Interest and optimization of genomic selection for rainbow trout

The French rainbow trout breeding programs are using since 2004 a family selection method based on pedigrees established by molecular markers and performance measured on one to two thousand sibs of the candidates for selection. Genomic selection (GS) is an obvious development for these companies if it can be efficiently implemented technically and economically. A first medium density (MD) genotyp-ing tool, including 57,000 markers, being available since 2015, the aim of the thesis was to assess the interest of SG for three French lines of rainbow trout. A prerequisite was to assess whether the density of the MD chip was sufficient for the linkage desequi-librium between successive markers to allow effec-tive GS with regard to the genetic diversity of the French lines. The effective sizes of the lines were estimated at values from 50 to 70 and the linkage desequilibrium between successive markers at val-ues (r² ~ 0.30) compatible with a good GS efficiency. These results were confirmed by cross-validation studies of GS for various traits of reproduction, resistance to infectiouspancreatic necrosis (IPN), growth, cutting yield and quality flesh. With estimat-ed heritability at low (NPI) to high values (eviscer-ated and headed weight), all these traits are very polygenic: few genomic regions explain more than 1% of the genetic variance and no major effect re-gion (> 10% of the variance) has not been identified. The results obtained make it possible to conclude on an accuracy of GS superior of 10 to 30% (ac-cording to the studied traits) compared to the selec-tion based on the pedigree. They open up avenues for reflection on an economic optimization of GS by using low density chips to obtain a GS accuracy close to that obtained in MD (loss of efficiency 10% de la variance) n’a été identifiée. Les résultats obtenus permettent de conclure à une précision de la SG supérieure de 10 à 37% (selon les caractères étudiés) par rapport à la sélection basée sur le pedigree. Ils ouvrent des pistes pour l’optimisation économique de l’investissement en SG par utilisation de puces à plus basse densité permettant d’obtenir une précision de la SG proche de celle obtenue en MD (perte d’efficacité < 5% au-delà de 6000 marqueurs) et par réduction du nombre de collatéraux phénotypés. En effet, 700 individus collatéraux des candidats à la sélection semblent suffisants pour obtenir des précisions de la SG supérieures à celles d’une sélection sur pedigree. Les entreprises de sélection françaises peuvent donc mettre en œuvre une SG techniquement efficace et rapidement rentable en truite arc-en-ciel.