Where Did the Heritability Go? Genetics' Most Stubborn Unsolved Problem
For much of the past two decades, the scientific community operated under a reasonable expectation: sequence enough human genomes, compare enough patients to enough healthy controls, and the genetic underpinnings of common diseases would reveal themselves. Genome-wide association studies, or GWAS, were the instrument designed to fulfill that promise. They have delivered important findings. They have also exposed a profound and still-unresolved puzzle that sits at the center of modern genetics.
The puzzle has a name: missing heritability. And understanding it matters not just for researchers, but for anyone who has ever wondered what a genetic test can—and cannot—tell them about their future health.
What Heritability Actually Means
Before exploring what is missing, it helps to clarify what heritability measures in the first place. In genetics, heritability refers to the proportion of variation in a given trait—height, blood pressure, risk of schizophrenia—that can be attributed to genetic differences among individuals within a population. It is not a fixed property of a trait itself, but a statistical estimate that depends heavily on the population and environment being studied.
Twin studies have long been the gold standard for estimating heritability. Because identical twins share nearly all of their DNA while fraternal twins share roughly half, comparing how similar each pair is for a given trait provides a window into genetic influence. By these measures, height is approximately 80 percent heritable. Type 2 diabetes carries a heritability estimate around 50 percent. Schizophrenia approaches 80 percent. These figures suggest that genes play a substantial role in shaping these conditions.
The Promise of GWAS—and Its Limits
When researchers began conducting large-scale GWAS in the mid-2000s, the expectation was that scanning hundreds of thousands of genetic variants across the genome would identify the specific locations responsible for the heritability twin studies had documented. What emerged was a more complicated picture.
For height—among the most studied traits in human genetics—early GWAS identified dozens of associated variants. Yet when researchers summed up the collective contribution of those variants, they explained only about five percent of the trait's variation, despite an estimated heritability of 80 percent. Similar gaps appeared across nearly every complex disease studied, from cardiovascular disease to autism spectrum disorder. The variants that GWAS found were real, reproducible, and statistically robust. They simply could not account for most of the heritability that family and twin data predicted should exist.
This discrepancy became known as the missing heritability problem, a term popularized in a landmark 2008 commentary in Nature by Maynard Olson and colleagues. The phrase captured the frustration of a field that had expected a cleaner story.
Where Researchers Are Looking
Scientists have pursued several explanations for the gap, and the current consensus is that no single factor is responsible. Instead, the missing heritability likely reflects a convergence of several distinct phenomena.
Rare variants with large effects. Standard GWAS are designed to detect common genetic variants—those present in at least one to five percent of the population. Variants that are rarer than this threshold are statistically difficult to detect without enormous sample sizes. Yet rare variants can carry substantial effects on disease risk. Whole-genome and whole-exome sequencing studies have begun cataloging these rarer contributors, and they appear to account for a meaningful portion of the gap in certain conditions, including intellectual disabilities and some cancers.
Gene-gene interactions, or epistasis. Traditional heritability models assume that genetic variants act largely independently and their effects can be added together. In biological reality, genes interact. One variant may amplify or suppress the effect of another, and these interactions are notoriously difficult to detect statistically without sample sizes that dwarf even the largest current studies. Epistasis may not explain the majority of missing heritability, but researchers increasingly believe it contributes meaningfully to complex traits.
Structural and copy number variation. Much of the initial GWAS era focused on single-letter changes in the genetic code, known as single nucleotide polymorphisms. But the genome also varies at a larger scale: entire segments can be duplicated, deleted, inverted, or rearranged. These structural variants are harder to detect and were underrepresented in early association studies. Growing evidence suggests they play a significant role in neurodevelopmental conditions and other complex disorders.
Epigenetic contributions. Perhaps the most conceptually challenging piece of the missing heritability puzzle involves mechanisms that sit above the DNA sequence itself. Epigenetic modifications—chemical tags on DNA and its associated proteins that influence whether genes are active or silent—can be influenced by environment and, in some cases, transmitted across generations. If some of what twin studies measure as heritability is actually transmitted through epigenetic rather than sequence-level mechanisms, conventional GWAS would be blind to it entirely.
Overestimates in heritability calculations. A quieter but important thread of the scientific debate concerns whether twin-based heritability estimates are themselves inflated. Some researchers argue that identical twins share not just DNA but also more similar prenatal environments and social experiences than fraternal twins, potentially confounding the genetic signal. If heritability estimates have been systematically overstated, the gap that needs explaining may be smaller than it appears.
Why It Matters for Medicine
The missing heritability problem is not an abstract academic concern. It has direct consequences for the clinical promise of personalized medicine. Polygenic risk scores—tools that aggregate the effects of many genetic variants to estimate an individual's disease susceptibility—are increasingly being introduced into clinical settings, including cardiovascular disease screening and breast cancer risk assessment. Their predictive power depends directly on how completely known variants capture true genetic risk.
When a substantial fraction of heritability remains unaccounted for, polygenic scores underperform. A patient assessed as low-risk based on known variants may carry rare variants, structural changes, or epigenetic patterns that push their actual risk considerably higher. Conversely, apparent high-risk scores may not translate into disease if protective interactions elsewhere in the genome are ignored.
For clinicians and patients navigating genetic testing in the United States today, this means that a negative or reassuring result from a polygenic risk tool should be interpreted with appropriate caution. The science is advancing rapidly, but it has not yet closed the gap.
A Richer Model of Inheritance
Perhaps the most lasting contribution of the missing heritability debate is the way it has complicated—and ultimately enriched—the scientific understanding of how traits are inherited. The initial hope that GWAS would yield a relatively small number of high-impact variants, each explaining a meaningful slice of heritability, has given way to a far more intricate picture: thousands of variants of tiny individual effect, rare variants of large effect, structural complexity, molecular layers beyond the sequence, and environmental interactions woven throughout.
This complexity does not diminish the power of genetics as a scientific discipline. It deepens it. As sequencing technologies become more affordable and study cohorts grow into the millions, researchers are steadily recovering pieces of the missing signal. The picture that emerges is one of inheritance as a biological system rather than a simple ledger—dynamic, layered, and still yielding new surprises to those willing to look carefully.
