CONSTRUCTING A PREDICTIVE MODEL FOR SCN8A DEVELOPMENTAL AND EPILEPTIC ENCEPHALOPATHY

Abstract

SCN8A Developmental and Epileptic Encephalopathy (SCN8A-DEE) is a rare pediatric neurological disease caused by mutations in the SCN8A gene, which encodes for the voltage-gated sodium channel Nav1.6. The severity of this disease ranges broadly, from benign infantile epilepsy to debilitating epileptic encephalopathy. Due to the recent discovery of the cause of the disease and its low incidence, a standard of care among clinicians treating SCN8A-related epilepsy has not been established. To fully describe the disease spectrum and its natural history, detailed caregiver-reported data was collected through an online Registry established by the Hammer lab in 2014. Regions of interest within the SCN8A gene were identified using a patient mutation and public mutation database. Regions were classified as pathogenic, benign, complex, and lethal. Further genotype-phenotype and phenotype-phenotype associations were studied, focusing on the role that seizures and development have in determining disease outcome. Following these studies, an ordinal logistic regression model was made using a cohort of 154 patients with de novo gain of function mutations. The regression model identified four features that were highly predictive on their own and an additional five feature combinations that significantly contributed to predictive accuracy in the model. The final version of the predictive regression model was able to predict patient outcome with an accuracy of 78%, significantly higher than the expected 33% if randomly predicted in the cohort. The ordinal logistic regression model also confirms the presence of three distinct severity rankings within the SCN8A-DEE population.