Purchase this article with an account.
Xiaohui Li, Yelena Bykhovskaya, Talin Haritunians, David Siscovick, Anthony Aldave, Loretta Szczotka-Flynn, Sudha Iyengar, Jerome Rotter, Kent Taylor, Yaron Rabinowitz; Use of the polygenic model to predict risk of keratoconus using GWAS data. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5305. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The genetic contribution of individual SNPs to keratoconus susceptibility is usually modest and cannot be identified without large sample sizes. Thus, to begin to estimate the genetic contribution to keratoconus, we used polygenic modeling to generate genetic scores, utilizing combination of SNPs each with a small genetic effect, estimated from our genome-wide association study (GWAS) discovery cohort and tested in our GWAS replication cohort.
A discovery panel of 222 Caucasian keratoconus patients and 3324 controls was genotyped using the Illumina 370K beadchip. Selected SNPs identified by GWAS at p<=0.05 significance level (n=1189) were further genotyped in an independent replication case-control panel of 304 cases and 518 controls. 612 SNPs with low linkage disequilibrium (r2<0.2) were selected for the polygenic analysis. Genetic scores were generated for each individual in the replication panel, accounting for multiple SNPs based on the significant thresholds using the original score weighting from GWAS discovery data (implemented in PLINK). Association between these scores and keratoconus was tested by the logistic regression model. The keratoconus variance explained by the score was estimated by the Nagelkerke pseudo R-squared.
We scored SNPs identified in the GWAS discovery set based on four thresholds of significance: p<10-4 (n=2), p<10-3 (n=41), p<0.01 (n=279), and p<=0.05 (n=612). We next tested these GWAS-derived scores for potential association with keratoconus risk in the independent subjects from the replication cohort. As expected, we observed more significant association with keratoconus when including more SNPs. The explained phenotypic variance (r2) estimated by the score increased consistently with more alleles included in the model: 1.3%, 2.6%, 6.7%, and 7.1% for each threshold above, respectively (p=0.08, 1×10-6, 6×10-10, 2×10-10, respectively). Two of the polygenic scores reached Bonferroni corrected genome-wide significance of p<2.5×10-8, thus confirming the appropriateness of the modeling strategy. Thus, a genetic score, which accounted for a subset of GWAS SNPs with significance level of <=0.05, could explain at least 7% of phenotypic variance of keratoconus.
Using a polygenic model, we identified strong evidence that risk of keratoconus is well explained by hundreds of common alleles with small effect.
This PDF is available to Subscribers Only