Abstracts

Rationale: Clinical genetic discoveries aim to deepen biomedical knowledge of the mechanisms underlying the epilepsies, ultimately to enable advances in treatment and counseling. Rare genetic variants in established epilepsy genes can be sufficient to cause distinctive syndromes, recurring in families or arising de novo. Rare variants are also found in people with common epilepsies. Their scarcity in unselected populations suggests that these variants are likely to be detrimental biologically, and consequently, we hypothesize, to influence clinical manifestations. However, while the scale of modern genomic studies indicates vast potential for genotype-phenotype discovery, it also burdens manual phenotypic interpretation, typically constraining resolution to epilepsy types and broad syndromes.

Methods: We formalized the comparison of people with epilepsy by their clinical features, modeling clinical reasoning computationally for large cohorts. We optimized the Human Phenotype Ontology (HPO) and similarity algorithms for application to the epilepsies. The HPO represents the relationships between clinical concepts, allowing harmonization and analysis of clinical data. Extending our previous work, we collected further clinical data (including free text comments) from the Epi25 Collaborative and mapped these to HPO concepts manually. We reanalyzed established epilepsy genes by semantic similarity analysis and Fisher’s exact test before shifting focus to variants in new candidate genes and sets of biologically related genes.

Results: We analyzed the clinical and whole exome data available from people with epilepsy but without clinical genetic diagnosis recruited to the Epi25 Collaborative study. After quality control the cohort included 11,544 people with both genetic and clinical data, the latter yielding 286,490 HPO annotations using a repertoire of 1,579 clinical concepts. Established epilepsy genes had high semantic similarity, validating our approach. For example, 111 people carrying variants in SCN1A were similar (unadjusted p < 0.004) and had strong associations with focal hemiclonic (odds ratio, OR = 9.7) and febrile seizures (OR = 4.7), pyramidal signs (OR = 15), and encephalopathy (OR = 2.7). Exploration of novel genes identified several candidates with significant phenotypic similarity, which may hint at novel genetic etiologies. Individuals with variants in axon initiation segment genes were similar (p < 0.01), as were those with truncating variants in presynaptic genes (p < 0.05).

Conclusions: In this largest ever genotype-phenotype analysis in epilepsy, we demonstrated phenotypic evidence implicating particular genes and gene sets in clinical epilepsies. We confirmed classical associations with established monogenic epilepsy genes in people without the respective clinical diagnosis. In addition, our exploratory analysis has prioritized candidates for testing in independent cohorts. Recognition of the clinical associations of molecular features facilitates their use in classifications that integrate clinical and biological perspectives, paving the way for stratified natural history studies and clinical trials.

Funding: Please list any funding that was received in support of this abstract.: Wellcome, The Hartwell Foundation, NHGRI, NINDS.