1 Quantitative trait prediction based on SNP-array data, a simula- tion study.

Motivation: Using simulation studies for quantitative trait loci, we evaluate the prediction quality of regression models that include as covariates SNPs which did not achieve genome-wide significance in the original genome-wide association study, but were among the SNPs with the smallest p-value for the selected association test. We compare the results of such regression models to the standard approach which is to include only SNPs that achieve genome-wide significance. Using mean square prediction error (MSPE) as the model metric, our simulation results suggest that by using the coeffi-cient of determination (R2) value as a guideline to increase or reduce the number of SNPs included in the regression model, we can achieve better prediction quality than the standard approach. How-ever, important parameters such as trait heritability, the approximate number of QTLs, etc have to be determined from previous studies or have to be estimated accurately.