Accuracy of scarcely recorded wood traits in a Eucalyptus grandis population is improved by combining genomic selection and predictor traits

JURCIC, ESTEBAN J.; VILLALBA, PAMELA V.; DUTOUR, JOAQUÍN; CENTURIÓN, CARMELO; MUNILLA, SEBASTIÁN; CAPPA, EDUARDO P.

Harbin

Conferencia; The 20th IUFRO Tree Biotech and the 2nd FTMB (Forest Tree Molecular Biology and Biotechnology) Conference; 2022

The genomic selection methodology is particularly relevant for traits that are difficult or expensive to measure. In this study, we investigated the impact of using genomic information and/or records on predictor growth traits to increase the breeding value accuracies of scarcely recorded target wood quality traits. The scarcely recorded traits (n=1,214) were pulp yield, cellulose, extractive, and wood density, while the predictor traits (n=3,159) were diameter at breast height and total height. Data were obtained from an open-pollinated progeny trial of Eucalyptus grandis (Hill ex Maiden). A total of 548 trees were genotyped with 37,229 single nucleotide polymorphisms (SNPs) using the Axiom Euc72K. The performance of single- (ST) and multiple- (MT) trait single-step genomic best linear unbiased prediction (ssGBLUP) and conventional pedigree-based (ABLUP) models were compared. Theoretical accuracies for estimated breeding values were calculated by ten-fold cross-validation on all the scarcely recorded traits. Consistently, the ssGBLUP approach outperformed the ABLUP model, with accuracies across traits ranging from 5.99% and 8.02% above the latter. When ST and MT models were compared, generally large and significant increments of accuracies (up to 19.60%) in all the target traits were observed when records on predictor traits were available for both, the training and validation populations. On the other hand, when records of predictor traits were only available for the training population, the accuracies generally showed a smaller increase (or no increase at all; from 0.00% to 14.77%). The largest increments in accuracy (up to 27.58%) were achieved when genomic information and records on both predictor traits were included in the analysis. We conclude that the inclusion of predictor traits in the training and validation populations coupled with a multiple-trait ssGBLUP model is a promising breeding tool to improve the accuracy of breeding values in trees that have not been phenotyped for wood quality traits.