Discovering the genetic basis of epilepsies through computational analysis of clinical phenotype

Abstract

The epilepsies are complex, primarily because of their diverse seizure types, neurodevelopmental features, and comorbidities. High throughput sequencing and genotyping has accelerated the pace of discovery for new genetic causes and risk factors. The rate at which these biological discoveries have been translated into clinically actionable knowledge has been slower because of the burden of collecting, harmonizing, and analyzing precise clinical data manually, which prohibits detailed phenotypic analysis at scale. In this thesis, I report some of my contribution to the opening of this phenotyping bottleneck by modeling clinical neurological reasoning for computational scaling of phenotypic analysis to increase the clinical value of genetic discoveries. I present a conceptual model for harmonizing seizure descriptions based on contemporary classifications. I critically appraise methods of associating clinical and genetic features in epileptology and propose a new phenotypic similarity measure, SimMinTO, based on the communication of clinical features between experts that models the traditional notion of recognizing a genetically associated syndrome. Applying these tools in two studies of over 10,000 people, I characterize the clinical features of carriers of copy number variants and ultra-rare sequence variants with reference to other people with presumed genetic epilepsies, reflecting the situation in an epilepsy clinic. Finally, I present a detailed analysis of data from 67 adults with CDKL5 Deficiency Disorder, a developmental and epileptic encephalopathy in which prior knowledge has been limited to children. I show how harmonization of historical multisource data can enable retrospective natural history studies of rare disorders to inform prognostication and the design of expensive and lengthy prospective studies including precision medicine trials. I hope that this work will introduce new standards, promoting more formal clinical characterization of complex disorders that will facilitate the rapid discovery and validation of associations, improving classification for stratified care and research both within neurology and beyond.