With the advancement in Genomic Medicine and increasing adoption of genetic tests in clinical practice, an information model is needed to support representation of diverse kinds of statements made about genetic variations to enable standardized and efficient data exchange. The Scientific Evidence and Provenance Information Ontology (SEPIO) is a modeling framework designed to represent scientific assertions, and related evidence and provenance information. The GA4GH Variation Annotation (VA) group is developing a framework, based on SEPIO, to link variant data to structured annotations. The framework supports a wide variety of variant annotations, including causal association to disease/phenotype and interpretations of clinical relevance and actionability, and it supports existing clinical lab standards such as variant interpretation guidelines.
In this work, we set out to develop a model for variant annotation statements that describe therapeutic responses. To accomplish this, we examined 22,884 statements related to therapeutic response from three knowledgebases: PharmGKB, ClinVar, and CIViC. Natural language processing (NLP) and human curation were used to analyze the statements and extract subcomponents. Key elements in those statements were identified and the VA framework was utilized to create a core information model for therapeutic response statements for genetic variations.
The analysis of the statements resulted in 88,981 entities that were mapped to the VA modeling framework and used to create a model for therapeutic response statements. The proposed core information model is built around a statement comprised of a subject (‘genetic variant’), a predicate (‘has_response’) and a descriptor (‘therapeuticIntervention’). Statements may include one or more qualifiers that refine the meaning of the core triple by describing the variant’s origin (somatic or germline), the treated genetic condition, a ‘comparator’ against which the predicted effect of the subject variant is compared, or the population from which evidence was drawn.
Our proposed information model is derived from statements in knowledgebases that are used to aid the interpretation of clinical genomic test results. This model will enable the capture of structured knowledge related to therapeutic response, its utilization in clinical applications, and its dissemination throughout the clinical genomics community.