A deep learning algorithm to translate and classify cardiac electrophysiology

Abstract

The development of induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) has been a critical in vitro advance in the study of patient-specific physiology, pathophysiology and pharmacology. We designed a new deep learning multitask network approach intended to address the low throughput, high variability and immature phenotype of the iPSC-CM platform. The rationale for combining translation and classification tasks is because the most likely application of the deep learning technology we describe here is to translate iPSC-CMs following application of a perturbation. The deep learning network was trained using simulated action potential (AP) data and applied to classify cells into the drug-free and drugged categories and to predict the impact of electrophysiological perturbation across the continuum of aging from the immature iPSC-CMs to the adult ventricular myocytes. The phase of the AP extremely sensitive to perturbation due to a steep rise of the membrane resistance was found to contain the key information required for successful network multitasking. We also demonstrated successful translation of both experimental and simulated iPSC-CM AP data validating our network by prediction of experimental drug-induced effects on adult cardiomyocyte APs by the latter.

Quantitative cross-species translators of cardiac myocyte electrophysiology: Model training, experimental validation, and applications

“…Discrepancy between highly nonlinear biophysical model responses and linear approximations increases with the severity of the perturbation and increases prediction uncertainty. Recently, a deep learning network approach was demonstrated for translation of immature hiPSC-CM to adult electrophysiology ( 68 ). Comparison of these methodologies may help to determine what portion of all the possible nonlinear electrophysiological mechanisms underlying these biophysical models (and actual myocyte responses in vitro) can be encoded via linear regression and when nonlinear approaches should be used.…”

Animal Disease Models and Patient-iPS-Cell-Derived In Vitro Disease Models for Cardiovascular Biology—How Close to Disease?

“…Interestingly, it has been reported that an ECGage that is >8 years greater than the chronological age is associated with a higher rate of mortality [ 82 ]. In 2000, an artificial neural network was used to detect activations during ventricular fibrillation, which resulted in a classification correctness of 92% in the test examples [ 83 ], with improvements having been made since [ 84 , 85 , 86 ].…”

Biological noise is a key determinant of the reproducibility and adaptability of cardiac pacemaking and EC coupling

“…The spectral power in Gaussian white noise is uniformly distributed across all frequencies and is normally distributed in the time domain. Gaussian noise is one of the most common descriptors of fluctuations in biological systems and is therefore implemented in simulations of ionic currents and EC coupling (e.g., Aghasafari et al, 2021 ; Sato et al, 2009 ; Krogh-Madsen et al, 2017 ; Guevara and Lewis, 1995 ). However, as discussed in further detail below, there is a paucity of analyses of [Ca 2+ ] i and electrical noise during EC coupling and therefore of computational models that incorporate experimentally determined noise; without inclusion of these data, the predictions of these models are altered.…”