Answers are grounded in this paper and cite their sources.
Genomic sequence data hold enduring, lifelong clinical value, yet the information systems that generate and store them were built for episodic care and transient results. As a result, most of a patient's genomic information never enters a computable, reusable form: results arrive in the electronic health record (EHR) as static reports focused on a single indication, structured variant data are rarely captured, and standards-based interfaces to retrieve genomic data for reinterpretation, decision support, or artificial intelligence (AI) remain inconsistently implemented across clinical systems. In April 2024, the National Human Genome Research Institute (NHGRI) convened informatics and genomics leaders to define a Clinical Data Ecosystem for Genomic Health and identify the standards, infrastructure, and policies required to realize it. Here we summarize the principal conclusions of that workshop: the longitudinal value of the germline genome, the primacy of standards-based application programming interfaces (APIs), the requirements for real-time clinical decision support (CDS) delivered in primary care, the central role of computable phenotypes, the growing dependence of AI on granular, computable genomic data, the promise and pitfalls of polygenic risk scores (PRS), the tradeoffs among centralized, decentralized, and federated data-sharing models, and the imperative to treat patients as sovereign stakeholders. Together, we argue for a federated, standards-first architecture: a system that keeps data local, preserves patient control over research use, and is validated across heterogeneous real-world health systems. This is the most credible path from workshop consensus to durable infrastructure making genomic data computable, portable, and continuously reinterpretable across the lifespan.
No takes yet. Share an insight, caveat, or question.
Walton et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: