Backgroud. In the last decades, the implementation of high-throughput next-generation (NGS) technologies has profoundly changed the landscape of human genome sequencing. The analysis of whole genome, exomes or deputed genes in the context of multigene panels, while being rapidly and extensively available in the routine practice, has raised new and difficult challenges in result interpretation. Among others, due to the increase of genetic specimen testing, novel sequence variants have become an increasingly frequent finding which, differing from the standard sequence of the referenced general population, remain of unknown/uncertain significance. Objective. We present hereby our single center experience over a 15-year period on a cohort of consecutive patients addressed for suspicion of cardiac hereditary disease to the specialized cardiogenetic outpatient clinic of the University Hospital of Lausanne and enrolled in the prospective registry of hereditary cardiac diseases of the Service of Medical Genetics. After the identification and careful analysis of likely pathogenic/pathogenic variants and of variants of uncertain significance, we applied the new computational predictive algorithm Mutscore, based on a machine-learning approach, to test, in a new dataset, Mutscore performance as well as its ability in reclassifying variants of uncertain significance. Methods. Our study implied a retrospective review of DNA sequencing results of patients tested for suspected cardiac hereditary disease and addressed at our center from July 1st 2007 through December 31st 2021. For each likely pathogenic/pathogenic variants and for variants of uncertain significance we then applied the previously described Mutscore algorithm using a random forest approach. Results. Among the 488 tested patients, the most frequent clinical indications were presymptomatic screening (37.3%) proposed in case of a known familial variant, followed by suspected arrhythmic syndrome (28.3%) and hereditary cardiomyopathy (26.8%). A likely pathogenic/pathogenic variant was found in 198 patients for an overall diagnostic yield of 40.6% and of 23.6% considering only index cases. The rate of identified mutation-carriers was higher among patients with cardiomyopathy (45.8%) than among those with inherited arrhythmic syndromes (37%). The pathogenic variants were mainly identified in genes with definitive scientific evidence of causal association with the disease. We demonstrated the excellent predictive performance of the Mutscore in our dataset, as attested by an AUC (Area Under the Curve) of 0.895 and by the statistically significant positive correlation (r = 0.65, p-value = 0.000) between the Mutscore and the variant interpretation reported in ClinVar. We finally documented, according to Mutscore, the possible reclassification into likely benign/pathogenic variants for the 47.2% of the variants of uncertain significance. Conclusions. The 15-year experience of our specialized cardiogenetic outpatient clinic has demonstrated the pivotal role of the genetic analysis in diagnostic confirmation (in 40% of the patients tested) and in therapeutic management of patients with suspected cardiac hereditary diseases. Based on this solid clinical experience, we have also demonstrated in our dataset the excellent predictive performance of the Mutscore computational algorithm and its accuracy in reclassifying variants of uncertain significance. Although further validation studies are still needed, the Mutscore algorithm seems to be a valuable tool and a powerful resource contributing to disambiguation of variants of uncertain significance and to the further build the pathway towards precision medicine.
Background. Negli ultimi decenni, l'introduzione delle tecnologie di sequenziamento genetico di nuova generazione (NGS) ha cambiato profondamente la pratica della genetica medica. Se da un lato l'analisi dell'intero genoma, ha permesso un più ampio accesso all’analisi genetica nella pratica di routine, dall’altro ha sollevato nuove e difficili sfide nell'interpretazione dei risultati. A causa infatti dell'aumentato numero di test genetici, è divenuto sempre più frequente il riscontro di nuove varianti di sequenza che rimangono di significato incerto. Obiettivi. Lo studio attuale presenta l'esperienza quindicinale della consultazione congiunta di cardiogenetica dei servizi di Cardiologia e di Medicina Genetica dell'Ospedale Universitario di Losanna, relativa alla coorte di pazienti indirizzati per sospetto di malattia ereditaria cardiaca e arruolati nel registro prospettico delle malattie cardiache ereditarie del Servizio di Genetica Medica. Previa identificazione e analisi delle varianti probabilmente patogene/patogene e di significato incerto riscontrate, lo studio prevede l’applicazione del nuovo algoritmo computazionale predittivo Mutscore, basato su un approccio di apprendimento automatico (“machine learning”), al fine di testare, su un nuovo set di dati, la performance predittiva dell’algoritmo e la capacità di riclassificazione delle varianti di significato incerto. Metodi. Il nostro studio ha comportato l’analisi retrospettiva dei risultati di sequenziamento del DNA dei pazienti testati per sospetta malattia ereditaria cardiaca e indirizzati al nostro centro dal 1° luglio 2007 al 31 dicembre 2021. Per ogni variante probabilmente patogena/patogena e per le varianti di significato incerto abbiamo quindi applicato l'algoritmo Mutscore precedentemente descritto, utilizzando un approccio random forest. Risultati. Sui 488 pazienti testati, l’indicazione clinica più frequente è rappresentata dallo screening presintomatico (37,3%) in caso di variante familiare conosciuta, seguita dall’indicazione per sospetta sindrome aritmica (28,3%) e per cardiomiopatia ereditaria diagnosticata o sospetta (26,8%). Una variante patogena/probabilmente patogena è stata riscontrata in 198 pazienti per una resa diagnostica globale del 40,6% e del 23,6% nei soli casi indice. Il tasso di mutation-carriers identificati è risultato maggiore (45,8%) nei pazienti con cardiomiopatia rispetto a quelli (37%) affetti da sindromi aritmiche ereditarie. Le varianti patogene sono state identificate principalmente nei geni con chiara evidenza scientifica di associazione causale alla malattia. Abbiamo dimostrato l’ottima capacità predittiva del Mutscore sul nostro set di dati, come attestato da una AUC (Area Under the Curve) di 0,895 e dalla correlazione positiva statisticamente significativa (r= 0,65, p-value = 0.000) tra il Mutscore e l’interpretazione della variante annotata in ClinVar. Abbiamo infine documentato, sulla base del Mutscore, la possibile riclassificazione in varianti probabilmente benigne/patogene per il 47,2% delle varianti di significato incerto. Conclusione. Lo studio ha permesso di dimostrare, attraverso l’esperienza di 15 anni di consultazioni specializzate congiunte in Cardiogenetica, il ruolo fondamentale dell’analisi genetica nella conferma diagnostica (nel 40% dei pazienti testati) e nella presa in carico dei pazienti con sospetta malattia cardiaca ereditaria. Sulla base di questa solida esperienza clinica, abbiamo inoltre dimostrato, nella nostra coorte di pazienti, l’ottima performance predittiva dell’algoritmo computazionale Mutscore e la sua accuratezza nell’aiutare a riclassificare le varianti di significato incerto. Sebbene ulteriori studi di validazione siano ancora necessari, il Mutscore sembra essere un valido strumento ed una potente risorsa in grado di contribuire alla riclassificazione delle varianti di significato incerto.
Intelligenza Artificiale in Cardiogenetica: utilizzo dell’algoritmo predittivo Mutscore in 15 anni di pratica clinica
PORRETTA, ALESSANDRA PIA
2023-04-20
Abstract
Backgroud. In the last decades, the implementation of high-throughput next-generation (NGS) technologies has profoundly changed the landscape of human genome sequencing. The analysis of whole genome, exomes or deputed genes in the context of multigene panels, while being rapidly and extensively available in the routine practice, has raised new and difficult challenges in result interpretation. Among others, due to the increase of genetic specimen testing, novel sequence variants have become an increasingly frequent finding which, differing from the standard sequence of the referenced general population, remain of unknown/uncertain significance. Objective. We present hereby our single center experience over a 15-year period on a cohort of consecutive patients addressed for suspicion of cardiac hereditary disease to the specialized cardiogenetic outpatient clinic of the University Hospital of Lausanne and enrolled in the prospective registry of hereditary cardiac diseases of the Service of Medical Genetics. After the identification and careful analysis of likely pathogenic/pathogenic variants and of variants of uncertain significance, we applied the new computational predictive algorithm Mutscore, based on a machine-learning approach, to test, in a new dataset, Mutscore performance as well as its ability in reclassifying variants of uncertain significance. Methods. Our study implied a retrospective review of DNA sequencing results of patients tested for suspected cardiac hereditary disease and addressed at our center from July 1st 2007 through December 31st 2021. For each likely pathogenic/pathogenic variants and for variants of uncertain significance we then applied the previously described Mutscore algorithm using a random forest approach. Results. Among the 488 tested patients, the most frequent clinical indications were presymptomatic screening (37.3%) proposed in case of a known familial variant, followed by suspected arrhythmic syndrome (28.3%) and hereditary cardiomyopathy (26.8%). A likely pathogenic/pathogenic variant was found in 198 patients for an overall diagnostic yield of 40.6% and of 23.6% considering only index cases. The rate of identified mutation-carriers was higher among patients with cardiomyopathy (45.8%) than among those with inherited arrhythmic syndromes (37%). The pathogenic variants were mainly identified in genes with definitive scientific evidence of causal association with the disease. We demonstrated the excellent predictive performance of the Mutscore in our dataset, as attested by an AUC (Area Under the Curve) of 0.895 and by the statistically significant positive correlation (r = 0.65, p-value = 0.000) between the Mutscore and the variant interpretation reported in ClinVar. We finally documented, according to Mutscore, the possible reclassification into likely benign/pathogenic variants for the 47.2% of the variants of uncertain significance. Conclusions. The 15-year experience of our specialized cardiogenetic outpatient clinic has demonstrated the pivotal role of the genetic analysis in diagnostic confirmation (in 40% of the patients tested) and in therapeutic management of patients with suspected cardiac hereditary diseases. Based on this solid clinical experience, we have also demonstrated in our dataset the excellent predictive performance of the Mutscore computational algorithm and its accuracy in reclassifying variants of uncertain significance. Although further validation studies are still needed, the Mutscore algorithm seems to be a valuable tool and a powerful resource contributing to disambiguation of variants of uncertain significance and to the further build the pathway towards precision medicine.File | Dimensione | Formato | |
---|---|---|---|
PORRETTA.pdf
accesso aperto
Descrizione: Artificial Intelligence in Cardiogenetics: the Mutscore Algorithm in a single center 15-year experience
Tipologia:
Tesi di dottorato
Dimensione
14.36 MB
Formato
Adobe PDF
|
14.36 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.