82. Enabling large scale precision oncology research with a new standard for genomic variants: OMOP Genomic

Abstract

Precision Oncology requires the application of genomic variants to epidemiological methods used in clinical research. Variants are detected through traditional immunohistochemistry, microarray or PCR panels, or large-scale next generation sequencing, either for purposes of clinical care or understanding of biological processes of carcinogenesis and cancer treatment. However, we see few variants used in standard epidemiological research in the context of rich phenotypes: longitudinal observational patient data of treatment, progression and survival. The reason for this lies in the incompatibility of data representations and methods. In a typical observational study, cohorts, exposures, and outcomes are defined through clinical events encoded by predefined concepts. Epidemiological methods use this closed system model to measure the statistical change in rates, which are calculated with the denominator of all possible clinical states. The same is very challenging for genomic variants, which work like open systems: The representation of possibly infinite variants, whether previously observed or not. The Oncology Working Group of OHDSI has been working on the development of a canonical, comprehensive, and non-redundant representation of genomic variants that are clinically relevant for cancer. The resulting reference list, provided as part of the OMOP Standardized Vocabularies, is called 'OMOP Genomic'. It was constructed by consolidating genomic variants from public somatic cancer variant knowledgebases and contains +95,000 variations from 575 cancer genes. Assessment of the current gap, integration of other genomic knowledgebases are essential to improve the coverage of important and clinically relevant mutations implicated in cancer. Precision Oncology requires the application of genomic variants to epidemiological methods used in clinical research. Variants are detected through traditional immunohistochemistry, microarray or PCR panels, or large-scale next generation sequencing, either for purposes of clinical care or understanding of biological processes of carcinogenesis and cancer treatment. However, we see few variants used in standard epidemiological research in the context of rich phenotypes: longitudinal observational patient data of treatment, progression and survival. The reason for this lies in the incompatibility of data representations and methods. In a typical observational study, cohorts, exposures, and outcomes are defined through clinical events encoded by predefined concepts. Epidemiological methods use this closed system model to measure the statistical change in rates, which are calculated with the denominator of all possible clinical states. The same is very challenging for genomic variants, which work like open systems: The representation of possibly infinite variants, whether previously observed or not. The Oncology Working Group of OHDSI has been working on the development of a canonical, comprehensive, and non-redundant representation of genomic variants that are clinically relevant for cancer. The resulting reference list, provided as part of the OMOP Standardized Vocabularies, is called 'OMOP Genomic'. It was constructed by consolidating genomic variants from public somatic cancer variant knowledgebases and contains +95,000 variations from 575 cancer genes. Assessment of the current gap, integration of other genomic knowledgebases are essential to improve the coverage of important and clinically relevant mutations implicated in cancer.