Genomic selection using a fast EM algorithm. Part 2, Analysis of simulated data

The paper reports on a fast EM algorithm for genomic selection by mapping QTL in genome-wide dense SNP marker data. The algorithm called emBayesB was used to analyse a 6000 SNP dataset simulated for the QTLMAS XII workshop. True breeding value was accurately predicted by GEBV with a correlation of 0.85 in the validation data, while the regression coefficient of 0.97 indicated unbiased predictions of breeding value. The results were similar to Bayesian MCMC estimates but were calculated in a fraction of the time. emBayesB was also able to accurately map the location of individual QTL which explain more than 1% of the total genetic variation.