Genomic Prediction of Host Resistance to Sea Lice (L. salmonis) in Atlantic Salmon (S. salar)

Sea lice have a large negative economic and animal welfare impact for farmed Atlantic salmon. We tested the utility of genomic prediction using genome-wide SNP markers to predict breeding values for lice resistance and growth traits in commercial Atlantic salmon populations. The specific aims were (i) to estimate the heritability of lice resistance and growth traits using genomic and pedigree based methods, (ii) to assess the accuracy of genomic prediction using various SNP marker densities, and (iii) to test prediction accuracies in closely and distantly related training and validation populations.

Two commercial populations from different year groups (n = 1,119) were measured for weight, length and counts of lice 7 days after challenge. All animals were genotyped for ~33K SNPs, and trait heritabilities were estimated using a mixed model fitting the pedigree-based and genomic relationship matrix respectively. Genomic prediction accuracies were tested using five-fold cross validation analysis. The accuracy was calculated as the correlation between the EBVs and the adjusted-phenotypes divided by the heritability(h-square) [~r(y₁,y₂)/h].

Heritability estimates were ~0.25 for lice resistance and ~0.5 for growth traits.  Accurate genomic predictions of breeding values were possible for lice resistance (~0.6) and growth (~0.7) within populations; an improvement on using pedigree records alone. As few as 5K SNPs gave close to maximal accuracy, with little improvement using a higher marker density. Prediction accuracy was higher when training and validation populations were closely related, but less so with more distant relationships. In summary, genomic prediction is a useful tool for improving host resistance to sea lice in commercial salmon breeding schemes.