Toward the precision diagnosis and prevention of acquired and congenital long QT arrhythmias

The National Heart, Lung, and Blood Institute awarded The Leland Stanford Junior University $160,448 on September 30, 2025, under the Cardiovascular Diseases Research program (CFDA 93.837) to develop a deep learning artificial intelligence electrocardiogram model predicting patient-specific risk of drug-induced long QT interval prolongation during initiation of Class III anti-arrhythmic drugs such as sotalol and dofetilide. The model will enable precise identification of individual patients who will develop long QT, reducing the need for extended hospitalization and serial QT interval monitoring during drug loading while improving clinical safety in patients paradoxically at higher risk of sudden cardiac death from these arrhythmia-prevention drugs. The award also funds development of targeted therapeutics for preventing torsades de pointes in long QT patients using modern gene editing techniques, addressing diagnostic and therapeutic gaps that currently cause new drug candidates to fail FDA approval due to strict thorough QT study requirements. Performance occurs in Palo Alto, California through September 28, 2028.

Abstract

Torsades DE pointes (TDP) and subsequent sudden cardiac death (SCD) is a severe life-threatening arrhythmia that occurs in conditions that prolong the QT interval of the electrocardiogram (ECG). These conditions include both congenital LQT which can be caused by genetic mutations in various cardiac ION channels, as well as acquired LQT which is most commonly drug-induced, particularly through herg potassium channel blockade. The prevention of life-threatening TDP in drug-induced acquired LQT is of particular importance as many new drug candidates fail to make IT to market due to the FDA's strict requirements for a "thorough QT study," rejecting drug candidates that prolong the QT interval significantly. Thus, the development of both new precision diagnostics for predicting individual patient risk of acquired LQT as well as new targeted therapeutics for preventing TDP in LQT patients are greatly needed. Given the importance of QT prolongation on the FDA drug approval process, these advances would have broad implications not only on congenital LQT patients, but on the entire drug development industry. The goal of this proposal is to address these areas of need using artificial intelligence computational approaches and modern gene editing techniques. To this end, the applicant will develop a deep learning artificial intelligence ECG model to predict patient- specific risk of drug-induced LQT during the loading of class III anti-arrhythmic drugs such as sotalol and dofetilide. These drugs are used to prevent arrhythmias but paradoxically also can increase the risk of sudden cardiac death if a patient experiences QT prolongation. Therefore, patients must be hospitalized for serial QT interval monitoring during initiation. A deep learning model that can predict in advance from the baseline ECG which individual patients will develop LQT would enable US to precisely select only patients with a high confidence of success to attempt a costly and inconvenient inpatient drug load. The applicant will also address the need for new targeted therapies to prevent sudden cardiac death in both acquired and congenital LQT. Prior work has implicated the l-type calcium channel window current in the mechanism of TDP arrhythmia initiation. The applicant will investigate these predicted arrhythmia suppression effects in patient-derived congenital LQT induced pluripotent stem cells (ipsc) using crispr gene editing to reduce the window current. Improved understanding of this mechanism will be important in the pursuit of a universal therapy to prevent arrhythmia in all LQT conditions. The training in this fellowship will be instrumental in the applicant's career development as a physician-scientist in electrophysiology.