The production of rainbow trout (Oncorhynchus mykiss ) in France is increasingly focused on large-sized fish intended for smoking. The presence of red muscle (RM) , easily identifiable by its brown color , leads to flavor defects . Some specifications require processors to remove this portion of fillet , resulting in a direct economic loss . One of the objectives of the study is to quantify the genetic component associated with red muscle, through the estimation of genetic parameters including correlations with growth and processing traits and comparison of performances between crosses of two distinct strains .
An experimental cohort was produced by crossing 4 0 neomales and 4 0 females from the 2 commercial lines of rainbow trout from Aqualande group, according to 4 genotypes (AA, AB, BA and BB). At 18 months , 1436 fishes were slaughtered and measured. A steak before the dorsal fin and the caudal part of each individual were frozen and a DNA sample was collected for genotyping. The relative area of subcutaneous fat was estimated by Magnetic Resonance Imaging (MRI) on the steak. A thin transversal slice of the caudal part was photographed to quantify the area of red muscle using artificial intelligence (AI) . After quality control for phenotypes , genotypes , and pedigree reconstruction through parentage assignment , data from 1415 individuals were finally used for analysis . An animal model with genomic relationship matrix was used to estimate the genetic parameters related to the area and proportion of RM. P erformances of the 4 genotypes were compared with Tukey test to investigate heterosis and dominance effects.
The mean proportion of RM relative to the area of the steak was 8.52% in the study dataset. Heritability estimates were 0.24 for RM area and 0.32 for RM proportion. The genetic correlation between the two traits wa s positive but moderate (0.52). The proportion of RM measured by AI wa s genetically uncorrelated of other processing traits, whether it was with weight (0.09 ), carcass yield (0.09), eviscerated headless carcass yield (0.08) and relative surface of subcutaneous fat (0.09). Mean performances of hybrid crosses were systematically positioned between performances of pure lines. No dominance or heterosis effect was observed.
In brief, a new phenotyping method to measure RM surface by IA was developed in this study. The proportion of red muscle is heritable and genetically uncorrelated to growth and main processing traits . Reducing red muscle proportion in commercial stocks through its integration into breeding program is therefore feasible.
European Maritime and Fishery Fund and the French Ministery of Agriculture co-funded this work within the RedOUT project (FEAMP PFEA470019FA1000007).