Résumé
Fillet yield (i.e. the proportion of edible flesh from a given amount of fish) is a trait of high economic interest for species sold as fillets. Improving it by selective breeding is not an easy task, as it is a ratio trait, which causes mathematical difficulties. It is moreover a specific ratio where numerator ( fillet weight) and denominator (body weight) are very strongly correlated (genetic and phenotypic correlations in the range 0.89 – 0.99). This led some authors to conclude that they have the same genetic and phenotypic basis, precluding selection for improved fillet yield. In this study, we propose to study fillet yield as a component of two traits, fillet weight and waste weight (waste weight = body weight- fillet weight, so the sum of head, bones, fins and viscera weight), which we expect to be less correlated. Using data from 5 batches of fish from 3 species (sea bass, sea bream, rainbow trout), we show that as expected, fillet weight and waste weight are less correlated than fillet weight and body weight (on average, r A = 0.91 and r P = 0.88 vs. r A = 0.987, r P = 0.981). We used stochastic simulation to generate genotypes and phenotypes of fish using those genetic parameters for fillet weight and waste weight. We simulated 10 generations of selection for increased fillet yield using nine selection indices. Five indices had rather similar performances, residual fillet weight (the residual of the regression of fillet weight to body weight), fillet yield, fillet to waste ratio, restricted selection index (a linear index aimed at improving fillet weight while keeping waste weight constant) and linear index (optimized to improve fillet/waste ratio). With these indices, the average selection gain was +0.66% of fillet yield per generation (range 0.30 to 0.95% using real genetic parameters from 5 fish batches). Selection for the difference between fillet weight and waste weight was 30% less efficient, while selection for increased fillet weight or increased body weight was 55 – 65% less efficient. Selection against waste weight had a null or even negative impact on fillet yield. Factors favorable to higher selection gains are low initial fillet yield, different heritabilities and CVs of fillet weight and waste weight, low genetic correlation and high phenotypic correlation of fillet weight and waste weight. These results suggest that although fillet weight and body weight are strongly correlated and proportional to each other, moderate selection gains on fillet yield are possible. We consider it would be useful to add waste weight in the parameters recorded in future genetic studies on fillet yield.