T-World: A highly general computational model of a human ventricular myocyte

13 Cardiovascular disease is the leading cause of death, demanding new tools to improve 14 mechanistic understanding and overcome limitations of stem cell and animal-based research. We 15 introduce T-World, a highly general virtual model of human ventricular cardiomyocyte suitable for 16 multiscale studies. T-World shows comprehensive agreement with human physiology, from 17 electrical activation to contraction, and is the first to replicate all key cellular mechanisms driving 18 life-threatening arrhythmias. Extensively validated on unseen data, it demonstrates strong 19 predictivity across applications and scales. Using T-World we revealed a likely sex-specific 20 arrhythmia risk in females related to restitution properties, identified arrhythmia drivers in type 2 21 diabetes, and describe unexpected pro-arrhythmic role of NaV1.8 in heart failure. T-World 22 demonstrates strong performance in predicting drug-induced arrhythmia risk and opens new 23 opportunities for predicting and explaining drug efficacy, demonstrated by unpicking effects of 24 mexiletine in Long QT syndrome 2. T-World will be available as open-source code and an online 25 app. 26 27 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 2

1 Computational modelling and simulations of cardiac cellular and organ physiology have become 2 an integral part of contemporary cardiovascular research, providing insights into basic 3 physiological mechanisms1, mechanisms of arrhythmia1, therapy guidance2, and drug safety 4 assessment3,4. Digital evidence increasingly influences real-world applications in the 5 pharmaceutical industry5 and regulatory bodies such as the FDA6 and EMA7. Combined with recent 6 advances in data availability, hardware, and software, these models are driving the vision of the 7 digital twin technology 8, producing a virtual tool that integrates clinical data acquired for an 8 individual and that enables personalised diagnosis and treatment strategies. 9 Computational modelling and simulations also hold tremendous potential to contribute to 10 reducing, replacing, and refining the use of animals in research (‘3R principles’). While some 11 animal-based studies in cardiac research are indispensable, simulations can minimise animal use 12 by guiding experimental design, predicting outcomes, and aiding interpretation. Human-specific 13 virtual cells can also predict functional implications of animal data in the context of human 14 physiology, addressing critical species differences that may, e.g., make a drug safe in mice but 15 dangerous in humans9. Correspondingly, the European Medicines Agency has recognised 16 computational modelling as a key trend in advancing 3R principles10. 17 Multiple successful models of human ventricular cardiomyocytes have been developed to 18 investigate specific mechanisms of cardiac (patho)physiology and arrhythmia. Rudy-family models 19 (ORd and ToR-ORd) excel at predicting drug responses and generating early afterdepolarisations in 20 realistic conditions, making them valuable for drug studies comprising safety and efficacy 21 assessment 3,4. Bers/Grandi-family models are known for realistic calcium handling11–13. The Ten 22 Tusscher 2006 (TP06) model is widely used to study arrhythmia related to restitution properties14. 23 Despite their strengths, each model family lacks generality, capturing only a small subset of 24 arrhythmic behaviours and manifesting important discrepancies with experimental data on 25 fundamental physiology. This limits their utility for mechanistic studies, analysing multifactorial 26 drug effects, modelling complex diseases such as Type-2 diabetes (T2D) and heart failure, or 27 integrative arrhythmia studies. Cells and their models are highly complex and include numerous 28 components connected through non-linear feedback loops. As a result, flaws in one model 29 component can cascade, leading to incorrect predictions in other components and behaviors. This 30 limits a model’s predictive power and usefulness beyond its original focus. At the same time, the 31 most innovative and relevant applications often arise precisely in these out-of-domain contexts. 32 The lack of generality is in part also why different cellular models were typically used to study 33 aspects of arrhythmogenesis at cellular versus organ level1. 34 The absence of a comprehensive and physiologically accurate virtual cardiomyocyte impedes 35 progress toward translational applications and expanding the context of use of cardiac simulations. 36 To bridge this gap and unlock the full potential of cardiac simulations in research, industry, and 37 clinic, we sought to develop a unified highly general virtual cell model. The generality should 38 include 1) accurate recapitulation of human cellular cardiac physiology and its modulation by drugs 39 or physiological changes, 2) the capability to manifest all key arrhythmogenic behaviours in 40 conditions used to provoke them experimentally. This comprises early and delayed 41 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 3 afterdepolarisations (EADs and DADs)15,16, alternans17, and steep restitution of action potentials 1 (APs)18, 3) show suitability for multiscale modelling, enabling organ-level simulations. 2 Here, we present T-World, a novel virtual human cardiomyocyte that reproduces for the first time all 3 key cellular arrhythmic mechanisms and shows comprehensive agreement with data on human 4 cardiac physiology. Comparison to data not used in model creation demonstrates robust predictive 5 accuracy at a broad range of tasks. T-World integrates electrophysiology, calcium handling, 6 cardiomyocyte contraction, sympathetic stimulation, and sex differences, enabling comprehensive 7 studies of their interactions. T-World is freely available via Matlab, CellML, C, and CUDA, with a 8 free-to-use online graphical interface for non-coders (also runnable in Python). This model 9 advances our understanding of cardiac electrophysiology and arrhythmogenesis by 1) identifying a 10 likely sex-specific arrhythmia risk in females linked to restitution properties, 2) showing 11 applicability to analysis of drug efficacy through analysis of mechanism of action of anti-arrhythmic 12 drugs, and showing excellent performance in drug safety testing and, 3) identifying causes of high 13 arrhythmia risk in T2D, and 4) suggesting NaV1.8 as a relevant treatment target in heart failure. 14

15 T-World is a virtual cell model using sets of ordinary differential equations to describe, based on 16 experimental data, the dynamics of ionic currents, fluxes, and subcellular signalling. The overall 17 cell architecture and calcium handling were mainly inspired by the Bers/Grandi family of models11–18 13, with most ionic current formulations being inspired by the ToR-ORd model4. In order to enable all 19 key arrhythmic behaviours in relevant conditions, and to avoid limitations of these frameworks with 20 regards to basic physiological behaviours and response to (patho)physiological changes, we 21 introduced numerous innovations, such as a new hybrid approach to coupling L-type calcium 22 current and ryanodine receptors, new L-type calcium current model, heavily revised model of the 23 ryanodine receptor, re-developed model of sodium-potassium pump, and a wide array of changes 24 to most cell components. The ‘World’ in the model’s name reflects the fact that model designs and 25 expertise from the whole world were essential in its creation, and it goes beyond outputs of a single 26 group. 27 Please see Supplementary Methods for a detailed description of the following: 28 1. Model architecture 29 2. Calibration and validation criteria for T-World development and evaluation 30 3. Description of equations describing the ionic currents and fluxes 31 4. Contractility representation 32 5. CaMKII and β-adrenergic signalling 33 6. Sex differences 34 7. Organ-level simulation methodology 35 8. Methodology for studies on arrhythmogenic behaviours. 36 9. Methodology for sample applications: in silico trials, type 2 modelling, and NaV1.8 current 37 investigation 38 10. Graphical user interface 39 11. Notes on implementation 40 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 4 12. Sources of implementation of other models 1 T-World is distributed as open-source code and is available at , including sample scripts demonstrating its functionality. An online graphical user 3 interface enabling running T-World simulations is available at . Background of the T-World development, including the description of various dead 5 ends that we encountered during development, will be provided at the blog underlid.blogspot.com. 6 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 5

1 Representation and validation of cell and organ physiology 2 3 Figure 1. Cell and organ physiology. A) Conceptual diagram of the T-World model and its potential applications. See 4

for a detailed diagram including all ionic currents and cellular compartments. B) Endocardial action potential of 5 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 6 T-World versus experimental ranges 19. Slightly higher peak in simulation versus data was chosen, given that our model is 1 a single cell, whereas the experimental data are in small tissue samples which show a reduced peak due to cell-to-cell 2 coupling. C) Calcium transient (CaT) of T-World with highlighted biomarkers versus experimental ranges for: CaT duration 3 at 90% recovery level (CaTD90, shown in green), time to peak (ttp, shown in yellow), and calcium transient amplitude 4 (max, shown in red), based on standard error of mean ranges data by Coppini et al.20 D) Active tension developed by T-5 World with highlighted biomarkers versus experimental ranges for: time from peak to 95% recovery (rt95, shown in green), 6 time to peak (ttp, shown in yellow), and maximum active tension (max, shown in red), based on Margara et al.21. E) 7 Independent validation of the APD prolongation or shortening induced by 1 µM E-4031 (70% IKr block), 1 µM HMR-1556 8 (90% IKs block), 1 µM nisoldipine (90% ICaL block), and 10 µM mexiletine (54% INaL, 9% IKr, 20% ICaL block) at 0.5, 1.0 and 2.0 9 Hz pacing in the four models. Drug concentrations and their effects on channel blocks are based on O’Hara et al.19 Please 10 note the distinct y-axes for the four drugs. F) Healthy ventricular model constructed from clinical MRI data and ECG 11 simulation (solid line) compared with the ECG record from the patient used for the ventricular anatomy. G) Ventricular 12 fibrillation simulation in the setting of acute ischemia, when stimulation rate is progressively increased. Heart snapshots 13 above the ECG illustrate different stages of progression towards fibrillation. 14 Based on extensive experimental data, the T-World model represents a broad range of ionic 15 currents and fluxes across distinct cellular compartments, as well as subcellular signalling 16 pathways and contractility (see Figure 1A for a high-level overview). Distinct model components 17 are described by sets of ordinary differential equations, constructed to recapitulate baseline 18 experimental data on single ionic currents and other cellular elements. Coupling all those 19 components together yields a virtual cardiomyocyte with a high degree of biological detail and 20 realism, which can be used as a model system in cardiac research. 21 The three key outputs of a cardiomyocyte model are its AP, calcium transient (CaT), and the 22 resulting active tension during contraction. T-World shows a very strong agreement with human AP 23 data19 with regards to AP duration (APD), resting membrane potential, and the overall shape of the 24 AP during plateau and recovery (Figure 1B). It is in better agreement with human AP shape than 25 most prior state-of-the-art models (Supplementary Note 1). The CaT is also in excellent agreement 26 with human data on time to peak, duration, and amplitude20 (Figure 1C). T-World incorporates the 27 Land model of contraction22 as in the work of Margara et al.21, and its outputs are fully consistent 28 with experimental data on time to peak force, amplitude, and time to 95% recovery of contractility 29 in human myocardium 21 (Figure 1D). 30 The cardiac AP is determined by the specific mixture of ionic currents, and a given AP shape can be 31 achieved through various combinations and balances of currents23. To verify that the balance of key 32 ionic currents in T-World is human-like, we validated it by simulating its exposure to four simulated 33 channel-blocking drugs at three pacing rates (Figure 1E-H). The strong predictive performance 34 predisposes T-World to applications in safety pharmacology. See Supplementary note 2 for 35 comparison to other models (noting, in particular, problematic performance of TP06). 36 Excitation-contraction coupling (ECC) in cardiomyocytes is a process that ensures that electrical 37 signals translate into muscle contraction and pumping of the heart. It involves 1) electrical 38 activation of the cell, 2) consequent opening of L-type calcium channels, 3) triggering intracellular 39 calcium release from the sarcoplasmic reticulum (SR), 4) binding of the released calcium to the 40 contractile apparatus and resulting physical contraction. To correctly represent ECC, we introduced 41 numerous new developments in T-World compared to prior models, yielding a virtual cell that is in 42 excellent agreement with available data, while being mechanistically realistic. See Supplementary 43 note 3 for details of ECC development and validation. Briefly, the model maintains the realism of 44 calcium handling from Bers/Grandi formulations, while improving upon several important 45 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 7

of the framework. First, the model shows more physiological timing of calcium release 1 in response to L-type calcium channel opening with implications for better AP shape and overall 2 model plausibility. Second, T-World correctly responds to changes in SERCA pump function, which 3 enables plausible representation of disease and sympathetic nervous activity. The model shows 4 comprehensive agreement with heart-rate-dependence of various components of ECC, such as 5 calcium transient amplitude, contraction force, or SR calcium content; something not achieved in 6 prior models. Sodium concentration rate-dependence is also captured well, as is the relative 7 contribution of SERCA, NCX, and sarcolemmal calcium pump to clearance of calcium from cytosol 8 within an AP. Finally, T-World has human-like properties of the L-type calcium current (essential in 9 ECC), including current-voltage relationship, recovery from refractoriness, and relative contribution 10 of voltage- and calcium-dependent inactivation. 11 ECC and electrophysiology are strongly modulated by the β-adrenergic (βAR) signalling pathway, 12 which mediates the myocardial response to sympathetic nervous stimulation. Our model includes 13 the Heijman et al.24 βAR description, with modifications to account for updates to ionic currents 14 and inclusion of the contractile apparatus in the model (see Supplementary Methods). The 15 integrated model was calibrated based on human AP data, with subsequent validation 16 demonstrating a correct effect on CaT and contraction dynamics (Supplementary Note 4). 17 Pronounced differences exist between hearts from females and males, which subsequently 18 translate into differential risk of various adverse cardiac outcomes25. Given the extent and 19 importance of sex differences in cardiovascular physiology, we constructed a male and a female 20 version of T-World, based on available experimental data and prior simulation approaches26–28, 21 showing correctly longer APD and slightly reduced CaT amplitude and contraction in female 22 myocytes (Supplementary Note 5), supporting the utility of T-World for studies on sex differences 23 in cardiac (patho)physiology. There are also T-World versions for endocardial, midmyocardial, and 24 epicardial myocytes. 25 The virtual cell model can be used to build a virtual organ based on clinical MRI data, which enables 26 organ-level studies and reconstruction of ECG. We assessed the model’s performance across 27 scales by building a 3D model based on a patient’s anatomy, with the simulation yielding a human-28 like ECG signal (Figure 1F) 29,30. A major application of whole-organ models is the study of 29 arrhythmia such as ventricular fibrillation (VF), where the impact of tissue-level phenomena such 30 as fibrosis and conduction heterogeneities can be considered. However, numerous advanced 31 models like ToR-ORd or ORd struggled to produce VF dynamics unless their parameters were 32 specifically tuned for this purpose (in addition to imposing a pro-arrhythmic substrate such as 33 localised ischemia)30,31. Importantly, T-World does reproduce VF , as shown in Figure 1G, where VF 34 appears in the setting of acute anteroseptal ischemia and progressively increasing rate of 35 stimulation. Initially, the electrical propagation is stable, only manifesting ST segment elevation (a 36 hallmark of acute ischemia), but as the stimulation rate is increased, re-entrant wavefronts appear 37 (Figure 1G, snapshots 2,3), and gradually progress to spiral wave breakup and VF (snapshot 4). 38 39 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 8 Cellular arrhythmic behaviours in T-World 1 2 Figure 2. EADs, DADs, and alternans in T-World. A) Experimental data showing EADs at 0.25 Hz pacing, with 85% block 3 of IKr with dofetilide32. B) EADs evoked in T-World under corresponding conditions. C) Demonstration of differential EAD 4 formation under varying degrees of IKr availability in the following types of myocytes: male, female, and female + increased 5 ICaL, reflecting the basal part of the heart in a rabbit study33. The y-axis shows action potential duration for a range of IKr 6 scaling factors (fraction of current versus baseline) on the x-axis, with sharp transitions corresponding to changes in the 7 number of EADs. Insets show APs at corresponding dashed lines. D) Examples of triggered activity resulting from DADs. 8 The end of the pre-pacing train is shown in blue, with the spontaneous activity given in red. E) Illustration of concurrent 9 oscillations in CaT and APD. LL = large/long CaT and APD respectively, SS = small/short. F) Modulation of calcium 10 alternans by reduced and increased SERCA activity, as well as by βAR stimulation. 11 Early afterdepolarisations 12 EADs, extrasystolic depolarisations during an AP, contribute to arrhythmogenesis and are 13 commonly linked to drug-induced cardiotoxicity and long QT syndromes, being typically driven by 14 the reactivation of ICa,L during prolonged APD15 (Figure 2A). T-World replicates EADs under realistic 15 conditions of drug-induced long QT (Figure 2B), similar to ToR-ORd and ORd models 4,19, with a 13-16 mV amplitude, similar to experimental observations 32. In contrast, the TP06 model requires nearly 17 tripled ICaL to manifest EADs 34 , likely due to excessive IKs providing strong repolarisation reserve. The 18 Morotti2021 model13 (the most recent human model from the Bers/Grandi family) similarly 19 necessitates a +150% ICaL increase to induce EADs (Supplementary Figure S23). 20 T-World also highlights sex differences in EAD vulnerability. The female T-World variant requires less 21 IKr inhibition to induce EADs compared to the male variant, indicating greater EAD vulnerability 22 (Figure 2C), supporting data showing higher risk of drug-induced arrhythmia in female hearts 27,35. 23 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 9 Additionally, female-specific apicobasal ICaL gradients linked to estrogen36 suggest an additional risk 1 of EADs in basal regions of the heart in female models. Spatially constrained EADs resulting from 2 such gradients are likely to increase dispersion of repolarisation and may promote 3 arrhythmogenesis at the tissue level beyond the EAD risk itself 37. 4 Delayed afterdepolarisations 5 DADs are arrhythmia triggers occurring during diastole. They result from spontaneous SR calcium 6 release, generating inward currents (primarily via NCX) that depolarise the cell38, and are 7 particularly prominent in diseased hearts, such as in heart failure39. DADs arise from stochastic 8 subcellular calcium sparks best simulated by models with spatial calcium handling and stochastic 9 gating 40. However, given the high computational cost of such models, ‘common pool’ models with 10 similar complexity as T-World are often used to emulate DAD generation efficiently, especially 11 Bers/Grandi-like models such as Morotti202111,13. By contrast, ToR-ORd cannot produce any DADs 12 due to its RyR activation mechanism, while TP06 can yield DADs following parametric changes, but 13 these differ substantially from experimental recordings41. 14 T-World manifests spontaneous calcium releases and DADs (Figure 2D), and it can generate DAD 15 trains, as observed in certain experiments42 (Supplementary Figure S24). Spontaneous releases 16 are terminated when the SR content becomes sufficiently low following the spontaneous releases, 17 similar to experimental parallel measurements of intracellular and SR calcium43. In this regard, our 18 model differs from Morotti2021, where DADs stop occurring even when SR calcium keeps 19 increasing (Supplementary Figure S25). 20 To validate DADs in the model, we confirmed that faster pre-pacing and RyR sensitisation promote 21 DADs in T-World, as seen experimentally (Supplementary Figure S26). Furthermore, for 22 applications requiring stochasticity of DADs, we developed a version of T-World that includes 23 stochastic store-overload-dependent RyR release akin to the method by Colman et al.44 24 (Supplementary Figure S27). 25 Calcium and action potential alternans 26 Cardiac alternans, a periodic oscillation between long and short APDs, creates a pro-arrhythmic 27 substrate, promoting conduction block 45 and increasing arrhythmia risk46. APD alternans is 28 typically driven by underlying CaT oscillations and occurs at rapid heart rates47. While common at 29 high pacing rates in living hearts, many computer models do not recapitulate it, including the 30 Bers/Grandi family on which most of T-World calcium handling is originally based. 31 Thanks to its improved calcium handling, T-World produces AP and CaT alternans at realistic 32 frequencies48, with mild alternans at 260 ms and pronounced alternans at 240–250 ms (Figure 2E, 33 Supplementary Figure S28). Alternans is electromechanically concordant (long APD corresponds 34 to large CaT), matching experimental data in human-relevant species49–51. T-World shows CaT 35 alternans even when a fixed AP shape is imposed, confirming calcium oscillations as the primary 36 driver (Supplementary Figure S29). 37 A major improvement of T-World compared to prior state of the art is its correct response of 38 alternans to SERCA pump changes. Conditions like heart failure or pharmacological or 39 transcriptional SERCA reduction increase alternans vulnerability52–54, with alternans appearing at 40 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 10 slower pacing rates. T-World accurately reflects this by showing alternans at slower rates with 1 SERCA reduction (Figure 2F). This is in contrast with ToR-ORd, where SERCA inhibition suppresses 2 alternans (Supplementary Figure S30). Finally, we validated T-World by showing that an increase in 3 SERCA function or βAR activation suppress alternans (Figure 2F), in line with experimental data55,56. 4 Steep S1-S2 restitution 5 Steep APD restitution promotes arrhythmias by facilitating reentry and spiral wave breakup in 6 cardiac tissue14,57. It is measured using an S1-S2 protocol, where a premature stimulus (termed S2) 7 follows a train of stimuli (termed S1), generating a curve relating APD to the preceding diastolic 8 interval (Figure 3A). Slopes of the restitution curve greater than 1 facilitate proarrhythmia, and 9 human studies show that maximum curve slopes slightly above 1 are not uncommon 58,59. The TP06 10 model has been historically popular, because of its steep restitution properties. At the same time, 11 e.g. the ToR-ORd model, on which most of T-World’s electrophysiology is based, has a relatively flat 12 restitution (peak slope ~0.5), limiting its utility in these aspects. However, our revised L-type 13 calcium current and other developments lead T-World to exhibit good agreement with experimental 14 restitution data (Figure 3A) and S1S2 slope >1 in a part of the curve for S1 interval of 1000 ms 15 (Figure 3B). The importance of steep restitution is supported by the fact that T-World can reproduce 16 VF (Figure 1G), unlike the prior ToR-ORd model, which required substantial adaptations to achieve 17 VF . 18 In 2017, Shattock et al.18 showed that the maximum slope of the restitution curve is largely 19 determined by the steady-state APD of a cell. Specifically, the longer the APD, the steeper the 20 restitution (Figure 3C), which was corroborated by multiple studies using different means of 21 changing APD 60–62. Importantly, T-World is the only model among those capable of steep restitution 22 that recapitulates this feature (Figure 3D), with the TP06 model showing a weakly inverse APD-23 slope relationship, and the Morotti2021 a strongly inverse one (Supplementary Figure S31). This 24 makes T-World uniquely suitable for studying how APD changes due to disease or drugs modulate 25 arrhythmic risk via restitution changes. 26 One notable exception to the observation by Shattock et al. is the effect of βAR stimulation, which 27 shortens APD, but steepens the S1S2 slope in humans58. As an independent validation, we 28 simulated the effect of βAR activation in the T-World model, which correctly predicted the 29 phenotype (Supplementary Figure S32). Furthermore, we validated that shortening of the S1 30 interval correctly flattens the S1S2 restitution (Supplementary Figure S33). 31 Measuring restitution slope separately for the male and female versions of T-World, we observed a 32 steeper slope in the female myocyte (Figure 3E, Supplementary Figure S34). This would point to 33 an increased risk of arrhythmia in female hearts through steeper restitution, but intriguingly, we 34 were unable to find any experimental study addressing this hypothesis. However, we were able to 35 obtain human ventricular data from the study by Lovas et al. 63 and re-analysed them for sex 36 differences in peak slope. Mean (SD) peak slope in males was 1.54 (±0.63), increasing to a mean of 37 2.38 (±0.92) in females (p=0.069, t-test) (Figure 3F). This suggests that females may have steeper 38 restitution properties, a previously underappreciated sex-specific hazard. 39 40 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 11 1 Figure 3. S1-S2 restitution and stability of arrhythmic behaviours. A) S1-S2 restitution curve in the T-World, ToR-ORd, 2 Morotti2021, and TP06 models, and a range of human studies 58,63–65. B) Comparison of S1-S2 restitution slopes across T-3 World, ToR-ORd, Morotti2021, and TP06 models. C) Positive relationship between APD of a cell and its peak S1-S2 4 restitution slope, observed experimentally 18. D) Corresponding simulation in T-World showing that when a range of ion 5 channel conductances are varied (see Methods for details), cells with longer APD generally show a steeper slope of 6 restitution. E) Steepening of restitution in female versus male myocytes in baseline T-World. F) Experimental human data 7 comparing peak restitution slope in males vs females (N=7 in both groups, p-value obtained using unpaired t-test). 8 Stability of arrhythmic behaviours 9 To validate generality and robustness of T-World, we investigated the stability of the cellular 10 arrhythmic behaviours under parameter perturbation using a population-of-models approach. 11 While it is natural for cells, living and simulated alike, to manifest arrhythmogenic behaviours at 12 slightly different conditions, the majority of cells should be fundamentally capable of manifesting 13 them. In Supplementary Note 6, we demonstrate that T-World is highly robust with regards to its 14 arrhythmia precursor capabilities, which are its intrinsic properties, rather than phenomena that 15 only occur for highly specific sets of distinct parameters for each property. 16 Using T-World to predict drug effects and elucidate arrhythmia 17 mechanisms in disease 18 The robust representation of many cellular arrhythmia mechanisms makes T-World highly suitable 19 to facilitate a better understanding of their role in (patho)physiological conditions. Here, we show 20 how T-World can advance assessment of drug effects and provide insight into arrhythmogenic 21 mechanisms in diseases that may include diabetes, heart failure, or monogenic arrhythmia 22 disorders (e.g., the Long QT syndrome). 23 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 12 Drug safety and efficacy assessment 1 Prediction of drug safety through in silico trials is a successful translational application of 2 mechanistic computer models, with considerable uptake by industry and regulators 3. This is 3 crucial as cardiac side effects are a major cause of drug attrition and market withdrawal 66. To 4 demonstrate utility of T-World for drug safety, we conducted an in silico trial using populations-of-5 models 3,4, comparing predictions against clinical risk data for 61 drugs (Figure 4A). Compared to 6 prior works, we updated drug safety annotation based on the most recent version of the 7 Crediblemeds classification 67, and pharmacological data for several compounds (see Methods). 8 A population of 343 models, constrained by experimentally informed human reference ranges for 9 APD, CaT, and contraction biomarkers (Supplementary Figure S35) was exposed to 61 drugs at 10 doses up to 100 times therapeutic levels. The model correctly predicted all no-risk drugs as safe 11 and all high-risk drugs as unsafe. Classifying the drugs into two categories of safe (no risk) and 12 unsafe (high, possible, or conditional risk) yielded a prediction accuracy of 87%, with 79% 13 sensitivity and 100% specificity (Figure 4B). This represents an improvement over the prior ToR-ORd 14 model, highlighting the robustness of T-World despite its entirely different calcium-handling system 15 and revised ion current formulations. 16 A drug effect prediction can be reliable only when the underlying drug description data are 17 accurate. T-World can identify incorrect pharmacological descriptions of drugs, which can limit 18 prediction accuracy. When a drug with known phenotypic effect (e.g. changes to APD or 19 contractility) is simulated, a discrepancy between the simulation and known reality indicates that 20 an important effect of the drug is not included in the drug description data. We illustrate this using 21 lidocaine, a safe sodium-channel blocker, where one of two available descriptions was excluded a 22 priori during data curation. Both versions block peak INa and weakly block IKr, with one additionally 23 potently blocking INaL. Lidocaine is known to shorten APD68 , and this is recapitulated only by the 24 version including the INaL block, indicating its superiority (Figure 4C,D). Interestingly, the incorrect 25 description generates EADs and falsely indicated arrhythmic risk (Figure 4C,D), highlighting the 26 need to exclude incomplete drug descriptions. In this case, exclusion of the non-INaL formulation is 27 independently supported by studies directly demonstrating INaL inhibition by lidocaine68. 28 Similarly, we also identified an inaccuracy in the description of cilostazol, with the original drug 29 description failing to predict the effect of the drug on contractility. This resulted from its arguably 30 main effect of PDE3 inhibition not being included in the pharmacological data, which focused on 31 ion channel blockade only (Supplementary Figure S36). 32 Finally, T-World can be used for studies on drug efficacy, either by identifying promising 33 combinations of single channel blocks, or by disentangling different pro- and anti-arrhythmic 34 effects of drugs with complex multi-channel profiles. Recently, the multichannel blocker mexiletine 35 was proposed against Long QT syndrome 2 (LQTS2) 69 caused by APD prolongation due to loss-of-36 function mutations in IKr. In Supplementary Note 7, we unpick the positive effect of mexiletine in a 37 LQTS2 version of T-World, linking it to dual inhibitory effect of the drug on INaL and ICaL, which 38 outweigh its IKr-blocking effect. Further research is required to assess whether the drug blocks IKs, 39 which could be problematic during βAR activity. 40 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 13 1 2 Figure 4 In silico drug safety and efficacy assessment. A) Schematic of the in silico drug trial procedure, showing how 3 pharmacological data on dose-dependent inhibition of various currents by cardioactive drugs are applied to calibrated 4 populations of models. Subsequently, arrhythmogenic behaviours are detected, scored, and the prediction can be 5 compared to reference clinical risk. B) Predicted risk scores for 61 drugs with, with color-coded clinical risk, as in 3,70. 6 Tables of true/false classifications are provided in the right part for T-World and ToR-ORd. Please see Methods for a 7 summary of how several data updates lead to a subtly different performance of ToR-ORd in our study compared to the 8 original publication4. C) Effect of the first lidocaine description (with INaL effect) on AP to the left, showing overall safety to 9 the right (no model in the model manifests an EAD). D) Similar plot for the second lidocaine description available in the 10 database, showing gradual dose-dependent APD prolongation to the left and repolarisation abnormalities to the right. In 11 C, D, lidocaine effect is shown at the maximum concentration of 100x. 12 Assessing arrhythmogenesis in type-2 diabetes 13 The generality of T-World enables the creation of predictive disease-specific models. T2D is a major 14 21st-century epidemic linked to increased mortality, with cardiovascular disease as the leading 15 cause of death. Sudden cardiac death from ventricular arrhythmia is the primary driver, yet the 16 mechanisms behind ventricular arrhythmogenesis in T2D remain poorly understood 71. Limited 17 human data on ionic currents and calcium-handling proteins 72 show only partial alignment with 18 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 14 heterogeneous animal studies. Progressive cardiac remodelling in T2D further complicates 1 consistent disease characterisation. 2 A significant portion of sudden cardiac deaths in T2D occur in patients using potentially 3 proarrhythmic drugs71. This suggests hidden cardiotoxicity in T2D, with usually safe drugs (or safe 4 drug concentrations) becoming dangerous. To investigate this, we created a range of models (D1-5 D6) reflecting key T2D phenotypes from literature, primarily using human data supplemented by 6 animal studies (Supplementary Methods – T-World applications). All models exhibited AP 7 prolongation (Figure 5A), consistent with clinical QT prolongation in T2D 73–75. All diabetic model 8 variants are more vulnerable to EADs (Figure 5B), requiring less IKr inhibition to trigger EADs. This 9 includes those with reduced ICaL (D3-D6), which could be thought to be more protected. Key drivers 10 of EAD risk were IKr reduction and INaCa increase, further heightened by CaMKII hyperactivity and 11 increased ICaL in D1-D2 (Figure 5C). ICaL reduction alone (a component of D3-D6) showed reduced 12 risk but not enough to offset other remodelling effects. This suggests that across different 13 formulations of T2D remodelling, T2D patients require less IKr inhibition to manifest EADs, therefore 14 facing higher arrhythmia risk at drug doses considered safe for non-diabetic individuals. 15 A different arrhythmogenic behaviour that is markedly increased in T2D patients is alternans76. We 16 used the D1 version of T-World, which has recapitulated the clinical observation, showing alternans 17 at slower pacing compared to non-diabetic versions (Figure 5D). Interestingly, Bonapace et al. 18 furthermore observed that alternans vulnerability is positively associated with and diastolic 19 dysfunction in T2D patients77. To investigate this phenomenon in T-World, we correlated the slowest 20 pacing rate for CaT alternans with diastolic function (tau of relaxation) in a population of models 21 with perturbed parameters. Models prone to alternans exhibited impaired relaxation, consistent 22 with clinical data (Figure 5E). We hypothesised and subsequently confirmed that reduced SERCA 23 pump function, crucial for relaxation and calcium clearance, can causally drive this relationship 24 (Supplementary Note 8). 25 NaV1.8 can drive EADs in the failing heart 26 Cardiac disease may remodel ionic currents active under physiological conditions, but it can also 27 involve expression of nonstandard ionic currents, absent in a healthy heart. We employed T-World 28 to investigate the role of NaV1.8, a primarily neuronal sodium channel subtype with recently much 29 debated functionality in the heart. While NaV1.8 is minimally expressed in healthy hearts78, it 30 appears in hypertrophic or failing hearts, and may contribute disproportionately to the late sodium 31 current INaL79,80. Increased INaL can in general promote arrhythmias by prolonging APD (leading to 32 EADs) or increasing sodium influx, reducing NCX calcium efflux and causing DADs. However, given 33 NaV1.8’s unique biophysical properties, including right-shifted activation and inactivation 34 compared to NaV1.5 (Supplementary Figure S37), we hypothesised it could directly generate EADs 35 by providing depolarising current during the late AP plateau. 36 37 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 15 1 Figure 5. Arrhythmogenesis promotion by Type 2 diabetes and NaV1.8 current. A) Comparison of APs of six distinct 2 formulations of T2D remodelling (see Methods for details. B) Action potentials of control and T2D models across a range 3 of IKr multipliers (on y-axis). Sharp transitions between colours indicate a change in the presence of an EAD. Specifically, 4 the lowest teal point on the y-axis (APD of ca. 900) indicates the highest IKr availability which supports EAD formation. C) 5 The same type of visualisation showing how single elements of T2D remodelling considered throughout D1-D6 models, 6 added to the control model, alter EAD vulnerability. The ‘+CaMKII’ column also involves an increase in INaL, as described in 7 Methods, and ‘-ICaL’ corresponds to 80% ICaL density compared to control model. D) Comparing alternans vulnerability 8 (slowest pacing rate which induces CaT alternans) between population of control vs T2D models. The calibrated 9 population of 796 models used in Results: Stability of arrhythmic behaviours was used as the control population, with T2D 10 models created by adding diabetic remodelling to each of those models. E) A scatterplot of tau of mechanical relaxation 11 versus alternans threshold in the simulated T2D population F) Comparing control T-World model AP to APs obtained when 12 two different amounts of NaV1.8 current are added (expressed as relative percentage of peak INa). 13 Introducing a small NaV1.8 current (~0.3% of peak INa) to T-World prolonged APD (Figure 5F), 14 consistent with its role as an INaL source. Strikingly, increasing NaV1.8 by 2.75-fold (to only 0.8% of 15 peak INa) triggered EADs at 1 Hz pacing (Figure 5F). These EADs emerged at a take-off potential of -16 30 mV , clearly distinct from ICaL-driven EADs at -13 mV (Figure 4B). Simultaneous tracking of NaV1.8 17 current and ICaL during EADs revealed that NaV1.8 initiates depolarisation, subsequently activating 18 ICaL in a dual-current process (Supplementary Figure S38). Therefore, NaV1.8 can directly trigger 19 EADs rather than merely prolong APD to facilitate ICaL reactivation, potentially co-explaining 20 elevated arrhythmic risk in those patients81. This mechanism suggests NaV1.8 as a possible anti-21 arrhythmic target, e.g., providing additional rationale for the use of ranolazine, which is protective in 22 the hypertrophied heart20,82, and which blocks NaV1.883. Future work on NaV1.8 modelling is 23 suggested in Supplementary Note 9. 24

25 Here, we present the development, calibration, validation, and application of T-World, a novel 26 computer model of the human ventricular cardiomyocyte. This model addresses a longstanding but 27 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 16 unresolved need for a highly general virtual cardiac cell model with robust baseline physiology, 1 capable of replicating all key cellular mechanisms of arrhythmia. The model utilises a range of new 2 developments, as well as components and ideas from two influential modelling families from the 3 Rudy4,19 and Grandi and Bers labs11,12. It is the first model to unify those two different approaches to 4 modelling cardiac cells, improving upon their strengths, while resolving their known limitations. In 5 addition to good representation of DADs stemming from the Bers/Grandi framework11,12 and 6 presence of EAD in relevant conditions such as in ToR-ORd4, the model now manifests data-like 7 restitution properties and calcium-driven alternans which correctly responds to changes in SERCA 8 pumps. The model gains credibility through rigorous independent validation on unseen data, its 9 construction from well-characterised components, and its human-specific design, which bypasses 10 the species differences inherent to animal experiments (elaborated in Supplementary Note 10). Its 11 universality, predictivity, and adaptability make it an excellent tool for basic cardiac research at 12 cellular and organ level, pharmaceutical applications, and development of patient virtual twins 8. 13 T-World presents an important step towards realising the vision of the 3Rs: reduction, refinement, 14 and (partial) replacement of animal use in research and industry. It can also work synergistically 15 with in vitro models such as induced pluripotent stem cell-derived cardiomyocytes, helping 16 interpret their so far still typically immature phenotype in the context of the adult heart84. T-World’s 17 human nature can also be leveraged to utilize animal-based measurements (such as protein level 18 changes) and predict their functional implications for the human heart, thereby “humanising” the 19 data. 20 The fact that T-World directly represents cellular biology means that it can be used to investigate 21 the modulation of its components by drugs and/or disease-related alterations. It can be for example 22 used to understand mechanisms of high arrhythmia risk in a given disease, and then help discover 23 drugs that can ameliorate such a risk. We used T-World to disentangle mexiletine’s antiarrhythmic 24 effects in Long QT syndrome type 2, showing it results from a dual INaL and ICaL blockade. Such 25 insights may be also used in future to explore new therapeutic combinations of distinct drugs. T-26 World is furthermore suitable for carrying drug arrhythmia studies in a sex-specific manner, having 27 reproduced the higher risk of drug-induced arrhythmia in females 27,35. Finally, with its 28 representation of contractility and improved excitation-contraction coupling, T-World is well-suited 29 for studying drug effects on contractility. This is another rapidly developing domain of applications 30 with high relevance for industry85. An advantage of the comprehensive nature of T-World over 31 single-purpose predictors is that it can be used to address compound queries, such as “find the 32 most anti-arrhythmic drug for a given condition without compromising contractility” . 33 T-World will be useful in preclinical drug safety testing, one of the most established translational 34 applications of non-animal methods, with significant industry adoption. We demonstrate T-World’s 35 excellent performance in drug safety testing through population-of-models in silico trials, slightly 36 surpassing the prior state-of-the-art ToR-ORd4. We believe that the main barrier to improved drug 37 safety prediction now lies in data quality rather than model quality, as shown by our data curation 38 process. Notably, we introduce a novel use of simulations to identify inaccuracies or missing data 39 in drug action datasets, based on the capability of the drug description data to reconstruct known 40 phenotype. Discrepancies between simulated and observed drug effects on e.g., action potential or 41 contractility can signal missing mechanisms in drug data description, guiding additional 42 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 17 measurements. Missing effects likely contribute to misclassification of several drugs in our study 1 (Supplementary Note 11). 2 Notably, while the drug-induced arrhythmia risk mostly results from EADs, our study indicates via 3 simulations and subsequently analysed human data that females are also at a higher risk of a steep 4 restitution slope, which may promote arrhythmia at the tissue level14,57. This finding highlights how 5 simulations can be used to drive discovery in biological data. However, given the small sample size 6 and exploratory nature of the data, further larger-scale studies are needed. Considering the 7 elevated risk of EADs and steep restitution in females, our study reinforces the urgency of sex-8 specific drug dosing and treatment guidelines, which is currently not sufficiently addressed86. 9 Arrhythmias pose a significant risk in heart disease, and T-World’s improved baseline physiology 10 and arrhythmic behaviour make it ideal for studying complex cardiac diseases. Using T-World, we 11 constructed a pilot cell model for Type 2 diabetes (T2D), a condition with high arrhythmic burden 12 but limited mechanistic understanding71. The model revealed increased risks of EADs and 13 alternans, with T2D patients facing heightened vulnerability to drug-induced arrhythmia, which can 14 explain elevated rates of sudden cardiac death in this population. The suggested strong 15 involvement of NCX in the elevated EAD risk may warrant investigation of therapeutic potential of 16 NCX blockers such as ORM-10962, which inhibit both NCX and ICaL (both pro-EAD factors) while not 17 compromising contractility 87. At the same time, ORM-10962 was shown to inhibit alternans 18 experimentally88, possibly targeting also the second pro-arrhythmic aspect in T2D. Despite 19 promising results achieved, we note the urgent need to collect new, high-quality human datasets to 20 characterise and understand how T2D dysregulates the heart, given the paucity of existing data. 21 T-World’s realistic calcium handling and ECC make it well-suited for diseases with significant 22 calcium remodelling, such as heart and post-infarction remodelling. Unlike models like ToR-ORd, T-23 World can produce DADs, important in such diseases 39. A particular strength pertaining to 24 arrhythmia mechanisms is that T-World exhibits increased alternans vulnerability with reduction in 25 SERCA pumps, both hallmarks of those diseases 52,53. This is an improvement over major prior 26 human models such as Grandi et al. 12 which lacked alternans, or ORd and ToR-ORd4,19, which do 27 not respond well to SERCA changes. 28 T-World can be applied to study the role of nonstandard channels absent in healthy hearts, but 29 present in disease. In our study, we show and explain how the “brain-type” NaV1.8 channel may 30 directly contribute to arrhythmia in the diseased heart through its unusual gating properties. Our 31

indicate NaV1.8 as a potential important target in heart failure. 32 The inclusion of βAR signalling and excitation-contraction coupling modulation makes T-World 33 highly suitable for exploring the neurocardiac axis in arrhythmia and sympathetic nervous system 34 studies89. It is particularly well applicable for studies on arrhythmia and sympathetic nervous 35 system, given that the validation has demonstrated strong predictive performance with regards to 36 modulation of multiple arrhythmic mechanisms by sympathetic nervous activity. 37 Several limitations of T-World, most of which are intrinsic to the level of detail modelled, are given in 38 Supplementary Note 12. 39 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 18 The universality of T-World and its open-source nature predisposes it to the development of derived 1 cell models and their integration in organ-scale models. This includes other excitable cells (e.g., 2 atrial, sinoatrial, Purkinje, or neuronal), inclusion or adaptation of ionic currents, and addition of 3 new signalling pathways. To expand the model generality, we anticipate it will be particularly 4 important to represent dynamic regulation of trafficking and transcription 90 enabling studies on 5 long-term remodelling, representation of metabolism and reactive oxygen species91, and 6 integration with AI-driven structural modelling92. 7 Acknowledgments 8 J Tomek is supported by the Sir Henry Wellcome Fellowship (222781/Z/21/Z). J Heijman was 9 supported by the Netherlands Heart Foundation (grant no. 01-002-2022-0118, EmbRACE: Electro-10 Molecular Basis and theRapeutic management of Atrial Cardiomyopathy, fibrillation and 11 associated) and the Netherlands Organization for Scientific Research (NWO/ZonMW Vidi 12 09150171910029). D Bers is supported by NIH grants P01-HL141084 and R01-HL092097. This work 13 was also supported by a Wellcome Trust Fellowship in Basic Biomedical Sciences to B Rodriguez 14 (214290/Z/18/Z) and the CompBioMedX project (to B Rodriguez., EP/X019446/1). This study used 15 high-performance computing resources from the Polaris supercomputer at the Argonne Leadership 16 Computing Facility (ALCF), Argonne National Laboratory, United States of America. The US 17 Department of Energy's (DOE) Innovative and Novel Computational Impact on Theory and 18 Experiment (INCITE) Program awarded access to Polaris. The ACLF is supported by the Office of 19 Science of the US DOE under Contract No. DE-AC02-06CH11357. The project was further 20 supported by the National Research Development and Innovation Office (NKFIH FK-142949 for N 21 Nagy). TM Bury is supported by a Fonds de Recherche du Québec - Nature et technologies (FRQNT) 22 postdoctoral fellowship. A Bueno-Orovio acknowledges support from the Innovate UK grant 23 10110728. M Colman is supported by Medical Research Council Career Development Award (Grant 24 Number MR/V010050/1). 25 We thank Eleonora Grandi, Stefano Morotti, Haibo Ni for useful discussions on how models derived 26 from Shannon et al. operate. We thank Dirk Gillespie, Dezso Boda, Pavel Jungwirth, and Geir Halnes 27 for their insights on how ionic driving force through open L-type calcium channels should or should 28 not be modelled. 29 For the purpose of open access, the authors have applied a Creative Commons Attribution (CC-BY-30 NC) public copyright licence to any Author Accepted Manuscript version arising from this 31 submission. 32

33 1. Natalia A Trayanova, Aurore Lyon, Julie Shade & Jordi Heijman. Computational modeling of 34 cardiac electrophysiology and arrhythmogenesis: toward clinical translation. Physiol Rev 35 104, 1265–1333 (2024). 36 2. Boyle, P . M. et al. Computationally guided personalized targeted ablation of persistent atrial 37 fibrillation. Nat Biomed Eng 3, 870 (2019). 38 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 19 3. Passini, E. et al. Human In Silico Drug Trials Demonstrate Higher Accuracy than Animal 1 Models in Predicting Clinical Pro-Arrhythmic Cardiotoxicity. Front Physiol 12, 668 (2017). 2 4. Tomek, J. et al. Development, calibration, and validation of a novel human ventricular 3 myocyte model in health, disease, and drug block. Elife 8, e48890 (2019). 4 5. Passini, E. et al. The virtual assay software for human in silico drug trials to augment drug 5 cardiac testing. J Comput Sci 52, 101202 (2021). 6 6. Credibility of Computational Models Program: Research on Computational Models and 7 Simulation Associated with Medical Devices | FDA. https://www.fda.gov/medical-8 devices/medical-device-regulatory-science-research-programs-conducted-osel/credibility-9 computational-models-program-research-computational-models-and-simulation-10 associated. 11 7. Musuamba, F . T. et al. Scientific and regulatory evaluation of mechanistic in silico drug and 12 disease models in drug development: Building model credibility. CPT Pharmacometrics Syst 13 Pharmacol 10, 804–825 (2021). 14 8. Corral-Acero, J. et al. The ‘Digital Twin’ to enable the vision of precision cardiology. Eur Heart 15 J 41, 4556–4564 (2020). 16 9. Varró, A. et al. Cardiac transmembrane ion channels and action potentials: cellular 17 physiology and arrhythmogenic behavior. Physiol Rev 101, 1083–1176 (2021). 18 10. Medicines Agency, E. EMA Regulatory Science to 2025 Strategic reflection. (2025). 19 11. Shannon, T. R., Wang, F ., Puglisi, J., Weber, C. & Bers, D. M. A mathematical treatment of 20 integrated Ca dynamics within the ventricular myocyte. Biophys J 87, 3351–3371 (2004). 21 12. Grandi, E., Pasqualini, F . S. & Bers, D. M. A novel computational model of the human 22 ventricular action potential and Ca transient. J Mol Cell Cardiol 48, 112–121 (2010). 23 13. Morotti, S. et al. Quantitative cross-species translators of cardiac myocyte 24 electrophysiology: Model training, experimental validation, and applications. Sci Adv 7, 927 25 (2021). 26 14. Tusscher, K. H. W. J. ten & Panfilov, A. V . Alternans and spiral breakup in a human ventricular 27 tissue model. Am J Physiol Heart Circ Physiol 291, H1088–H1100 (2006). 28 15. Weiss, J. N., Garfinkel, A., Karagueuzian, H. S., Chen, P . S. & Qu, Z. Early afterdepolarizations 29 and cardiac arrhythmias. Heart Rhythm 7, 1891–1899 (2010). 30 16. Liu, M. B., De Lange, E., Garfinkel, A., Weiss, J. N. & Qu, Z. Delayed afterdepolarizations 31 generate both triggers and a vulnerable substrate promoting reentry in cardiac tissue. Heart 32 rhythm : the official journal of the Heart Rhythm Society 12, 2115 (2015). 33 17. Edwards, J. N. & Blatter, L. A. Cardiac alternans and intracellular calcium cycling. Clin Exp 34 Pharmacol Physiol 41, 524–532 (2014). 35 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 20 18. Shattock, M. J. et al. Restitution slope is principally determined by steady-state action 1 potential duration. Cardiovasc Res 113, 817–828 (2017). 2 19. O’Hara, T., Virág, L., Varró, A. & Rudy, Y . Simulation of the undiseased human cardiac 3 ventricular action potential: Model formulation and experimental validation. PLoS Comput 4 Biol 7, e1002061 (2011). 5 20. Coppini, R. et al. Late sodium current inhibition reverses electromechanical dysfunction in 6 human hypertrophic cardiomyopathy. Circulation 127, 575–584 (2013). 7 21. Margara, F . et al. In-silico human electro-mechanical ventricular modelling and simulation 8 for drug-induced pro-arrhythmia and inotropic risk assessment. Prog Biophys Mol Biol 159, 9 (2021). 10 22. Land, S. et al. A model of cardiac contraction based on novel measurements of tension 11 development in human cardiomyocytes. J Mol Cell Cardiol 106, (2017). 12 23. Weiss, J. N. et al. ‘Good enough solutions’ and the genetics of complex diseases. Circ Res 13 111, 493–504 (2012). 14 24. Heijman, J., Volders, P . G. A., Westra, R. L. & Rudy, Y . Local control of β-adrenergic 15 stimulation: Effects on ventricular myocyte electrophysiology and Ca(2+)-transient. J Mol 16 Cell Cardiol 50, 863–871 (2011). 17 25. Linde, C. et al. Sex differences in cardiac arrhythmia: a consensus document of the 18 European Heart Rhythm Association, endorsed by the Heart Rhythm Society and Asia Pacific 19 Heart Rhythm Society. Europace 20, 1565–1565ao (2018). 20 26. Holmes, M. et al. Sex-specific human electromechanical multiscale in-silico models for 21 virtual therapy evaluation. bioRxiv 2025.03.14.643310 (2025) 22 doi:10.1101/2025.03.14.643310. 23 27. Yang, P . C. & Clancy, C. E. In silico Prediction of Sex-Based Differences in Human 24 Susceptibility to Cardiac Ventricular Tachyarrhythmias. Front Physiol 3, (2012). 25 28. Peirlinck, M., Sahli Costabal, F . & Kuhl, E. Sex Differences in Drug-Induced 26 Arrhythmogenesis. Front Physiol 12, 708435 (2021). 27 29. Martinez-Navarro, H., Mincholé, A., Bueno-Orovio, A. & Rodriguez, B. High arrhythmic risk in 28 antero-septal acute myocardial ischemia is explained by increased transmural reentry 29 occurrence. Scientific Reports 2019 9:1 9, 1–12 (2019). 30 30. Martinez-Navarro, H. et al. ECG analysis of ventricular fibrillation dynamics reflects 31 ischaemic progression subject to variability in patient anatomy and electrode location. Front 32 Cardiovasc Med 11, 1408822 (2024). 33 31. Martinez-Navarro, H., Zhou, X., Bueno-Orovio, A. & Rodriguez, B. Electrophysiological and 34 anatomical factors determine arrhythmic risk in acute myocardial ischaemia and its 35 modulation by sodium current availability. Interface Focus 11, 20190124 (2020). 36 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 21 32. Guo, D. et al. Electrophysiological properties of HBI-3000: A new antiarrhythmic agent with 1 multiple-channel blocking properties in human ventricular myocytes. J Cardiovasc 2 Pharmacol 57, 79–85 (2011). 3 33. Sims, C. et al. Sex, age, and regional differences in L-type calcium current are important 4 determinants of arrhythmia phenotype in rabbit hearts with drug-induced long QT type 2. 5 Circ Res 102, (2008). 6 34. Vandersickel, N., Van Nieuwenhuyse, E., Seemann, G. & Panfilov, A. V . Spatial Patterns of 7 Excitation at Tissue and Whole Organ Level Due to Early Afterdepolarizations. Front Physiol 8 8, 404 (2017). 9 35. Lehmann, M. H., Hardy, S., Archibald, D., Quart, B. & MacNeil, D. J. Sex difference in risk of 10 torsade de pointes with d,l-sotalol. Circulation 94, 2534–2541 (1996). 11 36. Papp, R. et al. Genomic upregulation of cardiac Cav1.2α and NCX1 by estrogen in women. 12 Biol Sex Differ 8, 1–11 (2017). 13 37. Qu, Z. et al. R-on-T and the initiation of reentry revisited: Integrating old and new concepts. 14 Heart Rhythm 19, 1369–1383 (2022). 15 38. Xie, L. H. & Weiss, J. N. Arrhythmogenic consequences of intracellular calcium waves. Am J 16 Physiol Heart Circ Physiol 297, 997–1002 (2009). 17 39. Hoeker, G. S., Katra, R. P ., Wilson, L. D., Plummer, B. N. & Laurita, K. R. Spontaneous calcium 18 release in tissue from the failing canine heart. Am J Physiol Heart Circ Physiol 297, H1235 19 (2009). 20 40. Colman, M. A. et al. Multi-Scale Computational Modeling of Spatial Calcium Handling From 21 Nanodomain to Whole-Heart: Overview and Perspectives. Front Physiol 13, (2022). 22 41. Roshan, N. & Pandit, R. Multiscale studies of delayed afterdepolarizations I: A comparison of 23 two biophysically realistic mathematical models for human ventricular myocytes. Preprint at 24 https://arxiv.org/abs/2307.10134v1 (2023). 25 42. Voigt, N. et al. Enhanced sarcoplasmic reticulum Ca 2+ Leak and increased Na +-Ca 2+ 26 exchanger function underlie delayed afterdepolarizations in patients with chronic atrial 27 fibrillation. Circulation 125, 2059–2070 (2012). 28 43. Domeier, T. L., Maxwell, J. T. & Blatter, L. A. β-Adrenergic stimulation increases the intra-29 sarcoplasmic reticulum Ca2+ threshold for Ca2+ wave generation. J Physiol 590, 6093 30 (2012). 31 44. Colman, M. A. Arrhythmia mechanisms and spontaneous calcium release: Bi-directional 32 coupling between re-entrant and focal excitation. PLoS Comput Biol 15, e1007260 (2019). 33 45. Weiss, J. N. et al. From pulsus to pulseless: The saga of cardiac alternans. Circ Res 98, 1244–34 1253 (2006). 35 46. Tanno, K. et al. Microvolt T-wave alternans as a predictor of ventricular tachyarrhythmias: a 36 prospective study using atrial pacing. Circulation 109, 1854–1858 (2004). 37 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 22 47. Pruvot, E. J., Katra, R. P ., Rosenbaum, D. S. & Laurita, K. R. Role of Calcium Cycling Versus 1 Restitution in the Mechanism of Repolarization Alternans. Circ Res 94, 1083–1090 (2004). 2 48. Koller, M. L. et al. Altered dynamics of action potential restitution and alternans in humans 3 with structural heart disease. Circulation 112, 1542–1548 (2005). 4 49. Fukaya, H. et al. Arrhythmogenic cardiac alternans in heart failure is suppressed by late 5 sodium current blockade by ranolazine. Heart Rhythm 16, 281–289 (2019). 6 50. Wan, X. et al. New experimental evidence for mechanism of arrhythmogenic membrane 7 potential alternans based on balance of electrogenic I(NCX)/I(Ca) currents. Heart Rhythm 9, 8 1698–1705 (2012). 9 51. Wang, L. et al. Optical Mapping of Sarcoplasmic Reticulum Ca2+ in the Intact Heart: 10 Ryanodine Receptor Refractoriness During Alternans and Fibrillation. Circ Res 114, 1410 11 (2014). 12 52. Koller, M. L., Riccio, M. L. & Gilmour, R. F . Dynamic restitution of action potential duration 13 during electrical alternans and ventricular fibrillation. Am J Physiol 275, H1635–H1642 14 (1998). 15 53. Wang, L., Myles, R. C., Lee, I. J., Bers, D. M. & Ripplinger, C. M. Role of Reduced Sarco-16 Endoplasmic Reticulum Ca2+-ATPase Function on Sarcoplasmic Reticulum Ca2+ Alternans 17 in the Intact Rabbit Heart. Front Physiol 12, 656516 (2021). 18 54. Wan, X., Laurita, K. R., Pruvot, E. J. & Rosenbaum, D. S. Molecular correlates of 19 repolarization alternans in cardiac myocytes. J Mol Cell Cardiol (2005) 20 doi:10.1016/j.yjmcc.2005.06.004. 21 55. Tomek, J. et al. Β-Adrenergic receptor stimulation and alternans in the border zone of a 22 healed infarct: An ex vivo Study and Computational Investigation of Arrhythmogenesis. Front 23 Physiol 29, (2019). 24 56. Cutler, M. J., Wan, X., Laurita, K. R., Hajjar, R. J. & Rosenbaum, D. S. Targeted SERCA2a gene 25 expression identifies molecular mechanism and therapeutic target for arrhythmogenic 26 cardiac alternans. Circ Arrhythm Electrophysiol 2, 686 (2009). 27 57. Child, N. et al. An activation-repolarization time metric to predict localized regions of high 28 susceptibility to re-entry. Heart Rhythm 12, 1644 (2015). 29 58. Taggart, P . et al. Effect of adrenergic stimulation on action potential duration restitution in 30 humans. Circulation 107, 285–289 (2003). 31 59. Nash, M. P . et al. Whole heart action potential duration restitution properties in cardiac 32 patients: a combined clinical and modelling study. Exp Physiol 91, 339–354 (2006). 33 60. Jing, L., Brownson, K. & Patwardhan, A. Role of slow delayed rectifying potassium current in 34 dynamics of repolarization and electrical memory in swine ventricles. J Physiol Sci 64, 185 35 (2014). 36 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 23 61. Tolkacheva, E. G., Anumonwo, J. M. B. & Jalife, J. Action Potential Duration Restitution 1 Portraits of Mammalian Ventricular Myocytes: Role of Calcium Current. Biophys J 91, 2735–2 2745 (2006). 3 62. Yamauchi, S. et al. Restitution Properties and Occurrence of Ventricular Arrhythmia in LQT2 4 Type of Long QT Syndrome. J Cardiovasc Electrophysiol 13, 910–914 (2002). 5 63. Árpádffy-Lovas, T. et al. Electrical Restitution and Its Modifications by Antiarrhythmic Drugs 6 in Undiseased Human Ventricular Muscle. Front Pharmacol 11, 1 (2020). 7 64. Orini, M., Taggart, P ., Srinivasan, N., Hayward, M. & Lambiase, P . D. Interactions between 8 Activation and Repolarization Restitution Properties in the Intact Human Heart: In-Vivo 9 Whole-Heart Data and Mathematical Description. PLoS One 11, e0161765 (2016). 10 65. Pak, H. N. et al. Spatial Dispersion of Action Potential Duration Restitution Kinetics Is 11 Associated with Induction of Ventricular Tachycardia/Fibrillation in Humans. J Cardiovasc 12 Electrophysiol 15, 1357–1363 (2004). 13 66. Mamoshina, P ., Rodriguez, B. & Bueno-Orovio, A. Toward a broader view of mechanisms of 14 drug cardiotoxicity. Cell Rep Med 2, 100216 (2021). 15 67. crediblemeds.org. (2024). 16 68. Johannesen, L. et al. Late sodium current block for drug-induced long QT syndrome: Results 17 from a prospective clinical trial. Clin Pharmacol Ther 99, 214–223 (2016). 18 69. Crotti, L. et al. Therapeutic Efficacy of Mexiletine for Long QT Syndrome Type 2: Evidence 19 From Human Induced Pluripotent Stem Cell–Derived Cardiomyocytes, Transgenic Rabbits, 20 and Patients. Circulation 150, (2024). 21 70. Tomek, J. et al. Development, calibration, and validation of a novel human ventricular 22 myocyte model in health, disease, and drug block. Elife 8, e48890 (2019). 23 71. Lynge, T. H. et al. Sudden cardiac death among persons with diabetes aged 1-49 years: a 10-24 year nationwide study of 14 294 deaths in Denmark. Eur Heart J 41, 2699–2706 (2020). 25 72. Ashrafi, R. et al. Arrhythmogenic gene remodelling in elderly patients with type 2 diabetes 26 with aortic stenosis and normal left ventricular ejection fraction. Exp Physiol 102, 1424–1434 27 (2017). 28 73. Veglio, M. et al. Prevalence of increased QT interval duration and dispersion in type 2 29 diabetic patients and its relationship with coronary heart disease: a population-based 30 cohort. J Intern Med 251, 317–324 (2002). 31 74. Ninkovic, V . M. et al. Prevalence and risk factors for prolonged QT interval and QT dispersion 32 in patients with type 2 diabetes. Acta Diabetol 53, 737–744 (2016). 33 75. Bertrand, A., Lewis, A., Camps, J., Grau, V . & Rodriguez, B. Multi-modal characterisation of 34 early-stage, subclinical cardiac deterioration in patients with type 2 diabetes. Cardiovasc 35 Diabetol 23, 371 (2024). 36 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 24 76. Perkiömäki, J. et al. Heart Rate Turbulence and T-Wave Alternans in Patients with Coronary 1 Artery Disease: The Influence of Diabetes. Ann Noninvasive Electrocardiol 20, 481–487 2 (2015). 3 77. Bonapace, S. et al. Relationship Between Early Diastolic Dysfunction and Abnormal 4 Microvolt T-Wave Alternans in Patients With Type 2 Diabetes. Circ Cardiovasc Imaging 4, 5 408–414 (2011). 6 78. Casini, S. et al. Absence of Functional Nav1.8 Channels in Non-diseased Atrial and 7 Ventricular Cardiomyocytes. Cardiovasc Drugs Ther 33, 649–660 (2019). 8 79. Ahmad, S. et al. The functional consequences of sodium channel NaV1.8 in human left 9 ventricular hypertrophy. ESC Heart Fail 6, 154 (2019). 10 80. Dybkova, N. et al. CaMKII delta interaction with neuronal sodium channel Nav1.8 11 contributes to arrhythmogenic triggers in failing human and mouse cardiomyocytes. Eur 12 Heart J 41, (2020). 13 81. Bengel, P . et al. Contribution of the neuronal sodium channel NaV1.8 to sodium- and 14 calcium-dependent cellular proarrhythmia. J Mol Cell Cardiol 144, 35–46 (2020). 15 82. Coleman, J. A. et al. Effects of ranolazine on the arrhythmic substrate in hypertrophic 16 cardiomyopathy. Front Pharmacol 15, 1379236 (2024). 17 83. Rajamani, S., Shryock, J. C. & Belardinelli, L. Block of tetrodotoxin-sensitive, Na(V)1.7 and 18 tetrodotoxin-resistant, Na(V)1.8, Na+ channels by ranolazine. Channels (Austin) 2, 449–460 19 (2008). 20 84. Ernst, P ., Bidwell, P . A., Dora, M., Thomas, D. D. & Kamdar, F . Cardiac calcium regulation in 21 human induced pluripotent stem cell cardiomyocytes: Implications for disease modeling 22 and maturation. Front Cell Dev Biol 10, 986107 (2023). 23 85. Trovato, C. et al. In silico predictions of drug-induced changes in human cardiac contractility 24 agree with experimental recordings. Front Pharmacol 16, 1500668. 25 86. DeFilippis, E. M. & Van Spall, H. G. C. Is it Time for Sex-Specific Guidelines for 26 Cardiovascular Disease? J Am Coll Cardiol 78, 189–192 (2021). 27 87. Oravecz, K. et al. Inotropic effect of NCX inhibition depends on the relative activity of the 28 reverse NCX assessed by a novel inhibitor ORM-10962 on canine ventricular myocytes. Eur J 29 Pharmacol 818, 278–286 (2018). 30 88. Szlovák, J. et al. Blockade of sodium-calcium exchanger via ORM-10962 attenuates cardiac 31 alternans. J Mol Cell Cardiol (2021) doi:10.1016/j.yjmcc.2020.12.015. 32 89. Shivkumar, K. et al. Clinical neurocardiology defining the value of neuroscience-based 33 cardiovascular therapeutics. J Physiol 594, 3911–3954 (2016). 34 90. Meier, S., Grundland, A., Dobrev, D., Volders, P . G. A. & Heijman, J. In silico analysis of the 35 dynamic regulation of cardiac electrophysiology by Kv 11.1 ion-channel trafficking. J Physiol 36 601, 2711–2731 (2023). 37 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint 25 91. Cortassa, S., Juhaszova, M., Aon, M. A., Zorov, D. B. & Sollott, S. J. Mitochondrial Ca2+, redox 1 environment and ROS emission in heart failure: Two sides of the same coin? J Mol Cell 2 Cardiol 151, 113–125 (2021). 3 92. Ngo, K., Yang, P .-C., Yarov-Yarovoy, V ., Clancy, C. E. & Vorobyov, I. Harnessing AlphaFold to 4 reveal hERG channel conformational state secrets. Elife 13, (2024). 5 6 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.24.645031doi: bioRxiv preprint