Accuracy of genomic prediction of host resistance to salmon lice in Atlantic salmon (Salmo salar) using imputed high-density genotypes

Abstract

Salmon lice (Lepeophtheirus salmonis) is a marine ectoparasite responsible for major losses to the salmon farming industry each year. Salmonids are the primary hosts of the parasite, including the widely farmed species Atlantic salmon (Salmo salar) and rainbow trout (Oncorhynchus mykiss). Improving resistance towards the parasite in farmed Atlantic salmon could decrease the need for treatments, increase the welfare of the fish, as well as reduce the infection pressure on wild populations. Phenotypic resistance can be recorded in controlled challenge-tests and has been found to be moderately heritable. The aim of the study was to compare three different genomic selection models with respect to within- and across-family prediction accuracy with both moderate and high SNP-chip densities (215 K and imputed 750 K). The models tested were: Genomic Best Linear Unbiased Prediction (GBLUP), BayesC and a model combining a polygenic term and a BayesC term (BayesGC). Predictive abilities of the models were compared using five-fold cross-validation.The trait was found to be highly polygenic. All three models had a similar predictive ability. The BayesGC model had a slight advantage over the GBLUP and BayesC models, however this difference was not significant. For within-family prediction there was no advantage from increasing the SNP density from 215 K to 750 K genotype density. However, for across-family prediction a slight improvement in predictive ability was observed at the higher density compared to the lower.