Genomic-assisted sparse multi-location testing to increase genetic gains in barley


Views
0% 0
Downloads
0 0%
CC-BY-SA-4.0

Thumbnail Image

Date

2022-09-19

Date Issued

2022-09-19

Contributes to SDGs

SDG 1 - No povertySDG 2 - Zero hungerSDG 3 - Good health and well-beingSDG 13 - Climate actionSDG 15 - Life on land

Citation

Julian Garcia Maldonado. (19/9/2022). Genomic-assisted sparse multi-location testing to increase genetic gains in barley. Beirut, Lebanon: International Center for Agricultural Research in the Dry Areas (ICARDA).
In recent years, breeders have been incorporating genomic predictions (GP) into their breeding programs as a useful, cost-effective, and versatile tool to increase genetic gain. Coupled with new experimental designs, genomic prediction maximizes the information quality and quantity per dollar invested. This is the case of genomic-assisted sparse multi-location testing, an experimental design that allows breeders to arrange multi-location trials having only a fraction of the genotypes present in all environments. In this design, the observed genotypes at each environment are used to produce genomic predictions of the non-planted ones and the overlapping genotypes across locations enable the genotype x environment (GxE) connectivity and its modeling. This results in an increase in selection accuracy and/or intensity and thereby higher genetic gain. This strategy has been adopted by the Global Barley Breeding Program of ICARDA as the standard approach for preliminary yield trials. To assess GP accuracies within and across locations, 1,000 new 2- and 6-row stage 1 entries of the Feed Barley for Arid and Semi-Arid Environments Mega-Product Line were assembled in a preliminary yield trial series with four locations. These lines were distributed in 340 plot sparse p-rep trials that included 212 unreplicated entries per location, 18 entries replicated within location (p-rep entries), 80 entries replicated across locations and 6 commercial checks replicated both within and across locations. These trials were planted in four diverse locations in Morocco and Lebanon, each identified as representative of a Target Population Environment. The phenotypic correlation among environments were all lower than r = 0.30. In order to maximize the kinship connectivity, the lines were distributed across locations based on the pedigree-based matrix of relationships. In addition, all 1,000 lines were genotyped using a marker diversity set of 96 SNP distributed across the genome and previously selected for their high minor allele frequency (MAF > 0.40) among the parents. Genomic predictions were calculated using GBLUP fitted using the ASReml-R, following a two-stages analysis. A scheme of 10-fold cross-validation was set up to calculate GP accuracies for the multi-environment trials (MET). Different variables were added to the genomic models to improve the accuracy of the predictions. These included, population structure, pedigree information (hybrid matrix) and row-type. These were added both independently or combined into the models. Best results were obtained when only the hybrid matrix was included in the models. Final GxE genomic prediction accuracies ranged between 0.11 to 0.40.