ABSTRACT. Three different procedures were implemented to calculate weights for a genomic relationship matrix to restrict the shrinkage along iterations of weighted single-step genomic BLUP (WssGBLUP). The procedures as well as BayesC were tested with 3 simulated data sets. Prediction accuracy for WssGBLUP improved at 2nd or 3rd iteration by updating only the top number of SNP equal to 1 × or 3 × the number of QTL; accuracy increased after 3rd iteration and remained stable by using weights proportional to 2pi(1- pi)ui2+ constant. Except in the 5 QTL scenario, accuracies with all WssGBLUP procedures were higher than with BayesC. Noise in Manhattan plots was small with 5 and 100 QTL but large with 500 QTL.