Comparison of testing designs for genetic evaluation of social effects in aquaculture species

Abstract

Social interactions among individuals are commonly observed in both plants and animals, and may be partly genetically determined. The social effect of an individual does not affect the phenotype of the individual itself, but rather the phenotypes of all other individuals sharing the environment. As an alternative to traditional breeding schemes solely focusing on the direct breeding value affecting the phenotype of the individual itself, selection may be focused on total breeding values, including both direct and social effects. The latter may give a substantial increase in the rate of genetic gain of the population, given existence of significant social genetic effects. However, social effects are only identifiable when individuals are allocated in several groups. Previously, it has been suggested that an optimum design should include many small groups consisting of unrelated individuals, e.g., by randomly allocating individuals to each group (RAN). However, we hypothesize that a design with few (three) families per group (3FAM), and where each family is tested repeatedly in three different groups would be more advantageous. Using stochastic simulations, assuming a full-sib data structure typical for aquaculture breeding schemes, we compared the two designs with respect to accuracy of breeding values and their ability to produce accurate estimates of (co)variance components of direct and social effects. The results showed that the 3FAM design was clearly superior with respect to precision of estimated genetic (co)variance components of social effects and had substantially higher accuracy of both social and total breeding values, compared with the RAN design (16 to 105% increase). The advantage was particularly expressed at scenarios with zero or negative genetic correlation between direct and social genetic effects.