PSVIII-20 Univariate Genomic Prediction with Different Heritability, QTL and SNP Panel Scenarios using Artificial Neural Network

Abstract

Abstract A simulation study was carried out to determine the likely accuracy of genomic prediction from univariate artificial neural network model with 1 to 10 neurons (ANN-1–10) using the SNPs on the first chromosome of Brangus beef cattle for 50% genetically correlated two traits with heritabilities of 25% and 50% (T1h2=0.25 and T2h2=0.5) determined either by 50, 100, 250 or 500 QTL. After QTL were created by randomly selecting 50, 100, 250 and 500 SNPs from the 3361 SNPs of 719 animals, their effects were sampled from a bivariate normal distribution. The breeding value of animal i in each QTL scenario was generated as Σgijβ j where gij is the genotype of animal i at QTL j and the vector of β j has the effects of QTL j from a bivariate normal distribution for T1h2=0.25 and T2h2=0.5. Phenotypic values (Σgijβ j+ei) of animal i for traits were generated by adding residuals (ei) from a bivariate normal distribution to the Σgijβ j of animal i. Genomic predictions for T1h2=0.25 and T2h2=0.5 were carried out by univariate Feed Forward MultiLayer Perceptron ANN-1–10 neurons. Three sets of SNP panels were used for genomic prediction: only QTL genotypes (Panel1), all SNP markers, including the QTL (Panel2), and all SNP markers, excluding the QTL (Panel3). Correlations between phenotypes and predicted phenotypes from 10-fold cross validation for T1h2=0.25 and T2h2=0.5 were used to assess predictive ability of univariate ANN-1–10 neurons based on 4 QTL scenarios with 3 Panels of SNP panels. Table 1 shows that genomic predictions from the trait with high heritability can achieve higher correlation than those from the trait with low heritability. Panels including SNP markers perform better prediction than Panel1 and have a greater chance of including markers in LD with QTL and allow the possibility of predicting each QTL from collective action of several markers.