Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings

Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings View ORCID Profile Aleksandr Khudiakov , View ORCID Profile Manuela Mura , View ORCID Profile Federica Giannetti , View ORCID Profile Vladislav Leonov , View ORCID Profile Chiara Alberio , View ORCID Profile Marem Eskandr , View ORCID Profile Paul A Brink , View ORCID Profile Lia Crotti , View ORCID Profile Massimiliano Gnecchi , View ORCID Profile Peter J Schwartz , View ORCID Profile Luca Sala doi: https://doi.org/10.1101/2025.03.25.25324187 Aleksandr Khudiakov a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Aleksandr Khudiakov Manuela Mura b Translational Cardiology Unit, Fondazione IRCCS Policlinico San Matteo , Pavia, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Manuela Mura Federica Giannetti a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Federica Giannetti Vladislav Leonov a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy c Department of Surgery , Dentistry, Pediatrics and Gynecology, Cardiovascular Science, The University of Verona, Policlinico G.B. Rossi , Verona, Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Vladislav Leonov Chiara Alberio a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Chiara Alberio Marem Eskandr a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy d Department of Biotechnology and Biosciences, University of Milano – Bicocca , Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Marem Eskandr Paul A Brink e Department of Medicine, University of Stellenbosch , Tygerberg, South Africa MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Paul A Brink Lia Crotti a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy f Department of Medicine and Surgery, University of Milano-Bicocca , Milan, Italy MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Lia Crotti Massimiliano Gnecchi b Translational Cardiology Unit, Fondazione IRCCS Policlinico San Matteo , Pavia, Italy h Department of Molecular Medicine, Unit of Cardiology, University of Pavia , Pavia, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Massimiliano Gnecchi Peter J Schwartz a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Peter J Schwartz Luca Sala a Istituto Auxologico Italiano IRCCS, Center for Cardiac Arrhythmias of Genetic Origin and Laboratory of Cardiovascular Genetics , Milan, Italy d Department of Biotechnology and Biosciences, University of Milano – Bicocca , Italy PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Luca Sala For correspondence: luca.sala{at}auxologico.it Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Background Long QT syndrome (LQTS) is a life-threatening genetic disorder characterized by prolonged QT intervals on electrocardiograms. Congenital forms are mostly associated with variants in the KCNQ1 and KCNH2 genes. Among pathogenic or likely pathogenic (P/LP) variants, some are associated with a significantly higher incidence of cardiac events compared to others. While therapies have significantly reduced mortality, some patients are unresponsive or intolerant to therapy, perpetuating their arrhythmic risk, including sudden cardiac death. Current approaches for risk stratification are insufficient, highlighting the critical need for more accurate identification and management of patients carrying high risk genetic variants. Objectives To develop a refined risk stratification model for P/LP variants by applying machine learning classification to electrophysiological data measured in human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs). Methods Eleven patient-specific hiPSC lines carrying six P/LP variants in KCNQ1 or KCNH2 were differentiated to cardiomyocytes (hiPSC-CMs). Electrophysiological responses from multielectrode array recordings at baseline and after application of selective ion channel blockers or pro-arrhythmic compounds were used to train a machine learning model to classify variant-specific risk levels based on in vitro electrophysiological readouts. Results Our findings revealed a correlation between variant risk level, hiPSC-CM electrophysiological profiles, and drug responses. The machine learning classifier, trained on multielectrode array recordings, achieved 89% accuracy in classification of P/LP genetic variants according to the associated risk levels. Conclusions This study demonstrates that integrating hiPSC-CM electrophysiological profiling with machine learning provides a robust method to improve variant-specific risk stratification for LQTS patients. Clinical Aspects Understanding which patients may be at risk of cardiac events or sudden cardiac death is crucial to implement appropriate preventive measures. This study leverages patient-specific in vitro models and machine learning to improve the risk stratification of pathogenic/likely pathogenic variants associated with LQTS, better supporting clinical decisions related to risk assessment and management of LQTS patients. This scalable approach can be implemented across multiple centres, enhancing the risk stratification of LQTS variants beyond what is currently possible when clinical data are limited. Translational Outlook Machine learning-based variant risk stratification is a novel approach for integrating hiPSC-CM-derived electrophysiological data into clinical workflows. While this study demonstrates the feasibility of our approach, further research is required to validate these findings across larger and more diverse patient cohorts. Additionally, efforts to standardize the pipeline and adapt it for multicentric implementation are necessary. Download figure Open in new tab Graphical Abstract Discriminating LQTS patients at high or low risk for sudden death is a clinical challenge. Improved stratification of pathogenic/likely pathogenic variants is achievable through machine learning classification on in vitro electrophysiological data. Integration of the clinical workflow with data from patient-specific in vitro models will enhance risk stratification. Introduction The long QT syndrome (LQTS) is a life-threatening disease of genetic origin characterized by a prolonged QT interval on the electrocardiogram, by propensity to lethal arrhythmias especially under stress, and by an increased sensitivity to drugs affecting cardiac repolarization through block of the I Kr potassium current 1 – 4 . Appropriate therapies based on beta-blockers 5 , left cardiac sympathetic denervation 6 , 7 and sodium channel blockers 8 , 9 have effectively reduced the risk for patients to develop potentially-lethal ventricular arrhythmias or sudden cardiac death (SCD), limiting the use of an implantable cardioverter defibrillator only to selected patients 5 , 10 ; one of the main challenges for clinicians dealing with LQTS is the identification, clinical management and protection of patients who are more at risk of developing SCD. Advancements in next-generation sequencing have led to the discovery of numerous variants in genes encoding cardiac ion channels, subunits, or associated proteins implicated in LQTS 11 , 12 . While these insights have guided gene-specific patient management 8 , current genetic interpretation guidelines 13 often do not help identify patients at higher risk, with a large number of variants classified as variants of uncertain significance (VUS). Even variants classified as pathogenic or likely pathogenic (P/LP) according to the American College of Medical Genetics and Genomics (ACMG) guidelines may exhibit divergent risks of life-threatening cardiac events, including SCD 14 , 15 , requiring profoundly different clinical management. Accurate clinical risk assessment for a variant requires data from many affected individuals and this is not feasible for most of the variants, limiting precision medicine and hampering the development of variant-specific therapies 16 , 17 . The classification of variant pathogenicity through in vitro or in vivo studies offers a promising solution to address this limitation 18 , 19 . Functional studies using patient-specific cardiomyocytes derived from human induced pluripotent stem cells (hiPSC-CMs) further refined this, providing high-quality functional data on the effect of the variants on action potentials, calcium transients or contractility. hiPSC-CM-based models have been particularly useful for disease modeling of congenital or acquired cardiac disorders such as LQTS 20 – 24 , Jervell and Lange-Nielsen syndrome (JLNS) 25 , Timothy syndrome 26 , cardiomyopathies 27 – 29 , congenital heart defects 30 , drug testing 31 – 33 and drug repurposing 34 . In the present study, using hiPSC-CMs from eleven patients, we focused on six well-characterized P/LP variants from the two most prevalent LQTS subtypes, LQT1 and LQT2, which together account for 90% of cases. These variants were identified in individuals with clinically heterogeneous presentations, from normal to prolonged QT intervals and displaying low or high incidence of life-threatening cardiac events per variant (defined here as Low and High Risk P/LP variants). Using patient-derived hiPSC-CMs and high-throughput multielectrode arrays (MEA), we investigated i) whether hiPSC-CMs carrying P/LP variants with differing clinical severities exhibit distinct drug responses in vitro ; ii) whether in vitro phenotypes match clinical records and could be used for a more accurate variant-associated risk stratification; iii) whether accurate P/LP variant risk stratification can be automated through machine learning on in vitro readouts. Methods Ethical statement This study was conducted in accordance with the Declaration of Helsinki and the ethics committee of Istituto Auxologico Italiano IRCCS gave ethical approval for this work (Approval number: 2020_10_20_07). Appropriate informed consents were obtained from all donors. Patient-specific hiPSC lines included in the study A total of eleven hiPSC lines were utilized in this study. HiPSC lines were derived from LQT1, JLNS, and LQT2 patients carrying the variants in KCNQ1 or KCNH2 gene (See Table 1 for detailed information). The line carrying the KCNH2 p.A561V variant was provided by Joseph C. Wu, MD, PhD, at the Stanford Cardiovascular Institute. As a wild-type control (WT), the WTC-11 line, generated from a healthy subject, was obtained from the Coriell Institute for Medical Research (catalog No. GM25256) and used as a bona fide wild-type reference 35 . Detailed characterization of each hiPSC line, including genetic and functional properties, is provided in Supplemental Table 1 and Supplemental Figure 1 and Supplemental Figure 2. View this table: View inline View popup Download powerpoint Table 1. Clinical characteristics of the patient cohort included in the study. Cardiac events frequency assessment Clinical data on documented cardiac events (syncope, sustained ventricular tachycardia, appropriate ICD shock, sudden cardiac arrest, sudden cardiac death), QTc values and Schwartz scores (when reported) were extracted from our internal database of LQTS patients, The Human Gene Mutation Database (HGMD) 36 and Variant Browser database 37 , 38 . The frequency of cardiac events was calculated as the proportion of patients with a history of such events relative to the total number of patients carrying the specific variant. hiPSC culture and differentiation to hiPSC-CMs HiPSCs were cultured on multiwell plates coated with recombinant human vitronectin (Gibco) in Essential 8 Flex medium kit (Gibco). HiPSCs were plated on Matrigel (Corning) coated plates for cardiac differentiation. hiPSCs were differentiated to hiPSC-CMs following a small-molecule based protocol 39 , purified through glucose starvation 40 , starting from day 7, and cryopreserved in Bambanker (Nippon Genetics) at days 9-16. For the subsequent experiments hiPSC-CMs were thawed, replated at low density and expanded as previously reported 41 . A list of main reagents used in the study is reported in Supplemental Table 2. Data were collected from at least three independent differentiations of each hiPSC line for each experiment. Patch-clamp For patch clamp experiments, metabolic maturation of hiPSC-CMs was performed as previously described 42 . Action potentials and I Ks currents were recorded at 37 °C from isolated hiPSC-CMs plated on glass coverslips as described in detail in the Supplemental Methods section. Action potentials were recorded in perforated patch mode under 1 Hz pacing. MultiElectrode Arrays HiPSC-CMs were plated on 24-well multiwell MEAs (MultiChannel Systems) coated with bovine fibronectin (Merck) as previously described 43 , 44 . The recordings were performed on spontaneously beating hiPSC-CMs at 37°C. The detailed description of drug treatments and analysis of the recordings is present in the Supplemental Methods section. Pro-arrhythmic compound selection The selection of pro-arrhythmic compounds was performed using the CredibleMeds® QT prolonging drugs database 45 intersected with OpenFDA database ( https://open.fda.gov/ ). OpenFDA data were used to prioritize compounds based on their occurrence in the FDA Adverse Event Reporting System. The following query terms were utilized to extract the top 100 compounds from the OpenFDA database: “ Cardiac failure ”, “ Torsade de pointes ”, “ Electrocardiogram QT prolonged ”, “ Cardiac arrest ”, “ Ventricular extrasystoles ”, “ Ventricular arrhythmia ”. The common compounds between databases were 46 (Supplemental Table 3). We excluded compounds that were unlikely to provide pro-arrhythmic effects in a pure culture of hiPSC-CMs (such as diuretics or drugs associated with chronic cardiotoxicity) as beyond the scope of this study. Seven drugs were then selected in the highest-risk category, i.e. those contraindicated in congenital LQTS and are known to pose a significant risk for developing drug-induced Torsades de Pointes (TdP). Detailed information on the compounds is provided in Supplemental Table 4. Data Analysis and Statistics RStudio (version 2023.09.1) and R (version 4.3.1) were used for data analysis. Multiple group comparisons were conducted using the Kruskal-Wallis test, followed by Dunn’s test with Benjamini-Hochberg correction. Pairwise comparisons were performed using the Wilcoxon test. Categorical variable comparisons were assessed with Fisher’s exact test. A statistical significance level of p ≤ 0.05 was used for all tests. Figures denote significance as *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001, while exact p-values are reported in text and tables. Numerical data are reported as mean ± standard error of the mean (SEM), and plots display mean ± SEM, combined with scattered individual data points when relevant. Categorical data are presented as bar plots. Machine learning model for variant risk prediction A random forest dichotomous classification model was implemented using the comprehensive tidymodels framework 46 . The dataset was stochastically divided into training and validation subsets with a ratio of 70:30, ensuring that drug readouts were evenly distributed between the two datasets. A v-fold (v = 10) cross validation was performed to evaluate the model performance. To further evaluate the classification performance of the model for each genetic variant, a leave-one-out cross validation was performed by recursively removing the data for each variant. The accuracy of variant prediction was calculated as the percentage of correctly predicted readouts (rows) on the total number of readouts of the dataset. Overall score of variant prediction was calculated as a sum of accuracies of each variant risk classification per number of variants. The datasets, models and R code used for the analyses are available in the lab’s GitHub repository: https://github.com/invitroheart . Results Patient cohort characteristics The study included ten patients, with their clinical information summarized in Table 1 . The study cohort comprised a LQT1 family trio (Supplemental Figure 3), which includes an asymptomatic mother carrying the KCNQ1 p.R190W variant, an asymptomatic father carrying the KCNQ1 p.R594Q variant, and their daughter affected by Jervell and Lange-Nielsen syndrome 47 (JLNS) carrying both variants in compound heterozygosity 48 . Four patients from a large South African family carrying the KCNQ1 p.A341V variant associated with a very severe LQT1 phenotype 49 – 51 ; two of these patients were asymptomatic while the other two were severely symptomatic, carrying polymorphisms in the modifier gene NOS1AP (one patient in heterozygous and one patient in homozygous state) associated with a prolonged QT interval and with an increased risk of life-threatening events 52 . Among the LQT2 cases, two patients carried the KCNH2 p.R366X variant: one symptomatic and one asymptomatic. Both were treated with beta-blockers. The symptomatic patient also underwent left cardiac sympathetic denervation due to difficulties in optimizing beta-blocker therapy because of asthenia. No recurrence of symptoms was observed on therapy. A third symptomatic LQT2 patient, carrying the de novo KCNH2 p.A561V variant, experienced cardiac arrests and was treated with beta-blockers and an ICD implantation. Genetic variant risk level assignment To define the threshold for categorizing variants associated with different risks of cardiac events (Low Risk vs High Risk) we analyzed the frequency distribution of cardiac events across reported P/LP genetic variants in KCNQ1 and KCNH2 genes (1099 patients carrying 162 variants in KCNQ1 and 623 patients carrying 175 variants in KCNH2 ) ( Figure 1 , A). For both genes, two distinct peaks in the distribution were identified: one at 0% frequency (variants with no reported cardiac events) and another at ≥25% frequency (variants where at least 25% of carriers experienced cardiac events). Based on this analysis, a threshold of 25% was selected to differentiate these populations. Using this threshold, the variants in our study were categorized as follows: High Risk ( KCNQ1 p.R594Q and KCNQ1 p.R190W & p.R594Q, KCNQ1 p.A341V, and KCNH2 p.A561V) and Low Risk (WT, KCNQ1 p.R190W, and KCNH2 p.R366X) ( Figure 1 , B). Download figure Open in new tab Figure 1. Clinical parameters and variant risk level assignment. A. Frequency of cardiac events across pathogenic/likely pathogenic genetic variants in KCNQ1 (blue) and KCNH2 (orange) genes. B. Percentage of patients with cardiac events for each variant, with assigned variant risk level indicated. C, D. Baseline QTc and Schwartz score values for each variant (N = 1–183). Consistent with the incidence of cardiac events, carriers of High Risk variants KCNQ1 p.A341V and KCNH2 p.A561V had significantly longer QTc values compared to carriers of the Low Risk variants KCNQ1 p.R190W and KCNQ1 p.R366X respectively ( Figure 1 , C, Supplemental Table 5). In addition, the Schwartz scores of KCNQ1 p.A341V carriers were higher than those of KCNQ1 p.R190W ( Figure 1 , C, Supplemental Table 5). Notably, the KCNQ1 p.R190W & p.R594Q variant, observed in a single carrier, presented with deafness and exhibited markedly prolonged QTc (578 ms) and elevated Schwartz score (7.5), consistent with the severe JLNS phenotype. In vitro characterisation of genetic variants selected for the study Four out of six variants had functional characterization from literature: KCNQ1 p.R594Q, KCNQ1 p.A341V, KCNH2 p.R366X, KCNH2 p.A561V leading to loss-of-function, with KCNQ1 p.R594Q and KCNH2 p.A561V being also trafficking-deficient (Supplemental Table 6). We complemented the characterization providing data for the hiPSC-CMs derived from a family trio ( KCNQ1 p.R190W, KCNQ1 p.R594Q, and KCNQ1 p.R190W & p.R594Q) ( Figure 2 ). Download figure Open in new tab Figure 2. Characterization of hiPSC-CMs carrying genetic variants KCNQ1 p.R190W and KCNQ1 p.R594Q. A. Representative action potential traces for each hiPSC-CM variant, elicited at 1 Hz stimulation. B. Action potential duration at 90% repolarization (APD90). C. Corrected field potential duration (cFPD) measurements in hiPSC-CMs for each variant (N = 157–180). D. Representative I Ks current traces (lower panel), obtained by subtracting HMR1556-insensitive current (middle panel) from the total inward potassium current recording (upper panel). E, F End-step and tail I Ks current-voltage (I-V) relationships. * denotes p < 0.05 for all groups vs WT, # for WT vs KCNQ1 p.R190W & p.R594Q, § for WT vs KCNQ1 p.R594Q and vs KCNQ1 p.R190W & p.R594Q. We observed the repolarization prolongation at single-cell level (APD90) and a reduced ability to follow high-frequency pacing in isolated hiPSC-CMs carrying KCNQ1 p.R190W, KCNQ1 p.R594Q, and KCNQ1 p.R190W & p.R594Q compared to WT ( Figure 2 , A, B; Supplemental Figure 4, A, B). HiPSC-CM monolayers carrying KCNQ1 p.R594Q and KCNQ1 p.R190W & p.R594Q demonstrated also corrected FPD prolongation in comparison to WT (422.2 ± 126.7 ms and 424.9 ± 78.7 ms vs 281.8 ± 77.4 ms respectively, p ≤ 0.001) ( Figure 2 , C). Measurement of I Ks with patch clamp demonstrated decreased end-step and tail current densities in hiPSC-CMs carrying KCNQ1 p.R190W, compared to WT, and almost absent I Ks in hiPSC-CMs carrying KCNQ1 p.R594Q and KCNQ1 p.R190W & p.R594Q, in line with their APD prolongation ( Figure 2 , D-F). Baseline electrophysiology of hiPSC-CM cohort Baseline hiPSC-CM electrophysiology assessed with MEA measurements accurately reflected the risk level assigned to each variant. Specifically, hiPSC-CMs from the High Risk group exhibited increased baseline FPD, corrected FPD (cFPD), and RR interval duration ( Figure 3 , A-D; Table 2 ) in comparison to wild type hiPSC-CMs from the Low Risk group. The propensity for irregular beating patterns at baseline was significantly higher in the High Risk group (20.2% of wells in High Risk group vs 8.7% in Low Risk group, p ≤ 0.001) ( Figure 3 , E). Download figure Open in new tab Figure 3. Baseline electrophysiology of hiPSC-CMs carrying High Risk and Low Risk genetic variants. A. Representative field potentials obtained for each variant. B-D. Baseline field potential duration (FPD), corrected FPD (cFPD), and beat-to-beat interval (RR) measurements of hiPSC-CMs for each variant; asterisks indicate statistical significance with the WT group. N = 154-547. E. Incidence of regular (gray) or irregular (yellow) beating patterns for hiPSC-CMs carrying Low Risk and High Risk variants (N = 589-1038). View this table: View inline View popup Download powerpoint Table 2. Baseline electrophysiological parameters obtained from MEA recording for each variant HiPSC-CMs from High Risk and Low Risk groups demonstrate distinct drug responses to ion channel blockers To further discriminate hiPSC-CMs carrying variants with different risk levels we obtained concentration-response curves for 11 selected drugs ( Figure 4 , 5; Supplemental Figures 5-16; Supplemental Table 7). Selective ion channel blockers E4031 (I Kr , Kv11.1), HMR-1556 (I Ks , Kv7.1), nifedipine (I CaL , CaV 1.2), and tetrodotoxin (I Na , Nav 1.5) were used to assess ion channel-specific drug responses as previously proposed by multiple ion channel effects (MICE) approach for drug testing assays 53 . Since I Ks blockers were previously reported to be more effective with reduced cardiac repolarization reserve caused by KCNH2 genetic variants or I Kr pharmacological block 54 , we also tested the effect of HMR1556 on hiPSC-CMs pre-treated with 20 nM I Kr blocker E4031. Download figure Open in new tab Figure 4: HiPSC-CMs from Low Risk and High Risk groups exhibit differential responses to ion channel blockers. A-E: Concentration-dependent response curves to ion channel blockers E4031, HMR1556, HMR1556 following E4031 pre-treatment, nifedipine, and tetrodotoxin. Responses are shown for normalized corrected field potential duration (normalized cFPD), normalized RR interval (normalized RR), and normalized peak-to-peak amplitude (normalized PtPA). Bar plots illustrate the distribution of beating pattern abnormalities. Green dashed lines indicate reported Cmax values for each drug. Asterisks indicate statistical significance of comparisons between the High Risk and Low Risk groups. N = 19-80 per group per drug concentration. Among pure ion channel blockers, we observed a differential response to the I Kr blocker E4031 and to I Ks blocker HMR1556 after E4031 pretreatment between hiPSC-CMs carrying High Risk and Low Risk variants. Specifically, hiPSC-CMs with High Risk variants showed a greater concentration-dependent prolongation of cFPD following treatment with E4031 (up to 350.0 ± 183.2% vs Vehicle for High Risk and up to 218.2 ± 69.9% vs Vehicle for Low Risk, p ≤ 0.001). Furthermore, E4031 increased the RR interval duration in hiPSC-CMs carrying High Risk variants more than in those carrying Low Risk variants ( Figure 4 , A). The High Risk group also demonstrated a higher incidence of beating abnormalities at high concentrations of E4031 ( Figure 4 , A). HMR1556 treatment alone resulted in a small cFPD prolongation that was similar between the High Risk and Low Risk groups ( Figure 4 , B). In contrast the combination of E4031 pre-treatment followed by HMR1556 concentration-response led to a significant cFPD prolongation (up to 188.4 ± 40.6% vs. Vehicle for High Risk and up to 169.2 ± 26.4% vs. Vehicle for Low Risk, p = 0.004) ( Figure 4 , C). Treatment with the I CaL blocker nifedipine led to a cFPD and RR interval decrease ( Figure 4 , D). High concentrations of the I Na blocker tetrodotoxin (10 µM and 30 µM) led to a decrease in peak-to-peak amplitude and prolongation of the RR interval, along with an increased incidence of beating abnormalities ( Figure 4 , E). The High Risk group exhibited significantly more beating abnormalities at 10 µM of tetrodotoxin compared to the Low Risk group (73.6% vs. 46.9%, respectively; p = 0.003). HiPSC-CMs carrying High Risk and Low Risk P/LP variants demonstrate distinct drug responses to proarrhythmic drugs In addition to pure ion channel blockers seven proarrhythmic compounds were selected using the OpenFDA and CredibleMeds databases, as described in the Methods: β2-adrenergic receptor agonist albuterol (salbutamol); the antibiotics clarithromycin, ciprofloxacin, and moxifloxacin; the antipsychotic drugs chlorpromazine and haloperidol; the antiarrhythmic drug dofetilide. Among these compounds dofetilide, haloperidol, and the β2-adrenergic receptor agonist salbutamol demonstrated differential concentration-response curves between High Risk and Low Risk groups ( Figure 5 ). Specifically, responses to the antibiotics clarithromycin and ciprofloxacin were similar between the two groups ( Figure 5 A, C). The highest concentration of clarithromycin tested (10 µM) led to arrhythmias and cessation of beating ( Figure 5 A). Another antibiotic, moxifloxacin, demonstrated a similar effect, with notable toxicity at a concentration of 100 µM, expressed by cFPD and RR increases, a drop in peak-to-peak amplitude, arrhythmias, and cessation of beating similar for both groups ( Figure 5 , F). Treatment of hiPSC-CMs with the antipsychotic drug chlorpromazine resulted in moderate cFPD prolongation that was similar between the two groups ( Figure 5 , B). hiPSC-CMs carrying High Risk variants showed a concentration-dependent increase in cFPD following treatment with the dofetilide (up to 317.5 ± 128.5% vs. Vehicle for High Risk and up to 208.2 ± 61.4% vs. Vehicle for Low Risk, p ≤ 0.001) ( Figure 5 , D) and haloperidol (up to 253.1 ± 134.0% vs. Vehicle for High Risk and up to 178.2 ± 50.6% vs. Vehicle for Low Risk, p ≤ 0.001) ( Figure 5 , E). Salbutamol decreased the RR interval of hiPSC-CMs carrying High Risk variants to a greater extent compared to those carrying Low Risk variants ( Figure 5 , G). Download figure Open in new tab Figure 5. HiPSC-CMs from Low Risk and High Risk groups exhibit differential responses to proarrhythmic drugs. A-G: Concentration-dependent response curves for proarrhythmic drugs clarithromycin, chlorpromazine, ciprofloxacin, dofetilide, haloperidol, moxifloxacin, and salbutamol. Responses are presented for normalized corrected field potential duration (normalized cFPD), normalized RR interval (normalized RR), and normalized peak-to-peak amplitude (normalized PtPA). Bar plots illustrate the distribution of beating pattern abnormalities. Green dashed lines indicate reported Cmax values for each drug. Asterisks indicate statistical significance of comparisons between the High Risk and Low Risk groups. N = 26-121 per group per drug concentration. Gene-specific drug response differences between in hiPSC-CMs carrying High Risk and Low Risk P/LP variants To evaluate gene-specific drug responses, we compared the concentration-response curves between High and Low Risk P/LP variants, separating hiPSC-CMs with KCNQ1 variants (LQT1, JLNS) from those with KCNH2 variants (LQT2). We analyzed responses to I Kr -blocking drugs – dofetilide, E4031, and haloperidol – which demonstrated the strongest discrimination between High Risk and Low Risk P/LP variants. Gene-specific features emerged in drug responses between hiPSC-CMs with High and Low Risk P/LP variants. High Risk hiPSC-CMs carrying KCNQ1 variants showed significant FPD prolongation upon drug treatment ( Figure 6A–C ; Supplemental Table 8), while, in contrast, High Risk KCNH2 variant carriers demonstrated cFPD prolongation similar to the Low Risk KCNH2 group, but exhibited a marked decrease in peak-to-peak amplitude and an increased incidence of abnormal beating patterns ( Figure 6D–F ; Supplemental Table 9). Download figure Open in new tab Figure 6. HiPSC-CMs carrying High Risk and Low Risk variants demonstrate gene-specific drug responses. A-F. Concentration-dependent response curves for dofetilide, E4031, and haloperidol on normalized corrected field potential duration (normalized cFPD), normalized RR interval (normalized RR), and normalized peak-to-peak amplitude (normalized PtPA). Bar plots illustrate the distribution of beating pattern abnormalities. Asterisks indicate statistical significance of comparisons between High Risk and Low Risk groups. N = 8-101 per group per drug concentration. Machine learning model stratifies P/LP variant risk from hiPSC-CM multielectrode array data We curated a MEA dataset from eleven patient-specific hiPSC lines carrying seven variants, exposed to a panel of 11 drugs. The complete dataset comprises 10,410 data points and it was used to train a machine learning classifier for stratifying P/LP variant risk based on in vitro drug responses in patient-specific hiPSC-CMs. To ensure robust and unbiased predictions, we deliberately included all drug-responses, even those from drugs that did not show apparent discrimination between High Risk and Low Risk variants according to concentration-response curves. The random forest model demonstrated robust classification performance with 89% accuracy and excellent discriminative ability, with an area under the receiver operating characteristic (ROC) curve (AUC) of 96% ( Figure 7 , A). The model also demonstrated a strong discriminative ability for LQTS subtype, achieving 79% classification accuracy and an AUC of 94% ( Figure 7 , B). Download figure Open in new tab Figure 7. Variant risk levels and disease type prediction using MEA measurements and machine learning model. A. Receiver Operating Characteristic (ROC) curve illustrating the performance of a classification model for variant risk prediction. B. ROC curves representing the model’s performance in predicting disease type. C. Model validation for single-variant risk level predictions. The proportion of readouts predicted as High Risk (red) and Low Risk (gray) is shown for each variant. The 50% threshold line delineates variants classified as High Risk (above) from those classified as Low Risk (below). Machine learning validation confirmed model accuracy for predicting individual variant risk To further evaluate the model’s generalization performance and its ability to predict risk levels for novel variants, we performed leave-one-out cross-validation by sequentially excluding each variant from the training dataset and predicting its risk level as if it were a novel variant. The model correctly classified 6 out of 7 variants into their appropriate risk categories, with one exception ( KCNH2 p.R366X), achieving an overall prediction accuracy of 72.6% ( Figure 7 , C, Table 3 ). View this table: View inline View popup Download powerpoint Table 3. LQTS variant risk level prediction accuracy for each genetic variant included in the study. Discussion Risk stratification for LQTS patients remains a significant clinical challenge, particularly for carriers of ultra-rare P/LP variants or for those identified in a single proband. These shortcomings impact our ability to identify patients who are at low or high risk of developing cardiac events that could potentially lead to SCD. Here, we developed a machine learning pipeline to classify P/LP variant risk based on in vitro electrophysiological readouts from patient-specific hiPSC-CM lines carrying genetic variants associated with varying frequencies of cardiac events (Low Risk or High Risk). Our findings support prior evidence demonstrating that hiPSC-CMs can effectively recapitulate the clinical phenotype of LQTS 20 – 24 . This was particularly evident in hiPSC-CMs carrying severe genetic variants, such as compound heterozygous KCNQ1 p.R190W & p.R594Q, KCNQ1 p.A341V, and KCNH2 p.A561V, where a clear prolongation in the repolarization duration was clearly observed in vitro . We further strengthened the dataset beyond baseline measurements by challenging each variant with ion channel blockers and pro-arrhythmic compounds, thus creating variant-specific drug response profiles not immediately obtainable in clinical practice. This is supported by previous research demonstrating excellent capacity of hiPSC-CMs to discriminate high/low responders to an acute sotalol administration 55 or to tyrosine kinase inhibitors 56 , 57 . To overcome the high baseline variability typical of hiPSC-CM measurements, often observed in cohort studies, we used relative readouts obtained from normalized drug responses to provide a more accurate discrimination between disease phenotypes. Indeed, High Risk variants were clearly discriminated from their exacerbated response to certain pro-arrhythmic drug treatments, leading to more severe changes in FPD, cFPD, and RR intervals or to a higher occurrence of beating pattern abnormalities. The panel of pure ion channel blockers targeting four key ion channels responsible for action potential regulation revealed that pharmacological I Kr blockade alone already provides effective variant risk discrimination. Conversely, treatment of hiPSC-CMs with the I Ks blocker HMR1556 caused only minimal FPD prolongation 58 , 59 . Reducing the repolarization reserve by pre-treating hiPSC-CMs with I Kr blocker E4031 54 , 60 increased the sensitivity to HMR1556 and discriminated between High Risk and Low Risk groups. As expected, the I CaL blocker nifedipine and the I Na blocker tetrodotoxin were not effective in distinguishing High Risk from Low Risk groups within our LQT1 and LQT2 cohort, which is expected since no pathogenic variants were identified in genes encoding for calcium and sodium channel subunits. These blockers could be useful for identifying Ca² [or Na [handling abnormalities to other LQTS subtypes beyond LQT1 and LQT2, e.g. LQT3 or CALM-LQTS 22 , 61 . Among the pro-arrhythmic drugs selected based on analyses of the OpenFDA and CredibleMeds databases, acute treatment of hiPSC-CMs with the potent I Kr blocker dofetilide and the antipsychotic drug haloperidol demonstrated clear concentration-dependent cFPD prolongation, consistently distinguishing between the High Risk and Low Risk groups. Acute treatment with salbutamol, a β 2 -adrenergic agonist commonly used as a bronchodilator and typically precluded to LQTS patients 62 , differentially increased the spontaneous beating frequency in hiPSC-CMs; this allowed the discrimination of High Risk variants, which revealed a higher sensitivity, from Low Risk variants and could be influenced by a difference in beating frequencies at baseline between these two groups. No arrhythmic events were observed supporting previous studies on commercially available hiPSC-CMs 63 . Treatment of hiPSC-CMs with ciprofloxacin, clarithromycin, and moxifloxacin within therapeutic concentration ranges resulted in moderate FPD prolongation as similarly emerged in commercially available wild-type hiPSC-CMs 31 , 64 . There is limited information available on gene-specific drug responses in LQT1 and LQT2 both in vivo and in vitro 65 – 67 . In our study, three drugs with increased potency to block hERG (E4031, dofetilide, and haloperidol) well discriminated diseased hiPSC-CMs carrying High Risk variants in either KCNQ1 or KCNH2 genes. This observation aligns with findings from adult transgenic rabbit models, where LQT2 rabbits exhibited greater susceptibility to drug-induced TdP than LQT1 rabbits after administration of the anesthetic propofol, with a known hERG channel blocking effect 66 . These results are consistent with the concept of a severely compromised repolarization reserve in LQT2 cardiomyocytes, which renders them more vulnerable to proarrhythmic triggers 60 . Variant risk prediction based on multiparametric readouts is a novel and rapidly developing field. A recent study combined clinical data, trafficking assays and automated patch clamp data from heterologous expression systems, to score the risk of KCNH2 variants associated with LQT2 68 . The use of hiPSC-CMs provides enhanced precision by addressing the limitations of heterologous systems that lack cardiac-specific context. Having demonstrated that hiPSC-CM drug responses effectively discriminate between High Risk and Low Risk P/LP variants, we developed the ML model trained on in vitro hiPSC-CM electrophysiological readouts able to predict retrospectively severity of variants with high precision – a methodology that, to our knowledge, has not been previously applied to LQTS. While we deliberately chose well-known P/LP genetic variants with a loss-of-function mechanism for our study, the model’s versatility extends beyond retrospective analyses to the classification of ultra-rare variants and VUS. Importantly, the ML model trained for disease discrimination predicted each disease subtype with a 79% accuracy. This robust performance indicates that disease-specific electrophysiological features are distinct from variant risk classification features, providing solid evidence to expand the model incorporating other LQTS subtypes. Successful application of the model to predict the risk level of single variants on which it was not originally trained represents a high-level validation, further demonstrating applicability of our approach for assessing novel, previously unreported variants. The observed discrepancy in predicting the risk level for the KCNH2 p.R366X variant underscores the impact of patient-specific drug responses. Although assigned as Low Risk due to a low cardiac event frequency (16.7%), two hiPSC-CM lines from different LQT2 patients exhibited divergent drug response patterns. One line, derived from a patient with no cardiac events, aligned with Low Risk variants, whereas the other, from a patient with documented cardiac events, markedly prolonged QTc (622 ms), and LSCD performed, with those of High Risk. This highlights the importance of separating the genetic background from the primary disease-causing variant, particularly for P/LP variants that fall within the intermediate range of the risk level distribution. Indeed, in the case of High Risk KCNQ1 p.A341V variant, despite differences in genetic background (such as NOS1AP genetic variants) the primary variant itself is so severe that the probability of misclassification is lower. These findings emphasize the necessity of integrating data from multiple sources for accurate variant risk classification, including multicentral in vitro data, clinical cohort records, structural variant pathogenicity predictions, and Bayesian penetrance estimates 37 . In conclusion, we established a robust method for stratifying P/LP variant risk by integrating ML classification to electrophysiological recordings in patient-specific hiPSC-CMs. This approach enhances the risk stratification of LQTS variants beyond what is currently possible from clinical data alone, providing value for variants with limited clinical information. Study limitations First, experiments were performed on hiPSC-CMs which, despite significant progress in maturation techniques, still exhibit an immature phenotype compared to adult cardiomyocytes 69 , 70 . This immaturity could partially explain blunted or exacerbated drug responses for specific compounds. Secondly, we focused on six LQTS-causing LP/P variants. While this demonstrates proof of principle for retrospective variant risk stratification, further investigations are necessary to extend this approach to a broader range of variants or LQTS subtypes. Third, while our model simplifies clinical risk into a binary classification, further evolutions of this approach could consider incorporating multiple classification states to accommodate intermediate disease phenotypes based on both qualitative and quantitative cardiac events associated with each variant. Funding Sources This project has received funding from the Horizon Europe EU (HORIZON-MSCA-2022-PF-01 PREPARE No. 101105561 to AK), Horizon 2020 (H2020-MSCA-IF-2017 No. 795209 to LS), Fondazione CARIPLO grant No. 2019-1691 to LS, Leducq Foundation grant 18CVD05 to PJS. Italian Ministry of University and Research within Mission 4, “Education and Research”, Component 2, “From Research to Business”, Investment 1.2, “Funding projects presented by young researchers” of the National Recovery and Resilience Plan. Project No. 2022-NAZ-0485 (H45E22001210006) to LS. Data Availability The datasets, models and R code used for the analyses are available in the lab GitHub repository. Abbreviations and acronyms APD Action Potential Duration FPD Field Potential Duration iPSC-CMs Cardiomyocytes derived from human induced pluripotent stem cells JLNS Jervell and Lange-Nielsen Syndrome LQTS Long QT Syndrome MEA Multielectrode Arrays ML Machine Learning P/LP Pathogenic/Likely Pathogenic TdP Torsades de Pointes QTc QT interval corrected for heart rate WT Wild type References 1. ↵ Schwartz PJ , Periti M , Malliani A . The long Q-T syndrome . Am Heart J . 1975 ; 89 : 378 – 390 . OpenUrl CrossRef PubMed Web of Science 2. Schwartz PJ , Priori SG , Spazzolini C , et al. Genotype-Phenotype Correlation in the Long-QT Syndrome . Circulation . 2001 ; 103 : 89 – 95 . OpenUrl Abstract / FREE Full Text 3. Behr ER , Roden D . Drug-induced arrhythmia: pharmacogenomic prescribing? Eur Heart J . 2013 ; 34 : 89 – 95 . OpenUrl CrossRef PubMed 4. ↵ Weeke PE , Kellemann JS , Jespersen CB , et al. Long-term proarrhythmic pharmacotherapy among patients with congenital long QT syndrome and risk of arrhythmia and mortality . Eur Heart J . 2019 ; 40 : 3110 – 3117 . OpenUrl PubMed 5. ↵ Zeppenfeld K , Tfelt-Hansen J , De Riva M , et al. 2022 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death . Eur Heart J . 2022 ; 43 : 3997 – 4126 . OpenUrl CrossRef PubMed 6. ↵ Schwartz PJ . Cardiac sympathetic denervation to prevent life-threatening arrhythmias . Nat Rev Cardiol . 2014 ; 11 : 346 – 353 . OpenUrl CrossRef PubMed 7. ↵ Dusi V , Pugliese L , De Ferrari GM , et al. Left Cardiac Sympathetic Denervation for Long QT Syndrome . JACC Clin Electrophysiol . 2022 ; 8 : 281 – 294 . OpenUrl PubMed 8. ↵ Schwartz PJ , Priori SG , Locati EH , et al. Long QT Syndrome Patients With Mutations of the SCN5A and HERG Genes Have Differential Responses to Na + Channel Blockade and to Increases in Heart Rate: Implications for Gene-Specific Therapy . Circulation . 1995 ; 92 : 3381 – 3386 . OpenUrl Abstract / FREE Full Text 9. ↵ Crotti L , Neves R , Dagradi F , et al. Therapeutic Efficacy of Mexiletine for Long QT Syndrome Type 2: Evidence From Human Induced Pluripotent Stem Cell–Derived Cardiomyocytes, Transgenic Rabbits, and Patients . Circulation . 2024 ; 150 : 531 – 543 . OpenUrl PubMed 10. ↵ Schwartz PJ , Ackerman MJ . The long QT syndrome: A transatlantic clinical approach to diagnosis and therapy . Eur Heart J . 2013 ; 34 : 3109 – 3116 . OpenUrl CrossRef PubMed Web of Science 11. ↵ Schwartz PJ , Ackerman MJ , George AL , Wilde AAM . Impact of Genetics on the Clinical Management of Channelopathies . J Am Coll Cardiol . 2013 ; 62 : 169 – 180 . OpenUrl FREE Full Text 12. ↵ Giudicessi JR , Wilde AAM , Ackerman MJ . The genetic architecture of long QT syndrome: A critical reappraisal . Trends Cardiovasc Med . 2018 ; 28 : 453 – 464 . OpenUrl CrossRef PubMed 13. ↵ Richards S , Aziz N , Bale S , et al. Standards and guidelines for the interpretation of sequence variants: A joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology . Genet Med . 2015 ; 17 : 405 – 424 . OpenUrl CrossRef PubMed 14. ↵ Aziz PF , Wieand TS , Ganley J , et al. Genotype– and Mutation Site–Specific QT Adaptation During Exercise, Recovery, and Postural Changes in Children With Long-QT Syndrome . Circ Arrhythm Electrophysiol . 2011 ; 4 : 867 – 873 . OpenUrl Abstract / FREE Full Text 15. ↵ Schwartz PJ , Moreno C , Kotta M-C , et al. Mutation location and I Ks regulation in the arrhythmic risk of long QT syndrome type 1: the importance of the KCNQ1 S6 region . Eur Heart J . 2021 ; 42 : 4743 – 4755 . OpenUrl PubMed 16. ↵ Ackerman MJ . Genetic purgatory and the cardiac channelopathies: Exposing the variants of uncertain/unknown significance issue . Heart Rhythm . 2015 ; 12 : 2325 – 2331 . OpenUrl CrossRef PubMed 17. ↵ Gnecchi M , Sala L , Schwartz PJ . Precision Medicine and cardiac channelopathies: when dreams meet reality . Eur Heart J . 2021 ; 42 : 1661 – 1675 . OpenUrl PubMed 18. ↵ Brnich SE , Abou Tayoun AN , Couch FJ , et al. Recommendations for application of the functional evidence PS3/BS3 criterion using the ACMG/AMP sequence variant interpretation framework . Genome Med . 2020 ; 12 : 3 . OpenUrl CrossRef 19. ↵ Lichou F , Trynka G . Functional studies of GWAS variants are gaining momentum . Nat Commun . 2020 ; 11 : 6283 . OpenUrl CrossRef PubMed 20. ↵ Moretti A , Bellin M , Welling A , et al. Patient-Specific Induced Pluripotent Stem-Cell Models for Long-QT Syndrome . N Engl J Med . 2010 ; 363 : 1397 – 1409 . OpenUrl CrossRef PubMed 21. Itzhaki I , Maizels L , Huber I , et al. Modelling the long QT syndrome with induced pluripotent stem cells . Nature . 2011 ; 471 : 225 – 229 . OpenUrl CrossRef PubMed Web of Science 22. ↵ Rocchetti M , Sala L , Dreizehnter L , et al. Elucidating arrhythmogenic mechanisms of long-QT syndrome CALM1-F142L mutation in patient-specific induced pluripotent stem cell-derived cardiomyocytes . Cardiovasc Res . 2017 ; 113 : 531 – 541 . OpenUrl CrossRef PubMed 23. Garg P , Garg V , Shrestha R , Sanguinetti MC , Kamp TJ , Wu JC . Human Induced Pluripotent Stem Cell–Derived Cardiomyocytes as Models for Cardiac Channelopathies . Circ Res . 2018 ; 123 : 224 – 243 . OpenUrl Abstract / FREE Full Text 24. ↵ Lee Y-K , Sala L , Mura M , et al. MTMR4 SNVs modulate ion channel degradation and clinical severity in congenital long QT syndrome: insights in the mechanism of action of protective modifier genes . Cardiovasc Res . 2021 ; 117 : 767 – 779 . OpenUrl CrossRef PubMed 25. ↵ Zhang M , D’Aniello C , Verkerk AO , et al. Recessive cardiac phenotypes in induced pluripotent stem cell models of Jervell and Lange-Nielsen syndrome: Disease mechanisms and pharmacological rescue . Proc Natl Acad Sci . 2014 ; 111 . 26. ↵ Yazawa M , Hsueh B , Jia X , et al. Using induced pluripotent stem cells to investigate cardiac phenotypes in Timothy syndrome . Nature . 2011 ; 471 : 230 – 234 . OpenUrl CrossRef PubMed Web of Science 27. ↵ Lan F , Lee AS , Liang P , et al. Abnormal Calcium Handling Properties Underlie Familial Hypertrophic Cardiomyopathy Pathology in Patient-Specific Induced Pluripotent Stem Cells . Cell Stem Cell . 2013 ; 12 : 101 – 113 . OpenUrl CrossRef PubMed Web of Science 28. Ma D , Wei H , Lu J , et al. Generation of patient-specific induced pluripotent stem cell-derived cardiomyocytes as a cellular model of arrhythmogenic right ventricular cardiomyopathy . Eur Heart J . 2013 ; 34 : 1122 – 1133 . OpenUrl CrossRef PubMed Web of Science 29. ↵ Sun N , Yazawa M , Liu J , et al. Patient-Specific Induced Pluripotent Stem Cells as a Model for Familial Dilated Cardiomyopathy . Sci Transl Med . 2012 ; 4 . 30. ↵ Miao Y , Tian L , Martin M , et al. Intrinsic Endocardial Defects Contribute to Hypoplastic Left Heart Syndrome . Cell Stem Cell . 2020 ; 27 : 574 – 589 .e8. OpenUrl CrossRef PubMed 31. ↵ Blinova K , Dang Q , Millard D , et al. International Multisite Study of Human-Induced Pluripotent Stem Cell-Derived Cardiomyocytes for Drug Proarrhythmic Potential Assessment . Cell Rep . 2018 ; 24 : 3582 – 3592 . OpenUrl CrossRef PubMed 32. Sala L , Bellin M , Mummery CL . Integrating cardiomyocytes from human pluripotent stem cells in safety pharmacology: has the time come? Br J Pharmacol . 2017 ; 174 : 3749 – 3765 . OpenUrl PubMed 33. ↵ Paik DT , Chandy M , Wu JC . Patient and Disease–Specific Induced Pluripotent Stem Cells for Discovery of Personalized Cardiovascular Drugs and Therapeutics Ohlstein EH, editor . Pharmacol Rev . 2020 ; 72 : 320 – 342 . OpenUrl CrossRef PubMed 34. ↵ Mehta A , Ramachandra CJA , Singh P , et al. Identification of a targeted and testable antiarrhythmic therapy for long-QT syndrome type 2 using a patient-specific cellular model . Eur Heart J . 2018 ; 39 : 1446 – 1455 . OpenUrl PubMed 35. ↵ Roberts B , Haupt A , Tucker A , et al. Systematic gene tagging using CRISPR/Cas9 in human stem cells to illuminate cell organization Drubin DG, editor . Mol Biol Cell . 2017 ; 28 : 2854 – 2874 . OpenUrl Abstract / FREE Full Text 36. ↵ Stenson PD , Ball EV , Mort M , et al. Human Gene Mutation Database (HGMD ® ): 2003 update: HGMD 2003 UPDATE . Hum Mutat . 2003 ; 21 : 577 – 581 . OpenUrl CrossRef PubMed Web of Science 37. ↵ O’Neill MJ , Sala L , Denjoy I , et al. Continuous Bayesian variant interpretation accounts for incomplete penetrance among Mendelian cardiac channelopathies . Genet Med . 2023 ; 25 : 100355 . OpenUrl PubMed 38. ↵ Kozek K , Wada Y , Sala L , et al. Estimating the Posttest Probability of Long QT Syndrome Diagnosis for Rare KCNH2 Variants . Circ Genomic Precis Med . 2021 ; 14 : e003289 . OpenUrl 39. ↵ Lian X , Hsiao C , Wilson G , et al. Robust cardiomyocyte differentiation from human pluripotent stem cells via temporal modulation of canonical Wnt signaling . Proc Natl Acad Sci U S A . 2012 ; 109 : E1848 – 57 . OpenUrl Abstract / FREE Full Text 40. ↵ Tohyama S , Hattori F , Sano M , et al. Distinct Metabolic Flow Enables Large-Scale Purification of Mouse and Human Pluripotent Stem Cell-Derived Cardiomyocytes . Cell Stem Cell . 2013 ; 12 : 127 – 137 . OpenUrl CrossRef PubMed Web of Science 41. ↵ Buikema JW , Lee S , Goodyer WR , et al. Wnt Activation and Reduced Cell-Cell Contact Synergistically Induce Massive Expansion of Functional Human iPSC-Derived Cardiomyocytes . Cell Stem Cell . 2020 ; 27 : 50 – 63 .e5. OpenUrl CrossRef PubMed 42. ↵ Feyen DAM , McKeithan WL , Bruyneel AAN , et al. Metabolic Maturation Media Improve Physiological Function of Human iPSC-Derived Cardiomyocytes . Cell Rep . 2020 ; 32 : 107925 . OpenUrl CrossRef PubMed 43. ↵ Sala L , Ward-van Oostwaard D , Tertoolen LGJ , Mummery CL , Bellin M . Electrophysiological Analysis of human Pluripotent Stem Cell-derived Cardiomyocytes (hPSC-CMs) Using Multi-electrode Arrays (MEAs) . J Vis Exp . 2017 : 55587 . 44. ↵ Sala L , Leonov V , Mura M , et al. Use of hiPSC-Derived Cardiomyocytes to Rule Out Proarrhythmic Effects of Drugs: The Case of Hydroxychloroquine in COVID-19 . Front Physiol . 2022 ; 12 : 730127 . OpenUrl PubMed 45. ↵ Woosley RL , Heise CW , Gallo T , Romero KA , Woosley RD , Lambson J . www.CredibleMeds.org , QTdrugs List 46. ↵ Kuhn M , Wickham H . Tidymodels: a collection of packages for modeling and machine learning using tidyverse principles . 2020 . 47. ↵ Schwartz PJ , Spazzolini C , Crotti L , et al. The Jervell and Lange-Nielsen Syndrome: Natural History, Molecular Basis, and Clinical Outcome . Circulation . 2006 ; 113 : 783 – 790 . OpenUrl Abstract / FREE Full Text 48. ↵ Mura M , Lee Y , Ginevrino M , et al. Generation of the human induced pluripotent stem cell (hiPSC) line PSMi002-A from a patient affected by the Jervell and Lange-Nielsen syndrome and carrier of two compound heterozygous mutations on the KCNQ1 gene . Stem Cell Res . 2018 ; 29 : 157 – 161 . OpenUrl CrossRef PubMed 49. ↵ Brink PA , Crotti L , Corfield V , et al. Phenotypic Variability and Unusual Clinical Severity of Congenital Long-QT Syndrome in a Founder Population . Circulation . 2005 ; 112 : 2602 – 2610 . OpenUrl Abstract / FREE Full Text 50. Crotti L , Spazzolini C , Schwartz PJ , et al. The Common Long-QT Syndrome Mutation KCNQ1/A341V Causes Unusually Severe Clinical Manifestations in Patients With Different Ethnic Backgrounds: Toward a Mutation-Specific Risk Stratification . Circulation . 2007 ; 116 : 2366 – 2375 . OpenUrl Abstract / FREE Full Text 51. ↵ Ronchi C , Bernardi J , Mura M , et al. NOS1AP polymorphisms reduce NOS1 activity and interact with prolonged repolarization in arrhythmogenesis . Cardiovasc Res . 2021 ; 117 : 472 – 483 . OpenUrl PubMed 52. ↵ Crotti L , Monti MC , Insolia R , et al. NOS1AP Is a Genetic Modifier of the Long-QT Syndrome . Circulation . 2009 ; 120 : 1657 – 1663 . OpenUrl Abstract / FREE Full Text 53. ↵ Kramer J , Obejero-Paz CA , Myatt G , et al. MICE Models: Superior to the HERG Model in Predicting Torsade de Pointes . Sci Rep . 2013 ; 3 : 2100 . OpenUrl CrossRef PubMed 54. ↵ Braam SR , Tertoolen L , Casini S , et al. Repolarization reserve determines drug responses in human pluripotent stem cell derived cardiomyocytes . Stem Cell Res . 2013 ; 10 : 48 – 56 . OpenUrl CrossRef PubMed 55. ↵ Stillitano F , Hansen J , Kong C-W , et al. Modeling susceptibility to drug-induced long QT with a panel of subject-specific induced pluripotent stem cells . eLife . 2017 ; 6 : e19406 . OpenUrl CrossRef PubMed 56. ↵ Golforoush P , Schneider MD . Intensive care for human hearts in pluripotent stem cell models . Npj Regen Med . 2020 ; 5 : 4 . 57. ↵ Shead KD , Huethorst E , Burton F , Lang NN , Myles RC , Smith GL . Human Induced Pluripotent Stem Cell-Derived Cardiomyocytes for Preclinical Cardiotoxicity Screening in Cardio-Oncology . JACC CardioOncology . 2024 : S2666087324002710 . 58. ↵ Wada Y , Wang L , Hall LD , et al. The electrophysiologic effects of KCNQ1 extend beyond expression of I Ks: evidence from genetic and pharmacologic block . Cardiovasc Res . 2024 ; 120 : 735 – 744 . OpenUrl PubMed 59. ↵ Zeng H , Wang J , Clouse H , Lagrutta A , Sannajust F . Human-induced pluripotent stem cell-derived cardiomyocytes have limited I Ks for repolarization reserve as revealed by specific KCNQ1/KCNE1 blocker . JRSM Cardiovasc Dis . 2019 ; 8 : 204800401985491 . OpenUrl 60. ↵ Varró A , Baczkó I . Cardiac ventricular repolarization reserve: a principle for understanding drug-related proarrhythmic risk: Repolarization reserve and proarrhythmia . Br J Pharmacol . 2011 ; 164 : 14 – 36 . OpenUrl CrossRef PubMed Web of Science 61. ↵ Ma D , Wei H , Zhao Y , et al. Modeling type 3 long QT syndrome with cardiomyocytes derived from patient-specific induced pluripotent stem cells . Int J Cardiol . 2013 ; 168 : 5277 – 5286 . OpenUrl CrossRef PubMed 62. ↵ Zreik J , LaPage MJ , Zreik H . Congenital Long QT Syndrome Unmasked by Albuterol in an Adolescent with Asthma . J Emerg Med . 2024 ; 67 : e446 – e450 . OpenUrl PubMed 63. ↵ Kopljar I , Lu HR , Van Ammel K , et al. Development of a Human iPSC Cardiomyocyte-Based Scoring System for Cardiac Hazard Identification in Early Drug Safety De-risking . Stem Cell Rep . 2018 ; 11 : 1365 – 1377 . OpenUrl 64. ↵ Ando H , Yoshinaga T , Yamamoto W , et al. A new paradigm for drug-induced torsadogenic risk assessment using human iPS cell-derived cardiomyocytes . J Pharmacol Toxicol Methods . 2017 ; 84 : 111 – 127 . OpenUrl CrossRef PubMed 65. ↵ Noda T . Gene-specific response of dynamic ventricular repolarization to sympathetic stimulation in LQT1, LQT2 and LQT3 forms of congenital long QT syndrome . Eur Heart J . 2002 ; 23 : 975 – 983 . OpenUrl CrossRef PubMed Web of Science 66. ↵ Odening KE , Hyder O , Chaves L , et al. Pharmacogenomics of anesthetic drugs in transgenic LQT1 and LQT2 rabbits reveal genotype-specific differential effects on cardiac repolarization . Am J Physiol-Heart Circ Physiol . 2008 ; 295 : H2264 – H2272 . OpenUrl CrossRef PubMed Web of Science 67. ↵ Yoshinaga D , Baba S , Makiyama T , et al. Phenotype-Based High-Throughput Classification of Long QT Syndrome Subtypes Using Human Induced Pluripotent Stem Cells . Stem Cell Rep . 2019 ; 13 : 394 – 404 . OpenUrl 68. ↵ O’Neill MJ , Ng C-A , Aizawa T , et al. Assays of Variant Effect and Automated Patch Clamping Improve KCNH2 –LQTS Variant Classification and Cardiac Event Risk Stratification . Circulation . 2024 :CIRCULATIONAHA.124.069828. 69. ↵ Karbassi E , Fenix A , Marchiano S , et al. Cardiomyocyte maturation: advances in knowledge and implications for regenerative medicine . Nat Rev Cardiol . 2020 ; 17 : 341 – 359 . OpenUrl CrossRef PubMed 70. ↵ Veerman CC , Kosmidis G , Mummery CL , Casini S , Verkerk AO , Bellin M . Immaturity of Human Stem-Cell-Derived Cardiomyocytes in Culture: Fatal Flaw or Soluble Problem? Stem Cells Dev . 2015 ; 24 : 1035 – 1052 . OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted March 26, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings Aleksandr Khudiakov , Manuela Mura , Federica Giannetti , Vladislav Leonov , Chiara Alberio , Marem Eskandr , Paul A Brink , Lia Crotti , Massimiliano Gnecchi , Peter J Schwartz , Luca Sala medRxiv 2025.03.25.25324187; doi: https://doi.org/10.1101/2025.03.25.25324187 Share This Article: Copy Citation Tools Genetic variants risk assessment for Long QT Syndrome through machine learning and multielectrode array recordings Aleksandr Khudiakov , Manuela Mura , Federica Giannetti , Vladislav Leonov , Chiara Alberio , Marem Eskandr , Paul A Brink , Lia Crotti , Massimiliano Gnecchi , Peter J Schwartz , Luca Sala medRxiv 2025.03.25.25324187; doi: https://doi.org/10.1101/2025.03.25.25324187 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cardiovascular Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15227) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6597) Geriatric Medicine (668) Health Economics (997) Health Informatics (4534) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9230) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a002e8710e4fe2c5',t:'MTc3OTUyNzIyMQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();