Large-Scale Genome Databases in Medicine: Resources, Applications, and Challenges

Abstract

The completed sequencing of the human genome in the first decade of the twenty-first century shifted the central problem of medical genetics from the acquisition of sequence to the interpretation of sequence. Interpretation requires scale: a single human genome contains roughly four to five million variants relative to the reference assembly, and distinguishing the few hundred that are medically actionable from the millions that are benign cannot be accomplished by reasoning from first principles. Large databases of human genomes, exomes, variant interpretations, tumour sequences, and linked clinical records have therefore become the operative infrastructure of genomic medicine, performing the same function for sequence data that randomized trials perform for therapeutics. This review describes the principal classes of large genomic databases used in clinical and translational medicine and evaluates their design, scale, data types, access models, and medical applications. Population reference databases, exemplified by the 1000 Genomes Project, the Exome Aggregation Consortium, the Genome Aggregation Database, and the Trans-Omics for Precision Medicine programme, provide allele frequency and mutational constraint estimates that constitute the primary evidence for variant filtering. Clinical variant knowledge bases, including ClinVar, ClinGen, the Human Gene Mutation Database, OMIM, and locus-specific databases, aggregate curated interpretations and phenotype–gene relationships. Population biobanks with linked electronic health records, including UK Biobank, All of Us, FinnGen, the Million Veteran Program, BioBank Japan, the Estonian Biobank, the China Kadoorie Biobank, the Taiwan Biobank, and deCODE, enable genome-wide association, phenome-wide association, polygenic risk estimation, and pharmacogenomic discovery at a scale that supports clinical translation. Cancer genome resources, including The Cancer Genome Atlas, the International Cancer Genome Consortium, AACR Project GENIE, COSMIC, and cBioPortal, define the somatic mutational landscape that underpins molecular oncology. Functional genomic atlases such as GTEx, ENCODE, the Human Cell Atlas, and the Cancer Dependency Map supply the regulatory and cellular context in which non-coding variation acquires meaning. The review then examines the access and governance architectures that determine who can use these data, the clinical applications they currently support, and the structural limitations that constrain them, principally the profound ancestry bias of most databases, the unresolved problem of variants of uncertain significance, the variable quality of phenotype and variant annotation, and the ethical and economic fragility of long-term data stewardship. It concludes that the next phase of development must prioritize globally representative sampling, machine-readable and evidence-graded interpretation, and sustainable public funding of openly accessible reference resources.