Economic optimization of full-sib test group size and genotyping effort in a breeding program for Atlantic salmon

Abstract

Abstract Background Breeding companies may want to maximize the rate of genetic gain from their breeding program within a limited budget. In salmon breeding programs, full-sibs of selection candidates are subjected to performance tests for traits that cannot be recorded on selection candidates. While marginal gains in the aggregate genotype from phenotyping and genotyping more full-sibs per candidate decrease, costs increase linearly, which suggests that there is an optimum in the allocation of the budget among these activities. Here, we studied how allocation of the fixed budget to numbers of phenotyped and genotyped test individuals in performance tests can be optimized. Methods Gain in the aggregate genotype was a function of the numbers of full-sibs of selection candidates that were (1) phenotyped in a challenge test for sea lice resistance (2) phenotyped in a slaughter test (3) genotyped in the challenge test, and (4) genotyped in the slaughter test. Each of these activities was subject to budget constraints. Using a grid search, we optimized allocation of the budget among activities to maximize gain in the aggregate genotype. We performed sensitivity analyses on the maximum gain in the aggregate genotype and on the relative allocation of the budget among activities at the optimum. Results Maximum gain in the aggregate genotype was €386/ton per generation. The response surface for gain in the aggregate genotype was rather flat around the optimum, but it curved strongly near the extremes. Maximum gain was sensitive to the size of the budget and the relative emphasis on breeding goal traits, but less sensitive to the accuracy of genomic prediction and costs of phenotyping and genotyping. The relative allocation of budget among activities at the optimum was sensitive to costs of phenotyping and genotyping and the relative emphasis on breeding goal traits, but was less sensitive to the accuracy of genomic prediction and the size of the budget. Conclusions There is an optimum allocation of budget to the numbers of full-sibs of selection candidates that are phenotyped and genotyped in performance tests that maximizes gain in the aggregate genotype. Although potential gains from optimizing group sizes and genotyping effort may be small, they come at no extra cost.