A systems medicine strategy to predict the efficacy of drugs for monogenic epilepsies.

Abstract

Monogenic epilepsies are rare but often severe. Because of their rarity, they are neglected by traditional drug developers. Hence, many lack effective treatments. Treatments for a disease can be discovered more quickly and economically by computationally predicting drugs that can be repurposed for it. We aimed to create a computational method to predict the efficacy of drugs for monogenic epilepsies, and to use the method to predict drugs for Dravet syndrome, as (1) it is the archetypal monogenic catastrophic epilepsy; (2) few antiseizure medications are efficacious in Dravet syndrome; and (3) predicting the effect of drugs on Dravet syndrome is challenging, because Dravet syndrome is typically caused by an SCN1A mutation, but some antiseizure medications that are efficacious in Dravet syndrome do not affect SCN1A, and some antiseizure medications that affect SCN1A aggravate seizures in Dravet syndrome. We have devised a computational method to predict drugs that could be repurposed for a monogenic epilepsy, based on a combined measure of drugs' effects upon (1) the function of the disease's causal gene and other genes predicted to influence its phenotype, (2) the transcriptomic dysregulation induced by the casual gene mutation, and (3) clinical phenotypes. Our method correctly predicts drugs that are more effective, less effective, ineffective, and aggravating for seizures in people with Dravet syndrome. Our method correctly predicts the positive "hits" from large-scale screening of compounds in an animal model of Dravet syndrome. We predict the relative efficacy of 1462 drugs. At least 38 drugs are ranked higher than one or more of the antiseizure drugs currently used for Dravet syndrome and have existing evidence of antiseizure efficacy in animal models. Our predictions are a novel resource for identifying new treatments for seizures in Dravet syndrome, and our method can be adapted for other monogenic epilepsies.