Genomic analysis of refractive error in 1.76 million people reveals the performance and population limits of genetic prediction of myopia
Background: How well can genetic risk for refractive error be predicted?
A study published in Nature Genetics identified 932 genetic variants associated with refractive error, including myopia and hyperopia, and evaluated models that use genetic information to distinguish levels of myopia risk. The discovery analysis included 1,761,068 people of European, East Asian, and African ancestry. A model incorporating functional annotations explained 21.4% of variation in refraction between individuals in cross-validation among people of European ancestry. Its performance was lower when applied to other ancestry groups.
Refractive error occurs when the eye’s length and the refractive power of the cornea and lens do not match, preventing light from focusing precisely on the retina. Many genetic variants are involved, but previous genetic studies have disproportionately represented people of European ancestry. The researchers therefore combined ancestry-specific analyses with cross-ancestry analyses to examine both shared genetic features and group-specific differences.
Key findings: 241 new associations and prediction based on functional annotations
The researchers combined genome-wide association study (GWAS) data from 10 cohorts. The study included 1,495,159 people of European ancestry, 121,172 of East Asian ancestry, and 144,737 of African ancestry. The analyses used measured mean spherical equivalent refraction (MSE), as well as refraction estimated from age at first wearing glasses, self-reported myopia, and information from medical records. In other words, not all participants underwent eye examinations using the same method.
The combined analysis identified 932 largely independent associated variants across 575 genomic regions; 241 of these associations were newly reported. Gene-prioritization analyses highlighted 23 genes involved in eye development. The study also identified 16 putative causal candidate variants supported by all six statistical fine-mapping strategies. Their actual causal roles still require experimental validation.
The researchers built a polygenic risk score (PRS) incorporating annotations that describe the functional characteristics of variants. In 10-fold cross-validation in European-ancestry UK Biobank participants, it explained 21.4% of the variance in MSE. This is the proportion of differences in refraction between people explained by the model; it does not represent an individual’s probability of developing myopia or the model’s prediction accuracy. The score explained 18.0% of the variance in the separate European-ancestry EPIC-Norfolk cohort.
In European-ancestry UK Biobank participants, the PRS had an area under the receiver operating characteristic curve (AUC) of 0.806 for distinguishing high myopia, defined as MSE of −6.00 diopters or less. AUC was 0.883 when only age at first wearing glasses was used, rising to 0.920 when that information was combined with the PRS. AUC measures how well the model distinguishes the presence or absence of high myopia; it should not be interpreted as accuracy in predicting general vision loss or blindness.
Implications and outlook: Prediction gaps across ancestry groups and the need for clinical validation
When a model trained on European-ancestry data was applied to other groups, it explained 8.8% of MSE variance among people of East Asian ancestry and 4.0% among people of African ancestry. In a separate comparison that excluded tuning samples, a multi-ancestry model increased the explained variance from 7.1% to 9.7% for East Asian ancestry and from 4.1% to 4.2% for African ancestry, compared with the European-ancestry model. The validation samples differed, so these figures cannot be directly compared with the earlier results.
The findings suggest that genetic information could serve as an additional measure for distinguishing levels of myopia risk. However, the predominance of European ancestry in the discovery sample and differences in measurement methods remain challenges. More balanced samples and precise refraction measurements are needed to determine whether performance holds in other groups. This study did not establish that examination or management strategies based on genetic scores reduce myopia progression or complications.
Sources: Nature Genetics paper, provided publicly available full text and methods.
Nature Genetics, Published online: 20 April 2026; doi:10.1038/s41588-026-02576-0Multi-ancestry and ancestry-stratified genome-wide analyses identify genetic variants associated with refractive error and enable construction of an enhanced polygenic predictor incorporating functional annotations.
One approach for future studies is to combine genetic scores with existing eye examinations and medical history to identify people for earlier examinations. The researchers also recommended earlier eye examinations for people at higher genetic risk of myopia, but this recommendation is based on risk-stratification results. The study did not show that a single genetic test can determine the optimal time for management or prevent vision loss. In particular, the substantial prediction gap limits direct application of the performance observed among people of European ancestry to people of East Asian or African ancestry. Prospective studies should assess prediction performance across groups and determine whether adding genetic information to examination and management strategies provides real benefits to patients compared with existing approaches.