The SCN1A -epilepsy prediction model calculates the probability of developing Dravet syndrome versus genetic epilepsy with febrile seizures plus (GEFS+) based on a given SCN1A variant and the age of seizure onset. The model considers the potential effect of the queried variant and compares it with an international database of 1,018 SCN1A patients with Dravet syndrome or GEFS+ from seven countries.
Amino acid position
Amino acid change
Age of onset
Pathogenic variants in the neuronal sodium-channel α1-subunit gene ( SCN1A ) are the most frequent monogenic cause of epilepsy. Phenotypes comprise a wide clinical spectrum including the severe childhood epilepsy, Dravet syndrome, characterized by drug-resistant seizures, intellectual disability and high mortality, and the milder genetic epilepsy with febrile seizures plus (GEFS+), characterized by normal cognition. Early recognition of a child's risk developing Dravet syndrome versus GEFS+ is key for implementing gene-specific, disease-modifying therapies before cognitive impairment emerges
Based on a well-phenotyped international cohort of 1,018 SCN1A -positive Dravet syndrome and GEFS+ patients, we used data from five countries (training cohort, N=743) to develop multiple prediction models. We employed data easily available for any young child presenting with recurrent seizures and a pathogenic SCN1A variant, including age of seizure onset and the genetic variant, combined with a newly-developed SCN1A genetic prediction score. We validated the model using two independent blinded cohorts from Australia (N=203) and Belgium (N=72).
A high SCN1A genetic score and young age of onset, were each associated with Dravet syndrome (P<0.001). A combined ' SCN1A genetic score & seizure onset' model separated Dravet syndrome from GEFS+ more effectively (area under the curve [AUC]=0.89) and outperformed all other models (AUC=0.79-0.85; P<0.001). Model performance was replicated in both validation cohorts (AUC-Australia=0.94; AUC-Belgium=0.92).
Study overview. We used a supervised machine learning approach to generate a prediction model using genetic data ( SCN1A genetic score) and clinical data (age of seizure onset in months) from 743 patients (training cohort). We tested the prediction model with two independent blinded validation cohorts (n= 275) and integrated it into the SCN1A -epilepsy prediction model.