Evaluation of Prediction Accuracy for Volume of Distribution in Rat and Human Using In Vitro, In Vivo, PBPK and QSAR Methods

Mathew, Shibin; Tess, David A.; Burchett, Woodrow; Chang, George; Woody, Nathaniel A.; Keefer, Christopher E.; Orozco, Christine C.; Lin, Jian; Jordan, Samantha; Yamazaki, Shinji; Jones, Rhys D.O.; Di, Li

doi:10.1016/j.xphs.2020.12.005

Cited by 17 publications

(8 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For ChemProp, test set performance metrics are represented by the green bars in Figure . We note production-ready performance for the mouse IV properties: With an AAFE < 3 and CCC > 0.5 for AUCiv, Vss, and CLp, our models match the level of performance typically achieved with conventional IVIVE approaches. ,− Properties associated with an oral administration remained more difficult to predict (AAFE > 3 for AUCpo and F). Predictions of rat PK properties performed significantly worse (AAFE > 4).…”

Section: Resultsmentioning

confidence: 61%

“…Similarly, Vss can be predicted based on experimental fu,p, fu,tissue, logD, and pK a values, often leading to AAFE values of 1.5−3.0, depending on the ionization class. 61 Compared to these prediction errors, our best models of mouse and rat Vss offered similar performance (AAFE of 1.96−2.83). Lastly, oral F can be predicted via an integration of various in vitro results, including solubility, permeability, metabolic CLint, and transporter liabilities, often within PBPK models.…”

Section: ■ Discussionmentioning

confidence: 89%

“…Compared to these published results, our ML models of CLp offer a comparable prediction performance (AAFEs of 2.05–2.66) and low bias (AFEs of 1.00–1.36), in the absence of any experimental input. Similarly, Vss can be predicted based on experimental fu,p, fu,tissue, logD, and p K a values, often leading to AAFE values of 1.5–3.0, depending on the ionization class . Compared to these prediction errors, our best models of mouse and rat Vss offered similar performance (AAFE of 1.96–2.83).…”

Section: Discussionmentioning

confidence: 91%

See 2 more Smart Citations

Computational Predictions of Nonclinical Pharmacokinetics at the Drug Design Stage

Stoyanova

Katzberger

Komissarov

et al. 2023

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Although computational predictions of pharmacokinetics (PK) are desirable at the drug design stage, existing approaches are often limited by prediction accuracy and human interpretability. Using a discovery data set of mouse and rat PK studies at Roche (9,685 unique compounds), we performed a proof-of-concept study to predict key PK properties from chemical structure alone, including plasma clearance (CLp), volume of distribution at steady-state (Vss), and oral bioavailability (F). Ten machine learning (ML) models were evaluated, including Single-Task, Multitask, and transfer learning approaches (i.e., pretraining with in vitro data). In addition to prediction accuracy, we emphasized human interpretability of outcomes, especially the quantification of uncertainty, applicability domains, and explanations of predictions in terms of molecular features. Results show that intravenous (IV) PK properties (CLp and Vss) can be predicted with good precision (average absolute fold error, AAFE of 1.96− 2.84 depending on data split) and low bias (average fold error, AFE of 0.98−1.36), with AutoGluon, Gaussian Process Regressor (GP), and ChemProp displaying the best performance. Driven by higher complexity of oral PK studies, predictions of F were more challenging, with the best AAFE values of 2.35−2.60 and higher overprediction bias (AFE of 1.45−1.62). Multi-Task approaches and pretraining of ChemProp neural networks with in vitro data showed similar precision to Single-Task models but helped reduce the bias and increase correlations between observations and predictions. A combination of GP-computed prediction variance, molecular clustering, and dimensionality-reduction provided valuable quantitative insights into prediction uncertainty and applicability domains. SHAPley Additive exPlanations (SHAPs) highlighted molecular features contributing to prediction outcomes of Vss, providing explanations that could aid drug design. Combined results show that computational predictions of PK are feasible at the drug design stage, with several ML technologies converging to successfully leverage historical PK data sets. Further studies are needed to unlock the full potential of this approach, especially with respect to data set sizes and quality, transfer learning between in vitro and in vivo data sets, model-independent quantification of uncertainty, and explainability of predictions.

show abstract

Section: Resultsmentioning

confidence: 61%

Section: ■ Discussionmentioning

confidence: 89%

Section: Discussionmentioning

confidence: 91%

See 1 more Smart Citation

Computational Predictions of Nonclinical Pharmacokinetics at the Drug Design Stage

Stoyanova

Katzberger

Komissarov

et al. 2023

J. Chem. Inf. Model.

View full text Add to dashboard Cite

show abstract

“…When predicting V ss , variability in the measurements of the input parameters (fraction unbound, logP, and pKa) could also lead to V ss prediction inaccuracy. For instance, the Rodgers and Rowland method can have abrupt changes in V ss prediction and a high dependency upon accurate estimation of pKa values around 7 (Mathew et al., 2021). Less reliable V ss predictions have also been found for compounds with logP >3.…”

Section: Discussionmentioning

confidence: 99%

Shared learning from a physiologically based pharmacokinetic modeling strategy for human pharmacokinetics prediction through retrospective analysis of Genentech compounds

Mao

et al. 2023

Biopharm & Drug Disp

View full text Add to dashboard Cite

The quantitative prediction of human pharmacokinetics (PK) including the PK profile and key PK parameters are critical for early drug development decisions, successful phase I clinical trials, and the establishment of a range of doses to enable phase II clinical dose selection. Here, we describe an approach employing physiologically based pharmacokinetic (PBPK) modeling (Simcyp) to predict human PK and to validate its performance through retrospective analysis of 18 Genentech compounds for which clinical data are available. In short, physicochemical parameters and in vitro data for preclinical species were integrated using PBPK modeling to predict the in vivo PK observed in mouse, rat, dog, and cynomolgus monkey. Through this process, the in vitro to in vivo extrapolation (IVIVE) was determined and then incorporated into PBPK modeling in order to predict human PK. Overall, the prediction obtained using this PBPK‐IVIVE approach captured the observed human PK profiles of the compounds from the dataset well. The predicted Cmax was within 2‐fold of the observed Cmax for 94% of the compounds while the predicted area under the curve (AUC) was within 2‐fold of the observed AUC for 72% of the compounds. Additionally, important IVIVE trends were revealed through this investigation, including application of scaling factors determined from preclinical IVIVE to human PK prediction for each molecule. Based upon the analysis, this PBPK‐based approach now serves as a practical strategy for human PK prediction at the candidate selection stage at Genentech.

show abstract

“…R 2 was calculated with eq 6. (6) Two methods were used to assess model prediction precision and accuracy: absolute average fold error (AAFE) and median absolute fold error (MAFE). AAFE provides the average absolute spread of the model prediction error from unity.…”

Section: ■ Experimental Sectionmentioning

confidence: 99%

The Comparison of Machine Learning and Mechanistic In Vitro–In Vivo Extrapolation Models for the Prediction of Human Intrinsic Clearance

Keefer,

Chang,

et al. 2023

Mol. Pharmaceutics

Self Cite

View full text Add to dashboard Cite

Accurate prediction of human pharmacokinetics (PK) remains one of the key objectives of drug metabolism and PK (DMPK) scientists in drug discovery projects. This is typically performed by using in vitro−in vivo extrapolation (IVIVE) based on mechanistic PK models. In recent years, machine learning (ML), with its ability to harness patterns from previous outcomes to predict future events, has gained increased popularity in application to absorption, distribution, metabolism, and excretion (ADME) sciences. This study compares the performance of various ML and mechanistic models for the prediction of human IV clearance for a large (645) set of diverse compounds with literature human IV PK data, as well as measured relevant in vitro end points. ML models were built using multiple approaches for the descriptors: (1) calculated physical properties and structural descriptors based on chemical structure alone (classical QSAR/QSPR); (2) in vitro measured inputs only with no structure-based descriptors (ML IVIVE); and (3) in silico ML IVIVE using in silico model predictions for the in vitro inputs. For the mechanistic models, well-stirred and parallel-tube liver models were considered with and without the use of empirical scaling factors and with and without renal clearance. The best ML model for the prediction of in vivo human intrinsic clearance (CL int ) was an in vitro ML IVIVE model using only six in vitro inputs with an average absolute fold error (AAFE) of 2.5. The best mechanistic model used the parallel-tube liver model, with empirical scaling factors resulting in an AAFE of 2.8. The corresponding mechanistic model with full in silico inputs achieved an AAFE of 3.3. These relative performances of the models were confirmed with the prediction of 16 Pfizer drug candidates that were not part of the original data set. Results show that ML IVIVE models are comparable to or superior to their best mechanistic counterparts. We also show that ML IVIVE models can be used to derive insights into factors for the improvement of mechanistic PK prediction.

show abstract

Evaluation of Prediction Accuracy for Volume of Distribution in Rat and Human Using In Vitro, In Vivo, PBPK and QSAR Methods

Cited by 17 publications

References 60 publications

Computational Predictions of Nonclinical Pharmacokinetics at the Drug Design Stage

Computational Predictions of Nonclinical Pharmacokinetics at the Drug Design Stage

Shared learning from a physiologically based pharmacokinetic modeling strategy for human pharmacokinetics prediction through retrospective analysis of Genentech compounds

The Comparison of Machine Learning and Mechanistic In Vitro–In Vivo Extrapolation Models for the Prediction of Human Intrinsic Clearance

Contact Info

Product

Resources

About