RT - Journal Article T1 - Evaluation of Genomic Prediction Accuracy in Different Genomic Architectures of Quantitative and Threshold Traits with the Imputation of Simulated Genomic Data using Random Forest Method JF - rap YR - 2018 JO - rap VO - 9 IS - 20 UR - http://rap.sanru.ac.ir/article-1-877-en.html SP - 129 EP - 138 K1 - Linkage disequilibrium K1 - Discrete traits K1 - Machine learning K1 - Imputation K1 - Genomic architecture AB - Genomic selection is a promising challenge for discovering genetic variants influencing quantitative and threshold traits for improving the genetic gain and accuracy of genomic prediction in animal breeding. Since a proportion of genotypes are generally uncalled, therefore, prediction of genomic accuracy requires imputation of missing genotypes. The objectives of this study were (1) to quantify imputation accuracy and to assess the factors affecting it; and (2) to evaluate the genomic accuracy of random forest (RF) algorithm to analyze binary threshold and quantitative traits. In the first phase, genomic data were simulated by QMSim software to reflect variations in heritability (h2 = 0.05 and 0.25), number of QTL (QTL=96 and 960) and linkage disequilibrium (LD=low and high) for 48 chromosomes. In the second phase, for real condition simulating, we randomly masked markers with 50% and 90% missing rate for each scenario; afterwards, hidden markers were imputed using FImpute software, and estimated imputation accuracy. In the third phase, to estimate genomic breeding values, we applied Random forest algorithm for original (before masking a proportion of SNPs) and imputed genotypes with quantitative and quality phenotypes. The accuracy of imputation was improved with increasing level of LD. With increase a major proportion of masked markers (90%), results of current study shed light on the effects of imputation accuracy on accuracy of genomic prediction. In the scenario combining the highest heritability, LD and QTL for threshold traits and in the scenario combining the highest heritability and LD and the least QTL for quantitative traits, random forest method had the best performance of genomic accuracy. Generally, accuracy of genomic prediction for threshold traits had more precise than quantitative trait when using the random forest method. LA eng UL http://rap.sanru.ac.ir/article-1-877-en.html M3 10.29252/rap.9.20.129 ER -