Machine Learning and Rare Variant Adjudication in Type 1 Long QT Syndrome.

Long QT syndrome (LQTS) is a clinically and genetically heterogenous disorder of myocardial repolarization that often manifests clinically as heart rate–corrected QT interval (QTc) prolongation on 12-lead ECG and increased risk of syncope and sudden cardiac death.1,2 Among phenotypically robust nonsyndromic LQTS cases (ie, persistent QTc prolongation ≥480 ms or Schwartz diagnostic score ≥3.5), ≈75% are anticipated to harbor a heterozygous pathogenic variant in 1 of the 3 major LQTS-susceptibility genes ( KCNQ1 /LQT1, ≈35%; KCNH2 /LQT2, ≈30%; and SCN5A /LQT3, ≈10%).2–4 As a result of established genotype-phenotype correlations, the identification of a putative pathogenic KCNQ1 , KCNH2 , or SCN5A genetic variant often enables use of genotype-guided approaches to risk stratification and clinical management.5 As such, current Heart Rhythm Society/European Heart Rhythm Association guidelines consider LQTS-specific genetic testing for individuals with a strong clinical suspicion of LQTS based on clinical/family history and electrocardiographic phenotype as a class I recommendation.6 See Article by Li et al However, even when potentially LQTS-causative rare variants are unearthed, the presence of a ≈3% to 8% background rate of rare KCNQ1 , KCNH2 , and SCN5A nonsynonymous variants in public exomes/genomes can make assignment of causation problematic.7–9 When coupled with the increased utilization of genetic testing to probe weak or nonexistent clinical phenotypes, whether out of concern for the marked incomplete penetrance and variable expressivity observed in most sudden cardiac death-predisposing genetic heart disorders10 or a failure to recognize the probablistic rather than binary nature of genetic testing, the net result has been a surge …