All of Us multi-omics widens the map of human genetic regulation
The All of Us Research Program has moved a step beyond broad population-scale genomics and into molecular mechanism. The new multi-omics cohort brings together proteomic data from 10,043 participants and transcriptomic data from 9,043 participants, all drawn from ancestrally diverse U.S. volunteers. In practical terms, this creates a large, layered reference for asking not only which variants are present, but how they alter RNA expression, splicing, and circulating protein levels across different ancestry groups.
The scale matters because gene discovery alone rarely explains disease biology. Variant interpretation becomes far more useful when a signal can be tracked across molecular layers and then anchored to a known disease gene or therapeutic target. Here, investigators assembled comprehensive expression, splicing, and protein quantitative trait locus analyses and identified 182,246 quantitative trait loci in total. A large share of those findings was then refined through fine-mapping, with 54,255 loci resolving to rare variants and 59.4% of those, or 32,235, narrowing to a single causal variant.
That level of resolution changes the evidentiary standard for mechanism.
Rare variant signals sharpen causal inference across molecular layers
Rare variants are often the hardest to interpret in clinical genetics because they can be population-specific, difficult to replicate, and sparsely represented in reference databases. The multi-omics framework in All of Us addresses that problem by linking rare variation to measurable molecular consequences. Signals that recur across RNA and protein layers are more credible than isolated associations, especially when they converge on the same gene and pathway.
The findings also showed that rare variant effects were more likely to be ancestrally restricted. That pattern is biologically and clinically important. In a cohort deliberately designed to reflect the diversity of the United States, ancestry-specific signals are less likely to be dismissed as noise and more likely to represent real population structure in human biology. For genetic diagnostics, this means a variant that appears rare or absent in one reference panel may still exert a substantial effect in another ancestry group, altering how pathogenicity is assessed and how missingness in existing databases is interpreted.
The study also showed replication across molecular layers. When a rare variant influences transcriptional output, affects splicing, and shifts protein abundance, the biological story becomes more coherent. Each layer supports the same underlying mechanism, whether the ultimate effect is gain or loss of function, altered transcript isoform usage, or a change in protein quantity. That layered concordance strengthens gene prioritization for disorders where the phenotype may be subtle, pleiotropic, or incompletely captured in routine clinical records.
Mendelian disease genes and drug targets gained stronger support
One of the most clinically consequential observations was the enrichment of rare variant QTLs in known Mendelian disease genes and approved drug targets. The odds ratio for Mendelian disease genes was 1.4, while approved drug targets carried an odds ratio of 1.3. These are not abstract genomic statistics; they point toward a resource that can improve both diagnosis and therapeutic biology.
A variant that lands in a Mendelian disease gene carries a different interpretive weight when it also alters RNA or protein output. For clinicians and translational researchers, that combination can help move a candidate variant from uncertain significance toward a more actionable classification. For drug development, the overlap with approved targets is equally relevant. Human genetic perturbation of a target that is already druggable can validate the pathway, clarify dose-response expectations, and reveal whether the biology is likely to be beneficial or risky before a compound ever reaches a large trial.
The multi-omic results also supported evidence for 5,186 variants, many of which are absent from ClinVar, while still affecting established disease genes. That gap matters in everyday genetics. Clinical databases remain powerful, but they are incomplete, especially for ancestrally diverse populations that have historically been underrepresented in discovery datasets. A variant missing from ClinVar is not necessarily benign; it may simply have escaped prior detection, prior testing, or prior aggregation at sufficient scale. By linking such variants to reproducible molecular effects, All of Us adds a functional layer that can help close that gap.
Why ancestry diversity changes the interpretation of genetic disease
The ancestry composition of the cohort is not a secondary design feature. It is the condition that makes many of these discoveries possible. Rare variants are often concentrated within specific ancestral groups, and those signals can be diluted or missed entirely in studies dominated by a narrow genetic background. In a more diverse cohort, the same variant can be observed with enough frequency to permit fine-mapping and molecular confirmation.
This has direct implications for equity in genomic medicine. A diagnostic pipeline built mostly on European reference data will systematically underperform when applied to patients from other ancestral backgrounds. The result is more variants of uncertain significance, fewer confident annotations, and lower clinical yield. The All of Us data show how multi-omics can reduce that blind spot by connecting genotype to function in a population that better reflects the country’s genetic diversity.
It also changes the way investigators think about causality. A single locus may look complex when examined only through association statistics. Once the same locus is interrogated through RNA and protein readouts, the underlying mechanism may become much clearer. Some variants will alter transcript abundance; others will change splicing; others still will affect protein levels without an obvious transcriptional signature. The multi-layer approach helps separate these mechanisms rather than forcing them into one generic category of “association.”
Functional genomics is becoming a clinical interpretation tool
The practical consequence of this cohort is the creation of a reference resource that supports both discovery and interpretation. For rare disease genetics, multi-omics can help prioritize candidate variants, especially when family data are limited or when standard variant annotation is inconclusive. For complex disease research, it can reveal shared pathways that connect common and rare variation to the same downstream biology.
The resource also has immediate translational value for target validation. When a signal repeatedly implicates an approved drug target, the biological credibility of that target rises. When a signal repeatedly implicates a Mendelian disease gene, the odds that the gene has true causal relevance also increase. When a variant affects multiple molecular layers, the mechanism is less likely to be a statistical artifact and more likely to reflect a real perturbation in human biology.
A single cohort does not solve variant interpretation across every disease state, and it does not eliminate the need for functional validation in model systems or clinical follow-up in patients. The present dataset is still limited by the molecular assays it includes and by the fact that even a large cohort cannot fully capture every rare allele present across the population. Some variants will remain difficult to classify until additional phenotyping, deeper sequencing, or experimental work is available. The most concrete next step is not broader rhetoric about precision medicine; it is continued expansion of ancestrally diverse multi-omic reference data so that rare variant interpretation becomes less speculative and more biologically grounded.