Purchase this article with an account.
Caroline C W Klaver, Milly S. Tedja, Xikun Han, Virginie JM Verhoeven, Nick Ericksson, Nick Furlotte, Najaf Amin, Cornelia van Duijn, Stuart MacGregor; Genetic risk scores of the largest multi-trait GWAS meta-analysis on refractive errors. Invest. Ophthalmol. Vis. Sci. 2019;60(9):3143. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Myopia is a complex genetic trait and its morbidity will become a public health issue in the near future. The explained phenotypic variance remains low (8% based on 161 loci). The aim of this study is to contribute to the missing heritability (h2) for RE by performing a multi-trait GWAS meta-analysis, and to elucidate the impact of the genetic factors by constructing genetic risk scores (GRS) for myopia risk prediction models in the case-control MYopia Study (MYST).
We performed a multi-trait GWAS meta-analysis on Spherical Equivalent (SphE), Age of Diagnosis of Myopia and Myopia Status in 620,035 participants from CREAM (N=44,710), UK Biobank (N=348,414) and 23andMe (N=134,145) using MTAG. Subsequently, we selected index variants from the genome wide significant SNPs on SphE using Plink --clump based on position (250 kb) and LD structure (>0.6). We then constructed GRS per individual in MYST (Ncases= 450; Ncontrol=591; cases with high myopia (HM) defined as spherical equivalent (SphE) ≤ -6 D and axial length (AL) ≥ 26mm): Σ(βSNP1*genotype + … + βSNP420*genotype). Finally, we calculated the variance explained (Nagelkerke r2) by GRS for HM and we constructed ROC curves comparing prediction models for HM using GRS, education level and parental myopia, with age and sex included in all models tested.
The multi-trait GWAS meta-analysis yielded 420 loci associated with RE and a replication of 94% (151/161) of the previously reported loci. The 420 loci explained 9.8% of the phenotypic variance (P=4.18E-17). All prediction models for HM, except the education model (P=0.08), were significantly different from the prediction model including only age and sex. The GRS-only model (AUC 0.67), the model including education level only (AUC 0.65) or parental myopia only (AUC 0.69), the model including education level and parental myopia without GRS (AUC 0.70) and with GRS (AUC 0.69), were not significantly different from each other.
This large dataset enabled further identification of novel loci associated with RE and thereby contributed to solving the missing heritability. The genetic risk prediction model predicting high myopia based on all 420 risk loci performed similar to prediction models including only nongenetic risk factors (education and parental myopia). This exemplifies the complex nature of the trait.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only