Résumé
Marker-based parentage assignment provides the opportunity to investigate factors of efficiency for mixed-family designs and factorial mating. In such designs, family size is both uncontrolled and small, which may be thought to limit the accuracy of estimated breeding values (EBVs). The objective of this work was to estimate the accuracy of EBVs of growth and quality traits in a large factorial mating design and in commercial breeding conditions. An expected six hundred full-sib families of rainbow trout Oncorhynchus mykiss (2042 fish in total) were produced by ten factorial matings of six dams with ten sires. Fish were phenotyped for body weight, carcass yield, fillet yield, fillet fat content and fi llet colour, and family information was recovered using microsatellite markers. The accuracy of EBVs was estimated using or removing individual performance to mimic combined family selection (with individual phenotype) or sib selection (without individual phenotype). The traits investigated had medium to high heritability (0.17–0.58). High to very high accuracy (0.630–0.817) was estimated for combined family selection. The accuracy of sib selection (not using individual phenotype) was 18–22% lower (0.542–0.638), but remained in the upper range reported for such traits. This level of accuracy was higher than those reported in conventional breeding programs using separate family rearing. This was true even for families with a very low number of full-sibs. Individual EBV accuracy was more closely linked to the total number of full- and half-sibs of each fish than to its number of full-sibs. We hypothesize that this was due to the factorial mating, which led to a high number of the genetic ties between sibs. These results highlight the possibility of introducing precise estimated breeding values for quality traits into combined or sib selection in breeding programs when using mixed families from factorial designs and marker- based parentage assignment in aquaculture species