A Phenome-Wide Association Study and the Discovery of a New Clinical Spectrum of Hereditary Cancer Genes
Menée aux Etats-Unis à partir de données 2016-2018 des dossiers de santé électroniques de 214 020 personnes, cette étude d'association à l'échelle du phénome identifie 9 maladies et affections associées à des variants de 13 gènes liés aux cancers à risque héréditaire
The transition from paper to electronic health records (EHRs) during the past 2 decades has fundamentally changed the way medicine is practiced. Electronic health records enable seamless integration of patient-centric health information from all aspects of health care, and the interactive interface to medical records for both patients and clinicians greatly facilitates efficient communication and decision-making. This shift has been accompanied by an exponential growth of clinical data stored in the EHR system. In the field of oncology, this explosion of health-related data has further accelerated with the application of next-generation sequencing (NGS) into clinical practice. Traditional approaches of genotype-phenotype association studies typically focus on 1 gene or trait at a time, using genetic information from an affected family or a particular patient cohort. Next-generation sequencing–based discovery for germline and somatic variants offers a unique opportunity for accurate identification of genes that predispose affected individuals to the development of cancer. Systematic incorporation of all genotype-phenotype associations has created encyclopedic platforms, such as the Online Mendelian Inheritance in Man (OMIM), which provide comprehensive curation of existing knowledge with the potential to inform clinical decision-making. Methods to effectively analyze the ever-increasing amount of health data and systematically link genetic information from NGS with clinical data from EHRs is becoming an active area of ongoing research. Recent development in cancer data science, particularly through the application of artificial intelligence and machine learning, greatly promotes the speed and scale of information analysis at an unprecedented rate. An integrated system of data collection, storage, analysis, and modeling will build a critical foundation for computational medicine, with a sustained impact in cancer care and beyond.
JAMA Oncology , éditorial, 2021