The concept of heritability sits at the heart of quantitative genetics. Twin and family studies have long established that traits ranging from height and body mass index to blood cell counts, cholesterol levels and susceptibility to schizophrenia are substantially inherited. Yet when geneticists sum up the effects of the millions of single-nucleotide polymorphisms captured in genome-wide association studies, or GWAS, they typically recover only a fraction of the heritability estimated from family data — a shortfall famously dubbed the missing heritability. Explanations have included rare variants, gene-gene interactions, gene-environment effects and imperfect measurement. The new work adds a powerful quantitative argument that a major piece of the missing puzzle has been hiding in plain sight, in the form of large DNA rearrangements that standard genotyping arrays and short-read sequencing struggle to detect.

Structural variants are, in many respects, the heavyweight class of genetic variation. A single deletion can remove an entire gene or a regulatory region controlling several genes; a duplication can multiply gene dosage; an inversion can disrupt gene order and alter how regulatory elements connect to their targets. Decades of cytogenetics and, more recently, long-read sequencing have shown that individual structural variants often exert effects on gene expression and phenotype that are disproportionately large compared with their frequency in the population. What remained unknown was whether this per-variant potency translates into a meaningful share of genome-wide heritability when aggregated across the whole genome and across the full spectrum of common human traits. Answering that question required a methodological innovation, because the standard statistical machinery of GWAS was never designed with structural variants in mind.

MiXeR-SV, the tool introduced by the Oslo-led team, bridges that gap by integrating long-read-derived structural variant catalogs — reference maps of deletions, duplications and other rearrangements assembled from advanced sequencing technologies — with GWAS summary statistics, the aggregated association data that most large genetic studies publish. Rather than requiring raw genome sequences from hundreds of thousands of individuals, the method works from the summary level, estimating how much of the genetic signal in a GWAS can be attributed to variants that are correlated, or in linkage disequilibrium, with structural variants. The framework extends the earlier MiXeR platform, which the group developed to quantify the number of causal variants underlying complex traits, and adds a partitioning layer that separates the heritability contribution of structural variants from that of ordinary single-nucleotide variants.

The researchers applied the method to 105 complex traits drawn from large-scale biobank GWAS resources. Using a stringent Bonferroni-corrected significance threshold of P less than 4.8 times ten to the minus four, they identified 31 traits showing significant enrichment of heritability in structural variants. The pattern was not uniform across biology. Enrichment was most extensive for hematological traits such as blood cell counts, for metabolic traits and biomarkers, and for cancer-related phenotypes. Neuropsychiatric and cardiometabolic traits showed enrichment as well, but in a more trait-specific fashion, with particular structural variant classes contributing in particular contexts. All anthropometric traits examined — measures of body size and proportions — showed evidence of structural variant contribution, though with modest effect sizes, consistent with the highly polygenic architecture in which thousands of variants each nudge the trait slightly.

A central concern for any study relying on reference catalogs of structural variants is whether the results depend on the particular technology used to build the catalog. Long-read sequencing platforms, such as those producing very long continuous DNA reads, and graph-based pangenome approaches detect overlapping but not identical sets of rearrangements, each with its own biases. To address this, the team repeated their analyses with two independent structural variant reference panels built from complementary long-read and graph-based catalogs. The results were remarkably consistent: 93.5 percent of the traits that reached significance with one panel were also significant with the other. This cross-technology agreement substantially strengthens the case that the observed enrichment reflects genuine biology rather than an artifact of any single variant-calling pipeline.

Perhaps the most consequential finding concerns the missing heritability problem directly. When the researchers incorporated structural variant architecture into their heritability models, the total heritability estimates for enriched phenotypes consistently increased. In other words, allowing for structural variants did not merely reshuffle a fixed pie of heritability — it enlarged the pie, recovering genetic signal that pure SNP-based models had left on the table. Moreover, the proportion of heritability attributable to structural variants correlated with the gap between SNP-based heritability estimates and the higher heritability estimates from twin studies across traits. This correlation suggests that a measurable portion of the long-standing discrepancy between family-based and molecular estimates of heritability can be explained by large rearrangements that conventional GWAS genotyping fails to capture well.

The team also tested whether the findings hold across human populations, an essential check in a field where most genetic research has historically focused on European ancestry cohorts. Cross-ancestry replication in Biobank Japan, one of the largest biobanks in Asia, confirmed the enrichment patterns observed in the discovery analyses, indicating that the disproportionate contribution of structural variants to complex trait heritability is not an artifact of one population’s demographic history or linkage disequilibrium structure. Supplementary analyses extended the validation further, including sensitivity tests of the linkage disequilibrium reference panels, simulations across varying GWAS sample sizes, and replication in the Taiwan Biobank, all of which supported the robustness of the core result.

The technical achievement underlying these findings is worth emphasizing. Structural variants are difficult to genotype at scale: their breakpoints vary between individuals, their sequence content can be complex, and short-read sequencing — the workhorse of most GWAS — cannot resolve many of them unambiguously. By leveraging the growing public catalogs produced by long-read sequencing consortia and pangenome projects, and by developing statistical machinery that operates on GWAS summary statistics rather than raw sequence data, MiXeR-SV makes it feasible to ask genome-wide questions about structural variant contributions using data that already exist. The authors note that the study used only publicly available data with existing ethical approvals, and that the framework’s estimates were calibrated through simulations covering linkage disequilibrium panel configurations and GWAS sample sizes.

The implications reach into two of the most active areas of human genetics. For genetic prediction, the results suggest that polygenic risk scores built exclusively from single-nucleotide variants systematically omit a component of inherited risk, and that incorporating structural variant information could improve prediction for traits where enrichment is strong, such as blood cell traits and certain biomarkers. For fine-mapping — the effort to pinpoint the causal variants behind association signals — the findings indicate that structural variants should be explicitly included as candidates, since a GWAS signal tagged by common SNPs may in fact be driven by a nearby deletion or duplication with a much larger effect on nearby genes. As long-read sequencing becomes cheaper and structural variant catalogs grow more complete across diverse ancestries, the quantitative framework introduced here offers a way to track how much of the genome’s influence on human health has been waiting in its largest and least-charted variations.

Subject of Research: Contribution of structural variants to the heritability of human complex traits

Article Title: Structural variants contribute substantially to complex trait heritability

Article References: Nguyen, D. T., Shadrin, A. A., Parker, N., Fuhrer, J., Vo, N. S., Dale, A. M., Andreassen, O. A., & Frei, O. (2026). Structural variants contribute substantially to complex trait heritability. Genome Biology. https://doi.org/10.1186/s13059-026-04248-y

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04248-y

Keywords: structural variants, heritability, GWAS, long-read sequencing, genetic architecture, complex traits, missing heritability, MiXeR-SV, polygenic traits, biobank, cross-ancestry replication, Genome Biology