Comparison of methods for the prediction of human clearance from hepatocyte intrinsic clearance for a set of reference compounds and an external evaluation set

Comparison of Human Long-Term Liver Models for Clearance Prediction of Slowly Metabolized Compounds

The accurate prediction of human clearance is an important task during drug development. The proportion of low clearance compounds has increased in drug development pipelines across the industry since such compounds may be dosed in lower amounts and at lower frequency. These type of compounds present new challenges to in vitro systems used for clearance extrapolation. In this study, we compared the accuracy of clearance predictions of suspension culture to four different long-term stable in vitro liver models, including HepaRG sandwich culture, the Hµrel stochastic co-culture, the Hepatopac micropatterned co-culture (MPCC), and a micro-array spheroid culture. Hepatocytes in long-term stable systems remained viable and active over several days of incubation. Although intrinsic clearance values were generally high in suspension culture, clearance of low turnover compounds could frequently not be determined using this method. Metabolic activity and intrinsic clearance values from HepaRG cultures were low and, consequently, many compounds with low turnover did not show significant decline despite long incubation times. Similarly, stochastic co-cultures occasionally failed to show significant turnover for multiple low and medium turnover compounds. Among the different methods, MPCCs and spheroids provided the most consistent measurements. Notably, all culture methods resulted in underprediction of clearance; this could, however, be compensated for by regression correction. Combined, the results indicate that spheroid culture as well as the MPCC system provide adequate in vitro tools for human extrapolation for compounds with low metabolic turnover. SIGNIFICANCE STATEMENT: In this study, we compared suspension cultures, HepaRG sandwich cultures, the Hµrel liver stochastic co-cultures, the Hepatopac micropatterned co-cultures (MPCC), and micro-array spheroid cultures for low clearance determination and prediction. Overall, HepaRG and suspension cultures showed modest value for the low determination and prediction of clearance compounds. The micro-array spheroid culture resulted in the most robust clearance measurements, whereas using the MPCC resulted in the most accurate prediction for low clearance compounds.

Recent Studies of Artificial Intelligence on In Silico Drug Distribution Prediction

Drug distribution is an important process in pharmacokinetics because it has the potential to influence both the amount of medicine reaching the active sites and the effectiveness as well as safety of the drug. The main causes of 90% of drug failures in clinical development are lack of efficacy and uncontrolled toxicity. In recent years, several advances and promising developments in drug distribution property prediction have been achieved, especially in silico, which helped to drastically reduce the time and expense of screening undesired drug candidates. In this study, we provide comprehensive knowledge of drug distribution background, influencing factors, and artificial intelligence-based distribution property prediction models from 2019 to the present. Additionally, we gathered and analyzed public databases and datasets commonly utilized by the scientific community for distribution prediction. The distribution property prediction performance of five large ADMET prediction tools is mentioned as a benchmark for future research. On this basis, we also offer future challenges in drug distribution prediction and research directions. We hope that this review will provide researchers with helpful insight into distribution prediction, thus facilitating the development of innovative approaches for drug discovery.

Translational PK/PD Modeling of Tumor Growth Inhibition and Target Inhibition to Support Dose Range Selection of the LMP7 Inhibitor M3258 in Relapsed/Refractory Multiple Myeloma

M3258 is an orally bioavailable, potent, selective, reversible inhibitor of the large multifunctional peptidase 7 (LMP7, β5i, PSMB8) proteolytic subunit of the immunoproteasome, a component of the cellular protein degradation machinery, highly expressed in malignant hematopoietic cells including multiple myeloma. Here we describe the fit-for-purpose pharmacokinetic (PK)/pharmacodynamic (PD)/efficacy modeling of M3258 based on preclinical data from several species. The inhibition of LMP7 activity (PD) and tumor growth (efficacy) were tested in human multiple myeloma xenografts in mice. PK and efficacy data were correlated yielding a free M3258 concentration of 45 nM for half-maximal tumor growth inhibition (KC₅₀). As M3258 only weakly inhibits LMP7 in mouse cells, both in vitro and in vivo bridging studies were performed in rats, monkeys, and dogs for translational modeling. These data indicated that the PD response in human xenograft models was closely reflected in dog PBMCs. A PK/PD model was established, predicting a free IC₅₀ value of 9 nM for M3258 in dogs in vivo, in close agreement with in vitro measurements. In parallel, the human PK parameters of M3258 were predicted by various approaches including in vitro extrapolation and allometric scaling. Using PK/PD/efficacy simulations, the efficacious dose range and corresponding PD response in human were predicted. Taken together, these efforts supported the design of a phase Ia study of M3258 in multiple myeloma patients (NCT04075721). At the lowest tested dose level, the predicted exposure matched well with the observed exposure while the duration of LMP7 inhibition was underpredicted by the model. SIGNIFICANCE STATEMENT: M3258 is a novel inhibitor of the immunoproteasome subunit LMP7. The human PK and human efficacious dose range of M3258 were predicted using in vitro-in vivo extrapolation and allometric scaling methods together with a fit-for-purpose PK/PD and efficacy model based on data from several species. A comparison with data from the Phase Ia clinical study showed that the human PK was accurately predicted, while the extent and duration of PD response were more pronounced than estimated.

In Silico Prediction of Human Clinical Pharmacokinetics with ANDROMEDA by Prosilico: Predictions for an Established Benchmarking Data Set, a Modern Small Drug Data Set, and a Comparison with Laboratory Methods

There is an ongoing aim to replace animal and in vitro laboratory models with in silico methods. Such replacement requires the successful validation and comparably good performance of the alternative methods. We have developed an in silico prediction system for human clinical pharmacokinetics, based on machine learning, conformal prediction and a new physiologically-based pharmacokinetic model, i.e. ANDROMEDA. The objectives of this study were: a) to evaluate how well ANDROMEDA predicts the human clinical pharmacokinetics of a previously proposed benchmarking data set comprising 24 physicochemically diverse drugs and 28 small drug molecules new to the market in 2021; b) to compare its predictive performance with that of laboratory methods; and c) to investigate and describe the pharmacokinetic characteristics of the modern drugs. Median and maximum prediction errors for the selected major parameters were ca 1.2 to 2.5-fold and 16-fold for both data sets, respectively. Prediction accuracy was on par with, or better than, the best laboratory-based prediction methods (superior performance for a vast majority of the comparisons), and the prediction range was considerably broader. The modern drugs have higher average molecular weight than those in the benchmarking set from 15 years earlier (ca 200 g/mol higher), and were predicted to (generally) have relatively complex pharmacokinetics, including permeability and dissolution limitations and significant renal, biliary and/or gut-wall elimination. In conclusion, the results were overall better than those obtained with laboratory methods, and thus serve to further validate the ANDROMEDA in silico system for the prediction of human clinical pharmacokinetics of modern and physicochemically diverse drugs.

Discovery and Characterization of Potent, Efficacious, Orally Available Antimalarial Plasmepsin X Inhibitors and Preclinical Safety Assessment of

Plasmepsin X (PMX) is an essential aspartyl protease controlling malaria parasite egress and invasion of erythrocytes, development of functional liver merozoites (prophylactic activity), and blocking transmission to mosquitoes, making it a potential multistage drug target. We report the optimization of an aspartyl protease binding scaffold and the discovery of potent, orally active PMX inhibitors with in vivo antimalarial efficacy. Incorporation of safety evaluation early in the characterization of PMX inhibitors precluded compounds with a long human half-life (t_1/2) to be developed. Optimization focused on improving the off-target safety profile led to the identification of UCB7362 that had an improved in vitro and in vivo safety profile but a shorter predicted human t_1/2. UCB7362 is estimated to achieve 9 log 10 unit reduction in asexual blood-stage parasites with once-daily dosing of 50 mg for 7 days. This work demonstrates the potential to deliver PMX inhibitors with in vivo efficacy to treat malaria.

Multi-Well Array Culture of Primary Human Hepatocyte Spheroids for Clearance Extrapolation of Slowly Metabolized Compounds

Accurate prediction of human pharmacokinetics using in vitro tools is an important task during drug development. Albeit, currently used in vitro systems for clearance extrapolation such as microsomes and primary human hepatocytes in suspension culture show reproducible turnover, the utility of these systems is limited by a rapid decline of activity of drug metabolizing enzymes. In this study, a multi-well array culture of primary human hepatocyte spheroids was compared to suspension and single spheroid cultures from the same donor. Multi-well spheroids remained viable and functional over the incubation time of 3 days, showing physiological excretion of albumin and α-AGP. Their metabolic activity was similar compared to suspension and single spheroid cultures. This physiological activity, the high cell concentration, and the prolonged incubation time resulted in significant turnover of all tested low clearance compounds (n = 8). In stark contrast, only one or none of the compounds showed significant turnover when single spheroid or suspension cultures were used. Using multi-well spheroids and a regression offset approach (log(CL_int) = 1.1 × + 0.85), clearance was predicted within 3-fold for 93% (13/14) of the tested compounds. Thus, multi-well spheroids represent a novel and valuable addition to the ADME in vitro tool kit for the determination of low clearance and overall clearance prediction. Graphical Abstract.

In silico prediction of volume of distribution of drugs in man using conformal prediction performs on par with animal data-based models

Volume of distribution at steady state (V_ss) is an important pharmacokinetic endpoint. In this study we apply machine learning and conformal prediction for human V_ss prediction, and make a head-to-head comparison with rat-to-man scaling, allometric scaling and the Rodgers-Lukova method on combined in silico and in vitro data, using a test set of 105 compounds with experimentally observed V_ss.The mean prediction error and % with <2-fold prediction error for our method were 2.4-fold and 64%, respectively. 69% of test compounds had an observed V_ss within the prediction interval at a 70% confidence level. In comparison, 2.2-, 2.9- and 3.1-fold mean errors and 69, 64 and 61% of predictions with <2-fold error was reached with rat-to-man and allometric scaling and Rodgers-Lukova method, respectively.We conclude that our method has theoretically proven validity that was empirically confirmed, and showing predictive accuracy on par with animal models and superior to an alternative widely used in silico-based method. The option for the user to select the level of confidence in predictions offers better guidance on how to optimise V_ss in drug discovery applications.

Comparison between lab variability and in silico prediction errors for the unbound fraction of drugs in human plasma

Variability of the unbound fraction in plasma (f_u) between labs, methods and conditions is known to exist. Variability and uncertainty of this parameter influence predictions of the overall pharmacokinetics of drug candidates and might jeopardise safety in early clinical trials. Objectives of this study were to evaluate the variability of human in vitro f_u-estimates between labs for a range of different drugs, and to develop and validate an in silico f_u-prediction method and compare the results to the lab variability.A new in silico method with prediction accuracy (Q²) of 0.69 for log f_u was developed. The median and maximum prediction errors were 1.9- and 92-fold, respectively. Corresponding estimates for lab variability (ratio between max and min f_u for each compound) were 2.0- and 185-fold, respectively. Greater than 10-fold lab variability was found for 14 of 117 selected compounds.Comparisons demonstrate that in silico predictions were about as reliable as lab estimates when these have been generated during different conditions. Results propose that the new validated in silico prediction method is valuable not only for predictions at the drug design stage, but also for reducing uncertainties of f_u-estimations and improving safety of drug candidates entering the clinical phase.

Recent developments in in vitro and in vivo models for improved translation of preclinical pharmacokinetics and pharmacodynamics data

Improved pharmacokinetics/pharmacodynamics (PK/PD) prediction in the early stages of drug development is essential to inform lead optimization strategies and reduce attrition rates. Recently, there have been significant advancements in the development of new in vitro and in vivo strategies to better characterize pharmacokinetic properties and efficacy of drug leads. Herein, we review advances in experimental and mathematical models for clearance predictions, advancements in developing novel tools to capture slowly metabolized drugs, in vivo model developments to capture human etiology for supporting drug development, limitations and gaps in these efforts, and a perspective on the future in the field.

Advances in Predictions of Oral Bioavailability of Candidate Drugs in Man with New Machine Learning Methodology

Oral bioavailability (F) is an essential determinant for the systemic exposure and dosing regimens of drug candidates. F is determined by numerous processes, and computational predictions of human estimates have so far shown limited results. We describe a new methodology where F in humans is predicted directly from chemical structure using an integrated strategy combining 9 machine learning models, 3 sets of structural alerts, and 2 physiologically-based pharmacokinetic models. We evaluate the model on a benchmark dataset consisting of 184 compounds, obtaining a predictive accuracy (Q²) of 0.50, which is successful according to a pharmaceutical industry proposal. Twenty-seven compounds were found (beforehand) to be outside the main applicability domain for the model. We compare our results with interspecies correlations (rat, mouse and dog vs. human) using the same dataset, where animal vs. human-correlations (R²) were found to be 0.21 to 0.40 and maximum prediction errors were smaller than maximum interspecies differences. We conclude that our method has sufficient predictive accuracy to be practically useful with applications in human exposure and dose predictions, compound optimization and decision making, with potential to rationalize drug discovery and development and decrease failures and overexposures in early clinical trials with candidate drugs.

Using Chemical Structure Information to Develop Predictive Models for In Vitro Toxicokinetic Parameters to Inform High-throughput Risk-assessment

The toxicokinetic (TK) parameters fraction of the chemical unbound to plasma proteins and metabolic clearance are critical for relating exposure and internal dose when building in vitro-based risk assessment models. However, experimental toxicokinetic studies have only been carried out on limited chemicals of environmental interest (~1000 chemicals with TK data relative to tens of thousands of chemicals of interest). This work evaluated the utility of chemical structure information to predict TK parameters in silico; development of cluster-based read-across and quantitative structure-activity relationship models of fraction unbound or fub (regression) and intrinsic clearance or Cl_int (classification and regression) using a dataset of 1487 chemicals; utilization of predicted TK parameters to estimate uncertainty in steady-state plasma concentration (C_ss); and subsequent in vitro-in vivo extrapolation analyses to derive bioactivity-exposure ratio (BER) plot to compare human oral equivalent doses and exposure predictions using androgen and estrogen receptor activity data for 233 chemicals as an example dataset. The results demonstrate that fub is structurally more predictable than Cl_int. The model with the highest observed performance for fub had an external test set RMSE/σ=0.62 and R²=0.61, for Cl_int classification had an external test set accuracy = 65.9%, and for intrinsic clearance regression had an external test set RMSE/σ=0.90 and R²=0.20. This relatively low performance is in part due to the large uncertainty in the underlying Cl_int data. We show that C_ss is relatively insensitive to uncertainty in Cl_int. The models were benchmarked against the ADMET Predictor software. Finally, the BER analysis allowed identification of 14 out of 136 chemicals for further risk assessment demonstrating the utility of these models in aiding risk-based chemical prioritization.

In Vitro Hepatic Uptake in Human and Monkey Hepatocytes in the Presence and Absence of Serum Protein and Its In Vitro to In Vivo Extrapolation

It is well documented that human hepatic clearance based on in vitro metabolism or transporter assays systematically resulted in underprediction; therefore, large empirical scalars are often needed in either static or physiologically based pharmacokinetic (PBPK) models to accurately predict human pharmacokinetics (PK). In our current investigation, we assessed hepatic uptake in hepatocyte suspension in Krebs-Henseleit buffer in the presence and absence of serum. The results showed that the unbound intrinsic active clearance (CL_u,int,active) values obtained by normalizing the unbound fraction in the buffer containing 10% serum were generally higher than the CL_u,int,active obtained directly from protein free buffer, suggesting "protein-facilitated" uptake. The differences of CL_u,int,active in the buffer with and without protein ranged from 1- to 925-fold and negatively correlated to the unbound serum binding of organic anion transporting polypeptide substrates. When using the uptake values obtained from buffer containing serum versus serum-free buffer, the median of scaling factors (SFs) for CL_u,int,active reduced from 24.2-4.6 to 22.7-7.1 for human and monkey, respectively, demonstrating the improvement of in vitro to in vivo extrapolation in a PBPK model. Furthermore, values of CL_u,int,active were significantly higher in monkey hepatocytes than that in human, and the species differences appeared to be compound dependent. Scaling up in vitro uptake values derived in assays containing species-specific serum can compensate for the species-specific variabilities when using cynomolgus monkey as a probe animal model. Incorporating SFs calibrated in monkey and together with scaled in vitro data can be a reliable approach for the prospective human PK prediction in early drug discovery. SIGNIFICANCE STATEMENT: We investigated the protein effect on hepatic uptake in human and monkey hepatocytes and improved the in vitro to in vivo extrapolation using parameters obtained from the incubation in the present of serum protein. In addition, significantly higher active uptake clearances were observed in monkey hepatocytes than in human, and the species differences appeared to be compound dependent. The physiologically based pharmacokinetic model that incorporates scaling factors calibrated in monkey and together with scaled in vitro human data can be a reliable approach for the prospective human pharmacokinetics prediction.

Prediction of active human dose: learnings from 20 years of Merck KGaA experience, illustrated by case studies

High-quality dose predictions based on a good understanding of target engagement is one of the main translational goals in drug development. Here, we systematically evaluate active human dose predictions for 15 Merck KGaA/EMD Serono assets spanning several modalities and therapeutic areas. Using case studies, we illustrate the value of adhering to the translational best practices of having an exposure-response relationship in an appropriate animal model; having validated, translatable pharmacodynamic (PD) biomarkers measurable in Phase I populations in the right tissue; having a deeper understanding of biology; and capturing uncertainties in predictions. Given the gap in publications on the subject, we believe that the learnings from this unique diverse data set, which are generic to the industry, will trigger actions to improve future predictions.

Species differences in metabolism of a new antiepileptic drug candidate, DSP-0565 [2-(2'-fluoro[1,1'-biphenyl]-2-yl)acetamide]

The metabolism and pharmacokinetics of DSP-0565 [2-(2'-fluoro[1,1'-biphenyl]-2-yl)acetamide], an antiepileptic drug candidate, was investigated in rats, dogs, and humans. In human hepatocytes, [¹⁴ C]DSP-0565 was primarily metabolized via amide bond hydrolysis to (2'-fluoro[1,1'-biphenyl]-2-yl)acetic acid (M8), while in rat and dog hepatocytes, it was primarily metabolized via both hydrolysis to M8 and hydroxylation at the benzene ring or the benzyl site to oxidized metabolites. After single oral administration of [¹⁴ C]DSP-0565 to rats and dogs, the major radioactivity fraction was recovered in the urine (71-72% of dose) with a much smaller fraction recovered in feces (23-25% of dose). As primary metabolites in their excreta, M8, oxidized metabolites, and glucuronide of DSP-0565 were detected. The contribution of metabolic pathways was estimated from metabolite profiles in their excreta: the major metabolic pathway was oxidation (57-62%) and the next highest was the hydrolysis pathway (23-33%). These results suggest that there are marked species differences in the metabolic pathways of DSP-0565 between humans and animals. Finally, DSP-0565 human oral clearance (CL/F) was predicted using in vitro-in vivo extrapolation (IVIVE) with/without animal scaling factors (SF, in vivo intrinsic clearance/in vitro intrinsic clearance). The SF improved the underestimation of IVIVE (fold error = 0.22), but the prediction was overestimated (fold error = 2.4-3.3). In contrast, the use of SF for hydrolysis pathway was the most accurate for the prediction (fold error = 1.0-1.4). Our findings suggest that understanding of species differences in metabolic pathways between humans and animals is important for predicting human metabolic clearance when using animal SF.

An examination of protein binding and protein-facilitated uptake relating to in vitro-in vivo extrapolation

As explained by the free drug theory, the unbound fraction of drug has long been thought to drive the efficacy of a molecule. Thus, the fraction unbound term, or f_u, appears in equations for fundamental pharmacokinetic parameters such as clearance, and is used when attempting in vitro to in vivo extrapolation (IVIVE). In recent years though, it has been noted that IVIVE does not always yield accurate predictions, and that some highly protein bound ligands have more efficient uptake than can be explained by their unbound fractions. This review explores the evolution of f_u terms included when implementing IVIVE, the concept of protein-facilitated uptake, and the mechanisms that have been proposed to account for facilitated uptake.

Determination of low intrinsic clearance in vitro: the benefit of a novel internal standard in human hepatocyte incubations

1. A novel method utilizing an internal standard in hepatocytes incubations has been developed and demonstrated to decrease the variability in the determination of intrinsic clearance (CL_int) in this system. The reduced variability was shown to allow differentiation of lower elimination rate constants from noise. 2. The suggested method was able to compensate for a small but systematic error (0.5 µL/min/10⁶ cells) caused by an evaporation of approximately 15% of the volume during the incubation time. 3. The approach was validated using six commercial drugs (ketoprofen, tolbutamide, phenacetin, etodolac and quinidine) which were metabolized by different pathways. 4. The suggested internal standard, MSC1815677, was extensively characterized and the acquired data suggest that it fulfills the requirements of an internal standard present during the incubation. The proposed internal standard was stable during the incubation and showed a low potential to inhibit drug metabolizing enzymes and transporters. With MSC1815677 we propose a novel simple, robust and cost-effective method to address the challenges in the estimation of low clearance in hepatocyte incubations.

Improving Prediction of Metabolic Clearance Using Quantitative Extrapolation of Results Obtained From Human Hepatic Micropatterned Cocultures Model and by Considering the Impact of Albumin Binding

The objective was to compare, with the same data set, the predictive performance of 3 in vitro assays of hepatic clearance (CL), namely, micropatterned cocultures (also referring to HepatoPac^®) and suspension as well as monolayer hepatocytes to define which assay is the most accurate. Furthermore, existing in vitro-to-in vivo extrapolation (IVIVE) methods were challenged to verify which method is the most predictive (i.e., direct scaling method without binding correction, conventional method based either on the unbound fraction in plasma (fu_p) according to the free-drug hypothesis, or based on an fu_p value adjusted for the albumin [ALB]-facilitated hepatic uptake phenomenon). Accordingly, the role of ALB binding was specifically challenged, and consequently, the ALB production was monitored in parallel to the metabolic stability. The ALB concentration data were used to compare the in vitro assays and to adjust the value of fu_p of each drug to mimic the ALB-facilitated hepatic uptake phenomenon. The results confirmed that the direct and conventional IVIVE methods generally overpredicted and underpredicted the CL in vivo in humans, respectively. However, the underprediction of the conventional IVIVE method based on fu_p was significantly reduced from data generated with the HepatoPac^® system compared with the 2 other in vitro assays, which is possibly because that system is producing ALB at a rate much closer to the in vivo condition in liver. Hence, these observations suggest that the presence of more ALB molecules per hepatocyte in that HepatoPac^® system may have facilitated the hepatic uptake of several bound drugs because their intrinsic CL was increased instead of being decreased by the ALB binding effect. Accordingly, the IVIVE method based on the fu_p value adjusted for the ALB-facilitated uptake phenomenon gave the lowest prediction bias from the statistical analyses. This study indicated that the HepatoPac^® system combined with the adjusted value of fu_p was the most reliable IVIVE method and revealed the importance of quantifying the in vitro-to-in vivo variation of ALB concentration to improve the CL predictions, which would help any future physiologically based pharmacokinetics modeling exercise.