Project Details
Projekt Print View

TAILSPIN - Translation of inotropic and lusitropic drug effects from rats to humans based on comprehensive in silico models.

Subject Area Anatomy and Physiology
Biophysics
Medical Physics, Biomedical Technology
Term from 2021 to 2024
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 468256475
 
Final Report Year 2024

Final Report Abstract

Impairments of myocardial contractility (inotropy) and relaxation (lusitropy) can have severe cardiovascular implications making inotropy and lusitropy of paramount importance in drug development. While the telemetered rat is a popular animal model for early screenings, notable physiological differences between the rat and human heart hinder the translation of findings. Starting at the cellular level, the first study investigated the capability of computer modelling to improve the translation of inotropic and lusitropic drug effects. To this end, rat and human ventricular cardiomyocyte models were constructed and comprehensive experimental data were used for calibration and the validation of physiological and pharmacological behaviour. Sensitivity analyses unveiled substantial differences in the inotropic and lusitropic response to various targeted proteins. To bridge this gap, an approach was developed that integrates computer modelling to predict drug effects in human ventricular cardiomyocytes based on measurements in rat ventricular cardiomyocytes. Good agreement between predicted and measured human drug effects, using synthetic and experimental data, demonstrated the potential of computer modelling to improve the drug effect translation. Future work could explore the extension to the whole heart level. Substantial progress has been made in modelling the 3D electromechanics at the whole heart level and current research focuses on the personalised calibration. Phenomenological models are preferred because the number of parameters to be inferred is small but pharmacological studies require biophysically detailed models with many more parameters making the calibration computationally very expensive. The second study introduced an automated multifidelity approach for efficient calibrations of active mechanics models. Evaluated on a cohort of seven human left ventricular electromechanics models, the results showcased not only good agreement between simulated and clinical pressure and volume transients but also low computational cost. High computational costs involved in simulations of cardiac function are increasingly addressed using surrogate models. Gaussian process emulators have become popular but struggle with handling bifurcations such as early afterdepolarizations in action potentials. This is a crucial drawback for pharmacological studies. The third study introduced a neural networkbased emulator, which was applied to the human ventricular cardiomyocyte action potential with major maximum conductances as inputs. Evaluations on synthetic data showed high accuracy of emulating normal action potentials and this held mostly true for action potentials exhibiting early afterdepolarisations. The inference of maximum conductances was also evaluated and while larger inaccuracies were observed when utilising experimental data - a limitation particularly inherent to the fact that small tissue preparations were studied - the accuracy on synthetic data remained high.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung