Statistical model comparison in genetic analysis of challenge test data on Streptococcus agalactiae resistance in Nile tilapia (Oreochromis niloticus)

Abstract

Five statistical models were used to analysis genetic variations, heritability and the ability to predict survival of Nile tilapia based on challenge test data. 12 full-sib Nile tilapia families (1870 individuals) with an average weight of 8g in young group (G8) and 35g in older group (G35) were tested. Sire-dam binary threshold (BTH_sire-dam) model and Animal model (BTH_Animal) were using test-period survival as the univariate trait definition. And sire-dam survival score (SS) model was applied to test-day survival (one record per fish per day). Also a family mean survival (FMS) model and a family mean area (FMA) model were tested with the mean survival rate and integral area per family as trait definitions, respectively. Highest heritability was found with the FMS model (0.56 ± 0.42), followed by the BTH_sire-dam model (0.45 ± 0.17) and the FMA model (0.42 ± 0.34). The BTH_Animal model got a relatively much lower value of heritability (0.27 ± 0.11), whereas the SS model (0.07 ± 0.03) was distinctly the lowest. The highest Pearson correlation coefficients between predicted breeding values for the various models were 0.97 for the BTH_sire-dam model and the BTH_Animal model, 0.96 for the SS model, 0.87 for the FMA model and the FMS model had the lowest prediction power with a correlation of 0.77. In conclusion, BTH_sire-dam model was the best fitted model with a high heritability and the highest selection accuracy among these five models.