From Organ-on-Chip Signals to Human Pharmacology through Mechanistic Translation, Uncertainty Propagation, and Multiscale Modeling Download PDF

Organ-on-chip systems can generate human-relevant observations of tissue exposure, transport, metabolism, injury, and pharmacodynamic response, yet these observations do not translate directly into predictions of organism-level pharmacology. A measured chip signal remains conditional on device geometry, materials, flow, cell provenance, tissue maturity, exposure history, sampling procedures, and assay interpretation. This article proposes a Mechanistic Translation and Uncertainty Architecture for connecting organ-on-chip evidence with human pharmacology without treating platform performance, biological resemblance, or model fit as sufficient evidence of translational validity. The architecture comprises a structured representation of device and tissue states, a measurement-interpretation layer, mechanistic interfaces linking cellular and tissue processes with physiologically based pharmacokinetic, pharmacokinetic–pharmacodynamic, and quantitative systems pharmacology models, a multimodal calibration layer, an uncertainty ledger, and a context-of-use qualification gate. Its central principle is that translation is a sequence of conditional transformations rather than a direct endpoint mapping. Uncertainty must therefore be identified at its source, propagated across interfaces, and interpreted relative to the intended prediction or decision. Clinical pharmacology observations, quantitative imaging, biomarkers, and omics measurements are positioned as complementary constraints that may expose model discrepancy as well as refine parameters. The proposed architecture is conceptual and has not been empirically validated, clinically qualified, or accepted for regulatory use. Its applicability will depend on organ system, compound properties, biological context, available evidence, and decision consequence. By separating observation from biological state, local mechanism from whole-body behavior, calibration from confirmation, and prediction from decision fitness, the architecture provides a structured basis for evaluating when organ-on-chip signals may contribute to defensible human-pharmacology inference.

Introduction

Organ-on-chip technologies have progressed from specialized microfluidic prototypes toward increasingly sophisticated experimental systems intended to reproduce selected aspects of human tissue physiology, disease, and drug response. Their scientific and commercial development has expanded the range of available tissue models, biological readouts, and potential pharmaceutical applications, but platform maturation does not by itself resolve how an observed response should be translated into human pharmacology [1-3]. A device may reproduce a barrier, metabolic pathway, contractile response, or inflammatory phenotype while still omitting organism-level determinants such as inter-organ exchange, systemic clearance, endocrine regulation, immune trafficking, compensatory physiology, and population variability. The central translational problem is therefore not whether a chip appears biologically realistic in isolation, but whether its outputs can be interpreted within a defensible chain of mechanistic and evidentiary relationships.

This problem is particularly important for computational pharmaceutical science. Quantitative clinical pharmacology offers mathematical structures for linking experimental measurements with human exposure and response, but those structures require clearly defined inputs, physiological assumptions, parameter provenance, and uncertainty characterization [4]. Organ-on-chip measurements cannot simply be inserted into a physiologically based pharmacokinetic or pharmacokinetic–pharmacodynamic model as though they were direct estimates of human organ function. Their meaning depends on how the device modifies transport, binding, residence time, metabolism, tissue accessibility, and signal generation. Computational integration therefore requires an explicit representation of the experimental system that produced each parameter or response trajectory.

Organ-specific microphysiological systems illustrate both the opportunity and the interpretive difficulty. Liver platforms, for example, may provide information about metabolism, accumulation, cellular injury, or functional response, yet the relevance of each output depends on tissue competence, exposure conditions, assay design, and the intended pharmacological question [5]. Similar constraints apply to intestinal, renal, cardiac, vascular, neural, and multi-organ systems. A model that performs adequately for mechanistic interpretation within one device may remain unsuitable for cross-platform comparison, human pharmacokinetic prediction, population extrapolation, or development decision support. Translational usefulness must consequently be evaluated as context-dependent rather than treated as an inherent property of the platform.

This article develops an original Mechanistic Translation and Uncertainty Architecture as a conceptual contribution for organizing organ-on-chip evidence within human-pharmacology modeling. The proposed architecture separates six functions: representation of device conditions, tissue states, and exposure histories; interpretation of measured signals; mechanistic connection across biological scales; calibration with clinical, imaging, biomarker, and omics evidence; propagation of uncertainty; and qualification for a predefined context of use. It does not claim that these components have been validated as a unified system. Instead, it specifies the relationships and boundary conditions that should be examined before chip-derived evidence is used for prediction, comparison, or decision support.

The Translation Gap between Organ-On-Chip Outputs and Human Pharmacology

The translation gap begins with a mismatch between the scale at which an organ-on-chip observation is generated and the scale at which a pharmaceutical inference is desired. A chip typically measures a local event under controlled conditions: disappearance of a compound from a perfusate, appearance of a metabolite, alteration of barrier permeability, release of a biomarker, change in contractility, or development of cellular injury. Human pharmacology, by contrast, concerns systemic exposure, tissue distribution, clearance, temporal response, interindividual variability, and the interaction of multiple organs and regulatory systems. Studies integrating vascularized organ chips, coupled gut–liver platforms, and time-resolved pharmacokinetic–pharmacodynamic measurements show that quantitative translation becomes possible only when inter-organ processing and temporal exposure are modeled explicitly [6-8].

The gap is therefore not a single scaling problem. It comprises at least four distinct inferential transitions: from nominal device input to effective tissue exposure; from measured signal to latent biological state; from local tissue process to organ contribution; and from organ contribution to whole-body behavior. Each transition requires its own assumptions, parameterization, and evidence. Reviews of organ-chip pharmacokinetic applications emphasize that platform configuration, readout selection, physiological relevance, and computational strategy must be aligned with the intended human prediction [9]. An apparently accurate chip endpoint can remain translationally ambiguous when the exposure experienced by the cells, the mechanism generating the signal, or the relationship between the represented tissue and the human organ is insufficiently characterized.

The proposed architecture treats this gap as a sequence of conditional transformations rather than a direct correspondence between chip and patient. The first transformation reconstructs the physical and biological state of the device. The second interprets raw observations through an explicit measurement model. The third introduces mechanistic representations of transport, metabolism, target interaction, tissue response, and inter-organ exchange. The fourth connects those representations with human physiology and population variability. At every stage, discrepancy may arise because the device lacks a relevant process, the experimental state is incompletely observed, the mathematical structure is inadequate, or the target population differs from the biological context represented on the chip. Translation should therefore preserve the provenance and conditionality of every intermediate quantity.

The practical significance of this formulation is that technological availability cannot be equated with translational readiness. Broad dissemination remains constrained by platform variability, standardization, reproducibility, usability, and the absence of universally accepted validation procedures [10]. The architecture accordingly requires evidence to be organized around the intended claim rather than around the sophistication of the device or model. Table 1 organizes the evidence, constructs, and interpretation boundaries needed to develop the translation gap between organ-on-chip outputs and human pharmacology within the article’s central argument.

Table 1. Evidence domains, core questions, scientific requirements, and interpretation boundaries for the translation gap between organ-on-chip outputs and human pharmacology

Representing Device Conditions, Tissue States, and Exposure Histories

The first operational requirement of the proposed architecture is a translational state representation that remains attached to every chip-derived observation. Device materials are especially important because polymers, tubing, adhesives, membranes, and coatings can absorb or adsorb compounds, producing substantial differences between nominal concentration and the exposure available to cells [11]. Geometry, surface area, compartment volume, recirculation, flow, shear, oxygenation, and sampling position likewise influence residence time and spatial concentration gradients. A chip result that is reported without these conditions loses information needed to reconstruct the physical process through which the tissue encountered the compound.

Representation must also include the design and biological choices that define the experimental system. Guidance for organ-chip development emphasizes that device architecture, cell sources, extracellular matrices, operating protocols, functional criteria, and intended applications should be described together rather than as separable technical details [12]. In the proposed architecture, these descriptors form a device-condition state rather than supplementary metadata. They are treated as model variables, covariates, or provenance attributes according to their expected influence. This approach does not imply that every descriptor must appear in every equation; it requires that potentially influential conditions remain visible, testable, and available for sensitivity or transportability analysis.

Tissue state must be represented as a dynamic biological condition rather than a fixed label such as “liver,” “kidney,” or “heart.” Multi-organ systems containing matured tissue niches demonstrate that cell composition, tissue organization, vascular coupling, and functional maturation can materially shape inter-tissue communication [13]. Tissue representation should therefore include donor provenance, cell-type composition, differentiation status, maturation, viability, barrier integrity, metabolic or secretory competence, inflammatory state, disease induction, and temporal stability. These properties determine which mechanisms are available to generate the observed response. Their documentation does not prove equivalence with a human organ, but it identifies the biological domain within which model parameters and conclusions may remain interpretable.

Exposure history completes the state representation by describing how the biological system was perturbed over time. Development guidelines for liver microphysiological systems support fit-for-purpose characterization of functional performance and operating conditions rather than reliance on an isolated endpoint [14]. Reproducibility studies further show that metabolism, accumulation, and toxicity can behave as distinguishable properties and should not be inferred from one another [15]. The proposed architecture consequently records dosing route, timing, duration, repeat exposure, washout, nominal and free concentrations, local concentration estimates, sampling losses, metabolite appearance, and recovery periods. This time-resolved representation is necessary because identical terminal measurements may arise from different exposure trajectories, adaptation states, or accumulated tissue burdens. It also provides the initial conditions required for the mechanistic bridges developed in the next section.

Mechanistic Bridges from Chip Responses to Organism-Level Behavior

A mechanistic bridge converts a chip-derived observation into an organism-level hypothesis by defining the processes that connect local tissue behavior with systemic exposure or response. Physiologically based pharmacokinetic modeling provides one candidate structure because it represents organs as physiologically parameterized compartments connected through blood flow, transport, metabolism, and elimination. When integrated cautiously with organ-on-chip evidence, such models may translate measured permeability, intrinsic clearance, secretion, uptake, or tissue-partitioning behavior into conditional whole-body scenarios [16]. The bridge is not a generic scaling coefficient. It is an explicit set of equations, assumptions, parameter transformations, and physiological constraints that identifies what is transferred from the device, what is supplied from independent human knowledge, and what remains uncertain.

The required bridge depends on the pharmacological question. A renal chip intended to inform drug disposition, for example, may provide evidence about epithelial transport, secretion, reabsorption, or metabolism, but those measurements must still be normalized, connected to renal physiology, and integrated with distribution and nonrenal clearance. Modeling of opioid disposition using a kidney proximal-tubule microphysiological system illustrates how an organ-specific experimental system can contribute to a bounded mechanistic extrapolation without representing the entire organism [17]. The appropriate interpretation is therefore that the device informs selected model components under specified conditions, not that the chip independently predicts human pharmacokinetics.

The proposed architecture organizes mechanistic translation through cross-scale interface contracts. Each interface specifies the exchanged variable, unit, temporal resolution, spatial interpretation, parameter provenance, and applicable conservation rule. Device-scale transport may supply local exposure to a cellular uptake model; the cellular model may supply metabolic or target-engagement rates to a tissue model; the tissue model may contribute organ-level parameters to a PBPK or PK–PD representation. Computational simulation can also be used in the opposite direction to examine whether chip geometry, sampling schedules, or perturbations are capable of distinguishing competing mechanistic explanations [18]. These bidirectional relationships make experimental design and model construction mutually informative rather than treating computation as a final extrapolation step.

Mechanistic detail alone does not establish credibility. A biologically elaborate model can reproduce observed data through compensating parameter errors, unidentifiable mechanisms, or assumptions that do not remain valid outside the calibration domain. Confidence in PBPK applications depends on the intended purpose, physiological plausibility, data quality, verification of implementation, parameter support, evaluation against relevant observations, and transparent characterization of uncertainty [19]. The proposed mechanistic bridge should therefore remain modular and challengeable. Alternative structures should be retained when evidence cannot distinguish them, and each human-level output should preserve a traceable connection to the chip states, external physiological evidence, and assumptions from which it was derived.

Uncertainty Propagation across Biological and Modeling Scales

Uncertainty enters before a chip experiment begins and changes as evidence moves through the translation architecture. Experimental uncertainty includes device-to-device variability, donor heterogeneity, assay error, batch effects, incomplete exposure measurement, and temporal sampling limitations. Modeling introduces additional uncertainty in parameters, initial conditions, physiological scaling, structural assumptions, population distributions, and scenario definitions. Bayesian PBPK methods provide a formal means of representing parameter knowledge through prior and posterior distributions and of carrying that uncertainty into predicted outputs [20]. However, a posterior interval is only as comprehensive as the model structures, likelihood assumptions, and discrepancy terms used to construct it.

Sensitivity analysis serves a related but distinct purpose. It asks which inputs and assumptions drive variation in model outputs and whether influential parameters interact nonlinearly across biological scales. Global sensitivity analysis can reveal that uncertainty attributed to one organ process depends on simultaneous variation in transport, binding, metabolism, or physiology elsewhere in the model [21]. Sensitivity should not be interpreted as proof that an influential parameter is biologically causal or accurately estimated. Instead, it identifies where additional evidence, better measurement, or structural revision may have the greatest effect on the output relevant to the intended use.

The proposed architecture therefore uses an uncertainty ledger that accompanies each transformation from device signal to human-pharmacology output. The ledger distinguishes measurement uncertainty, biological variability, parameter uncertainty, structural uncertainty, cross-scale uncertainty, transportability uncertainty, scenario uncertainty, and decision uncertainty. Systems-model evaluation frameworks support assessing purpose, structure, data, behavior, and prediction as separate dimensions rather than compressing credibility into one performance statistic [22]. Parameter interpretation must also be conditioned on structural and practical identifiability, because apparently precise estimates may remain nonunique or weakly informed by the available evidence [23]. Figure 1 presents the organ-on-chip-to-human translation bridge, showing how the article’s key components and boundaries are connected within uncertainty propagation across biological and modeling scales.

The figure is an original conceptual synthesis that organizes the article’s central contribution across translation gap between organ-on-chip outputs and human pharmacology, representing device conditions, tissue states, and exposure histories, representing device conditions, tissue states, exposure histories, mechanistic bridges from chip responses to organism-level behavior. Arrows and grouping indicate proposed or evidence-supported relationships rather than measured effect sizes. The figure does not represent a validated predictive model, regulatory determination, clinical recommendation, or deployment-ready system.

The ledger should record both the magnitude and ownership of uncertainty. Measurement uncertainty may belong primarily to an assay model, whereas scale-transfer uncertainty may arise at the interface between an in-vitro rate and a human organ parameter. Structural uncertainty may require alternative models rather than wider parameter distributions, and decision uncertainty may depend on available alternatives and the consequences of error rather than on prediction variance alone. These categories can interact: incomplete tissue-state measurement can create parameter uncertainty, which may then be misinterpreted as biological variability or hidden within an apparently well-calibrated population model. The architecture therefore treats uncertainty propagation as a scientific argument about what is known, assumed, and unresolved, not merely as the numerical production of confidence intervals.

Calibration with Clinical, Imaging, and Omics Evidence

Calibration connects the mechanistic architecture with observations that are closer to the human pharmacological context of interest. Clinical pharmacokinetic and pharmacodynamic data can update parameters, challenge scaling assumptions, and reveal population-specific discrepancies. Translational learning across clinical studies shows how mechanistic models may be revised as evidence accumulates across patient populations rather than being treated as fixed extrapolation devices [24]. In the proposed architecture, clinical calibration is therefore iterative. It can strengthen a chip-informed mechanism, expose missing physiology, or demonstrate that a parameter derived from one experimental state does not transport to the intended population.

Quantitative imaging provides an orthogonal source of information because it can constrain anatomical distribution, tissue exposure, target engagement, or temporal occupancy in vivo. Positron emission tomography has been used in drug development to connect molecular properties, central nervous system exposure, and target-related measurements with clinical pharmacology questions [25]. Imaging signals nevertheless require their own observation models, tracer assumptions, spatial resolution limits, and corrections. They should not be treated automatically as direct measurements of therapeutic drug concentration or pharmacological effect. Their value lies in constraining selected model components that cannot be identified from plasma concentrations or chip measurements alone.

Patient-derived experimental systems, biomarkers, and clinically grounded phenotypes may further improve the relevance of organ-chip interpretation. Human biomimetic liver systems have been discussed as potential bridges between drug development and precision-medicine questions, particularly when biological variability and patient context are represented explicitly [26]. The architecture uses such evidence bidirectionally. Clinical observations can guide the choice of chip phenotype or perturbation, while chip experiments may help interpret why a clinical subgroup differs from an average response. This relationship remains hypothesis-generating unless matched or prospectively aligned evidence demonstrates that the experimental mechanism transports to the human context.

Omics evidence can support mechanistic parameterization when it measures quantities that correspond to model components. Quantitative proteomics, for example, may inform the abundance of drug-metabolizing enzymes and transporters used in absorption, distribution, metabolism, and excretion models [27]. Yet abundance is not equivalent to catalytic activity, membrane localization, regulatory state, or net physiological contribution. Transcriptomic, proteomic, metabolomic, and related measurements must therefore be connected to the model through explicit biological assumptions and observation functions. The architecture treats omics as a constraint on plausible parameter states and mechanisms, not as an automatic substitute for functional evidence or clinical confirmation.

Qualification for Prediction, Comparison, and Decision Support

Qualification asks whether the integrated evidence is adequate for a stated use. It is distinct from calibration, because a model may fit available observations while remaining unsuitable for an external prediction or consequential decision. It is also distinct from general platform validation, because the same chip and modeling system may be adequate for interpreting a controlled experiment but inadequate for predicting human tissue exposure, comparing compounds across devices, or supporting a development decision. The context of use must specify the scientific question, target output, comparator, relevant population, user, decision consequence, and acceptable residual uncertainty.

Credibility assessment in model-informed drug development is risk-informed and context-dependent. It requires consideration of verification, validation evidence, applicability, uncertainty, biological plausibility, and the consequence of an incorrect model-informed conclusion [28-30]. The proposed architecture consequently rejects a binary distinction between “validated” and “unvalidated” models. Qualification is represented instead as a bounded argument linking the intended use to evidence that the device state is adequately characterized, the measurement model is defensible, the mechanistic bridge is appropriate, uncertainty has been propagated, and relevant external evidence has challenged the resulting output.

The depth of evaluation should increase with model influence and decision consequence. A model used to select additional sampling times may require evidence that it reproduces device transport and identifies informative parameters. A model used to compare compounds requires stronger evidence of comparable exposure, tissue state, assay behavior, and rank robustness. A model used to generate human predictions requires external clinical constraints, population-relevant physiology, and calibrated predictive uncertainty. Decision support adds the need to compare the model with existing alternatives and to examine the consequences of false reassurance, false rejection, or incorrect prioritization. Risk and evidentiary cost must therefore be balanced rather than subjected to one universal validation recipe [31].

The proposed qualification gate produces a use-specific credibility statement rather than an approval label. Its output should identify the supported claim, applicable conditions, evaluation evidence, known failure modes, unresolved uncertainties, human-review requirements, and prohibited interpretations. Table 2 organizes the evidence, constructs, and interpretation boundaries needed to develop qualification for prediction, comparison, and decision support within the article’s central argument.

Table 2. Evaluation, implementation, and research priorities arising from From Organ-on-Chip Signals to Human Pharmacology through Mechanistic Translation, Uncertainty Propagation, and Multiscale Modeling

Limitations and Boundary Conditions

The proposed Mechanistic Translation and Uncertainty Architecture is a conceptual organization of translational modeling requirements rather than an empirically validated platform. It does not demonstrate that organ-on-chip systems improve compound selection, predict clinical pharmacokinetics, identify effective doses, reduce development failure, or improve patient outcomes. Its components are supported individually by selected organ-chip, pharmacometric, imaging, omics, and credibility literature, but their integration into one architecture remains proposed. The architecture should not be interpreted as evidence that all relevant biology can be represented mathematically or that uncertainty can be reduced to a complete numerical distribution.

The available evidence is heterogeneous across organs, devices, compounds, biological sources, and intended applications. Liver, intestinal, renal, vascular, cardiac, and neural systems differ in maturity, readout quality, scaling requirements, and access to corresponding human observations. Platform reproducibility and standardization remain unresolved in many settings [10]. PBPK and systems models may also contain plausible but non-identifiable structures, and confidence principles developed for established pharmacometric applications cannot be transferred automatically to a new chip-informed use [19]. Identifiability assessment may reveal that existing experiments cannot distinguish the parameters or mechanisms required for the intended translation [23].

Misuse could occur if conceptual completeness is mistaken for prediction readiness. A well-documented state representation does not establish physiological equivalence; a mechanistic model does not establish causality; calibration does not establish external validity; and quantified uncertainty does not guarantee that all important sources of error have been included. Credibility frameworks can organize evidence, but qualification remains case-specific and dependent on decision consequence [29]. The architecture therefore cannot establish regulatory acceptability, clinical utility, individualized treatment suitability, universal platform comparability, or operational deployment readiness. Such determinations require independent evidence and accountable review within the relevant scientific, clinical, and regulatory context.

Research Agenda and Implementation Priorities

The first research priority is to determine the minimum translational state representation required for different pharmacological questions. Prospective studies should vary device materials, flow conditions, tissue maturity, exposure schedules, and measurement strategies to identify which descriptors materially affect translation. Reporting practices should attach machine-readable device, tissue, exposure, assay, and provenance records to every modeled output, extending current guidance on organ-chip development and characterization [12]. Multisite experiments should then examine whether the same state representation enables reconstruction and comparison across laboratories rather than merely improving documentation within one platform.

The second priority is to design experiments around mechanistic discrimination and uncertainty reduction. Simulation should be used before experimentation to identify sampling schedules, perturbations, and measurements capable of distinguishing alternative transport, metabolic, or response mechanisms [18]. Bayesian estimation can represent prior knowledge and update uncertainty as evidence accumulates [20], while global sensitivity analysis can identify influential parameters and interactions [21]. These tools should be combined with identifiability analysis, explicit model-discrepancy terms, and prospective external evaluation. The aim should not be to maximize retrospective fit, but to determine whether chip-derived evidence adds identifiable and transportable information beyond existing in-vitro and clinical sources.

The third priority is staged implementation tied to context of use. Early applications should focus on reconstructing device exposure, interpreting within-chip mechanisms, and improving experimental design. More demanding stages should evaluate compound comparison, cross-platform transportability, and externally tested human predictions. Clinical, imaging, biomarker, and omics evidence should be incorporated through prespecified observation models rather than informal comparison, with translational learning used to update mechanistic assumptions as new human evidence becomes available [24]. Quantitative omics should be connected to functional parameters only when the abundance-to-activity relationship is justified [27]. Progression toward decision support should require a risk-proportionate credibility argument, explicit human oversight, version control, and reassessment whenever the model, device, target population, or intended use changes [28].

Conclusion

The proposed Mechanistic Translation and Uncertainty Architecture reframes organ-on-chip translation as a chain of conditional, testable transformations rather than a direct correspondence between a chip endpoint and human pharmacology. Its contribution is to keep device conditions, tissue states, exposure histories, measurement processes, mechanistic models, calibration evidence, uncertainty, and decision context connected within one traceable structure. The architecture separates biological resemblance from functional adequacy, association from mechanism, model fit from external prediction, and prediction from decision fitness. It is intended to guide theory building, experimental design, model evaluation, reporting, and future validation, not to claim that any chip-informed model is clinically, regulatorily, or operationally qualified. Its value will ultimately depend on prospective evidence showing that explicit state representation, mechanistic interfacing, multimodal calibration, and uncertainty propagation improve the reliability and interpretability of specific translational uses.

Acknowledgments: None

Conflict of interest: None

Financial support: None

Ethics statement: None

1. Low LA, Mummery C, Berridge BR, Austin CP, Tagle DA. Organs-on-chips: Into the next decade. Nat Rev Drug Discov. 2021;20(5):345-61. doi:10.1038/s41573-020-0079-3
2. Ingber DE. Human organs-on-chips for disease modelling, drug development and personalized medicine. Nat Rev Genet. 2022;23(8):467-91. doi:10.1038/s41576-022-00466-9
3. Zhang B, Radisic M. Organ-on-a-chip devices advance to market. Lab Chip. 2017;17(14):2395-420. doi:10.1039/c6lc01554a
4. Isoherranen N, Madabushi R, Huang SM. Emerging role of organ-on-a-chip technologies in quantitative clinical pharmacology evaluation. Clin Transl Sci. 2019;12(2):113-21. doi:10.1111/cts.12627
5. Ribeiro AJS, Yang X, Patel V, Madabushi R, Strauss DG. Liver microphysiological systems for predicting and evaluating drug effects. Clin Pharmacol Ther. 2019;106(1):139-47. doi:10.1002/cpt.1458
6. Herland A, Maoz BM, Das D, Somayaji MR, Prantil-Baun R, Novak R, et al. Quantitative prediction of human pharmacokinetic responses to drugs via fluidically coupled vascularized organ chips. Nat Biomed Eng. 2020;4(4):421-36. doi:10.1038/s41551-019-0498-9
7. Tsamandouras N, Chen WLK, Edington CD, Stokes CL, Griffith LG, Cirit M. Integrated gut and liver microphysiological systems for quantitative in vitro pharmacokinetic studies. AAPS J. 2017;19(5):1499-512. doi:10.1208/s12248-017-0122-4
8. McAleer CW, Pointon A, Long CJ, Brighton RL, Wilkin BD, Bridges LR, et al. On the potential of in vitro organ-chip models to define temporal pharmacokinetic-pharmacodynamic relationships. Sci Rep. 2019;9(1):9619. doi:10.1038/s41598-019-45656-4
9. Keuper-Navis M, Walles M, Poller B, Myszczyszyn A, van der Made TK, Donkers J, et al. The application of organ-on-chip models for the prediction of human pharmacokinetic profiles during drug development. Pharmacol Res. 2023;195:106853. doi:10.1016/j.phrs.2023.106853
10. Alver CG, Drabbe E, Ishahak M, Agarwal A. Roadblocks confronting widespread dissemination and deployment of organs on chips. Nat Commun. 2024;15(1):5118. doi:10.1038/s41467-024-48864-3
11. Grant J, Özkan A, Oh C, Mahajan G, Prantil-Baun R, Ingber DE. Simulating drug concentrations in PDMS microfluidic organ chips. Lab Chip. 2021;21(18):3509-19. doi:10.1039/d1lc00348h
12. Rogal J, Schlünder K, Loskill P. Developer’s guide to an organ-on-chip model. ACS Biomater Sci Eng. 2022;8(11):4643-7. doi:10.1021/acsbiomaterials.1c01536
13. Ronaldson-Bouchard K, Teles D, Yeager K, Tavakol DN, Zhao Y, Chramiec A, et al. A multi-organ chip with matured tissue niches linked by vascular flow. Nat Biomed Eng. 2022;6(4):351-71. doi:10.1038/s41551-022-00882-6
14. Baudy AR, Otieno MA, Hewitt P, Gan J, Roth A, Keller D, et al. Liver microphysiological systems development guidelines for safety risk assessment in the pharmaceutical industry. Lab Chip. 2020;20(2):215-25. doi:10.1039/c9lc00768g
15. Rubiano A, Indapurkar A, Yokosawa R, Miedzik A, Rosenzweig B, Arefin A, et al. Characterizing the reproducibility in using a liver microphysiological system for assaying drug toxicity, metabolism, and accumulation. Clin Transl Sci. 2021;14(3):1049-61. doi:10.1111/cts.12969
16. Yang Y, Chen Y, Wang L, Xu S, Fang G, Guo X, et al. PBPK modeling on organs-on-chips: An overview of recent advancements. Front Bioeng Biotechnol. 2022;10:900481. doi:10.3389/fbioe.2022.900481
17. Imaoka T, Huang W, Shum S, Hailey DW, Chang SY, Chapron A, et al. Bridging the gap between in silico and in vivo by modeling opioid disposition in a kidney proximal tubule microphysiological system. Sci Rep. 2021;11(1):21356. doi:10.1038/s41598-021-00338-y
18. Milani N, Parrott N, Galetin A, Fowler S, Gertz M. In silico modeling and simulation of organ-on-a-chip systems to support data analysis and a priori experimental design. CPT Pharmacometrics Syst Pharmacol. 2024;13(4):524-43. doi:10.1002/psp4.13110
19. Peters SA, Dolgos H. Requirements to establishing confidence in physiologically based pharmacokinetic (PBPK) models and overcoming some of the challenges to meeting them. Clin Pharmacokinet. 2019;58(11):1355-71. doi:10.1007/s40262-019-00790-0
20. Elmokadem A, Zhang Y, Knab T, Jordie E, Gillespie WR. Bayesian PBPK modeling using R/Stan/Torsten and Julia/SciML/Turing.Jl. CPT Pharmacometrics Syst Pharmacol. 2023;12(3):300-10. doi:10.1002/psp4.12926
21. Najjar A, Hamadeh A, Krause S, Schepky A, Edginton A. Global sensitivity analysis of Open Systems Pharmacology Suite physiologically based pharmacokinetic models. CPT Pharmacometrics Syst Pharmacol. 2024;13(12):2052-67. doi:10.1002/psp4.13256
22. Braakman S, Pathmanathan P, Moore H. Evaluation framework for systems models. CPT Pharmacometrics Syst Pharmacol. 2022;11(3):264-89. doi:10.1002/psp4.12755
23. van Noort M, Ruppert M, DeJongh J, Marostica E, Bosch R, Mešić E, et al. Navigating the landscape of parameter identifiability methods: A workflow recommendation for model development. CPT Pharmacometrics Syst Pharmacol. 2024;13(7):1170-9. doi:10.1002/psp4.13148
24. Krauss M, Hofmann U, Schafmayer C, Igel S, Schlender J, Mueller C, et al. Translational learning from clinical studies predicts drug pharmacokinetics across patient populations. NPJ Syst Biol Appl. 2017;3(1):11. doi:10.1038/s41540-017-0012-5
25. Varrone A, Bundgaard C, Bang-Andersen B. PET as a translational tool in drug development for neuroscience compounds. Clin Pharmacol Ther. 2022;111(4):774-85. doi:10.1002/cpt.2548
26. Gough A, Soto-Gutierrez A, Vernetti L, Ebrahimkhani MR, Stern AM, Taylor DL. Human biomimetic liver microphysiology systems in drug development and precision medicine. Nat Rev Gastroenterol Hepatol. 2021;18(4):252-68. doi:10.1038/s41575-020-00386-1
27. Ahire D, Kruger L, Sharma S, Mettu VS, Basit A, Prasad B. Quantitative proteomics in translational absorption, distribution, metabolism, and excretion and precision medicine. Pharmacol Rev. 2022;74(3):769-96. doi:10.1124/pharmrev.121.000449
28. Kuemmel C, Yang Y, Zhang X, Florian J, Zhu H, Tegenge M, et al. Consideration of a credibility assessment framework in model-informed drug development: Potential application to physiologically-based pharmacokinetic modeling and simulation. CPT Pharmacometrics Syst Pharmacol. 2020;9(1):21-8. doi:10.1002/psp4.12479
29. Musuamba FT, Skottheim Rusten I, Lesage R, Russo G, Bursi R, Emili L, et al. Scientific and regulatory evaluation of mechanistic in silico drug and disease models in drug development: Building model credibility. CPT Pharmacometrics Syst Pharmacol. 2021;10(8):804-25. doi:10.1002/psp4.12669
30. Kirouac DC. How do we “validate” a QSP model? CPT Pharmacometrics Syst Pharmacol. 2018;7(9):547-8. doi:10.1002/psp4.12310
31. Lemaire V, Hu C, van der Graaf PH, Chang S, Wang W. No recipe for quantitative systems pharmacology model validation, but a balancing act between risk and cost. Clin Pharmacol Ther. 2024;115(1):25-8. doi:10.1002/cpt.3082