Digital Twins from the Atom to the Rhythm

PROJECT SUMMARY: Our overarching goal is to develop a transformative integrative clinical, experimental and in silico-based pipeline to create a digital twin technology for patient-specific prediction. Digital twin technology holds the promise of the development and application of virtual models that replicate physiological

processes and characteristics of diseases to reveal mechanisms, simulate disease progression, identify potential drug targets and simultaneously predict drug efficacy. While our planned approach is broadly applicable, here we will apply digital twins to the problem of identification of cardiac drug targets and prediction of the efficacy or

cardiotoxicity of drugs in individuals. A major strength of our digital twin approach is that it incorporates data from the atomic structure to the cardiac rhythm, allowing the inclusion of individual differences that affect individual protein structure, cellular electrophysiology and electrocardiograms. Digital twins will allow for improved

understanding of how variation between individuals modifies disease severity and drug cardiotoxicity risks. Such a technology is possible now due to the maturity of deep-learning based modeling and simulation approaches in conjunction with the increasing availability of ion channel protein structures in physiologically relevant states,

and the development of patient-specific induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs). Combining these developments will allow for the realization of high throughput testing for individuals to determine their disease- and drug-related risks. Indeed, our preliminary data indicate the promise of a new deep learning

method to extract in silico representations of individual cellular electrophysiology and Ca2+ handling to digitally replicate the mechanistic cellular fingerprint. We will build a new digital twin framework across multiple system scales by bringing together new methods in atomistic scale simulation with recently developed cellular level

models and deep learning networks to discover new protocols to extract needed model parameters from data and for translating from iPSC-CM to adult cardiac myocyte electrophysiology. We will develop and test an experimental and computational digital twin platform applied to problem of personal cardiac disease expression

and drug-induced cardiotoxicity via a combined computational-experimental approach that will allow the construction, prediction and validation of patient-specific digital twin cardiac cells. We aim to 1) Develop cardiac ion channel protein digital twins for structure and function prediction, 2) Develop cardiac myocyte digital twins,

and 3) Predict the patient-specific cardiac safety pharmacology of individual drugs and combined therapeutics. We are bringing together model simulations at the level of the atom in a totally new way to include genetic mutations spanning benign variants to ones with known arrhythmia risks (from which all other models can be

developed by extension) and predict their impact on drug interactions and biological function modulations at different scales. The proposed studies have the potential to conceptually transform the field by generating an integrative, high-throughput framework that predicts individual responses to disease and drugs.