Optimizing genomic prediction using low-density marker panels for streptococcosis resistance in red tilapia (Oreochromis spp.).

Streptococcosis is a major disease that causes huge economic losses to tilapia farming in Thailand. Breeding for Streptococcosis agalactiae resistant strains using the pedigree BLUP method has proven an effective approach to control the disease in red tilapia, but the accuracy of selection is relatively low. Genomic selection, which is based on genome-wide markers to predict genomic breeding values of selection candidates, provides a powerful approach for accelerating genetic progress and producing permanent gains in the population. We evaluated the implementation of four genomic prediction models, GBLUP, ssGBLUP, BayesB and BayesC, using 19 sets of SNP markers (ranging from 500 to 24 582 SNPs) in 886 fish challenged with S. agalactiae. The accuracy of prediction was estimated using a five-fold cross-validation approach, with 708 and 178 individuals sampled for the training and validation sets respectively. Prediction of the accuracy of each of the models was improved substantially compared with PBLUP (10%) using 1000 informative SNPs. The GBLUP model (65%), which required less computing time, outperformed the remaining models - ssGBLUP (53%), BayesB (47%) and BayesC (42%).