ROC curve analyses of prior complex diseases have shown modest predictive ability of genetic sum scores, with AUCs of 0.54 for diabetes for a genetic risk score created based on previously associated variants (Talmud et al., 2010) to 0.65 corresponding to the 3% of variance in schizophrenia risk explained by a risk score created based on a large number of SNPs that met less stringent p-value thresholds in GWAS (Purcell et al., 2009; Jostins and Barrett, 2011). Most of the genetic variants contributing to AD have small effect sizes. This, along with the fact that AD has both genetic and environmental risk factors, means that any one SNP alone is not expected to be a good predictor of AD. This study aims to explore the aggregate impact of multiple genetic variants with small effect sizes on risk prediction in order to test whether known genetic contributions to AD can be an effective predictor.
