Multiscale Modeling of Drug Induced Liver Injury from to Organ to Lobule

Multiscale Modeling of Drug Induced Liver Injury from to Organ to Lobule | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Multiscale Modeling of Drug Induced Liver Injury from to Organ to Lobule Alon Malka-Markovitz, Stelian Camara Dit Pinto, Mohammed Cherkaoui, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5876496/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 23 Jun, 2025 Read the published version in npj Digital Medicine → Version 1 posted 6 You are reading this latest preprint version Abstract Drug-induced liver injury poses significant challenges in drug development and in clinical care. This study builds on prior work developing a Human Liver Virtual Twin by creating a Multiscale Computational Fluid Dynamics framework that integrates patient-specific anatomical data to predict acetaminophen-induced liver injury as a demonstration of its capability. The model bridges vascular, lobular, and cellular scales to simulate dynamic blood flow, drug transport, and injury mechanisms that accurately reflect clinically observed spatial heterogeneity. Results demonstrate accurate blood flow dynamics, predictions of hepatocellular damage, and demonstrate a scalable framework for studying spatial heterogeneity applicable to other hepatic pathologies. This work establishes the foundational principles for a whole-organ virtual liver simulation methodology, potentially becoming a powerful tool to guide safety in therapeutic development and clinical treatment strategies, ultimately reducing reliance translation from animal models for preclinical drug testing. Biological sciences/Computational biology and bioinformatics/Computational models Health sciences/Medical research/Translational research Physical sciences/Mathematics and computing/Computational science Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Introduction Drug-induced liver injury (DILI) is a significant risk to patient health and a major burden on our healthcare system by negatively impacting both drug development, regulatory approval and post-market clinical management. DILI can either be intrinsic, which is dose-dependent and predictable or idiosyncratic, which is unpredictable and not dose-dependent. Idiosyncratic DILI can be mediated by immunological or non-immunological mechanisms and the diagnosis often requires exclusion of other etiologies of liver disease 1 . DILI is a leading cause of attrition during clinical drug development and a primary reason for post-approval withdrawal of drugs from the market 2 – 4 . The infrequency and unpredictability of idiosyncratic DILI is especially problematic because of the low probability of detection during preclinical and early clinical testing 2 – 4 . The inability of in vitro and animal models to identify human-specific toxic responses prevents detection idiosyncratic DILI until hepatotoxic adverse events occur in clinical stages of development 5 – 10 . One novel solution to the inadequacies of in vitro and animal testing during preclinical drug development would be a human liver virtual twin (HLVT) that can accurately depict human hepatic physiology and dynamic responses involving drug Absorption, Distribution, Metabolization and Excretion (ADME). Such a HLVT would be comprised of coupled computational models that mimic the structure, function, and responses of a normal human liver. Combining data from imaging, genetics, and clinical data, HLVTs could provide patient or population specific surrogate livers suitable for functional simulations, enabling more precise diagnosis, treatment planning, and predictable drug testing 11 . Indeed, other organ-level VTs have shown promise as replacement for animal or organoid models by providing additional human-specific insights on liver dynamics for the development of novel therapies in various areas of medicine 12 – 19 , stimulating multiple ongoing efforts toward building and validating human organ VTs 20 – 25 . Through our development plan for the Living Liver 11 virtual twin, our approach is to systematically build a HLVT capable of predicting idiosyncratic DILI in a stepwise fashion, validating its accuracy at each step. At this time, a focus directly on idiosyncratic DILI is premature given the incomplete understanding of the pathogenesis of specific idiosyncratic drugs. A clearer clinical foundation will be required to model and validate the physiological and biochemical phenomena across scales ranging from the cellular, tissue, and organ levels 11 , 26 , 27 . Thus, we have sought to build a strong methodological foundation by first modeling a well-characterized yet clinically significant dose-dependent DILI. Since hemodynamic variations are believed to be a major source of DILI's spatial heterogeneity, we will use this as a proxy for propensity for DILI. It is believed that heterogeneous blood flow contributes to difference in drug distribution and hepatotoxicity distribution over the whole organ 28 . Such hemodynamic-driven heterogeneity has been observed in various liver diseases 29 – 32 and during targeted drug delivery applications 33 – 35 . EASL clinical guidelines for DILI reviewed in 2022 highlight how perfusion abnormalities due to previous injury may relate to the zonal heterogeneity of DILI, particularly in conditions where liver architecture and blood flow influence the distribution of hepatotoxic agents 36 . The guidelines suggest that abnormal perfusion, due to altered blood supply or vascular changes, contributes to the distribution of damage across different liver zones. Altered perfusion impacts the zonal distribution of Kupffer cells and monocyte-derived macrophages, leading to localized areas of increased inflammation or fibrosis, typically in regions with specific perfusion characteristics, such as pericentral zones 37 . Dose-dependent APAP-induced liver injury (AILI) was chosen for this study due to its clinical and economic significance, as it is the most common cause of both acute liver injury and potentially fatal acute liver failure 1 , 38 . Currently in the U.S., AILI causes > 50% of cases of acute liver failure 39 , 40 , accounts for 20% of liver transplants performed for acute liver failure 40 , and costs U.S. healthcare system an estimated $ 86.9 million 41 . In addition, studies of AILI provide an abundant database of clinical and pathological findings and details about the sequential mechanisms of APAP hepatotoxicity. We note that it is well established that APAP-induced hepatotoxicity is primarily driven by metabolic activation and glutathione depletion, rather than vascular perfusion patterns. While hemodynamic factors may influence drug delivery, they do not play a direct role in the initiation of APAP-induced injury. The vascular model presented here provides a framework for studying DILI scenarios where perfusion dynamics are more relevant, such as fibrosis-associated changes in blood flow resistance. Our prior publication focused on a virtual model of AILI within the hepatic lobule, the fundamental functional unit of the liver. This is needed to create an accurate model that integrates the kinetics of APAP metabolism in different zones of the lobule and multi-spatial scales encompassing molecular, cellular, tissue, and organ levels 11 . Pathological data from clinical cases of human AILI was used and accurately predicted the kinetics of AILI, as measured by the onset of hepatocellular necrosis in central lobular zone 3 hepatocytes. The model assumed necrosis was the result of hemodynamic and metabolic differences along the portal vein-central vein axis of cords of hepatocytes 38 . This study builds on that prior work by coupling macro-scale portal hemodynamics with the previously developed virtual hepatic lobule 11 . This is sufficient to begin exploration of patient-specific spatial heterogeneity of DILI within the whole organ. As such, it provides the foundation for predicting severity in individuals with AILI by capturing both the multi-scale, 3D spatial organization of the liver and resulting liver hemodynamics. This is achieved by bridging the organ-level phenomena with their cellular and sub-cellular origins that are necessary for predicting hepatotoxic severity. This novel approach complements and extends existing statistical or Physiologically Based Pharmacokinetic (PBPK) approaches that cannot capture the multi-scale structure and physiological hepatic processes. Modeling liver dynamics at the lobular scale has significantly advanced our understanding of DILI. These efforts have primarily focused on simulating the microcirculation within liver sinusoidal networks to approximate localized hepatic function. Hexagonal porous media models are commonly used for simulating blood flow within lobules, evolving from simple 2D geometries to more complex representations incorporating isotropic and anisotropic permeabilities 42 – 47 . Studies of metabolic zonation have demonstrated how differences in enzymatic activity across periportal and pericentral zones influence injury patterns 48 – 50 . Additionally, research on blood flow obstructions has emphasized the importance of lobular network connectivity in shaping patterns of injury 48 . We recently published the first iteration of the virtual lobule, that integrates hemodynamics spanning from the portal triad to the central vein with APAP metabolism and hepatocyte injury 11 . When applied to a retrospective patient cohort, it demonstrated injury patterns consistent with general histopathological observations. However, while these models offer valuable insights into lobular-scale phenomena, lack of integration with the broader vascular dynamics limits their ability to simulate organ-level processes comprehensively. Separately, modeling hepatic perfusion at the organ scale have employed Computational Fluid Dynamics (CFD) to explore the effect of anatomical variations on blood flow and drug distribution. For instance, studies have demonstrated how vascular septa and anisotropic properties of hepatic microvasculature influence portal vein hemodynamics 51 , 52 . Additionally, porous media models have further captured dynamic flow patterns within microcirculatory networks, enabling improved simulations of drug absorption 53 , 54 . In vivo validation studies, such as 4D MRI studies, are in good agreement with macro-scale CFD predictions but face challenges in resolving smaller vascular segments due to imaging resolution limitations 55 . Together, these approaches offer insights into perfusion dynamics yet their reliance on static boundary conditions and lack of integration with finer-scale processes limits their applicability in modeling localized DILI. Efforts to bridge the gap between organ-level and lobular-scale models have focused on multiscale integration frameworks. Techniques such as the Biot-Darcy-Brinkman (BDB) model use homogenization approaches to simulate microcirculation while ensuring computational scalability 56 . Modular systems like those based on the Systems Biology Markup Language (SBML) enable tissue-specific biological parameterization for simulating xenobiotic metabolism 57 . These frameworks have begun to combine systemic pharmacokinetic (PK) models with spatially resolved lobular simulations 51 , 52 , 58 . Building on this body of work, the goal of this work was to develop a multiscale liver virtual twin that can integrate patient-specific portal vein anatomy to predict drug diffusion, metabolism, and hepatotoxicity across the liver's vascular hierarchy. To achieve this, we implement a novel integration of large-scale portal vein models derived from MR angiography with surrogate capillary networks and lobule-level sinusoidal blood flow simulations, capturing drug transport dynamics at multiple scales. The mesoscale capillary network provides critical boundary conditions for approximately 100,000 microscale lobule models, enabling spatially resolved predictions of localized drug exposure and damage. By exceeding the resolution limits of standard CT and MRI imaging, this approach delivers precise inlet conditions for each lobule and offers a new level of accuracy in predicting DILI within patient-specific liver models A key innovation of our approach is the comprehensive coupling of these multi-scale components, enabling the use of boundary conditions that dynamically adapt across scales. This framework allows for the incorporation of damage-induced vascular resistance changes, reflecting the evolving hemodynamic impact of DILI. The ability to simulate how vascular boundary conditions evolve in response to changes in lobule health is critical for accurately modeling patient-specific liver injury. By explicitly coupling systemic pharmacokinetics, organ-level perfusion, and cellular-level physiology, our model advances beyond conventional statistical and PBPK methods to provide a mechanistic, spatially resolved framework for predicting localized hepatotoxicity and its potential systemic implications. This work establishes a scalable foundation for future extensions that will incorporate progressive liver injury modeling and dynamically evolving vascular properties, enhancing its potential for personalized medicine applications. Results Portal vein simulations were run based on patient specific portal vein geometry generated from segmented images (see Fig. 5 ). The simulations demonstrated inlet pressures of 1150 ± 50 Pa with inlet flow rates ranging from 9.48 to 11.43 cm³/s (see Fig. 4 ), this is with good agreement to the results of Ma et al. 52 . The simulations results can be seen in Fig. 6 . Velocity patterns showed non-uniform flow due to complex vein geometry (Fig. 6 B), with higher velocities in the central vein bulk and minimal velocities near the wall (Fig. 6 C), this is consistent with in vivo measurements using 4D MRI flow, which similarly show irregular flow dynamics in realistic patient models 55 shown in Fig. 7 . Outlets exhibited flow rates from 0.6 to 2.1 cm³/s, with velocities varying between 0.035 ± 0.007 m/s and 0.10 ± 0.02 m/s. The results showed expected APAP transport whole 0.3e-3 mass fraction across all 22 outlets. The effect of drug concentration on flow dynamics was considered to be minimal in our simulations, as changes in concentration do not significantly impact the overall flow behavior. To confirm this, sensitivity simulations at both upper and lower concentration limits were performed and found no noticeable influence on outcomes. Density differences between blood (1.025 g/mL) and APAP (1.260 g/mL) induced minor pressure changes (~ 10%), deemed negligible under current conditions. The suggested 2D Capillary Surrogate Model simulations (CSM) were run with the CCO inferred model to join the macro portal veins simulation, and the lobule small size inlet capillaries as illustrated in Fig. 3 . As explained above, two consecutive CSM simulations were conducted from Large CSM (from 1.5mm inlet to 0.3mm) and small CSM (from 0.3mm inlet to 0.05mm outlet). Large CSM simulations were based on the Portal vein outlet results. High, average and low resulted inlet velocity showed a similar velocity pattern (see Fig. 8 ) with different extremum values. The inlet velocities dropped by 16–9% across outlets due to bifurcation and structural changes. All 64 outlets in the simulation received APAP concentration illustrating a healthy capillary structure of smaller veins. Small CSM followed the same method, the inlet velocities boundary conditions given by the Large CSM. In the small CSM simulations, the outlets boundary conditions are modeled as constant pressure of 800 ± 50 [Pa]. High, average and low resulted inlet velocity showed similar patterns (see Fig. 9 ). However, in the case of the Small CSM, the outlets’ velocities dropped by 12 − 8% across outlets due to bifurcation and structural changes. Similarly to the Large CSM, the Small CSM simulation shows that APAP is transported across all outlets. These results of both Large and Small CSM validated the model’s scalability and accuracy in capturing microvascular dynamics. The lobule simulations were divided into two aspects: (1) drug flow delivery simulation and (2) hepatocyte damage model. Lobular drug flow simulations showed inlet velocities from 140 to 1260 µm/s, derived from the CSM outlets, as seen in Fig. 10 . Porous media permeability values were fitted and were consistent with literature, averaging 1.25 × 10⁻¹⁴ m², reflecting a porous medium with 15% sinusoidal tissue and 85% hepatocyte tissue 59 . The average results closely align with those obtained using an 800 Pa pressure inlet in the earlier lobule model. Velocity profiles across the lobule displayed linear relationships with inlet velocities, ranging from 60 to 600 µm/s, in agreement with experimental data 53 . Comparison to Nishii et al 60 computational validated lobule model achieved similar results as illustrated in Fig. 11 . Drug concentrations reached 100% across all nodes, ensuring accurate representation of transport into hepatocytes. Cellular damage model results for average lobule velocity damage induced on the lobular cells following an obtained for a 27.1g overdose are presented in Fig. 12 . The obtained results are consistent with those previously obtained 11 since no change has been made to the lobule mechanical properties (porous media permeability or other) and shown here for completeness. To further explore the hemodynamic predicting capabilities of the model, we expanded our simulations from acetaminophen-induced liver injury (AILI) to recreating patient specific condition. The new simulations explored a model of hepatocellular necrosis with injury to liver sinusoidal endothelial cells and/or endothelia of the central veins (i.e., sinusoidal obstructive syndrome) that does result in clinically significant increased lobular resistance to sinusoidal blood flow. This was examined through synthetic cases simulations having different conditions (increasing vascular outlet pressure to simulate higher blood flow resistance within the lobule). Results for the first scenario show that a 2.5% pressure increase in vascular outlets (from 800 Pa to 820 Pa) is needed to reduce the flow outlet to a negligible velocity (see Fig. 13 B) compared to the healthy one (see Fig. 13 A). Drug transport becomes negligeable in the involved lobules. Conversely, in the second scenario, applying the same 2.5% pressure increase across all outlets (from 800 Pa to 820 Pa) restored drug flow throughout the liver but induces elevated inlet pressures of all the vascular veins by a similar 2.5% (20Pa) as seen in Fig. 13 C. This phenomenon was investigated with various higher pressures (up to 20% increase) and showed similar results. These findings underscore differences in intricate balance between vascular resistance and drug distribution at the lobular level depending on the type and extent of injury and necrosis. Discussion This study presents a comprehensive multiscale model that integrates CFD simulations across the liver's vascular hierarchy, from the lobular to organ scale. Using a dual CSM scheme and tailored boundary conditions, the framework bridges macro- and microscale dynamics, enabling patient-specific predictions of drug distribution, metabolism, and hepatotoxicity. By simulating APAP absorption and flow heterogeneity, the model offers new spatially and temporally resolved insights previously unavailable into drug transport and tissue damage, with potential implications for liver function in both health and disease once rigorous clinical validation is performed. Portal vein simulations demonstrated strong agreement with previous in silico models 52 , 61 as well as in vivo measurements using 4D MRI flow 55 , 62 , with velocities around 0.15 m/s. While this validation supports the model's ability to reproduce known conditions, resolution limits of CT and MRI prevented validation in smaller vascular segments. Simulations also revealed spatial heterogeneity, with lower pressures and velocities observed in the right liver lobe. These findings highlight the importance of accounting for regional variability to predict drug intake and flow distribution accurately. The CSM successfully generated detailed flow data for approximately 100,000 lobule outlets, connecting macro portal vein dynamics to micro sinusoidal flows. This scalability allows the integration of numerous lobules, enabling predictions of how localized lobular damage impacts overall liver health. Expending to the general scope of DILI, the model captures critical feedback loops, where damaged lobules can increase local vascular resistance, which leads to blood flow redistribution. By incorporating spatial heterogeneity, the CSM offers valuable insights into liver dynamics under stress conditions, including portal hypertension and DILI. Lobule simulations provided detailed insights into how drug flow, porous media permeability, and velocity affect drug delivery and tissue damage within the lobules. Simulations revealed a linear relationship between inlet velocity and flow within the porous lobule tissue, with velocities ranging from 60 to 600 µm/s, consistent with experimental data 53 , 63 . The alignment of the porous media permeability parameters with published results validated the model's accuracy. The virtual lobule damage model, based on prior work 11 , also reproduced patterns of APAP-induced injury. Furthermore, the simulations demonstrated that even with the velocity spatial heterogeneity, a single episode of AILI does not significantly alter drug distribution dynamics, as hepatocytes have taken up nearly all the APAP before injury and necrosis occur. By incorporating spatial variability and feedback mechanisms, this model advances understanding of liver injury progression and provides a robust foundation for further work toward a goal of personalized medicine. The integration of patient-specific anatomical and clinical data enables tailored simulations, dynamically predicting liver responses to drugs based on individual physiology. This approach offers significant potential for improving therapeutic strategies for DILI, other liver diseases, and the genesis and progression of portal hypertension due to cirrhosis. Despite its strengths, the current model has limitations that can be addressed by future refinements. Static boundary conditions may not fully capture dynamic changes in blood flow and vascular resistance during injury progression. Incorporating adaptive boundary conditions that evolve in real time could address this limitation. Additionally, the model employs a 2D planar capillary surrogate to simplify bifurcation structures and steady-state flow. While this preserves key geometric features, it does not fully capture three-dimensional microvascular dynamics or local permeability variations due to fibrosis or inflammation. Drug transport is also simplified, excluding active transport mechanisms. Future refinements will explore 3D reconstructions and dynamic permeability changes to enhance predictive accuracy. While the model effectively captures vascular inflow heterogeneity from the portal vein, it needs to be further refined to capture the hepatic artery (approximately 20% of the total lobular flow) and hepatic vein flow to fully address inflow and especially drainage of the HLVT. To enhance physiological realism, future extensions will also incorporate lymph formation and flow. These future refinements will allow us to capture crucial cellular-level processes - inflammation, immune responses, and endothelial injury – and hold the potential to leverage the HLVT beyond DILI toward liver disease. This study demonstrates the successful integration of multiscale simulations, bridging the gap between organ- and lobule-level dynamics. The model accurately captures blood flow across the portal vein up to its smallest capillaries and predicts drug absorption processes across scales. We address the limitations of earlier single-scale models—some focused on large-scale vascular hemodynamics, others on small-scale lobular transport—by unifying macro (portal vein), mesoscale (capillaries), and micro (lobules) within one patient-specific framework. This integration enables the first CFD-based simulation of drug-induced liver injury that simultaneously spans these critical scales. It provides a robust and scalable framework for investigating liver function and drug-induced damage. Incorporating patient-specific data further enhances its relevance for clinical and pharmaceutical applications, paving the way for personalized treatment strategies and advancing the understanding of liver disease and therapy. As the model evolves, and the Living Liver initiative grows, it holds the potential to transform the development and clinical application of novel therapeutics, improving precision, efficacy and safety. Material and Methods MRI Data A standard 1.5 Tesla magnetic MRI scan of the patient’s liver was performed with gadolinium contrast enhancement. Fast breath-hold T1 and T2 sequences are obtained, with additional diffusion-weighted imaging used for further structural characterization. The images are obtained in Digital Imaging and Communications in Medicine (DICOM) standard and exported to a separate semi-autonomous segmentation software (LiverVision™, Medivision Ankara, Türkiye). LiverVision segments different structures within the DICOM images using sensitivity thresholding based on pixel brightness and avidity. The post-segmentation images are then rendered into 3D images using stereolithography or standard tessellation language (STL) format. The STL surfaces were smoothed and joined to a full 3D solid model to avoid small elements and all inlets and outlets were cut perpendicular to the vein diameter to obtain accurate results. Ethics statement: Patient consent was not obtained because data was retrospective, based on standard care, and was analyzed anonymously 64 . The protocol was approved by the Institutional Review Board of the National University Hospital of Singapore in accordance with the Declaration of Helsinki (reference number 2020/01044). Multiscale modeling We developed a methodology to address the variations in both time and spatial scales from the liver (organ), lobule (unit) and hepatocyte (cellular) levels (Figure 1). These models span several orders of magnitude, both temporally as APAP absorption occurs over hours to days while blood flow and drug transport occur in seconds or less, and spatially with the organ level (10 cm) to lobules (1 mm) and hepatocytes (microns). Bridging these differences requires a multiscale integration framework, detailed below. The liver’s spatial and temporal scales present challenges for multiscale modeling of DILI. At the organ scale (Figure 2A), the liver geometry is derived from patient-specific CT/MRI scans. Blood flow in major vessels, ranging from 100 mm to 1.5 mm in diameter, is modeled using CFD simulations. The drug input was defined to represent drug diffusion in the blood after absorption by the guts. This provides the model with clinically relevant to APAP concentration evolution. The mesoscale (Figure 2B), spanning 1.5 mm to 50 microns, focuses on capillaries and smaller vessels, where resolution limitations of CT/MRI necessitate the use of a surrogate CFD model. This mesoscale model generates 100,000 data points of outlet velocity and concentration, which serve as inputs for the lobule scale (Figure 2C). At the lobule scale, the liver’s functional units are modeled to incorporate blood flow and APAP transport into hepatocytes, as described by Camara Dit Pinto et al. 11 . These simulations refine predictions of cellular damage caused by APAP (Figure 2D). The process of linking models with different timescales is called temporal integration. At the organ level, drug absorption happens over hours or days, while blood flow simulations focus on changes that occur in seconds. To connect these scales, drug concentrations entering the blood vessels are broken into small intervals and analyzed in 5-second steps. These velocities and concentrations are then used to set boundary conditions for smaller-scale models of the liver, which capture APAP transport and hepatocyte damage dynamics to simulate how the drug spreads and causes cell damage, using 20-minute steps. Spatial integration ensures that information flows smoothly between models of different sizes. Large-scale models of the liver provide data on blood flow and drug concentration directly to medium-scale simulations, which then refine this information for smaller-scale models of the liver’s functional units. Boundary conditions at each scale are iteratively updated based on feedback from the others (Figure 3). This approach connects the big picture (how the liver absorbs drugs) with the details (how liver cells are damaged). The multiscale drug pipeline integrates blood flow and cellular injury models to simulate APAP-induced liver damage (Figure 2), building on previously reported methods 11 with a key modification: boundary conditions from the portal vein are now introduced via CFD simulations. The pipeline has two main components: 1) blood flow simulation with APAP transport and 2) a cellular injury model, which run iteratively to simulate the progression of injury within the lobule. First, healthy blood flow is modeled using a species transport and advection-diffusion framework to track APAP concentration throughout the lobular blood flow. The cellular injury model then evaluates the impact of the drug at each node of the computational mesh, determining health status node by node. Simulations are conducted using a transient, pressure-based Navier-Stokes solver, enhanced with Python scripts to incorporate specific parameters outlined in Table 1. While the current pipeline is parameterized for APAP, it can be adapted to assess the effects of other drugs and treatments. For this model, drug diffusion from the gut into the bloodstream is specified to follow a predictable timeline: diffusion begins immediately after ingestion, reaching peak concentrations at 3 hours, averaging at 6 hours, and dropping to undetectable levels by 8 hours. This structured framework enables the dynamic simulation of APAP transport and injury progression within the liver. Blood flow simulation with drug transport and Boundary conditions Blood flow simulations across three scales were performed using a transient, pressure-based solver. Portal venous flow is typically modeled as steady-state; however, this study employs a transient solver to capture pulsatile hepatic inflow, better reflecting physiological dynamics, especially in pathological states like fibrosis, cirrhosis, or drug-induced vascular changes. While drug transport was modeled using a species transport framework to ensure accurate mass fractions of drug to blood within each element. To account for the non-Newtonian behavior of blood, the Carreau model for shear-thinning viscosity was applied at the capillary and lobule scales 65 . This model captures the variation in blood viscosity with shear rate, particularly in regions with high velocity gradients, enabling realistic simulation of pressure differences and velocity distributions. Parameters for the Carreau model were drawn from established blood flow modeling literature 66 and are detailed in Table 1. The viscous model ensures that shear-rate-dependent viscosity is accurately represented in areas with varying velocities. Boundary conditions were tailored to maintain fluid dynamic continuity across scales. Inlet velocities, fitted from Ma et al. 52 , determined the inlet pressures, as shown in Figure 4. Outlet boundary conditions at each scale were calculated to match the average outlet velocity from the previous scale to the resulting inlet pressure of the current scale. This approach ensured consistent transitions in fluid dynamics down to the lobule level, where inlet pressure was set at approximately 800 Pa 53 . The outlet pressures for each scale are summarized in Table 1, with a pressure span of ±50 Pa based on initial conditions in Figure 4. This span remained consistent across boundary conditions, and velocity spans were verified to align with the resulting pressure spans, ensuring accurate simulation across all scales. Simulation Parameter Value Reference All Simulations Permeability 1.0x10 -14 to 1.0x10 -15 [m 2 ] 53 Blood density 1050 [kg/m 3 ] 53 APAP drug density 1260 [kg/m 3 ] 67 Blood Mw 64,500 [kg/kmol] 68 APAP drug Mw 151.1 [Kg/kmol] 67 Mixture mass diffusivity 2.22x10 -10 [m 2 /s] 69 Viscosity model (Blood/APAP) Zero-shear viscosity 0.056 [Pa·s] 66 Infinite-shear viscosity 0.00345 [Pa·s] Time constant 1.902 [s] Power-law index(n) 0.22 Organ Scale Portal Vein Simulation Inlet diameter 10.8[mm] Patient data Outlet diameter 1.5-3.5 [mm] Patient data Flow rate inlet 9.48-11.43[cm 3 /s] 52 Pressure outlet 1100±50 [Pa] Tailored to fit Capillary Surrogate Model (CSM) Simulation Diameter Large SCM 1.5[mm] to 0.3[mm] From upper scale Diameter Small SCM 0.3[mm] to 50[µm] From upper scale Flow rate inlet From Upper scale From Upper scale Large CSM Pressure outlet 820±50 [Pa] Tailored to fit Small CSM Pressure outlet 800±50 [Pa] 53 Virtual Lobule Simulation Portal Track diameter 0.05 [mm] 59 Central Vein diameter 0.15 [mm] 53 Lobule diameter 1.5 [mm] 53 Flow rate inlet From upper scale From upper scale Pressure outlet 660 [Pa] 53 Table 1- blood flow and drug transport model parameters for each scale To ensure accurate inflow and outflow boundary conditions, the portal vasculature geometry was trimmed at the inlet and outlet to create smooth surfaces. A 3D simulation was constructed based on the refined geometry, capturing the complex vein structure and varying diameters, as shown in Figure 5. The mesh used an element size of 5x10⁻⁴ m, resulting in approximately 1 million elements. The vascular outlet diameters ranged from 3.5 mm to 1.5 mm, matching typical MR angiography resolution limit. The portal vein inlet flow rate varied between 9.48 cm³/s and 11.43 cm³/s in a one-second cycle, as shown in Figure 4, replicating the dynamics of heart-driven blood flow into the portal vein 52 . All 24 outlets were assigned a default pressure of 1100 ± 50 Pa to ensure compatibility with the models of veins below 1.5 mm in diameter. Simulations were run until a dynamic equilibrium of outlet velocities and drug concentration was achieved. A fixed time step of 0.001 seconds was applied after an initial relaxation phase with five steps of 1x10⁻⁵ seconds. The simulation duration was set to 100 seconds, sufficient to ensure stabilization of both drug transport and outlet velocities. Results were verified for convergence with respect to both mesh size and time step. APAP drug concentrations varied to follow evolution clinically observed during drug ingestion, with simulations exploring concentrations from 0 to 8e-3 kg/L to model different dosages over specific time intervals 70 . It is important to note that the simulation included only blood flow, drug transport, and their mixture, without modeling any biological or chemical reactions. The Capillary Surrogate Model (CSM) was developed to bridge the spatial gap between MR resolution limit and the liver lobule. Using the Constrained Constructive Optimization (CCO) approach 71 , the CSM generates approximately 10,000 individual outlets to the lobule scale. The model begins with an inlet diameter of 1.5 mm and extends to outlets as small as 50 microns, effectively connecting the macro (organ-level) and micro (lobule-level) scales. To enable hundreds of computational runs while preserving critical vascular structures, the Capillary Surrogate Model (CSM) was implemented as a 2D planar simulation. This approach preserves key structural details such as bifurcation angles, segment lengths, and diameters while allowing for sensitivity analyses and efficient flow dynamic studies. However, a key limitation of this 2D approach is that it prevents vessel crossings, which can affect local perfusion distributions compared to an in-vivo 3D vascular network. While this simplification does not fully replicate complex 3D hepatic vasculature, it allows for controlled sensitivity testing of hemodynamic parameters and can be extended to 3D models where computational feasibility permits. The CSM includes two configurations: the Large CSM and the Small CSM. The Large CSM represents a vascular segment with inlets of 1.5 mm and outlets of 0.3 mm, while the Small CSM scales down to inlets of 0.33 mm and outlets of 50 microns. Both models approximate blood flow behavior rather than providing precise, patient-specific results, serving as an efficient surrogate to capture flow dynamics. This is readily extendable to 3D models, provided the impact of various configurations on simulation outcomes and sensitivity analyses justify to additional computation burden. For the Large CSM, a mesh element size of 1.1x10⁻⁴ m was used, generating 70,000 elements. Inlet velocities ranged from 0.04 to 0.12 m/s, based on outputs from the 3D portal vein simulation. The outlets were assigned a default pressure of 830 ± 20 Pa, ensuring compatibility with downstream models for veins below 0.3 mm. In the Small CSM, a finer mesh element size of 2x10⁻⁵ m was applied, resulting in the same number of elements. Inlet velocities ranged from 0.004 to 0.04 m/s, with outlets set to a default pressure of 800 ± 20 Pa, aligning with lobule vein inlet conditions 53 . Both models were run until dynamic equilibrium of outlet velocities and drug concentrations was achieved. Simulations used a fixed timestep of 0.001 seconds, following an initial relaxation phase with five timesteps of 1×10⁻⁵ seconds. Simulation durations ranged from 10 to 100 seconds, ensuring stabilization of both velocity and APAP concentration throughout the veins and outlets. Convergence was verified with respect to both mesh size and time step, confirming the reliability of the results. This 2D surrogate modeling framework efficiently captures small vein flow and drug transport dynamics, providing a scalable tool to bridge the macroscopic and microscopic scales in liver modeling. While simplified, the CSM effectively complements lobule-level models and lays the groundwork for future exploration of 3D implementations and patient-specific configurations. The lobule simulations were based on a 2D representation of the liver lobule, following methodologies from previous work 11 . A mesh element size of 1.75x10⁻⁵ m was used, resulting in approximately 60,000 elements. Boundary conditions were derived from capillary-scale simulations, with inlet velocities corresponding to portal vein outlet data and outlet pressure set to 500 ± 20 Pa 53 . These conditions modeled normal blood flow across various portal vein outlets. A critical factor in the lobule simulation was the permeability of the porous media, which directly influenced internal velocities and pressures. Transitioning from the smooth, nonporous vascular system to the porous lobule, which encompasses both vasculature and tissue, required careful tailoring of boundary conditions. The permeability was adjusted to align the inlet pressure of 800 Pa with the outlet pressure of 500 Pa, ensuring consistent fluid dynamics across scales 51,53 . The simulation ran until dynamic equilibrium of outlet velocities and drug concentrations was achieved. A fixed timestep of 1 second was used, and simulations were run for 20 minutes to evaluate the progression of APAP-induced damage. Each cell in the lobule model represented hepatic arterial tissue and computed the concentration of APAP, providing data for assessing cellular injury. To model APAP-induced injury, a modified MALD damage model was employed, consistent with previous lobule simulation studies 11 . This model linked drug concentration in tissue cells to corresponding levels of cellular damage, enabling an evaluation of injury evolution within the lobule. The integration of these simulations allowed for detailed insight into the dynamics of APAP transport and hepatocyte damage at the cellular level. Modeling patient conditions The developed multiscale liver platform presented in this work opens the door to predicting hemodynamic changes based on patient pathologies. To showcase those capabilities, two scenarios were created. The first scenario recreates the effects of hepatocellular necrosis with injury to liver sinusoidal endothelial cells and/or endothelia of the central veins (i.e., sinusoidal obstructive syndrome). In this first scenario, an outlet pressure increase in a specific section of the small CSM is applied with aim to reduce the blood flow to a negligeable value (inducing limited drug transport in the affected lobules). The second scenario aims to recreate the effects of veinous pressure increase due to vena cava obstruction. In this second scenario, an outlet pressure increase is applied to the whole small CSM outlets. These scenarios do not represent specific known or observed clinical conditions, but are fictive scenarios created to tests the robustness of the model and explore possible theories. Declarations Data Availability: Segmented vasculature files can be made available upon reasonable request. Code Availability: ANSYS Fluent simulation files can be made available upon reasonable request. Acknowledgments: The authors would like to thank the Long Island University-Dassault Systèmes Center of Excellence and the BCM Learning Health System for their financial support. Author contributions: N.R.G. supervised and led the project conceptualization and administration. A.M.M performed the simulation setup, data collection, and curation. S.C.D.P. helped to implement parts of the computational pipeline, data collection, and data analysis. N.R.G., A.M.M. and S.C.D.P. analyzed the results and wrote the original draft. M.C., S.M.L., G.K.S., S.A., S.D., G.Y. and J.M.V. provided important clinical and computational intellectual content for the study design and results analysis. M.C., S.M.L., G.K.S., S.A., S.D., G.Y. and J.M.V. contributed to writing the final draft of the manuscript. All authors critically reviewed the paper and contributed important intellectual content. All authors have read and approved the manuscript. Competing interests: S.M.L. is a Dassault Systemes employee. J.M.V. is a member of IQDILI and the NIH NIDDK Drug-Induced Liver Injury Network Data Safety, and Management Board. N.R.G. has served as a consultant for Dassault Systemes. The remaining authors declare no competing interests. Additional information: Correspondence and requests for materials should be addressed to Nicolas R. Gallo. References Hosack, T., Damry, D. & Biswas, S. Drug-induced liver injury: a comprehensive review. Therap Adv Gastroenterol 16 , 17562848231163410 (2023). Andrade, R. J. et al. Drug-induced liver injury. Nat Rev Dis Primers 5 , 58 (2019). Mostafa, F. & Chen, M. Computational models for predicting liver toxicity in the deep learning era. Front Toxicol 5 , 1340860 (2023). Ellison, C., Hewitt, M. & Przybylak, K. In Silico Models for Hepatotoxicity. Methods Mol Biol 2425 , 355–392 (2022). Speid, L. H., Lumley, C. E. & Walker, S. R. Harmonization of guidelines for toxicity testing of pharmaceuticals by 1992. Regulatory Toxicology and Pharmacology 12 , 179–211 (1990). Mulliner, D. et al. Computational Models for Human and Animal Hepatotoxicity with a Global Application Scope. Chem. Res. Toxicol. 29 , 757–767 (2016). Xu, Z. et al. Research on the Species Difference of the Hepatotoxicity of Medicine Based on Transcriptome. Front. Pharmacol. 12 , (2021). Mitchell, S. J. & Hilmer, S. N. Drug-induced liver injury in older adults. Therapeutic Advances in Drug Safety 1 , 65–77 (2010). Shanks, N., Greek, R. & Greek, J. Are animal models predictive for humans? Philosophy, Ethics, and Humanities in Medicine 4 , 2 (2009). Olson, H. et al. Concordance of the Toxicity of Pharmaceuticals in Humans and in Animals. Regulatory Toxicology and Pharmacology 32 , 56–67 (2000). Camara Dit Pinto, S. et al. A virtual scalable model of the Hepatic Lobule for acetaminophen hepatotoxicity prediction. npj Digit. Med. 7 , 340 (2024). Esmailie, F. et al. Biomechanics of Transcatheter Aortic Valve Replacement Complications and Computational Predictive Modeling. Structural Heart 6 , 100032 (2022). Dowling, C., Gooley, R., McCormick, L., Firoozi, S. & Brecker, S. J. Patient-specific Computer Simulation: An Emerging Technology for Guiding the Transcatheter Treatment of Patients with Bicuspid Aortic Valve. Interv Cardiol 16 , e26 (2021). Finotello, A., Morganti, S. & Auricchio, F. Finite element analysis of TAVI: Impact of native aortic root computational modeling strategies on simulation outcomes. Medical Engineering & Physics 47 , 2–12 (2017). Astudillo, P., De Beule, M., Dambre, J. & Mortier, P. Towards safe and efficient preoperative planning of transcatheter mitral valve interventions. Morphologie 103 , 139–147 (2019). Baillargeon, B., Rebelo, N., Fox, D. D., Taylor, R. L. & Kuhl, E. The Living Heart Project: A robust and integrative simulator for human heart function. Eur J Mech A Solids 48 , 38–47 (2014). Flynn, D., Palmer, M., Schiestl, R., Levine, S. & Maeder, T. MDIC Landscape Report on Medical Device Computational Modeling and Simulation. (2023). Morrison, T. M., Dreher, M. L., Nagaraja, S., Angelone, L. M. & Kainz, W. The Role of Computational Modeling and Simulation in the Total Product Life Cycle of Peripheral Vascular Devices. J Med Device 11 , 024503 (2017). Badano, A. In silico imaging clinical trials: cheaper, faster, better, safer, and more scalable. Trials 22 , 64 (2021). Avanzato, R., Beritelli, F., Lombardo, A. & Ricci, C. Lung-DT: An AI-Powered Digital Twin Framework for Thoracic Health Monitoring and Diagnosis. Sensors 24 , 958 (2024). Zhu, L., Lu, W., Soleimani, M., Li, Z. & Zhang, M. Electrical Impedance Tomography Guided by Digital Twins and Deep Learning for Lung Monitoring. IEEE Transactions on Instrumentation and Measurement 72 , 1–9 (2023). Hurtado, D. E., Avilés-Rojas, N. & Concha, F. Multiscale modeling of lung mechanics: From alveolar microstructure to pulmonary function. Journal of the Mechanics and Physics of Solids 179 , 105364 (2023). Weickenmeier, J., Butler, C. A. M., Young, P. G., Goriely, A. & Kuhl, E. The mechanics of decompressive craniectomy: Personalized simulations. Computer Methods in Applied Mechanics and Engineering 314 , 180–195 (2017). Xiong, H. et al. The Digital Twin Brain: A Bridge between Biological and Artificial Intelligence. Intelligent Computing 2 , 0055 (2023). Sasikaladevi, N. & Revathi, A. Digital twin of renal system with CT-radiography for the early diagnosis of chronic kidney diseases. Biomedical Signal Processing and Control 88 , 105632 (2024). National Academies of Sciences, Engineering, and Medicine. Opportunities and Challenges for Digital Twins in Biomedical Research: Proceedings of a Workshop-in Brief. in 26922 (National Academies Press, Washington, D.C., 2023). doi:10.17226/26922. Ho, H., Rezania, V. & Schwen, L. O. Editorial: Multiscale modeling for the liver. Frontiers in Bioengineering and Biotechnology 11 , (2023). Schwen, L. O., Kuepfer, L. & Preusser, T. Modeling approaches for hepatic spatial heterogeneity in pharmacokinetic simulations. Drug Discovery Today: Disease Models 22 , 35–43 (2016). Karcaaltincaba, M. & Akhan, O. Imaging of hepatic steatosis and fatty sparing. European Journal of Radiology 61 , 33–43 (2007). Goldstein, N. S., Hastah, F., Galan, M. V. & Gordon, S. C. Fibrosis Heterogeneity in Nonalcoholic Steatohepatitis and Hepatitis C Virus Needle Core Biopsy Specimens. American Journal of Clinical Pathology 123 , 382–387 (2005). Avila, M. A. et al. Reduced mRNA abundance of the main enzymes involved in methionine metabolism in human liver cirrhosis and hepatocellular carcinoma. Journal of Hepatology 33 , 907–914 (2000). Xie, C. et al. Quantification of Hepatic Vascular and Parenchymal Regeneration in Mice. PLOS ONE 11 , e0160581 (2016). Gasselhuber, A. et al. Targeted drug delivery by high intensity focused ultrasound mediated hyperthermia combined with temperature-sensitive liposomes: Computational modelling and preliminary in vivovalidation. International Journal of Hyperthermia 28 , 337–348 (2012). Tanaka, T. et al. Radiologic Placement of Side-hole Catheter with Tip Fixation for Hepatic Arterial Infusion Chemotherapy. Journal of Vascular and Interventional Radiology 14 , 63–68 (2003). Gasselhuber, A. et al. Mathematical spatio-temporal model of drug delivery from low temperature sensitive liposomes during radiofrequency tumour ablation. International Journal of Hyperthermia 26 , 499–513 (2010). Brennan, P. N., Cartlidge, P., Manship, T. & Dillon, J. F. Guideline review: EASL clinical practice guidelines: drug-induced liver injury (DILI). Frontline Gastroenterol 13 , 332–336 (2022). Guillot, A. & Tacke, F. Spatial dimension of macrophage heterogeneity in liver diseases. egastro 1 , (2023). Jaeschke, H. & Ramachandran, A. Acetaminophen Hepatotoxicity: Paradigm for Understanding Mechanisms of Drug-Induced Liver Injury. Annual Review of Pathology: Mechanisms of Disease 19 , 453–478 (2024). Chidiac, A. S., Buckley, N. A., Noghrehchi, F. & Cairns, R. Paracetamol (acetaminophen) overdose and hepatotoxicity: mechanism, treatment, prevention measures, and estimates of burden of disease. Expert Opinion on Drug Metabolism & Toxicology 19 , 297–317 (2023). Li, X., Tang, J. & Mao, Y. Incidence and risk factors of drug-induced liver injury. Liver International 42 , 1999–2014 (2022). Mehrpour, O., Saeedi, F., Hadianfar, A., Mégarbane, B. & Hoyte, C. Prognostic factors of acetaminophen exposure in the United States: An analysis of 39,000 patients. Hum Exp Toxicol 40 , S814–S825 (2021). Coombe, D., Rezania, V. & Tuszynski, J. A. Dual continuum upscaling of liver lobule flow and metabolism to the full organ scale. Frontiers in Systems Biology 2 , (2022). Rezania, V., Marsh, R., Coombe, D. & Tuszynski, J. A physiologically-based flow network model for hepatic drug elimination I: regular lattice lobule model. Theor Biol Med Model 10 , 52 (2013). Rezania, V., Marsh, R., Coombe, D. & Tuszynski, J. A physiologically-based flow network model for hepatic drug elimination II: variable lattice lobule models. Theor Biol Med Model 10 , 53 (2013). Rezania, V., Coombe, D. & Tuszynski, J. A. A physiologically-based flow network model for hepatic drug elimination III: 2D/3D DLA lobule models. Theor Biol Med Model 13 , 9 (2016). Bonfiglio, A., Leungchavaphongse, K., Repetto, R. & Siggers, J. H. Mathematical Modeling of the Circulation in the Liver Lobule. Journal of Biomechanical Engineering 132 , (2010). Mosharaf-Dehkordi, M. A fully coupled porous media and channels flow approach for simulation of blood and bile flow through the liver lobules. Computer Methods in Biomechanics and Biomedical Engineering 22 , 901–915 (2019). Lambers, L. et al. Quantifying Fat Zonation in Liver Lobules: An IntegratedMultiscale In-silico Model Combining DisturbedMicroperfusion and Fat Metabolism via aContinuum-Biomechanical Bi-scale, Tri-phasic Approach. (2023). Gerhäusser, S. et al. Simulation of zonation-function relationships in the liver using coupled multiscale models: Application to drug-induced liver injury. 2024.03.26.586870 Preprint at https://doi.org/10.1101/2024.03.26.586870 (2024). Schliess, F. et al. Integrated metabolic spatial-temporal model for the prediction of ammonia detoxification during liver damage and regeneration. Hepatology 60 , 2040–2051 (2014). Debbaut, C. et al. A 3D porous media liver lobule model: the importance of vascular septa and anisotropic permeability for homogeneous perfusion. Comput Methods Biomech Biomed Engin 17 , 1295–1310 (2014). Ma, R. et al. Anatomically based simulation of hepatic perfusion in the human liver. International Journal for Numerical Methods in Biomedical Engineering 35 , e3229 (2019). Hu, J., Lü, S., Feng, S. & Long, M. Flow dynamics analyses of pathophysiological liver lobules using porous media theory. Acta Mech. Sin. 33 , 823–832 (2017). Schwen, L. O. et al. Spatio-Temporal Simulation of First Pass Drug Perfusion in the Liver. PLoS Comput Biol 10 , e1003499 (2014). Huang, A. et al. Reference values for 4D flow magnetic resonance imaging of the portal venous system. Abdom Radiol 48 , 2049–2059 (2023). Rohan, E., Turjanicová, J. & Lukeš, V. Multiscale modelling and simulations of tissue perfusion using the Biot-Darcy-Brinkman model. Computers & Structures 251 , 106404 (2021). Sluka, J. P. et al. A Liver-Centric Multiscale Modeling Framework for Xenobiotics. PLOS ONE 11 , e0162428 (2016). Diaz Ochoa, J. G. et al. A Multi-Scale Modeling Framework for Individualized, Spatiotemporal Prediction of Drug Effects and Toxicological Risk. Front. Pharmacol. 3 , (2013). Siggers, J. H., Leungchavaphongse, K., Ho, C. H. & Repetto, R. Mathematical model of blood and interstitial flow and lymph production in the liver. Biomech Model Mechanobiol 13 , 363–378 (2014). Nishii, K., Reese, G., Moran, E. C. & Sparks, J. L. Multiscale computational model of fluid flow and matrix deformation in decellularized liver. Journal of the Mechanical Behavior of Biomedical Materials 57 , 201–214 (2016). George, S. M., Eckert, L. M., Martin, D. R. & Giddens, D. P. Hemodynamics in Normal and Diseased Livers: Application of Image-Based Computational Models. Cardiovasc Eng Tech 6 , 80–91 (2015). Roldán-Alzate, A. et al. Characterization of mesenteric and portal hemodynamics using 4D flow MRI: the effects of meals and diurnal variation. Abdom Radiol 47 , 2106–2114 (2022). Nakata, M. et al. Alterations to hepatic microcirculation in thioacetamide-induced cirrhotic livers of rats. Osaka City Med J 48 , 1–8 (2002). Balci, D., Kirimker, E. O., Raptis, D. A., Gao, Y. & Kow, A. W. C. Uses of a dedicated 3D reconstruction software with augmented and mixed reality in planning and performing advanced liver surgery and living donor liver transplantation (with videos). Hepatobiliary & Pancreatic Diseases International 21 , 455–461 (2022). Yasuda, K., Armstrong, R. C. & Cohen, R. E. Shear flow properties of concentrated solutions of linear and star branched polystyrenes. Rheol Acta 20 , 163–178 (1981). Weddell, J. C., Kwack, J., Imoukhuede, P. I. & Masud, A. Hemodynamic Analysis in an Idealized Artery Tree: Differences in Wall Shear Stress between Newtonian and Non-Newtonian Blood Models. PLoS ONE 10 , e0124575 (2015). PubChem. Acetaminophen. https://pubchem.ncbi.nlm.nih.gov/compound/1983. Billett, H. H. Hemoglobin and Hematocrit. in Clinical Methods: The History, Physical, and Laboratory Examinations (eds. Walker, H. K., Hall, W. D. & Hurst, J. W.) (Butterworths, Boston, 1990). Ribeiro, A. et al. Diffusion coefficients of paracetamol in aqueous solutions. The Journal of Chemical Thermodynamics 54 , 97–99 (2012). Remien, C. H., Adler, F. R., Waddoups, L., Box, T. D. & Sussman, N. L. Mathematical modeling of liver injury and dysfunction after acetaminophen overdose: Early discrimination between survival and death. Hepatology 56 , 727–734 (2012). Whitehead, J. F., Laeseke, P. F., Periyasamy, S., Speidel, M. A. & Wagner, M. G. In silico simulation of hepatic arteries: An open‐source algorithm for efficient synthetic data generation. Medical Physics 50 , 5505–5517 (2023). Additional Declarations Competing interest reported. S.M.L. is a Dassault Systemes employee. J.M.V. is a member of IQDILI and the NIH NIDDK Drug-Induced Liver Injury Network Data Safety and Management Board. N.R.G. has served as a consultant for Dassault Systemes. The remaining authors declare no competing interests. Cite Share Download PDF Status: Published Journal Publication published 23 Jun, 2025 Read the published version in npj Digital Medicine → Version 1 posted Editorial decision: Accepted 19 May, 2025 Reviews received at journal 16 May, 2025 Reviewers agreed at journal 14 May, 2025 Reviewers invited by journal 13 May, 2025 Submission checks completed at journal 30 Apr, 2025 First submitted to journal 26 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5876496","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":450010784,"identity":"fd8ff621-d765-4121-bce0-18559d14d350","order_by":0,"name":"Alon Malka-Markovitz","email":"","orcid":"","institution":"Long Island University","correspondingAuthor":false,"prefix":"","firstName":"Alon","middleName":"","lastName":"Malka-Markovitz","suffix":""},{"id":450010785,"identity":"dad9d235-2163-4599-9d80-17e5f836b80c","order_by":1,"name":"Stelian Camara Dit Pinto","email":"","orcid":"","institution":"Long Island University","correspondingAuthor":false,"prefix":"","firstName":"Stelian","middleName":"Camara Dit","lastName":"Pinto","suffix":""},{"id":450010786,"identity":"c0bed58d-96ca-46dc-8733-b3d191afaa1c","order_by":2,"name":"Mohammed Cherkaoui","email":"","orcid":"","institution":"Long Island University","correspondingAuthor":false,"prefix":"","firstName":"Mohammed","middleName":"","lastName":"Cherkaoui","suffix":""},{"id":450010787,"identity":"40d52dd8-2afb-480e-b9c5-73a4363b4ee6","order_by":3,"name":"Steven M. Levine","email":"","orcid":"","institution":"Dassault Systèmes (France)","correspondingAuthor":false,"prefix":"","firstName":"Steven","middleName":"M.","lastName":"Levine","suffix":""},{"id":450010788,"identity":"2a1b44df-db6f-4967-a29b-55c2737ad54a","order_by":4,"name":"Sharmila Anandasbapathy","email":"","orcid":"","institution":"Baylor College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Sharmila","middleName":"","lastName":"Anandasbapathy","suffix":""},{"id":450010789,"identity":"5afa13dd-0f46-4d51-89ad-64794e36821f","order_by":5,"name":"Gagan K. Sood","email":"","orcid":"","institution":"Baylor College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Gagan","middleName":"K.","lastName":"Sood","suffix":""},{"id":450010790,"identity":"17563d36-2cd8-4826-9152-7e857687c36c","order_by":6,"name":"Sadhna Dhingra","email":"","orcid":"","institution":"Houston Methodist Hospital","correspondingAuthor":false,"prefix":"","firstName":"Sadhna","middleName":"","lastName":"Dhingra","suffix":""},{"id":450010791,"identity":"8af688e0-fbf2-499f-b85b-8814069bd25a","order_by":7,"name":"Gao Yujia","email":"","orcid":"","institution":"National University of Singapore","correspondingAuthor":false,"prefix":"","firstName":"Gao","middleName":"","lastName":"Yujia","suffix":""},{"id":450010792,"identity":"c05ab73f-1df0-4aa4-8648-8ff5017a3f35","order_by":8,"name":"John M. Vierling","email":"","orcid":"","institution":"Baylor College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"M.","lastName":"Vierling","suffix":""},{"id":450010793,"identity":"8fc4f2cb-f443-4497-ab9e-6572d8847b1d","order_by":9,"name":"Nicolas R. Gallo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIie3PPQrCMBTA8SeBuERdI4q9QorgBwqOXqPFSzhIiQiOuiqIx3BOCOgiuAouze6gi6OYVgSntKNg/hB4hPcjLYDL9Yshc+L3NBUw7pBkpJkkSKcCF3CkOQh8k8L8vWwl7SLScTCJwBtKLu5bWmcCyQuxkO4MN1mwV+AfQy7XO0qYwKOejTBFMA2wAJ+HXJVSQlq1bPKMwF9qQzYJqTyySThH4NHkFZ6+gjOI+ZdwoQijmsvVnpKquelubOSkdHx7RA1vOZLxbdIflA8zfb5ayCfzPZ8R5VhP83jeTZfL5fq7XqnLSWSzTTglAAAAAElFTkSuQmCC","orcid":"","institution":"Long Island University","correspondingAuthor":true,"prefix":"","firstName":"Nicolas","middleName":"R.","lastName":"Gallo","suffix":""}],"badges":[],"createdAt":"2025-01-22 01:23:02","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5876496/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5876496/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41746-025-01736-6","type":"published","date":"2025-06-23T15:57:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":82070931,"identity":"8a50ccb8-feae-4cea-9610-c524bd401686","added_by":"auto","created_at":"2025-05-06 13:19:11","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":168214,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eVisualization of the different scales of the human liver. \u003c/strong\u003eThe organ scale (cm) with \u003cstrong\u003ea\u003c/strong\u003e) the liver and \u003cstrong\u003eb\u003c/strong\u003e) the portal vasculature. \u003cstrong\u003ec\u003c/strong\u003e) The functional scale (mm), the lobule and \u003cstrong\u003ed\u003c/strong\u003e) the cellular scale (μm) with hepatocytes in beige, stellate cells in blue, Kupfer cells in orange and endothelial cells in pink. Figure adapted from Camara Dit Pinto et. al 2024 with authorization.\u003c/p\u003e","description":"","filename":"Figure1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/0120eff1c7bf432b1051e7e0.jpg"},{"id":82073178,"identity":"151975af-2c76-418e-95df-010b0a7aaeb9","added_by":"auto","created_at":"2025-05-06 13:27:11","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":177380,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMultiscale methodology.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) Introduction of patient-specific Portal Vien from MRI and APAP drug time concertation from PBPK models \u003cstrong\u003eb\u003c/strong\u003e) CFD simulations of Portal Vien \u003cstrong\u003ec\u003c/strong\u003e) CFD simulation of Large and Small capillaries model \u003cstrong\u003ed\u003c/strong\u003e) CFD Lobule models and \u003cstrong\u003ee\u003c/strong\u003e)Schematic view of lobule Damage model.\u003c/p\u003e","description":"","filename":"Figure2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/d0ccefff7d66ee5357dee802.jpg"},{"id":82073177,"identity":"ff57710f-8dff-437c-af27-d9ed1f2626ac","added_by":"auto","created_at":"2025-05-06 13:27:11","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":468047,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA multiscale method scheme.\u003c/strong\u003e Illustration of the multiscale hierarchical integration of models used to simulate drug transport and toxicity across physiological scales. The framework spans from the full-body PBPK model (bottom) to the lobule model (top), which represents the cellular microenvironment where tissue damage occurs.\u003c/p\u003e","description":"","filename":"Figure3.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/9da48a57643e791ef09d6836.jpg"},{"id":82073186,"identity":"5bae9ba4-c127-41e0-8c24-0cb0cd7816bc","added_by":"auto","created_at":"2025-05-06 13:27:11","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":374413,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePortal vein Blood flow rate and pressure inlet simulation.\u003c/strong\u003eTime-dependent simulation of portal vein inflow showing pulsatile pressure (blue) and flow rate (dashed orange) over a 2-second cardiac cycle. These dynamic inlet conditions are used to drive downstream liver simulations.\u003c/p\u003e","description":"","filename":"Figure4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/55cda4ad0f3e9617b022e8d3.jpg"},{"id":82070935,"identity":"37233800-0d5b-4e00-9593-4e43a2479298","added_by":"auto","created_at":"2025-05-06 13:19:11","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":127203,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePortal vein geometry and meshing.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) full patient-specific liver geometry. \u003cstrong\u003eb\u003c/strong\u003e) geometry of the patient-specific Portal Vein CFD simulation. \u003cstrong\u003ec\u003c/strong\u003e) Mesh cross-section.\u003c/p\u003e","description":"","filename":"Figure5.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/1484b25f44146063d55d8518.jpg"},{"id":82073180,"identity":"26d1d2e9-24ff-4796-8cbe-e500adf0cf96","added_by":"auto","created_at":"2025-05-06 13:27:11","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":447989,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePortal vein CFD Simulation results.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) pressure (Pa), \u003cstrong\u003eb\u003c/strong\u003e) velocity streamlines (m/s) \u003cstrong\u003ec\u003c/strong\u003e) representative cross-section of the velocity gradient in the portal vein.\u003c/p\u003e","description":"","filename":"Figure6.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/b771a2dd606e278172893ab5.jpg"},{"id":82070938,"identity":"571290e7-929e-421c-9caa-64037e210286","added_by":"auto","created_at":"2025-05-06 13:19:11","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":122731,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparison of simulation CFD results of the PV with 4D MRI flow.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) PV velocity magnitude (m/s) predicted with multi-scale pipeline. \u003cstrong\u003eb\u003c/strong\u003e) PV velocity magnitude (m/s) predicted measured with 4D MRI flow from Huang et al., 2023.\u003c/p\u003e","description":"","filename":"Figure7.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/b22703555a3529652dbcb03d.jpg"},{"id":82073877,"identity":"9374a753-4e44-4395-83e8-50b7bebc9543","added_by":"auto","created_at":"2025-05-06 13:35:11","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":253199,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLarge CSM CFD simulation results.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) pressure scheme (Pa) \u003cstrong\u003eb\u003c/strong\u003e) average inlet Velocity of 0.078 (m/s) \u003cstrong\u003ec\u003c/strong\u003e) high inlet velocity of 0.12 (m/s) and \u003cstrong\u003ed\u003c/strong\u003e) low inlet velocity of 0.035 (m/s).\u003c/p\u003e","description":"","filename":"Figure8.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/60177704de108d2067d8e836.jpg"},{"id":82070949,"identity":"c5f13c42-786f-43d1-a036-50c6a2228e63","added_by":"auto","created_at":"2025-05-06 13:19:11","extension":"jpg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":253620,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSmall CSM CFD simulation results.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) pressure scheme (Pa) \u003cstrong\u003eb\u003c/strong\u003e) average inlet Velocity of 7.2 (mm/s). \u003cstrong\u003ec\u003c/strong\u003e) high inlet velocity of 12.5 (mm/s) and \u003cstrong\u003ed\u003c/strong\u003e) low inlet velocity of 2 (mm/s).\u003c/p\u003e","description":"","filename":"Figure9.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/c13a62272924601b70986cc9.jpg"},{"id":82070951,"identity":"455cd316-f235-44cf-8f21-991b9123fee4","added_by":"auto","created_at":"2025-05-06 13:19:12","extension":"jpg","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":200350,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLobule CFD simulation results.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) High velocity \u003cstrong\u003eb\u003c/strong\u003e) average velocity and \u003cstrong\u003ec\u003c/strong\u003e) low velocity inlet of both pressure and velocity inside the lobule.\u003c/p\u003e","description":"","filename":"Figure10.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/c339eea62a48b9f62f9dcd32.jpg"},{"id":82075498,"identity":"9356979f-6091-4299-a69a-bf5e166ae7ba","added_by":"auto","created_at":"2025-05-06 13:43:11","extension":"jpg","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":67240,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eValidation of lobular velocity profile against literature data.\u003c/strong\u003e Velocity along the portal-to-central vein axis within the lobule is shown for the multiscale model (dashed orange) and experimental data from Nishii et al. (solid blue).\u003c/p\u003e","description":"","filename":"Figure11.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/d9a0e86fb10f443555ad6c93.jpg"},{"id":82070950,"identity":"522bb121-b71e-41f4-9b30-b35976e1f916","added_by":"auto","created_at":"2025-05-06 13:19:11","extension":"jpg","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":590513,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResults obtained for a 27.1g overdose.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e) Lobule cellular APAP input obtained for a 27.1g overdose. \u003cstrong\u003eb\u003c/strong\u003e) Induced proportion of cellular states in the lobule during overdose. \u003cstrong\u003ec\u003c/strong\u003e) Resulting lobular damage map for the scenario.\u003c/p\u003e","description":"","filename":"Figure12.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/416bc42442c319b8f04ce8e8.jpg"},{"id":82070953,"identity":"efcd0ebf-c009-4444-aa90-ce3bf54e12be","added_by":"auto","created_at":"2025-05-06 13:19:12","extension":"jpg","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":2337021,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimulated DILI synthetic scenarios progression under varying lobular damage conditions.\u003c/strong\u003e 2mm/s velocity inlet of L-CSM model results \u003cstrong\u003ea\u003c/strong\u003e) normal healthy capillary, \u003cstrong\u003eb\u003c/strong\u003e) 1-4 outlets section (marked in red) that has damaged lobules with 820Pa high-pressure, \u003cstrong\u003ec\u003c/strong\u003e) All lobules are damaged with 820Pa high pressure All 820Pa.\u003c/p\u003e","description":"","filename":"Figure13.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/9c55f5b1f844754189747e11.jpg"},{"id":85686340,"identity":"a4a65c3d-475a-4822-9075-02008246a1ae","added_by":"auto","created_at":"2025-06-30 16:05:37","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6628685,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5876496/v1/9405be3f-ae0b-4160-8a36-14a463a76079.pdf"}],"financialInterests":"Competing interest reported. S.M.L. is a Dassault Systemes employee. J.M.V. is a member of IQDILI and the NIH NIDDK Drug-Induced Liver Injury Network Data Safety and Management Board. N.R.G. has served as a consultant for Dassault Systemes. The remaining authors declare no competing interests.","formattedTitle":"Multiscale Modeling of Drug Induced Liver Injury from to Organ to Lobule","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDrug-induced liver injury (DILI) is a significant risk to patient health and a major burden on our healthcare system by negatively impacting both drug development, regulatory approval and post-market clinical management. DILI can either be intrinsic, which is dose-dependent and predictable or idiosyncratic, which is unpredictable and not dose-dependent. Idiosyncratic DILI can be mediated by immunological or non-immunological mechanisms and the diagnosis often requires exclusion of other etiologies of liver disease\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. DILI is a leading cause of attrition during clinical drug development and a primary reason for post-approval withdrawal of drugs from the market\u003csup\u003e\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. The infrequency and unpredictability of idiosyncratic DILI is especially problematic because of the low probability of detection during preclinical and early clinical testing\u003csup\u003e\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. The inability of \u003cem\u003ein vitro\u003c/em\u003e and animal models to identify human-specific toxic responses prevents detection idiosyncratic DILI until hepatotoxic adverse events occur in clinical stages of development\u003csup\u003e\u003cspan additionalcitationids=\"CR6 CR7 CR8 CR9\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOne novel solution to the inadequacies of \u003cem\u003ein vitro\u003c/em\u003e and animal testing during preclinical drug development would be a human liver virtual twin (HLVT) that can accurately depict human hepatic physiology and dynamic responses involving drug Absorption, Distribution, Metabolization and Excretion (ADME). Such a HLVT would be comprised of coupled computational models that mimic the structure, function, and responses of a normal human liver. Combining data from imaging, genetics, and clinical data, HLVTs could provide patient or population specific surrogate livers suitable for functional simulations, enabling more precise diagnosis, treatment planning, and predictable drug testing\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Indeed, other organ-level VTs have shown promise as replacement for animal or organoid models by providing additional human-specific insights on liver dynamics for the development of novel therapies in various areas of medicine\u003csup\u003e\u003cspan additionalcitationids=\"CR13 CR14 CR15 CR16 CR17 CR18\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, stimulating multiple ongoing efforts toward building and validating human organ VTs\u003csup\u003e\u003cspan additionalcitationids=\"CR21 CR22 CR23 CR24\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThrough our development plan for the Living Liver\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e virtual twin, our approach is to systematically build a HLVT capable of predicting idiosyncratic DILI in a stepwise fashion, validating its accuracy at each step. At this time, a focus directly on idiosyncratic DILI is premature given the incomplete understanding of the pathogenesis of specific idiosyncratic drugs. A clearer clinical foundation will be required to model and validate the physiological and biochemical phenomena across scales ranging from the cellular, tissue, and organ levels\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Thus, we have sought to build a strong methodological foundation by first modeling a well-characterized yet clinically significant dose-dependent DILI.\u003c/p\u003e \u003cp\u003eSince hemodynamic variations are believed to be a major source of DILI's spatial heterogeneity, we will use this as a proxy for propensity for DILI. It is believed that heterogeneous blood flow contributes to difference in drug distribution and hepatotoxicity distribution over the whole organ\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Such hemodynamic-driven heterogeneity has been observed in various liver diseases\u003csup\u003e\u003cspan additionalcitationids=\"CR30 CR31\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e and during targeted drug delivery applications\u003csup\u003e\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. EASL clinical guidelines for DILI reviewed in 2022 highlight how perfusion abnormalities due to previous injury may relate to the zonal heterogeneity of DILI, particularly in conditions where liver architecture and blood flow influence the distribution of hepatotoxic agents\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. The guidelines suggest that abnormal perfusion, due to altered blood supply or vascular changes, contributes to the distribution of damage across different liver zones. Altered perfusion impacts the zonal distribution of Kupffer cells and monocyte-derived macrophages, leading to localized areas of increased inflammation or fibrosis, typically in regions with specific perfusion characteristics, such as pericentral zones\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDose-dependent APAP-induced liver injury (AILI) was chosen for this study due to its clinical and economic significance, as it is the most common cause of both acute liver injury and potentially fatal acute liver failure \u003csup\u003e \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e \u003c/sup\u003e. Currently in the U.S., AILI causes\u0026thinsp;\u0026gt;\u0026thinsp;50% of cases of acute liver failure\u003csup\u003e \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e \u003c/sup\u003e, accounts for 20% of liver transplants performed for acute liver failure\u003csup\u003e \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e \u003c/sup\u003e, and costs U.S. healthcare system an estimated \u003cspan\u003e$\u003c/span\u003e86.9 million\u003csup\u003e \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e \u003c/sup\u003e. In addition, studies of AILI provide an abundant database of clinical and pathological findings and details about the sequential mechanisms of APAP hepatotoxicity. We note that it is well established that APAP-induced hepatotoxicity is primarily driven by metabolic activation and glutathione depletion, rather than vascular perfusion patterns. While hemodynamic factors may influence drug delivery, they do not play a direct role in the initiation of APAP-induced injury. The vascular model presented here provides a framework for studying DILI scenarios where perfusion dynamics are more relevant, such as fibrosis-associated changes in blood flow resistance.\u003c/p\u003e \u003cp\u003eOur prior publication focused on a virtual model of AILI within the hepatic lobule, the fundamental functional unit of the liver. This is needed to create an accurate model that integrates the kinetics of APAP metabolism in different zones of the lobule and multi-spatial scales encompassing molecular, cellular, tissue, and organ levels\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Pathological data from clinical cases of human AILI was used and accurately predicted the kinetics of AILI, as measured by the onset of hepatocellular necrosis in central lobular zone 3 hepatocytes. The model assumed necrosis was the result of hemodynamic and metabolic differences along the portal vein-central vein axis of cords of hepatocytes\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study builds on that prior work by coupling macro-scale portal hemodynamics with the previously developed virtual hepatic lobule\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. This is sufficient to begin exploration of patient-specific spatial heterogeneity of DILI within the whole organ. As such, it provides the foundation for predicting severity in individuals with AILI by capturing both the multi-scale, 3D spatial organization of the liver and resulting liver hemodynamics. This is achieved by bridging the organ-level phenomena with their cellular and sub-cellular origins that are necessary for predicting hepatotoxic severity. This novel approach complements and extends existing statistical or Physiologically Based Pharmacokinetic (PBPK) approaches that cannot capture the multi-scale structure and physiological hepatic processes.\u003c/p\u003e \u003cp\u003eModeling liver dynamics at the lobular scale has significantly advanced our understanding of DILI. These efforts have primarily focused on simulating the microcirculation within liver sinusoidal networks to approximate localized hepatic function. Hexagonal porous media models are commonly used for simulating blood flow within lobules, evolving from simple 2D geometries to more complex representations incorporating isotropic and anisotropic permeabilities\u003csup\u003e\u003cspan additionalcitationids=\"CR43 CR44 CR45 CR46\" citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Studies of metabolic zonation have demonstrated how differences in enzymatic activity across periportal and pericentral zones influence injury patterns\u003csup\u003e\u003cspan additionalcitationids=\"CR49\" citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. Additionally, research on blood flow obstructions has emphasized the importance of lobular network connectivity in shaping patterns of injury\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. We recently published the first iteration of the virtual lobule, that integrates hemodynamics spanning from the portal triad to the central vein with APAP metabolism and hepatocyte injury\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. When applied to a retrospective patient cohort, it demonstrated injury patterns consistent with general histopathological observations. However, while these models offer valuable insights into lobular-scale phenomena, lack of integration with the broader vascular dynamics limits their ability to simulate organ-level processes comprehensively.\u003c/p\u003e \u003cp\u003eSeparately, modeling hepatic perfusion at the organ scale have employed Computational Fluid Dynamics (CFD) to explore the effect of anatomical variations on blood flow and drug distribution. For instance, studies have demonstrated how vascular septa and anisotropic properties of hepatic microvasculature influence portal vein hemodynamics\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. Additionally, porous media models have further captured dynamic flow patterns within microcirculatory networks, enabling improved simulations of drug absorption\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e,\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. \u003cem\u003eIn vivo\u003c/em\u003e validation studies, such as 4D MRI studies, are in good agreement with macro-scale CFD predictions but face challenges in resolving smaller vascular segments due to imaging resolution limitations\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. Together, these approaches offer insights into perfusion dynamics yet their reliance on static boundary conditions and lack of integration with finer-scale processes limits their applicability in modeling localized DILI.\u003c/p\u003e \u003cp\u003eEfforts to bridge the gap between organ-level and lobular-scale models have focused on multiscale integration frameworks. Techniques such as the Biot-Darcy-Brinkman (BDB) model use homogenization approaches to simulate microcirculation while ensuring computational scalability\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. Modular systems like those based on the Systems Biology Markup Language (SBML) enable tissue-specific biological parameterization for simulating xenobiotic metabolism\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. These frameworks have begun to combine systemic pharmacokinetic (PK) models with spatially resolved lobular simulations\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBuilding on this body of work, the goal of this work was to develop a multiscale liver virtual twin that can integrate patient-specific portal vein anatomy to predict drug diffusion, metabolism, and hepatotoxicity across the liver's vascular hierarchy. To achieve this, we implement a novel integration of large-scale portal vein models derived from MR angiography with surrogate capillary networks and lobule-level sinusoidal blood flow simulations, capturing drug transport dynamics at multiple scales. The mesoscale capillary network provides critical boundary conditions for approximately 100,000 microscale lobule models, enabling spatially resolved predictions of localized drug exposure and damage. By exceeding the resolution limits of standard CT and MRI imaging, this approach delivers precise inlet conditions for each lobule and offers a new level of accuracy in predicting DILI within patient-specific liver models\u003c/p\u003e \u003cp\u003eA key innovation of our approach is the comprehensive coupling of these multi-scale components, enabling the use of boundary conditions that dynamically adapt across scales. This framework allows for the incorporation of damage-induced vascular resistance changes, reflecting the evolving hemodynamic impact of DILI. The ability to simulate how vascular boundary conditions evolve in response to changes in lobule health is critical for accurately modeling patient-specific liver injury. By explicitly coupling systemic pharmacokinetics, organ-level perfusion, and cellular-level physiology, our model advances beyond conventional statistical and PBPK methods to provide a mechanistic, spatially resolved framework for predicting localized hepatotoxicity and its potential systemic implications. This work establishes a scalable foundation for future extensions that will incorporate progressive liver injury modeling and dynamically evolving vascular properties, enhancing its potential for personalized medicine applications.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003ePortal vein simulations were run based on patient specific portal vein geometry generated from segmented images (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The simulations demonstrated inlet pressures of 1150\u0026thinsp;\u0026plusmn;\u0026thinsp;50 Pa with inlet flow rates ranging from 9.48 to 11.43 cm\u0026sup3;/s (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e4\u003c/span\u003e), this is with good agreement to the results of Ma et al. \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. The simulations results can be seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e6\u003c/span\u003e. Velocity patterns showed non-uniform flow due to complex vein geometry (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e6\u003c/span\u003eB), with higher velocities in the central vein bulk and minimal velocities near the wall (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e6\u003c/span\u003eC), this is consistent with \u003cem\u003ein vivo\u003c/em\u003e measurements using 4D MRI flow, which similarly show irregular flow dynamics in realistic patient models\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e7\u003c/span\u003e. Outlets exhibited flow rates from 0.6 to 2.1 cm\u0026sup3;/s, with velocities varying between 0.035\u0026thinsp;\u0026plusmn;\u0026thinsp;0.007 m/s and 0.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02 m/s. The results showed expected APAP transport whole 0.3e-3 mass fraction across all 22 outlets.\u003c/p\u003e\u003cp\u003eThe effect of drug concentration on flow dynamics was considered to be minimal in our simulations, as changes in concentration do not significantly impact the overall flow behavior. To confirm this, sensitivity simulations at both upper and lower concentration limits were performed and found no noticeable influence on outcomes. Density differences between blood (1.025 g/mL) and APAP (1.260 g/mL) induced minor pressure changes (~\u0026thinsp;10%), deemed negligible under current conditions.\u003c/p\u003e \u003cp\u003eThe suggested 2D Capillary Surrogate Model simulations (CSM) were run with the CCO inferred model to join the macro portal veins simulation, and the lobule small size inlet capillaries as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003e. As explained above, two consecutive CSM simulations were conducted from Large CSM (from 1.5mm inlet to 0.3mm) and small CSM (from 0.3mm inlet to 0.05mm outlet).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eLarge CSM simulations were based on the Portal vein outlet results. High, average and low resulted inlet velocity showed a similar velocity pattern (see Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e8\u003c/span\u003e) with different extremum values. The inlet velocities dropped by 16\u0026ndash;9% across outlets due to bifurcation and structural changes. All 64 outlets in the simulation received APAP concentration illustrating a healthy capillary structure of smaller veins.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSmall CSM followed the same method, the inlet velocities boundary conditions given by the Large CSM. In the small CSM simulations, the outlets boundary conditions are modeled as constant pressure of 800\u0026thinsp;\u0026plusmn;\u0026thinsp;50 [Pa]. High, average and low resulted inlet velocity showed similar patterns (see Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e9\u003c/span\u003e). However, in the case of the Small CSM, the outlets\u0026rsquo; velocities dropped by 12\u0026thinsp;\u0026minus;\u0026thinsp;8% across outlets due to bifurcation and structural changes. Similarly to the Large CSM, the Small CSM simulation shows that APAP is transported across all outlets. These results of both Large and Small CSM validated the model\u0026rsquo;s scalability and accuracy in capturing microvascular dynamics.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe lobule simulations were divided into two aspects: (1) drug flow delivery simulation and (2) hepatocyte damage model. Lobular drug flow simulations showed inlet velocities from 140 to 1260 \u0026micro;m/s, derived from the CSM outlets, as seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e10\u003c/span\u003e. Porous media permeability values were fitted and were consistent with literature, averaging 1.25 \u0026times; 10⁻\u0026sup1;⁴ m\u0026sup2;, reflecting a porous medium with 15% sinusoidal tissue and 85% hepatocyte tissue\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. The average results closely align with those obtained using an 800 Pa pressure inlet in the earlier lobule model. Velocity profiles across the lobule displayed linear relationships with inlet velocities, ranging from 60 to 600 \u0026micro;m/s, in agreement with experimental data\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. Comparison to Nishii et al \u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e computational validated lobule model achieved similar results as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e11\u003c/span\u003e. Drug concentrations reached 100% across all nodes, ensuring accurate representation of transport into hepatocytes. Cellular damage model results for average lobule velocity damage induced on the lobular cells following an obtained for a 27.1g overdose are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e12\u003c/span\u003e. The obtained results are consistent with those previously obtained\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e since no change has been made to the lobule mechanical properties (porous media permeability or other) and shown here for completeness.\u003c/p\u003e\u003cp\u003eTo further explore the hemodynamic predicting capabilities of the model, we expanded our simulations from acetaminophen-induced liver injury (AILI) to recreating patient specific condition. The new simulations explored a model of hepatocellular necrosis with injury to liver sinusoidal endothelial cells and/or endothelia of the central veins (i.e., sinusoidal obstructive syndrome) that does result in clinically significant increased lobular resistance to sinusoidal blood flow. This was examined through synthetic cases simulations having different conditions (increasing vascular outlet pressure to simulate higher blood flow resistance within the lobule). Results for the first scenario show that a 2.5% pressure increase in vascular outlets (from 800 Pa to 820 Pa) is needed to reduce the flow outlet to a negligible velocity (see Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e13\u003c/span\u003eB) compared to the healthy one (see Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e13\u003c/span\u003eA). Drug transport becomes negligeable in the involved lobules. Conversely, in the second scenario, applying the same 2.5% pressure increase across all outlets (from 800 Pa to 820 Pa) restored drug flow throughout the liver but induces elevated inlet pressures of all the vascular veins by a similar 2.5% (20Pa) as seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e13\u003c/span\u003eC. This phenomenon was investigated with various higher pressures (up to 20% increase) and showed similar results. These findings underscore differences in intricate balance between vascular resistance and drug distribution at the lobular level depending on the type and extent of injury and necrosis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study presents a comprehensive multiscale model that integrates CFD simulations across the liver's vascular hierarchy, from the lobular to organ scale. Using a dual CSM scheme and tailored boundary conditions, the framework bridges macro- and microscale dynamics, enabling patient-specific predictions of drug distribution, metabolism, and hepatotoxicity. By simulating APAP absorption and flow heterogeneity, the model offers new spatially and temporally resolved insights previously unavailable into drug transport and tissue damage, with potential implications for liver function in both health and disease once rigorous clinical validation is performed.\u003c/p\u003e \u003cp\u003ePortal vein simulations demonstrated strong agreement with previous \u003cem\u003ein silico\u003c/em\u003e models\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e as well as \u003cem\u003ein vivo\u003c/em\u003e measurements using 4D MRI flow\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e, with velocities around 0.15 m/s. While this validation supports the model's ability to reproduce known conditions, resolution limits of CT and MRI prevented validation in smaller vascular segments. Simulations also revealed spatial heterogeneity, with lower pressures and velocities observed in the right liver lobe. These findings highlight the importance of accounting for regional variability to predict drug intake and flow distribution accurately.\u003c/p\u003e \u003cp\u003eThe CSM successfully generated detailed flow data for approximately 100,000 lobule outlets, connecting macro portal vein dynamics to micro sinusoidal flows. This scalability allows the integration of numerous lobules, enabling predictions of how localized lobular damage impacts overall liver health. Expending to the general scope of DILI, the model captures critical feedback loops, where damaged lobules can increase local vascular resistance, which leads to blood flow redistribution. By incorporating spatial heterogeneity, the CSM offers valuable insights into liver dynamics under stress conditions, including portal hypertension and DILI.\u003c/p\u003e \u003cp\u003eLobule simulations provided detailed insights into how drug flow, porous media permeability, and velocity affect drug delivery and tissue damage within the lobules. Simulations revealed a linear relationship between inlet velocity and flow within the porous lobule tissue, with velocities ranging from 60 to 600 \u0026micro;m/s, consistent with experimental data\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e,\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. The alignment of the porous media permeability parameters with published results validated the model's accuracy. The virtual lobule damage model, based on prior work\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, also reproduced patterns of APAP-induced injury. Furthermore, the simulations demonstrated that even with the velocity spatial heterogeneity, a single episode of AILI does not significantly alter drug distribution dynamics, as hepatocytes have taken up nearly all the APAP before injury and necrosis occur. By incorporating spatial variability and feedback mechanisms, this model advances understanding of liver injury progression and provides a robust foundation for further work toward a goal of personalized medicine. The integration of patient-specific anatomical and clinical data enables tailored simulations, dynamically predicting liver responses to drugs based on individual physiology. This approach offers significant potential for improving therapeutic strategies for DILI, other liver diseases, and the genesis and progression of portal hypertension due to cirrhosis.\u003c/p\u003e \u003cp\u003eDespite its strengths, the current model has limitations that can be addressed by future refinements. Static boundary conditions may not fully capture dynamic changes in blood flow and vascular resistance during injury progression. Incorporating adaptive boundary conditions that evolve in real time could address this limitation. Additionally, the model employs a 2D planar capillary surrogate to simplify bifurcation structures and steady-state flow. While this preserves key geometric features, it does not fully capture three-dimensional microvascular dynamics or local permeability variations due to fibrosis or inflammation. Drug transport is also simplified, excluding active transport mechanisms. Future refinements will explore 3D reconstructions and dynamic permeability changes to enhance predictive accuracy.\u003c/p\u003e \u003cp\u003eWhile the model effectively captures vascular inflow heterogeneity from the portal vein, it needs to be further refined to capture the hepatic artery (approximately 20% of the total lobular flow) and hepatic vein flow to fully address inflow and especially drainage of the HLVT. To enhance physiological realism, future extensions will also incorporate lymph formation and flow. These future refinements will allow us to capture crucial cellular-level processes - inflammation, immune responses, and endothelial injury \u0026ndash; and hold the potential to leverage the HLVT beyond DILI toward liver disease.\u003c/p\u003e \u003cp\u003eThis study demonstrates the successful integration of multiscale simulations, bridging the gap between organ- and lobule-level dynamics. The model accurately captures blood flow across the portal vein up to its smallest capillaries and predicts drug absorption processes across scales. We address the limitations of earlier single-scale models\u0026mdash;some focused on large-scale vascular hemodynamics, others on small-scale lobular transport\u0026mdash;by unifying macro (portal vein), mesoscale (capillaries), and micro (lobules) within one patient-specific framework. This integration enables the first CFD-based simulation of drug-induced liver injury that simultaneously spans these critical scales. It provides a robust and scalable framework for investigating liver function and drug-induced damage. Incorporating patient-specific data further enhances its relevance for clinical and pharmaceutical applications, paving the way for personalized treatment strategies and advancing the understanding of liver disease and therapy. As the model evolves, and the Living Liver initiative grows, it holds the potential to transform the development and clinical application of novel therapeutics, improving precision, efficacy and safety.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003ch2\u003eMRI Data\u003c/h2\u003e\n\u003cp\u003eA standard 1.5 Tesla magnetic MRI scan of the patient\u0026rsquo;s liver was performed with gadolinium contrast enhancement. Fast breath-hold T1 and T2 sequences are obtained, with additional diffusion-weighted imaging used for further structural characterization. The images are obtained in Digital Imaging and Communications in Medicine (DICOM) standard and exported to a separate semi-autonomous segmentation software (LiverVision\u0026trade;, Medivision Ankara, T\u0026uuml;rkiye). LiverVision segments different structures within the DICOM images using sensitivity thresholding based on pixel brightness and avidity. The post-segmentation images are then rendered into 3D images using stereolithography or standard tessellation language (STL) format. The STL surfaces were smoothed and joined to a full 3D solid model to avoid small elements and all inlets and outlets were cut perpendicular to the vein diameter to obtain accurate results.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cu\u003eEthics statement:\u003c/u\u003e\u003c/strong\u003e Patient consent was not obtained because data was retrospective, based on standard care, and was analyzed anonymously\u003csup\u003e64\u003c/sup\u003e. The protocol was approved by the Institutional Review Board of the National University Hospital of Singapore in accordance with the Declaration of Helsinki (reference number 2020/01044).\u003c/p\u003e\n\u003ch2\u003eMultiscale modeling\u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eWe developed a methodology to address the variations in both time and spatial scales from the liver (organ), lobule (unit) and hepatocyte (cellular) levels (Figure 1). These models span several orders of magnitude, both temporally as APAP absorption occurs over hours to days while blood flow and drug transport occur in seconds or less, and spatially with the organ level (10 cm) to lobules (1 mm) and hepatocytes (microns). Bridging these differences requires a multiscale integration framework, detailed below.\u003c/p\u003e\n\u003cp\u003eThe liver\u0026rsquo;s spatial and temporal scales present challenges for multiscale modeling of DILI. At the organ scale (Figure 2A), the liver geometry is derived from patient-specific CT/MRI scans. Blood flow in major vessels, ranging from 100 mm to 1.5 mm in diameter, is modeled using CFD simulations. The drug input was defined to represent drug diffusion in the blood after absorption by the guts. This provides the model with clinically relevant to APAP concentration evolution. The mesoscale (Figure 2B), spanning 1.5 mm to 50 microns, focuses on capillaries and smaller vessels, where resolution limitations of CT/MRI necessitate the use of a surrogate CFD model. This mesoscale model generates 100,000 data points of outlet velocity and concentration, which serve as inputs for the lobule scale (Figure 2C). At the lobule scale, the liver\u0026rsquo;s functional units are modeled to incorporate blood flow and APAP transport into hepatocytes, as described by Camara Dit Pinto et al.\u003csup\u003e11\u003c/sup\u003e. These simulations refine predictions of cellular damage caused by APAP (Figure 2D).\u003c/p\u003e\n\u003cp\u003eThe process of linking models with different timescales is called temporal integration. At the organ level, drug absorption happens over hours or days, while blood flow simulations focus on changes that occur in seconds. To connect these scales, drug concentrations entering the blood vessels are broken into small intervals and analyzed in 5-second steps. These velocities and concentrations are then used to set boundary conditions for smaller-scale models of the liver, which capture APAP transport and hepatocyte damage dynamics to simulate how the drug spreads and causes cell damage, using 20-minute steps. Spatial integration ensures that information flows smoothly between models of different sizes. Large-scale models of the liver provide data on blood flow and drug concentration directly to medium-scale simulations, which then refine this information for smaller-scale models of the liver\u0026rsquo;s functional units. Boundary conditions at each scale are iteratively updated based on feedback from the others (Figure 3). This approach connects the big picture (how the liver absorbs drugs) with the details (how liver cells are damaged).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe multiscale drug pipeline integrates blood flow and cellular injury models to simulate APAP-induced liver damage (Figure 2), building on previously reported methods\u003csup\u003e11\u003c/sup\u003e with a key modification: boundary conditions from the portal vein are now introduced via CFD simulations. The pipeline has two main components: 1) blood flow simulation with APAP transport and 2) a cellular injury model, which run iteratively to simulate the progression of injury within the lobule. First, healthy blood flow is modeled using a species transport and advection-diffusion framework to track APAP concentration throughout the lobular blood flow. The cellular injury model then evaluates the impact of the drug at each node of the computational mesh, determining health status node by node. Simulations are conducted using a transient, pressure-based Navier-Stokes solver, enhanced with Python scripts to incorporate specific parameters outlined in Table 1. While the current pipeline is parameterized for APAP, it can be adapted to assess the effects of other drugs and treatments. For this model, drug diffusion from the gut into the bloodstream is specified to follow a predictable timeline: diffusion begins immediately after ingestion, reaching peak concentrations at 3 hours, averaging at 6 hours, and dropping to undetectable levels by 8 hours. This structured framework enables the dynamic simulation of APAP transport and injury progression within the liver.\u003c/p\u003e\n\u003ch2\u003eBlood flow simulation with drug transport and Boundary conditions\u003c/h2\u003e\n\u003cp\u003eBlood flow simulations across three scales were performed using a transient, pressure-based solver. Portal venous flow is typically modeled as steady-state; however, this study employs a transient solver to capture pulsatile hepatic inflow, better reflecting physiological dynamics, especially in pathological states like fibrosis, cirrhosis, or drug-induced vascular changes. While drug transport was modeled using a species transport framework to ensure accurate mass fractions of drug to blood within each element. To account for the non-Newtonian behavior of blood, the Carreau model for shear-thinning viscosity was applied at the capillary and lobule scales \u003csup\u003e65\u003c/sup\u003e. This model captures the variation in blood viscosity with shear rate, particularly in regions with high velocity gradients, enabling realistic simulation of pressure differences and velocity distributions. Parameters for the Carreau model were drawn from established blood flow modeling literature\u003csup\u003e66\u003c/sup\u003e and are detailed in Table 1. The viscous model ensures that shear-rate-dependent viscosity is accurately represented in areas with varying velocities.\u003c/p\u003e\n\u003cp\u003eBoundary conditions were tailored to maintain fluid dynamic continuity across scales. Inlet velocities, fitted from Ma \u003cem\u003eet al.\u003c/em\u003e\u003csup\u003e52\u003c/sup\u003e, determined the inlet pressures, as shown in Figure 4. Outlet boundary conditions at each scale were calculated to match the average outlet velocity from the previous scale to the resulting inlet pressure of the current scale. This approach ensured consistent transitions in fluid dynamics down to the lobule level, where inlet pressure was set at approximately 800 Pa\u003csup\u003e53\u003c/sup\u003e. The outlet pressures for each scale are summarized in Table 1, with a pressure span of \u0026plusmn;50 Pa based on initial conditions in Figure 4. This span remained consistent across boundary conditions, and velocity spans were verified to align with the resulting pressure spans, ensuring accurate simulation across all scales.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"\" width=\"661\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 112px;\"\u003e\n \u003cp\u003eSimulation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003eValue\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"6\" style=\"width: 112px;\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003cp\u003eSimulations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003ePermeability\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1.0x10\u003csup\u003e-14\u003c/sup\u003e to 1.0x10\u003csup\u003e-15\u003c/sup\u003e [m\u003csup\u003e2\u003c/sup\u003e]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eBlood density\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1050 [kg/m\u003csup\u003e3\u003c/sup\u003e]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eAPAP drug density\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1260 [kg/m\u003csup\u003e3\u003c/sup\u003e]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e67\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eBlood Mw\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e64,500 [kg/kmol]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e68\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eAPAP drug Mw\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e151.1 [Kg/kmol]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e67\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eMixture mass diffusivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e2.22x10\u003csup\u003e-10\u003c/sup\u003e [m\u003csup\u003e2\u003c/sup\u003e/s]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e69\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 112px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 112px;\"\u003e\n \u003cp\u003eViscosity model\u003c/p\u003e\n \u003cp\u003e(Blood/APAP)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eZero-shear viscosity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.056 [Pa\u0026middot;s]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e66\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eInfinite-shear viscosity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.00345 [Pa\u0026middot;s]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eTime constant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1.902 [s]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003ePower-law index(n)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.22\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 112px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 112px;\"\u003e\n \u003cp\u003eOrgan Scale Portal Vein Simulation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eInlet diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e10.8[mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003ePatient data\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eOutlet diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1.5-3.5 [mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003ePatient data\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eFlow rate inlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e9.48-11.43[cm\u003csup\u003e3\u003c/sup\u003e/s]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e52\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003ePressure outlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1100\u0026plusmn;50 [Pa]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eTailored to fit\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 112px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 112px;\"\u003e\n \u003cp\u003eCapillary Surrogate Model (CSM) Simulation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eDiameter Large SCM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1.5[mm] to 0.3[mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eFrom upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eDiameter Small SCM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.3[mm] to \u0026nbsp;50[\u0026micro;m]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eFrom upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eFlow rate inlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003eFrom Upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eFrom Upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eLarge CSM Pressure outlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e820\u0026plusmn;50 \u0026nbsp;[Pa]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eTailored to fit\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eSmall CSM Pressure outlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e800\u0026plusmn;50 [Pa]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 112px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" style=\"width: 112px;\"\u003e\n \u003cp\u003eVirtual Lobule Simulation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003ePortal Track diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.05 [mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e59\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eCentral Vein diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.15 [mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eLobule diameter\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e1.5 [mm]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003eFlow rate inlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003eFrom upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003eFrom upper scale\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 199px;\"\u003e\n \u003cp\u003ePressure outlet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e660 [Pa]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 162px;\"\u003e\n \u003cp\u003e\u003csup\u003e53\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eTable 1- blood flow and drug transport model parameters for each scale\u003c/p\u003e\n\u003cp\u003eTo ensure accurate inflow and outflow boundary conditions, the portal vasculature geometry was trimmed at the inlet and outlet to create smooth surfaces. A 3D simulation was constructed based on the refined geometry, capturing the complex vein structure and varying diameters, as shown in Figure 5. The mesh used an element size of 5x10⁻⁴ m, resulting in approximately 1 million elements. The vascular outlet diameters ranged from 3.5 mm to 1.5 mm, matching typical MR angiography resolution limit. The portal vein inlet flow rate varied between 9.48 cm\u0026sup3;/s and 11.43 cm\u0026sup3;/s in a one-second cycle, as shown in Figure 4, replicating the dynamics of heart-driven blood flow into the portal vein\u003csup\u003e52\u003c/sup\u003e. All 24 outlets were assigned a default pressure of 1100 \u0026plusmn; 50 Pa to ensure compatibility with the models of veins below 1.5 mm in diameter. Simulations were run until a dynamic equilibrium of outlet velocities and drug concentration was achieved. A fixed time step of 0.001 seconds was applied after an initial relaxation phase with five steps of 1x10⁻⁵ seconds. The simulation duration was set to 100 seconds, sufficient to ensure stabilization of both drug transport and outlet velocities. Results were verified for convergence with respect to both mesh size and time step.\u003c/p\u003e\n\u003cp\u003eAPAP drug concentrations varied to follow evolution clinically observed during drug ingestion, with simulations exploring concentrations from 0 to 8e-3 kg/L to model different dosages over specific time intervals\u003csup\u003e70\u003c/sup\u003e. It is important to note that the simulation included only blood flow, drug transport, and their mixture, without modeling any biological or chemical reactions.\u003c/p\u003e\n\u003cp\u003eThe Capillary Surrogate Model (CSM) was developed to bridge the spatial gap between MR resolution limit and the liver lobule. Using the Constrained Constructive Optimization (CCO) approach\u0026nbsp;\u003csup\u003e71\u003c/sup\u003e, the CSM generates approximately 10,000 individual outlets to the lobule scale. The model begins with an inlet diameter of 1.5 mm and extends to outlets as small as 50 microns, effectively connecting the macro (organ-level) and micro (lobule-level) scales. To enable hundreds of computational runs while preserving critical vascular structures, the Capillary Surrogate Model (CSM) was implemented as a 2D planar simulation. This approach preserves key structural details such as bifurcation angles, segment lengths, and diameters while allowing for sensitivity analyses and efficient flow dynamic studies. However, a key limitation of this 2D approach is that it prevents vessel crossings, which can affect local perfusion distributions compared to an in-vivo 3D vascular network. While this simplification does not fully replicate complex 3D hepatic vasculature, it allows for controlled sensitivity testing of hemodynamic parameters and can be extended to 3D models where computational feasibility permits. The CSM includes two configurations: the Large CSM and the Small CSM. The Large CSM represents a vascular segment with inlets of 1.5 mm and outlets of 0.3 mm, while the Small CSM scales down to inlets of 0.33 mm and outlets of 50 microns. Both models approximate blood flow behavior rather than providing precise, patient-specific results, serving as an efficient surrogate to capture flow dynamics. This is readily extendable to 3D models, provided the impact of various configurations on simulation outcomes and sensitivity analyses justify to additional computation burden.\u003c/p\u003e\n\u003cp\u003eFor the Large CSM, a mesh element size of 1.1x10⁻⁴ m was used, generating 70,000 elements. Inlet velocities ranged from 0.04 to 0.12 m/s, based on outputs from the 3D portal vein simulation. The outlets were assigned a default pressure of 830 \u0026plusmn; 20 Pa, ensuring compatibility with downstream models for veins below 0.3 mm. In the Small CSM, a finer mesh element size of 2x10⁻⁵ m was applied, resulting in the same number of elements. Inlet velocities ranged from 0.004 to 0.04 m/s, with outlets set to a default pressure of 800 \u0026plusmn; 20 Pa, aligning with lobule vein inlet conditions\u003csup\u003e53\u003c/sup\u003e. Both models were run until dynamic equilibrium of outlet velocities and drug concentrations was achieved. Simulations used a fixed timestep of 0.001 seconds, following an initial relaxation phase with five timesteps of 1\u0026times;10⁻⁵ seconds. Simulation durations ranged from 10 to 100 seconds, ensuring stabilization of both velocity and APAP concentration throughout the veins and outlets. Convergence was verified with respect to both mesh size and time step, confirming the reliability of the results. This 2D surrogate modeling framework efficiently captures small vein flow and drug transport dynamics, providing a scalable tool to bridge the macroscopic and microscopic scales in liver modeling. While simplified, the CSM effectively complements lobule-level models and lays the groundwork for future exploration of 3D implementations and patient-specific configurations.\u003c/p\u003e\n\u003cp\u003eThe lobule simulations were based on a 2D representation of the liver lobule, following methodologies from previous work\u003csup\u003e11\u003c/sup\u003e. A mesh element size of 1.75x10⁻⁵ m was used, resulting in approximately 60,000 elements. Boundary conditions were derived from capillary-scale simulations, with inlet velocities corresponding to portal vein outlet data and outlet pressure set to 500 \u0026plusmn; 20 Pa\u003csup\u003e53\u003c/sup\u003e. These conditions modeled normal blood flow across various portal vein outlets. A critical factor in the lobule simulation was the permeability of the porous media, which directly influenced internal velocities and pressures. Transitioning from the smooth, nonporous vascular system to the porous lobule, which encompasses both vasculature and tissue, required careful tailoring of boundary conditions. The permeability was adjusted to align the inlet pressure of 800 Pa with the outlet pressure of 500 Pa, ensuring consistent fluid dynamics across scales\u003csup\u003e51,53\u003c/sup\u003e. The simulation ran until dynamic equilibrium of outlet velocities and drug concentrations was achieved. A fixed timestep of 1 second was used, and simulations were run for 20 minutes to evaluate the progression of APAP-induced damage. Each cell in the lobule model represented hepatic arterial tissue and computed the concentration of APAP, providing data for assessing cellular injury. To model APAP-induced injury, a modified MALD damage model was employed, consistent with previous lobule simulation studies\u003csup\u003e11\u003c/sup\u003e. This model linked drug concentration in tissue cells to corresponding levels of cellular damage, enabling an evaluation of injury evolution within the lobule. The integration of these simulations allowed for detailed insight into the dynamics of APAP transport and hepatocyte damage at the cellular level.\u003c/p\u003e\n\u003ch2\u003eModeling patient conditions\u003c/h2\u003e\n\u003cp\u003eThe developed multiscale liver platform presented in this work opens the door to predicting hemodynamic changes based on patient pathologies. To showcase those capabilities, two scenarios were created. The first scenario recreates the effects of hepatocellular necrosis with injury to liver sinusoidal endothelial cells and/or endothelia of the central veins (i.e., sinusoidal obstructive syndrome). In this first scenario, an outlet pressure increase in a specific section of the small CSM is applied with aim to reduce the blood flow to a negligeable value (inducing limited drug transport in the affected lobules). The second scenario aims to recreate the effects of veinous pressure increase due to vena cava obstruction. In this second scenario, an outlet pressure increase is applied to the whole small CSM outlets. These scenarios do not represent specific known or observed clinical conditions, but are fictive scenarios created to tests the robustness of the model and explore possible theories.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eData Availability:\u0026nbsp;Segmented vasculature files can be made available upon reasonable request.\u003c/p\u003e\n\u003cp\u003eCode Availability:\u0026nbsp;ANSYS Fluent simulation files can be made available upon reasonable request.\u003c/p\u003e\n\u003cp\u003eAcknowledgments:\u0026nbsp;The authors would like to thank the Long Island University-Dassault Syst\u0026egrave;mes Center of Excellence and the BCM Learning Health System for their financial support.\u003c/p\u003e\n\u003cp\u003eAuthor contributions:\u0026nbsp;N.R.G. supervised and led the project conceptualization and administration. A.M.M performed the simulation setup, data collection, and curation. S.C.D.P. helped to implement parts of the computational pipeline, data collection, and data analysis. N.R.G., A.M.M. and S.C.D.P. analyzed the results and wrote the original draft. M.C., S.M.L., G.K.S., S.A., S.D., G.Y. and J.M.V. provided important clinical and computational intellectual content for the study design and results analysis. M.C., S.M.L., G.K.S., S.A., S.D., G.Y. and J.M.V. \u0026nbsp;contributed to writing the final draft of the manuscript. All authors critically reviewed the paper and contributed important intellectual content. All authors have read and approved the manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting interests:\u0026nbsp;S.M.L. is a Dassault Systemes employee. J.M.V. is a member of IQDILI and the NIH NIDDK Drug-Induced Liver Injury Network Data Safety, and Management Board. N.R.G. has served as a consultant for Dassault Systemes. The remaining authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003eAdditional information: Correspondence and requests for materials should be addressed to Nicolas R. Gallo.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHosack, T., Damry, D. \u0026amp; Biswas, S. Drug-induced liver injury: a comprehensive review. \u003cem\u003eTherap Adv Gastroenterol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 17562848231163410 (2023).\u003c/li\u003e\n\u003cli\u003eAndrade, R. J. \u003cem\u003eet al.\u003c/em\u003e Drug-induced liver injury. \u003cem\u003eNat Rev Dis Primers\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 58 (2019).\u003c/li\u003e\n\u003cli\u003eMostafa, F. \u0026amp; Chen, M. Computational models for predicting liver toxicity in the deep learning era. \u003cem\u003eFront Toxicol\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1340860 (2023).\u003c/li\u003e\n\u003cli\u003eEllison, C., Hewitt, M. \u0026amp; Przybylak, K. In Silico Models for Hepatotoxicity. \u003cem\u003eMethods Mol Biol\u003c/em\u003e \u003cstrong\u003e2425\u003c/strong\u003e, 355\u0026ndash;392 (2022).\u003c/li\u003e\n\u003cli\u003eSpeid, L. H., Lumley, C. E. \u0026amp; Walker, S. R. Harmonization of guidelines for toxicity testing of pharmaceuticals by 1992. \u003cem\u003eRegulatory Toxicology and Pharmacology\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 179\u0026ndash;211 (1990).\u003c/li\u003e\n\u003cli\u003eMulliner, D. \u003cem\u003eet al.\u003c/em\u003e Computational Models for Human and Animal Hepatotoxicity with a Global Application Scope. \u003cem\u003eChem. Res. Toxicol.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 757\u0026ndash;767 (2016).\u003c/li\u003e\n\u003cli\u003eXu, Z. \u003cem\u003eet al.\u003c/em\u003e Research on the Species Difference of the Hepatotoxicity of Medicine Based on Transcriptome. \u003cem\u003eFront. Pharmacol.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eMitchell, S. J. \u0026amp; Hilmer, S. N. Drug-induced liver injury in older adults. \u003cem\u003eTherapeutic Advances in Drug Safety\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 65\u0026ndash;77 (2010).\u003c/li\u003e\n\u003cli\u003eShanks, N., Greek, R. \u0026amp; Greek, J. Are animal models predictive for humans? \u003cem\u003ePhilosophy, Ethics, and Humanities in Medicine\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 2 (2009).\u003c/li\u003e\n\u003cli\u003eOlson, H. \u003cem\u003eet al.\u003c/em\u003e Concordance of the Toxicity of Pharmaceuticals in Humans and in Animals. \u003cem\u003eRegulatory Toxicology and Pharmacology\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 56\u0026ndash;67 (2000).\u003c/li\u003e\n\u003cli\u003eCamara Dit Pinto, S. \u003cem\u003eet al.\u003c/em\u003e A virtual scalable model of the Hepatic Lobule for acetaminophen hepatotoxicity prediction. \u003cem\u003enpj Digit. Med.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 340 (2024).\u003c/li\u003e\n\u003cli\u003eEsmailie, F. \u003cem\u003eet al.\u003c/em\u003e Biomechanics of Transcatheter Aortic Valve Replacement Complications and Computational Predictive Modeling. \u003cem\u003eStructural Heart\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 100032 (2022).\u003c/li\u003e\n\u003cli\u003eDowling, C., Gooley, R., McCormick, L., Firoozi, S. \u0026amp; Brecker, S. J. Patient-specific Computer Simulation: An Emerging Technology for Guiding the Transcatheter Treatment of Patients with Bicuspid Aortic Valve. \u003cem\u003eInterv Cardiol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, e26 (2021).\u003c/li\u003e\n\u003cli\u003eFinotello, A., Morganti, S. \u0026amp; Auricchio, F. Finite element analysis of TAVI: Impact of native aortic root computational modeling strategies on simulation outcomes. \u003cem\u003eMedical Engineering \u0026amp; Physics\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 2\u0026ndash;12 (2017).\u003c/li\u003e\n\u003cli\u003eAstudillo, P., De Beule, M., Dambre, J. \u0026amp; Mortier, P. Towards safe and efficient preoperative planning of transcatheter mitral valve interventions. \u003cem\u003eMorphologie\u003c/em\u003e \u003cstrong\u003e103\u003c/strong\u003e, 139\u0026ndash;147 (2019).\u003c/li\u003e\n\u003cli\u003eBaillargeon, B., Rebelo, N., Fox, D. D., Taylor, R. L. \u0026amp; Kuhl, E. The Living Heart Project: A robust and integrative simulator for human heart function. \u003cem\u003eEur J Mech A Solids\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 38\u0026ndash;47 (2014).\u003c/li\u003e\n\u003cli\u003eFlynn, D., Palmer, M., Schiestl, R., Levine, S. \u0026amp; Maeder, T. MDIC Landscape Report on Medical Device Computational Modeling and Simulation. (2023).\u003c/li\u003e\n\u003cli\u003eMorrison, T. M., Dreher, M. L., Nagaraja, S., Angelone, L. M. \u0026amp; Kainz, W. The Role of Computational Modeling and Simulation in the Total Product Life Cycle of Peripheral Vascular Devices. \u003cem\u003eJ Med Device\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 024503 (2017).\u003c/li\u003e\n\u003cli\u003eBadano, A. In silico imaging clinical trials: cheaper, faster, better, safer, and more scalable. \u003cem\u003eTrials\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 64 (2021).\u003c/li\u003e\n\u003cli\u003eAvanzato, R., Beritelli, F., Lombardo, A. \u0026amp; Ricci, C. Lung-DT: An AI-Powered Digital Twin Framework for Thoracic Health Monitoring and Diagnosis. \u003cem\u003eSensors\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 958 (2024).\u003c/li\u003e\n\u003cli\u003eZhu, L., Lu, W., Soleimani, M., Li, Z. \u0026amp; Zhang, M. Electrical Impedance Tomography Guided by Digital Twins and Deep Learning for Lung Monitoring. \u003cem\u003eIEEE Transactions on Instrumentation and Measurement\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 1\u0026ndash;9 (2023).\u003c/li\u003e\n\u003cli\u003eHurtado, D. E., Avil\u0026eacute;s-Rojas, N. \u0026amp; Concha, F. Multiscale modeling of lung mechanics: From alveolar microstructure to pulmonary function. \u003cem\u003eJournal of the Mechanics and Physics of Solids\u003c/em\u003e \u003cstrong\u003e179\u003c/strong\u003e, 105364 (2023).\u003c/li\u003e\n\u003cli\u003eWeickenmeier, J., Butler, C. A. M., Young, P. G., Goriely, A. \u0026amp; Kuhl, E. The mechanics of decompressive craniectomy: Personalized simulations. \u003cem\u003eComputer Methods in Applied Mechanics and Engineering\u003c/em\u003e \u003cstrong\u003e314\u003c/strong\u003e, 180\u0026ndash;195 (2017).\u003c/li\u003e\n\u003cli\u003eXiong, H. \u003cem\u003eet al.\u003c/em\u003e The Digital Twin Brain: A Bridge between Biological and Artificial Intelligence. \u003cem\u003eIntelligent Computing\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 0055 (2023).\u003c/li\u003e\n\u003cli\u003eSasikaladevi, N. \u0026amp; Revathi, A. Digital twin of renal system with CT-radiography for the early diagnosis of chronic kidney diseases. \u003cem\u003eBiomedical Signal Processing and Control\u003c/em\u003e \u003cstrong\u003e88\u003c/strong\u003e, 105632 (2024).\u003c/li\u003e\n\u003cli\u003eNational Academies of Sciences, Engineering, and Medicine. Opportunities and Challenges for Digital Twins in Biomedical Research: Proceedings of a Workshop-in Brief. in 26922 (National Academies Press, Washington, D.C., 2023). doi:10.17226/26922.\u003c/li\u003e\n\u003cli\u003eHo, H., Rezania, V. \u0026amp; Schwen, L. O. Editorial: Multiscale modeling for the liver. \u003cem\u003eFrontiers in Bioengineering and Biotechnology\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eSchwen, L. O., Kuepfer, L. \u0026amp; Preusser, T. Modeling approaches for hepatic spatial heterogeneity in pharmacokinetic simulations. \u003cem\u003eDrug Discovery Today: Disease Models\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 35\u0026ndash;43 (2016).\u003c/li\u003e\n\u003cli\u003eKarcaaltincaba, M. \u0026amp; Akhan, O. Imaging of hepatic steatosis and fatty sparing. \u003cem\u003eEuropean Journal of Radiology\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 33\u0026ndash;43 (2007).\u003c/li\u003e\n\u003cli\u003eGoldstein, N. S., Hastah, F., Galan, M. V. \u0026amp; Gordon, S. C. Fibrosis Heterogeneity in Nonalcoholic Steatohepatitis and Hepatitis C Virus Needle Core Biopsy Specimens. \u003cem\u003eAmerican Journal of Clinical Pathology\u003c/em\u003e \u003cstrong\u003e123\u003c/strong\u003e, 382\u0026ndash;387 (2005).\u003c/li\u003e\n\u003cli\u003eAvila, M. A. \u003cem\u003eet al.\u003c/em\u003e Reduced mRNA abundance of the main enzymes involved in methionine metabolism in human liver cirrhosis and hepatocellular carcinoma. \u003cem\u003eJournal of Hepatology\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 907\u0026ndash;914 (2000).\u003c/li\u003e\n\u003cli\u003eXie, C. \u003cem\u003eet al.\u003c/em\u003e Quantification of Hepatic Vascular and Parenchymal Regeneration in Mice. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, e0160581 (2016).\u003c/li\u003e\n\u003cli\u003eGasselhuber, A. \u003cem\u003eet al.\u003c/em\u003e Targeted drug delivery by high intensity focused ultrasound mediated hyperthermia combined with temperature-sensitive liposomes: Computational modelling and preliminary in vivovalidation. \u003cem\u003eInternational Journal of Hyperthermia\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 337\u0026ndash;348 (2012).\u003c/li\u003e\n\u003cli\u003eTanaka, T. \u003cem\u003eet al.\u003c/em\u003e Radiologic Placement of Side-hole Catheter with Tip Fixation for Hepatic Arterial Infusion Chemotherapy. \u003cem\u003eJournal of Vascular and Interventional Radiology\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 63\u0026ndash;68 (2003).\u003c/li\u003e\n\u003cli\u003eGasselhuber, A. \u003cem\u003eet al.\u003c/em\u003e Mathematical spatio-temporal model of drug delivery from low temperature sensitive liposomes during radiofrequency tumour ablation. \u003cem\u003eInternational Journal of Hyperthermia\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 499\u0026ndash;513 (2010).\u003c/li\u003e\n\u003cli\u003eBrennan, P. N., Cartlidge, P., Manship, T. \u0026amp; Dillon, J. F. Guideline review: EASL clinical practice guidelines: drug-induced liver injury (DILI). \u003cem\u003eFrontline Gastroenterol\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 332\u0026ndash;336 (2022).\u003c/li\u003e\n\u003cli\u003eGuillot, A. \u0026amp; Tacke, F. Spatial dimension of macrophage heterogeneity in liver diseases. \u003cem\u003eegastro\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eJaeschke, H. \u0026amp; Ramachandran, A. Acetaminophen Hepatotoxicity: Paradigm for Understanding Mechanisms of Drug-Induced Liver Injury. \u003cem\u003eAnnual Review of Pathology: Mechanisms of Disease\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 453\u0026ndash;478 (2024).\u003c/li\u003e\n\u003cli\u003eChidiac, A. S., Buckley, N. A., Noghrehchi, F. \u0026amp; Cairns, R. Paracetamol (acetaminophen) overdose and hepatotoxicity: mechanism, treatment, prevention measures, and estimates of burden of disease. \u003cem\u003eExpert Opinion on Drug Metabolism \u0026amp; Toxicology\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 297\u0026ndash;317 (2023).\u003c/li\u003e\n\u003cli\u003eLi, X., Tang, J. \u0026amp; Mao, Y. Incidence and risk factors of drug-induced liver injury. \u003cem\u003eLiver International\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 1999\u0026ndash;2014 (2022).\u003c/li\u003e\n\u003cli\u003eMehrpour, O., Saeedi, F., Hadianfar, A., M\u0026eacute;garbane, B. \u0026amp; Hoyte, C. Prognostic factors of acetaminophen exposure in the United States: An analysis of 39,000 patients. \u003cem\u003eHum Exp Toxicol\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, S814\u0026ndash;S825 (2021).\u003c/li\u003e\n\u003cli\u003eCoombe, D., Rezania, V. \u0026amp; Tuszynski, J. A. Dual continuum upscaling of liver lobule flow and metabolism to the full organ scale. \u003cem\u003eFrontiers in Systems Biology\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eRezania, V., Marsh, R., Coombe, D. \u0026amp; Tuszynski, J. A physiologically-based flow network model for hepatic drug elimination I: regular lattice lobule model. \u003cem\u003eTheor Biol Med Model\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 52 (2013).\u003c/li\u003e\n\u003cli\u003eRezania, V., Marsh, R., Coombe, D. \u0026amp; Tuszynski, J. A physiologically-based flow network model for hepatic drug elimination II: variable lattice lobule models. \u003cem\u003eTheor Biol Med Model\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 53 (2013).\u003c/li\u003e\n\u003cli\u003eRezania, V., Coombe, D. \u0026amp; Tuszynski, J. A. A physiologically-based flow network model for hepatic drug elimination III: 2D/3D DLA lobule models. \u003cem\u003eTheor Biol Med Model\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 9 (2016).\u003c/li\u003e\n\u003cli\u003eBonfiglio, A., Leungchavaphongse, K., Repetto, R. \u0026amp; Siggers, J. H. Mathematical Modeling of the Circulation in the Liver Lobule. \u003cem\u003eJournal of Biomechanical Engineering\u003c/em\u003e \u003cstrong\u003e132\u003c/strong\u003e, (2010).\u003c/li\u003e\n\u003cli\u003eMosharaf-Dehkordi, M. A fully coupled porous media and channels flow approach for simulation of blood and bile flow through the liver lobules. \u003cem\u003eComputer Methods in Biomechanics and Biomedical Engineering\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 901\u0026ndash;915 (2019).\u003c/li\u003e\n\u003cli\u003eLambers, L. \u003cem\u003eet al.\u003c/em\u003e Quantifying Fat Zonation in Liver Lobules: An IntegratedMultiscale In-silico Model Combining DisturbedMicroperfusion and Fat Metabolism via aContinuum-Biomechanical Bi-scale, Tri-phasic Approach. (2023).\u003c/li\u003e\n\u003cli\u003eGerh\u0026auml;usser, S. \u003cem\u003eet al.\u003c/em\u003e Simulation of zonation-function relationships in the liver using coupled multiscale models: Application to drug-induced liver injury. 2024.03.26.586870 Preprint at https://doi.org/10.1101/2024.03.26.586870 (2024).\u003c/li\u003e\n\u003cli\u003eSchliess, F. \u003cem\u003eet al.\u003c/em\u003e Integrated metabolic spatial-temporal model for the prediction of ammonia detoxification during liver damage and regeneration. \u003cem\u003eHepatology\u003c/em\u003e \u003cstrong\u003e60\u003c/strong\u003e, 2040\u0026ndash;2051 (2014).\u003c/li\u003e\n\u003cli\u003eDebbaut, C. \u003cem\u003eet al.\u003c/em\u003e A 3D porous media liver lobule model: the importance of vascular septa and anisotropic permeability for homogeneous perfusion. \u003cem\u003eComput Methods Biomech Biomed Engin\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 1295\u0026ndash;1310 (2014).\u003c/li\u003e\n\u003cli\u003eMa, R. \u003cem\u003eet al.\u003c/em\u003e Anatomically based simulation of hepatic perfusion in the human liver. \u003cem\u003eInternational Journal for Numerical Methods in Biomedical Engineering\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, e3229 (2019).\u003c/li\u003e\n\u003cli\u003eHu, J., L\u0026uuml;, S., Feng, S. \u0026amp; Long, M. Flow dynamics analyses of pathophysiological liver lobules using porous media theory. \u003cem\u003eActa Mech. Sin.\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 823\u0026ndash;832 (2017).\u003c/li\u003e\n\u003cli\u003eSchwen, L. O. \u003cem\u003eet al.\u003c/em\u003e Spatio-Temporal Simulation of First Pass Drug Perfusion in the Liver. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e1003499 (2014).\u003c/li\u003e\n\u003cli\u003eHuang, A. \u003cem\u003eet al.\u003c/em\u003e Reference values for 4D flow magnetic resonance imaging of the portal venous system. \u003cem\u003eAbdom Radiol\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 2049\u0026ndash;2059 (2023).\u003c/li\u003e\n\u003cli\u003eRohan, E., Turjanicov\u0026aacute;, J. \u0026amp; Luke\u0026scaron;, V. Multiscale modelling and simulations of tissue perfusion using the Biot-Darcy-Brinkman model. \u003cem\u003eComputers \u0026amp; Structures\u003c/em\u003e \u003cstrong\u003e251\u003c/strong\u003e, 106404 (2021).\u003c/li\u003e\n\u003cli\u003eSluka, J. P. \u003cem\u003eet al.\u003c/em\u003e A Liver-Centric Multiscale Modeling Framework for Xenobiotics. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, e0162428 (2016).\u003c/li\u003e\n\u003cli\u003eDiaz Ochoa, J. G. \u003cem\u003eet al.\u003c/em\u003e A Multi-Scale Modeling Framework for Individualized, Spatiotemporal Prediction of Drug Effects and Toxicological Risk. \u003cem\u003eFront. Pharmacol.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, (2013).\u003c/li\u003e\n\u003cli\u003eSiggers, J. H., Leungchavaphongse, K., Ho, C. H. \u0026amp; Repetto, R. Mathematical model of blood and interstitial flow and lymph production in the liver. \u003cem\u003eBiomech Model Mechanobiol\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 363\u0026ndash;378 (2014).\u003c/li\u003e\n\u003cli\u003eNishii, K., Reese, G., Moran, E. C. \u0026amp; Sparks, J. L. Multiscale computational model of fluid flow and matrix deformation in decellularized liver. \u003cem\u003eJournal of the Mechanical Behavior of Biomedical Materials\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 201\u0026ndash;214 (2016).\u003c/li\u003e\n\u003cli\u003eGeorge, S. M., Eckert, L. M., Martin, D. R. \u0026amp; Giddens, D. P. Hemodynamics in Normal and Diseased Livers: Application of Image-Based Computational Models. \u003cem\u003eCardiovasc Eng Tech\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 80\u0026ndash;91 (2015).\u003c/li\u003e\n\u003cli\u003eRold\u0026aacute;n-Alzate, A. \u003cem\u003eet al.\u003c/em\u003e Characterization of mesenteric and portal hemodynamics using 4D flow MRI: the effects of meals and diurnal variation. \u003cem\u003eAbdom Radiol\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 2106\u0026ndash;2114 (2022).\u003c/li\u003e\n\u003cli\u003eNakata, M. \u003cem\u003eet al.\u003c/em\u003e Alterations to hepatic microcirculation in thioacetamide-induced cirrhotic livers of rats. \u003cem\u003eOsaka City Med J\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 1\u0026ndash;8 (2002).\u003c/li\u003e\n\u003cli\u003eBalci, D., Kirimker, E. O., Raptis, D. A., Gao, Y. \u0026amp; Kow, A. W. C. Uses of a dedicated 3D reconstruction software with augmented and mixed reality in planning and performing advanced liver surgery and living donor liver transplantation (with videos). \u003cem\u003eHepatobiliary \u0026amp; Pancreatic Diseases International\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 455\u0026ndash;461 (2022).\u003c/li\u003e\n\u003cli\u003eYasuda, K., Armstrong, R. C. \u0026amp; Cohen, R. E. Shear flow properties of concentrated solutions of linear and star branched polystyrenes. \u003cem\u003eRheol Acta\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 163\u0026ndash;178 (1981).\u003c/li\u003e\n\u003cli\u003eWeddell, J. C., Kwack, J., Imoukhuede, P. I. \u0026amp; Masud, A. Hemodynamic Analysis in an Idealized Artery Tree: Differences in Wall Shear Stress between Newtonian and Non-Newtonian Blood Models. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e0124575 (2015).\u003c/li\u003e\n\u003cli\u003ePubChem. Acetaminophen. https://pubchem.ncbi.nlm.nih.gov/compound/1983.\u003c/li\u003e\n\u003cli\u003eBillett, H. H. Hemoglobin and Hematocrit. in \u003cem\u003eClinical Methods: The History, Physical, and Laboratory Examinations\u003c/em\u003e (eds. Walker, H. K., Hall, W. D. \u0026amp; Hurst, J. W.) (Butterworths, Boston, 1990).\u003c/li\u003e\n\u003cli\u003eRibeiro, A. \u003cem\u003eet al.\u003c/em\u003e Diffusion coefficients of paracetamol in aqueous solutions. \u003cem\u003eThe Journal of Chemical Thermodynamics\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 97\u0026ndash;99 (2012).\u003c/li\u003e\n\u003cli\u003eRemien, C. H., Adler, F. R., Waddoups, L., Box, T. D. \u0026amp; Sussman, N. L. Mathematical modeling of liver injury and dysfunction after acetaminophen overdose: Early discrimination between survival and death. \u003cem\u003eHepatology\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 727\u0026ndash;734 (2012).\u003c/li\u003e\n\u003cli\u003eWhitehead, J. F., Laeseke, P. F., Periyasamy, S., Speidel, M. A. \u0026amp; Wagner, M. G. In silico simulation of hepatic arteries: An open‐source algorithm for efficient synthetic data generation. \u003cem\u003eMedical Physics\u003c/em\u003e\u003cstrong\u003e50\u003c/strong\u003e, 5505\u0026ndash;5517 (2023).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5876496/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5876496/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDrug-induced liver injury poses significant challenges in drug development and in clinical care. This study builds on prior work developing a Human Liver Virtual Twin by creating a Multiscale Computational Fluid Dynamics framework that integrates patient-specific anatomical data to predict acetaminophen-induced liver injury as a demonstration of its capability. The model bridges vascular, lobular, and cellular scales to simulate dynamic blood flow, drug transport, and injury mechanisms that accurately reflect clinically observed spatial heterogeneity. Results demonstrate accurate blood flow dynamics, predictions of hepatocellular damage, and demonstrate a scalable framework for studying spatial heterogeneity applicable to other hepatic pathologies. This work establishes the foundational principles for a whole-organ virtual liver simulation methodology, potentially becoming a powerful tool to guide safety in therapeutic development and clinical treatment strategies, ultimately reducing reliance translation from animal models for preclinical drug testing.\u003c/p\u003e","manuscriptTitle":"Multiscale Modeling of Drug Induced Liver Injury from to Organ to Lobule","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-06 13:19:07","doi":"10.21203/rs.3.rs-5876496/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Accepted","date":"2025-05-20T00:11:31+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-16T15:40:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"197656002501084513306689989760931722982","date":"2025-05-14T06:53:57+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-05-13T18:35:52+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-30T05:41:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Digital Medicine","date":"2025-04-26T17:55:24+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f7285782-3545-4bfb-8564-119049454c8f","owner":[],"postedDate":"May 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":47886252,"name":"Biological sciences/Computational biology and bioinformatics/Computational models"},{"id":47886253,"name":"Health sciences/Medical research/Translational research"},{"id":47886254,"name":"Physical sciences/Mathematics and computing/Computational science"}],"tags":[],"updatedAt":"2025-06-30T16:04:53+00:00","versionOfRecord":{"articleIdentity":"rs-5876496","link":"https://doi.org/10.1038/s41746-025-01736-6","journal":{"identity":"npj-digital-medicine","isVorOnly":false,"title":"npj Digital Medicine"},"publishedOn":"2025-06-23 15:57:04","publishedOnDateReadable":"June 23rd, 2025"},"versionCreatedAt":"2025-05-06 13:19:07","video":"","vorDoi":"10.1038/s41746-025-01736-6","vorDoiUrl":"https://doi.org/10.1038/s41746-025-01736-6","workflowStages":[]},"version":"v1","identity":"rs-5876496","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5876496","identity":"rs-5876496","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}