Performance of multi-trait genomic selection for Eucalyptus robusta breeding program

Autor: Jean-Marc Bouvet, Lolona Ramamonjisoa, Jean-Michel Leong Pock Tsy, Garel Makouanzi, Tuong-Vi Cao-Hamadou, Laval Jacquin, Daniel Verhaegen, Tahina Rambolarimanana
Přispěvatelé: Centre de Coopération Internationale en Recherche Agronomique pour le Développement (Cirad), ESSA, Université d'Antananarivo, Amélioration génétique et adaptation des plantes méditerranéennes et tropicales (UMR AGAP), Centre de Coopération Internationale en Recherche Agronomique pour le Développement (Cirad)-Institut National de la Recherche Agronomique (INRA)-Centre international d'études supérieures en sciences agronomiques (Montpellier SupAgro)-Institut national d’études supérieures agronomiques de Montpellier (Montpellier SupAgro), Institut national d'enseignement supérieur pour l'agriculture, l'alimentation et l'environnement (Institut Agro)-Institut national d'enseignement supérieur pour l'agriculture, l'alimentation et l'environnement (Institut Agro), Université Marien Ngouabi, CIRAD
Rok vydání: 2018
Předmět:
Zdroj: Tree Genetics and Genomes
Tree Genetics and Genomes, Springer Verlag, 2018, 14 (5), ⟨10.1007/s11295-018-1286-5⟩
ISSN: 1614-2950
1614-2942
DOI: 10.1007/s11295-018-1286-5
Popis: International audience; In forest tree genetic improvement, multi-trait genomic selection (GS) may have advantages in improving the accuracy of the genotype estimation and shortening selection cycles. For the breeding of Eucalyptus robusta, one of the most exotic planted species in Madagascar, volume at 49months (V49), total lignin (TL), and holo-cellulose (Holo) were considered. For GS, 2919 single nucleotide polymorphisms (SNP) were used with the genomic best linear unbiased predictor (GBLUP) method, which was as efficient as the reproducing kernel Hilbert space (RKHS) and elastic net methods (EN), but more adapted to multi-trait modeling. The efficiency of individual I model, including the genomic data, was much higher than the provenance effect P model. For example, with V49, mean goodness-of-fit was: r(I_Full)=0.79, r(P_Full)=0.37 for I and P, respectively. The prediction accuracies using the cross-validation procedure were lower for V49: r(I)=0.29 r(P)=0.28. The genetic gains resulting from the indexes associating (V49, TL) and (V49, Holo) were higher using I than for the P model; for V49, the relative genetic gain was 37 and 20%, respectively, with 5% of selection intensity. The single-trait approach was as efficient as the multi-trait approach given the weak correlations between V49 and TL or Holo. The I model also brings greater diversity: for V49 the number of provenances represented in a selected population was two and three with the P model, and 6 and 16 with the I model.
Databáze: OpenAIRE
Externí odkaz: