The purpose of this study was to explore additional options for calculating weights in weighted single-step GBLUP (WssGBLUP). In GWAS by ssGBLUP, GEBV are converted to marker (SNP) effects. Unequal variances for markers are then derived from SNP solutions and subsequently incorporated into a weighted genomic relationship matrix. Improvements on the SNP weights were obtained iteratively by recomputing both the SNP effects and the GEBV. Six options were used to calculate the weights: (1) proportional to u_i² where u_i is the effect of the i-th SNP; (2) proportional to u_i² + constant; (3) weights as ν^|s−2|, where ν is a scale standing for the departure from normality, and s is number of standard deviation from mean for each u_i² where p_i is frequency of the second allele; (4) as the largest effect (u_i²) among every 20 SNP; (5) as the mean effect of every 20 SNP; (6) as the summation of effects of every 20 SNP. A simulated data set was used that included 15,600 animals in 5 generations, of which 1,540 were genotyped for 50k SNP. The simulation involved phenotypes for a trait with heritability of 0.5 and affected by 5, 100, and 500 QTL. Accuracy between TBV and GEBV for genotyped animals in the last generation was used for evaluation. Comparisons also involved BayesB and BayesC with deregressed proofs or EBV from BLUP, and π = 0.99, 0.9 or 0.5. In single-step, SNP effects were tracked along 10 iterations and weights were equal to 1.0 in the first iteration. Option 5 was the best in identifying simulated QTL without background noise and with precision in most of the regions. Option 2 kept accuracy of GEBV at the plateau after 2 iterations and was 0.81 as opposed to 0.70 for BayesC and 0.48 for BayesB under 500 QTL scenario. All methods reached better accuracies than BayesB and BayesC when number of QTL approached or exceeded 100 (0.2% of all SNP) due to automatically including PA in GEBV. Weights based on a sum of SNPs may be superior to those based on individual SNPs.