Simulated data from beef genotypes

We used the genotypes of 608,696 SNPs of 1217 cattle of our own population to simulate a quantitative trait. We used the mean squared error (MSE) to measure the accuracy of an estimated quantitative trait nucleotides (QTN) effect. Of the 608,696 SNPs, we assigned 20 loci as QTN. We evaluated the accuracies for all the 20 simulated QTNs. The QTNs with positions, effects, and MSE’s are demonstrated in Table ​Table2.2. We also assigned each of the 608,696 SNPs, a small effect randomly sampled from a normal distribution with mean 0 and a very small variance to mimic a polygenic effect. The 20 loci collectively contributed 50% of the phenotypic variance, and the polygenic effect contributed 25% of the phenotypic variance. The remaining 25% phenotypic variance was contributed by the residual errors. All the 608,696 SNPs were used to construct bins of the cattle genome under various thresholds. A series of threshold value c for constructing bins were examined, ranging from 0.1 to 1 incremented by 0.1, where c = 1 is equivalent to SNP-Lasso, i.e., using the original SNPs by LASSO.

Information about the 20 simulated QTNs

^aProportion of the phenotypic variance contributed by each individual QTN

To evaluate the performance of the four methods, we generated 100 simulated samples from the above set up. The statistical power for a method was defined as the total number of QTNs detected per sample divided by 20 and then averaged over the 100 replicated samples. Type 1 error of a method was defined as the total number of SNPs above the significant thresholds in regions beyond the ±10 kb reserved interval of a simulated true effect divided by the total number of markers above the significant threshold, and then averaged across the 100 replicated samples. All four methods were used for analysis of this simulated data.