Genomic prediction of health traits in humans: demonstrating the value of marker selection.

In this study, we explored prediction of human height, high-density lipoproteins (HDL) and body mass index (BMI) using SNPs within a Croatian (N=2,186) and into a UK population (N=810) using two methods from livestock breeding: Bayes-C (using Gibbs sampling) and G-BLUP. Correlation between predicted and observed trait values in 10-fold cross-validation was used to assess prediction accuracy. Using all available 263,357 SNPs, Bayes-C and G-BLUP had similar prediction accuracy across traits within the Croatian data, and for height and BMI when predicting into the UK population. However, Bayes-C outperformed G-BLUP in the prediction of less polygenic HDL into the UK population. Feature selection allowed G-BLUP to achieve equivalent predictive performance to Bayes-C across all traits with greatly reduced computational effort. Feature selection in the G-BLUP framework therefore provides an efficient alternative to computationally expensive Bayes-C for traits considered in this study.