Author
ASHRAF, BILAL - Aarhus University | |
EDRISS, VAHID - Cornell University | |
AKDEMIR, DENIZ - Cornell University | |
AUTRIQUE, ENRIQUE - International Maize & Wheat Improvement Center (CIMMYT) | |
BONNETT, DAVID - International Maize & Wheat Improvement Center (CIMMYT) | |
CROSSA, JOSE - International Maize & Wheat Improvement Center (CIMMYT) | |
JANSS, LUC - Aarhus University | |
SINGH, RAVI - International Maize & Wheat Improvement Center (CIMMYT) | |
Jannink, Jean-Luc |
Submitted to: Crop Science
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 11/10/2015 Publication Date: 4/29/2016 Citation: Ashraf, B., Edriss, V., Akdemir, D., Autrique, E., Bonnett, D., Crossa, J., Janss, L., Singh, R., Jannink, J. 2016. Genomic prediction using phenotypes from pedigreed lines with no marker data. Crop Science. 56(3):957-964. Interpretive Summary: Genomic prediction in plant breeding involves using information from DNA markers across the genome to predict the future performance of a breeding line. Genomic prediction thus far has only used information from individuals that have been genotyped. In practice, information from non-genotyped relatives of genotyped individuals can be used to improve the genomic prediction accuracy. The main objective of this study was to use information from non-genotyped lines to improve the genomic prediction accuracy in wheat breeding programs. We compared the performance of models that used only genotypic data, only pedigree data, or both. Data consisted of 1,176 genotyped and 11,131 non-genotyped wheat lines replicated in five management environments at the Centro Internacional de Mejoramiento de Maíz y Trigo (CIMMYT) experiment station in Obregon, Mexico. All wheat lines had pedigree information going back several generations. Prediction accuracies were calculated on the traits plant height, maturity, heading date and grain yield. Results indicate that models using both sources of information outperformed models that used only one source in all cases. In conclusion, the single-step procedure combining pedigree and genomic marker data should be favored where appropriate data is available for genomic prediction in wheat breeding programs. Technical Abstract: Until now genomic prediction in plant breeding has only used information from individuals that have been genotyped. In practice, information from non-genotyped relatives of genotyped individuals can be used to improve the genomic prediction accuracy. Single-step genomic prediction integrates all the marker and pedigree information of genotyped and non-genotyped individuals into a combined relationship matrix to perform genomic prediction. The main objective of this study was to use information from non-genotyped lines to improve the genomic prediction accuracy in wheat breeding programs. We compared the performance of three prediction models (HBLUP, GBLUP and ABLUP) and assessed the impact of different polygenic effect weights on prediction accuracy. Data consisted of 1,176 genotyped (via genotyping-by-sequencing, GBS) and 11,131 non-genotyped wheat lines replicated in five management environments at the Centro Internacional de Mejoramiento de Maíz y Trigo (CIMMYT) experiment station in Obregon, Mexico. Analyses were performed in three scenarios: (i) all lines had pedigree information but only some were genotyped, with phenotypes from one (plant height and maturity) or two (heading date and grain yield) environments in the 2011-2012 season, (ii) from genotyped lines with their pedigree information and their phenotypes in 4-5 environments in the 2012-2013 season was the only information used and (iii) the combination of Scenarios 1 and 2. Three prediction models were used i) GBLUP: only genomic information, ii) ABLUP: only pedigree information and iii) HBLUP: combined genomic and pedigree information. Prediction accuracies were calculated by 5-fold cross validation on the traits plant height, maturity, heading date and grain yield. Results indicate that single-step HBLUP outperformed GBLUP and pedigree-based ABLUP in all cases. In conclusion, the single-step procedure combining pedigree and genomic marker data should be favored where appropriate data is available for genomic prediction in wheat breeding programs. |