WSM = 0.5) compared with the miniPTM and ZLM (>0.68). Values of EH,WSM were elevated to those for EH, PTM and EH, ZLM when the Vmax s for sulfation and glucuronidation were raised 5.7- to 1.15-fold. The miniPTM is easily manageable mathematically and should be the new normal for liver/physiologic modeling. SIGNIFICANCE STATEMENT: Selection of the proper liver clearance model impacts strongly on CLH predictions. The authors recommend use of the tank-in-series miniPTM (3 compartments mini-parallel tube model), which displays similar properties as the dispersion model (DM) in relating CLint and [ S ] to CLH as a stand-in for the DM, which best describes the liver microcirculation. The miniPTM is readily modified to accommodate enzyme and transporter zonation.

A Data Driven Strategy and Case Study for Implementation of Singlicate Analysis in Ligand Binding Assays Used for PK Quantitation.

Duplicate analysis has been a conventional practice in the industry for ligand-binding assays (LBA), particularly for plate-based platforms like Enzyme-linked immunosorbent assay (ELISA) and Meso Scale Discovery (MSD) assays. Recent whitepapers and guidance have opened a door to exploring the implementation of single-well (singlicate) analysis approach for LBAs. Although the bioanalytical industry has actively investigated the suitability of singlicate analysis, applications in supporting regulated LBA bioanalysis are limited. The primary reason for this limitation is the absence of appropriate strategy to facilitate the transition from duplicate to singlicate analysis. In this paper we present the first case study with our data-driven approach to implement singlicate analysis in a clinical pharmacokinetics (PK) plate based LBA assay with ISR data. The central aspect of this strategy is a head-to-head comparison with Precision and Accuracy assessment in both duplicate and singlicate formats as the initial stage of assay validation. Subsequently, statistical analysis is conducted to evaluate method variability in both precision and accuracy. The results of our study indicated that there was no impactful difference between duplicate vs singlicate, affirming the suitability of singlicate analysis for the remaining steps of PK assay validation. The validation results obtained through singlicate analysis demonstrated acceptable assay performance characteristics across all validation parameters, aligning with regulatory guidance. The validated PK assay in singlicate has been employed to support a Phase I study. The appropriateness of singlicate analyses is further supported by initial Incurred Sample Reanalysis (ISR) data in which 90.1% of ISR samples fall within the acceptable criteria.

[P-glycoprotein activity in vivo in older adults: physiological, -pathophysiological and pharmacokinetic interplay at the blood-brain barrier]. / Activité de la p-glycoprotéine in vivo chez la personne âgée : intrications physiologique, physiopathologique et pharmacocinétique au niveau de la barrière hémato-encéphalique.

p-glycoprotein (P-gp) is an efflux transporter of xenobiotic and endogenous compounds across the blood-brain barrier (BBB). P-gp plays an essential role by limiting passage of these compounds into the brain tissue. It is susceptible to drug-drug interactions when interactors drugs are co-administrated. The efficiency of P-gp may be affected by the aging process and the development of neurodegenerative diseases. Studying this protein in older adults is therefore highly relevant for all these reasons. Understanding P-gp activity in vivo is essential when considering the physiological, pathophysiological, and pharmacokinetic perspectives, as these aspects seem to be interconnected to some extent. In vivo exploration in humans is based on neuroimaging techniques, which have been improving over the last years. The advancement of exploration and diagnostic tools is opening up new prospects for understanding P-gp activity at the BBB.

Distribution Clearance: Significance and Underlying Mechanisms.

PURPOSE: Evaluation of distribution kinetics is a neglected aspect of pharmacokinetics. This study examines the utility of the model-independent parameter whole body distribution clearance (CLD) in this respect. METHODS: Since mammillary compartmental models are widely used, CLD was calculated in terms of parameters of this model for 15 drugs. The underlying distribution processes were explored by assessment of relationships to pharmacokinetic parameters and covariates. RESULTS: The model-independence of the definition of the parameter CLD allowed a comparison of distributional properties of different drugs and provided physiological insight. Significant changes in CLD were observed as a result of drug-drug interactions, transporter polymorphisms and a diseased state. CONCLUSION: Total distribution clearance CLD is a useful parameter to evaluate distribution kinetics of drugs. Its estimation as an adjunct to the model-independent parameters clearance and steady-state volume of distribution is advocated.

Application of physiologically based pharmacokinetic modeling of novel drugs approved by the U.S. food and drug administration.

Physiologically based pharmacokinetic (PBPK) models which can leverage preclinical data to predict the pharmacokinetic properties of drugs rapidly became an essential tool to improve the efficiency and quality of novel drug development. In this review, by searching the Application Review Files in Drugs@FDA, we analyzed the current application of PBPK models in novel drugs approved by the U.S. Food and Drug Administration (FDA) in the past five years. According to the results, 243 novel drugs were approved by the FDA from 2019 to 2023. During this period, 74 Application Review Files of novel drugs approved by the FDA that used PBPK models. PBPK models were used in various areas, including drug-drug interactions (DDI), organ impairment (OI) patients, pediatrics, drug-gene interaction (DGI), disease impact, and food effects. DDI was the most widely used area of PBPK models for novel drugs, accounting for 74.2 % of the total. Software platforms with graphical user interfaces (GUI) have reduced the difficulty of PBPK modeling, and Simcyp was the most popular software platform among applicants, with a usage rate of 80.5 %. Despite its challenges, PBPK has demonstrated its potential in novel drug development, and a growing number of successful cases provide experience learned for researchers in the industry.

Prediction method of pharmacokinetic parameters of small molecule drugs based on GCN network model.

CONTEXT: Accurately predicting plasma protein binding rate (PPBR) and oral bioavailability (OBA) helps to better reveal the absorption and distribution of drugs in the human body and subsequent drug design. Although machine learning models have achieved good results in prediction accuracy, they often suffer from insufficient accuracy when dealing with data with irregular topological structures. METHODS: In view of this, this study proposes a pharmacokinetic parameter prediction framework based on graph convolutional networks (GCN), which predicts the PPBR and OBA of small molecule drugs. In the framework, GCN is first used to extract spatial feature information on the topological structure of drug molecules, in order to better learn node features and association information between nodes. Then, based on the principle of drug similarity, this study calculates the similarity between small molecule drugs, selects different thresholds to construct datasets, and establishes a prediction model centered on the GCN algorithm. The experimental results show that compared with traditional machine learning prediction models, the prediction model constructed based on the GCN method performs best on PPBR and OBA datasets with an inter-molecular similarity threshold of 0.25, with MAE of 0.155 and 0.167, respectively. In addition, in order to further improve the accuracy of the prediction model, GCN is combined with other algorithms. Compared to using a single GCN method, the distribution of the predicted values obtained by the combined model is highly consistent with the true values. In summary, this work provides a new method for improving the rate of early drug screening in the future.

Predicting the Intravenous Pharmacokinetics of Covalent Drugs in Animals and Humans.

30 covalent drugs were used to assess clearance (CL) prediction reliability in animals and humans. In animals, marked CL underprediction was observed using cryopreserved hepatocytes or liver microsomes (LMs) supplemented for cytochrome P450 activity. Improved quantitative performance was observed by combining metabolic stability data from LMs and liver S9 fractions, the latter supplemented with reduced glutathione for glutathione transferase activity. While human LMs provided reliable human CL predictions, prediction statistics were improved further by incorporating S9 stability data. CL predictions with allometric scaling were less robust compared to in vitro drug metabolism methods; the best results were obtained using the fu-corrected intercept model. Human volume of distribution (Vd) was well predicted using allometric scaling of animal pharmacokinetic data; the most reliable results were achieved using simple allometric scaling of unbound Vd values. These results provide a quantitative framework to guide appropriate method selection for human PK prediction with covalent drugs.

Dosing Adjustments in Cases of Altered Plasma Protein Binding are Most Needed for Drugs with a Volume of Distribution Below 1.3 L/kg.

BACKGROUND: The present literature offers conflicting views on the importance of changes in plasma protein binding in clinical therapeutics. Furthermore, there are no methods to calculate a new dosing regimen when such changes occur. METHODS: Previous models developed by Balaz et al. and Greenblat et al. were used to calculate a plasma protein binding (PPB) score for individual drugs based on the volume of distribution for total concentration and the bound fraction of drug. The models were further used to calculate a new drug dosing interval for cases of altered plasma protein binding. The equations apply best for drugs with fast absorption and fast distribution; they can be used as approximations for drugs with slow distribution by using the volume of distribution at steady state and the rate constant of the elimination phase. RESULTS: The newly developed equations show that changes in plasma protein binding are relevant only for drugs with a positive PPB score; such drugs must have a volume of distribution for total concentration below 1.3 L/kg and high protein binding. It is further shown that the drug dosing interval should be reduced when the remaining fraction of plasma protein binding is below the PPB score. CONCLUSION: A new method to rank drugs according to the impact of changes in plasma protein binding on their pharmacokinetic profile was developed. The new method was applied to show that drugs with high PPB scores need reductions in their dosing interval when the level of protein binding decreases.

SAAM II: A general mathematical modeling rapid prototyping environment.

Simulation Analysis and Modeling II (SAAM II) is a graphical modeling software used in life sciences for compartmental model analysis, particularly, but not exclusively, appreciated in pharmacokinetics (PK) and pharmacodynamics (PD), metabolism, and tracer modeling. Its intuitive "circles and arrows" visuals allow users to easily build, solve, and fit compartmental models without the need for coding. It is suitable for rapid prototyping of models for complex kinetic analysis or PK/PD problems, and in educating students and non-modelers. Although it is straightforward in design, SAAM II incorporates sophisticated algorithms programmed in C to address ordinary differential equations, deal with complex systems via forcing functions, conduct multivariable regression featuring the Bayesian maximum a posteriori, perform identifiability and sensitivity analyses, and offer reporting functionalities, all within a single package. After 26 years from the last SAAM II tutorial paper, we demonstrate here SAAM II's updated applicability to current life sciences challenges. We review its features and present four contemporary case studies, including examples in target-mediated PK/PD, CAR-T-cell therapy, viral dynamics, and transmission models in epidemiology. Through such examples, we demonstrate that SAAM II provides a suitable interface for rapid model selection and prototyping. By enabling the fast creation of detailed mathematical models, SAAM II addresses a unique requirement within the mathematical modeling community.

Beyond the basics: A deep dive into parameter estimation for advanced PBPK and QSP models.

Physiologically-based pharmacokinetic (PBPK) models and quantitative systems pharmacology (QSP) models have contributed to drug development strategies. The parameters of these models are commonly estimated by capturing observed values using the nonlinear least-squares method. Software packages for PBPK and QSP modeling provide a range of parameter estimation algorithms. To choose the most appropriate method, modelers need to understand the basic concept of each approach. This review provides a general introduction to the key points of parameter estimation with a focus on the PBPK and QSP models, and the respective parameter estimation algorithms. The latter part assesses the performance of five parameter estimation algorithms - the quasi-Newton method, Nelder-Mead method, genetic algorithm, particle swarm optimization, and Cluster Gauss-Newton method - using three examples of PBPK and QSP modeling. The assessment revealed that some parameter estimation results were significantly influenced by the initial values. Moreover, the choice of algorithms demonstrating good estimation results heavily depends on factors such as model structure and the parameters to be estimated. To obtain credible parameter estimation results, it is advisable to conduct multiple rounds of parameter estimation under different conditions, employing various estimation algorithms.

Pharmacokinetics-Pharmacodynamics Modeling for Evaluating Drug-Drug Interactions in Polypharmacy: Development and Challenges.

Polypharmacy is commonly employed in clinical settings. The potential risks of drug-drug interactions (DDIs) can compromise efficacy and pose serious health hazards. Integrating pharmacokinetics (PK) and pharmacodynamics (PD) models into DDIs research provides a reliable method for evaluating and optimizing drug regimens. With advancements in our comprehension of both individual drug mechanisms and DDIs, conventional models have begun to evolve towards more detailed and precise directions, especially in terms of the simulation and analysis of physiological mechanisms. Selecting appropriate models is crucial for an accurate assessment of DDIs. This review details the theoretical frameworks and quantitative benchmarks of PK and PD modeling in DDI evaluation, highlighting the establishment of PK/PD modeling against a backdrop of complex DDIs and physiological conditions, and further showcases the potential of quantitative systems pharmacology (QSP) in this field. Furthermore, it explores the current advancements and challenges in DDI evaluation based on models, emphasizing the role of emerging in vitro detection systems, high-throughput screening technologies, and advanced computational resources in improving prediction accuracy.

A Combination of Machine Learning and PBPK Modeling Approach for Pharmacokinetics Prediction of Small Molecules in Humans.

PURPOSE: Recently, there has been rapid development in model-informed drug development, which has the potential to reduce animal experiments and accelerate drug discovery. Physiologically based pharmacokinetic (PBPK) and machine learning (ML) models are commonly used in early drug discovery to predict drug properties. However, basic PBPK models require a large number of molecule-specific inputs from in vitro experiments, which hinders the efficiency and accuracy of these models. To address this issue, this paper introduces a new computational platform that combines ML and PBPK models. The platform predicts molecule PK profiles with high accuracy and without the need for experimental data. METHODS: This study developed a whole-body PBPK model and ML models of plasma protein fraction unbound ( f up ), Caco-2 cell permeability, and total plasma clearance to predict the PK of small molecules after intravenous administration. Pharmacokinetic profiles were simulated using a "bottom-up" PBPK modeling approach with ML inputs. Additionally, 40 compounds were used to evaluate the platform's accuracy. RESULTS: Results showed that the ML-PBPK model predicted the area under the concentration-time curve (AUC) with 65.0 % accuracy within a 2-fold range, which was higher than using in vitro inputs with 47.5 % accuracy. CONCLUSION: The ML-PBPK model platform provides high accuracy in prediction and reduces the number of experiments and time required compared to traditional PBPK approaches. The platform successfully predicts human PK parameters without in vitro and in vivo experiments and can potentially guide early drug discovery and development.

First-order compartment model solutions - Exponential sums and beyond.

First-order compartment models are common tools for modelling many biological processes, including pharmacokinetics. Given the compartments and the transfer rates, solutions for the time-dependent quantity (or concentration) curves can normally be described by a sum of exponentials. This paper investigates cases that go beyond simple sums of exponentials. With specific relations between the transfer rate constants, two exponential rate constants can be equal, in which case the normal solution cannot be used. The conditions for this to occur are discussed, and advice is provided on how to circumvent these cases. An example of an analytic solution is given for the rare case where an exact equality is the expected result. Furthermore, for models with at least three compartments, cases exist where the solution to a real-valued model involves complex-valued exponential rate constants. This leads to solutions with an oscillatory element in the solution for the tracer concentration, i.e., there are cases where the solution is not a simple sum of (real-valued) exponentials but also includes sine and cosine functions. Detailed solutions for three-compartment cases are given. As a tentative conclusion of the analysis, oscillatory solutions appear to be tied to cases with a cyclic element in the model itself.

An update on microfluidic multi-organ-on-a-chip systems for reproducing drug pharmacokinetics: the current state-of-the-art.

INTRODUCTION: Advances in the accessibility of manufacturing technologies and iPSC-based modeling have accelerated the overall progress of organs-on-a-chip. Notably, the progress in multi-organ systems is not progressing with equal speed, indicating that there are still major technological barriers to overcome that may include biological relevance, technological usability as well as overall accessibility. AREAS COVERED: We here review the progress in the field of multi-tissue- and body-on-a-chip pre and post- SARS-CoV-2 pandemic and review five selected studies with increasingly complex multi-organ chips aiming at pharmacological studies. EXPERT OPINION: We discuss future and necessary advances in the field of multi-organ chips including how to overcome challenges regarding cell diversity, improved culture conditions, model translatability as well as sensor integrations to enable microsystems to cover organ-organ interactions in not only toxicokinetic but more importantly pharmacodynamic and -kinetic studies.

Model-based bioequivalence approach for sparse pharmacokinetic bioequivalence studies: Model selection or model averaging?

Conventional pharmacokinetic (PK) bioequivalence (BE) studies aim to compare the rate and extent of drug absorption from a test (T) and reference (R) product using non-compartmental analysis (NCA) and the two one-sided test (TOST). Recently published regulatory guidance recommends alternative model-based (MB) approaches for BE assessment when NCA is challenging, as for long-acting injectables and products which require sparse PK sampling. However, our previous research on MB-TOST approaches showed that model misspecification can lead to inflated type I error. The objective of this research was to compare the performance of model selection (MS) on R product arm data and model averaging (MA) from a pool of candidate structural PK models in MBBE studies with sparse sampling. Our simulation study was inspired by a real case BE study using a two-way crossover design. PK data were simulated using three structural models under the null hypothesis and one model under the alternative hypothesis. MB-TOST was applied either using each of the five candidate models or following MS and MA with or without the simulated model in the pool. Assuming T and R have the same PK model, our simulation shows that following MS and MA, MB-TOST controls type I error rates at or below 0.05 and attains similar or even higher power than when using the simulated model. Thus, we propose to use MS prior to MB-TOST for BE studies with sparse PK sampling and to consider MA when candidate models have similar Akaike information criterion.

Development of a predictive algorithm for the efficacy of half-life extension strategies.

A challenge in development of peptide and protein therapeutics is rapid elimination from the body, necessitating frequent dosing that may lead to toxicities and/or poor patient compliance. To solve this issue, there has been great investment into half-life extension of rapidly eliminated drugs using approaches such as albumin binding, fusion to albumin or Fc, or conjugation to polyethylene glycol. Despite clinical successes of half-life extension products, no clear relationship has been drawn between properties of drugs and the pharmacokinetic parameters of their half-life extended analogues. In this study, non-compartmentally derived pharmacokinetic parameters (half-life, clearance, volume of distribution) were collected for 186 half-life extended drugs and their unmodified parent molecules. Statistical testing and regression analysis was performed to evaluate relationships between pharmacokinetic parameters and a matrix of variables. Multivariate linear regression models were developed for each of the three pharmacokinetic parameters and model predictions were in good agreement with observed data with r2 values for each parameter being: half-life: 0.879, clearance: 0.820, volume of distribution: 0.937. Significant predictors for each parameter included the corresponding pharmacokinetic parameter of the parent drug and descriptors related to molecular weight. This model represents a useful tool for prediction of the potential benefits of half-life extension.

Commentary: Pharmacokinetic Theory Must Consider Published Experimental Data.

Recently, we have proposed simple methodology to derive clearance and rate constant equations, independent of differential equations, based on Kirchhoff's Laws, a common methodology from physics used to describe rate-defining processes either in series or parallel. Our approach has been challenged in three recent publications, two published in this journal, but notably what is lacking is that none evaluate experimental pharmacokinetic data. As reviewed here, manuscripts from our laboratory have evaluated published experimental data, demonstrating that the Kirchhoff's Laws approach explains (1) why all of the experimental perfused liver clearance data appear to fit the equation that was previously believed to be the well-stirred model, (2) why linear pharmacokinetic systemic bioavailability determinations can be greater than 1, (3) why renal clearance can be a function of drug input processes, and (4) why statistically different bioavailability measures may be found for urinary excretion versus systemic concentration measurements. Our most recent paper demonstrates (5) how the universally accepted steady-state clearance approach used by the field for the past 50 years leads to unrealistic outcomes concerning the relationship between liver-to-blood Kpuu and hepatic availability FH , highlighting the potential for errors in pharmacokinetic evaluations based on differential equations. The Kirchhoff's Laws approach is applicable to all pharmacokinetic analyses of quality experimental data, those that were previously adequately explained with present pharmacokinetic theory, and those that were not. The publications that have attempted to rebut our position do not address unexplained experimental data, and we show here why their analyses are not valid. SIGNIFICANCE STATEMENT: The Kirchhoff's Laws approach to deriving clearance equations for linear systems in parallel or in series, independent of differential equations, successfully describes published pharmacokinetic data that has previously been unexplained. Three recent publications claim to refute our proposed methodology; these publications only make theoretical arguments, do not evaluate experimental data, and never demonstrate that the Kirchhoff methodology provides incorrect interpretations of experimental pharmacokinetic data, including statistically significant data not explained by present pharmacokinetic theory. We demonstrate why these analyses are invalid.