Setting up a strain-testing design for the seabass, Dicentrarchus labrax: a simulation study

Abstract

The seabass, Dicentrarchus labrax L., is a leading species in European marine aquaculture, but its genetic characteristics necessary to set up breeding programmes are still unknown. New tools, such as genomic markers, now allow to carry out strain testings and estimations of genetic parameters in reasonable scale experimental facilities. However, because of the high cost of genotyping, the total number of fish to be measured will be limited and mating designs (type of cross, number of sires and dams) must be optimised. An optimisation trial is done in this study by simulation of a quantitative trait for two different strains of fish according to a polygenic model with additive, dominance, maternal and residual effects. The phenotypic coefficient of variation and the difference between crosses means ( D=10%) were also included in the model. For technical limitations due to the species, the mating design chosen was a factorial cross between sires of both strains with dams from one strain (topcross). Different combinations of size of progeny sample (500–4000), number of sires (10–120/strain) and dams (4–16) were studied. Various mixed models (SAS®, proc Mixed, option REML) were used to estimate D, the power for detecting D, heritability ( h 2) and their standard error (S.E.) over 1000 repetitions. Incomplete factorials were also tested as well as the effect of differential survival between dam families. In strictly additive models (no dominance or maternal effects), D and h 2 estimates were unbiased. The number of dams did not noticeably affect the precision of estimates, whatever the number of sires. A joint optimisation of sample size and number of sires led to a consensus value of 40 sires/strain with a sample size of 2000 progenies, which gave good precisions for D and h 2. It was also shown that a mixed model (fixed strain effect+random sire effect) gave more accurate results than a simple fixed model (strain effect only). The inclusion of maternal and dominance effects did not markedly bias nor affect the precision of D and h 2 estimates in a full factorial design, but could lower the precision (up to 40% increase S.E.) in incomplete factorial design. Variance in dam family sizes (due to differential survivals) did not affect the results in full factorials when set at values observed in seabass. It is then concluded that a factorial cross of 40 sires/strain with 8 dams and a sample size of 2000 progenies (on average 1000/strain) is appropriate for comparing seabass strains (power >0.8 for detecting a 20% difference between strains means) and jointly estimating heritability.