Unlocking diversity in germplasm collections by genomic selection: a case study based on quantitative adult plant resistance to stripe rust (Puccinia striiformis f. sp. tritici) in spring wheat

Authors: MULETA, KEBEDE - Washington State University; BULLI, PETER - Washington State University; ZHANG, ZHIWU - Washington State University; Chen, Xianming; PUMPHREY, MICHAEL - Washington State University

Submitted to: The Plant Genome
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/9/2017
Publication Date: 8/17/2017
Citation: Muleta, K.T., Bulli, P., Zhang, Z., Chen, X., Pumphrey, M. 2017. Unlocking diversity in germplasm collections by genomic selection: a case study based on quantitative adult plant resistance to stripe rust (Puccinia striiformis f. sp. tritici) in spring wheat. The Plant Genome. https://doi:103835/plantgenome2016.12.0124.
DOI: https://doi.org/10.3835/plantgenome2016.12.0124

Interpretive Summary: Harnessing diversity from germplasm collections is more feasible than ever due to the development of lower-cost and higher-throughput genotyping methods, but the cost of phenotyping is still generally high, so efficient methods to sample and exploit useful diversity are needed. Genomic selection (GS) has the potential to enhance the use of desirable genetic variation in germplasm collections, through predicting genomic estimated breeding values (GEBVs) for as many traits as measured. In this study, we evaluate various population genetic properties and marker densities on the accuracy of GEBVs for applying GS for wheat germplasm utilization. Stripe rust resistance data of 1,163 globally sourced spring wheat accessions that were genotyped with the wheat 9K single nucleotide polymorphism (SNP) assay were used for conducting various genomic prediction tests. We found that prediction accuracy increased with increase in training population size and marker density. All 5,619 SNP markers detected in the assessions were not necessary to capture the trait variation in the germplasm collection, with no further gain in prediction accuracy beyond 1,850 markers (1 SNP per 3.2 cM), which is close to the linkage disequilibrium decay rate in this population. Collectively, our results suggest that larger germplasm collections may be efficiently sampled based on lower-density genotyping methods, while genetic relationships between the training and validation populations is critical when exploiting GS to select from germplasm collections.

Technical Abstract: Harnessing diversity from germplasm collections is more feasible than ever due to the development of lower-cost and higher-throughput genotyping methods. At the same time, the cost of phenotyping is still generally high, so efficient methods to sample and exploit useful diversity are needed. Genomic selection (GS) has the potential to enhance the utilization of desirable genetic variation in germplasm collections, through prediction of genomic estimated breeding values (GEBVs) for as many traits as have been measured. Here, we evaluate various scenarios of population genetic properties and marker density on the accuracy of GEBVs in the context of applying GS for wheat germplasm utilization. Empirical data for adult plant resistance to stripe rust in 1,163 globally sourced spring wheat accessions that were genotyped with the wheat 9K single nucleotide polymorphism (SNP) iSelect assay were used for conducting various genomic prediction tests. Not surprisingly, the results of the cross-validation tests demonstrated that prediction accuracy increased with increase in training population size and marker density. The ridge regression-Best Linear Unbiased Prediction (RR-BLUP) method of GS revealed that all available markers (5,619) were not necessary to capture the trait variation in the germplasm collection, with no further gain in prediction accuracy beyond 1 SNP per 3.2 cM (~1,850 markers), which is close to the linkage disequilibrium decay rate in this population. Collectively, our results suggest that larger germplasm collections may be efficiently sampled based on lower-density genotyping methods, while genetic relationships between the training and validation populations remains critical when exploiting GS to select from germplasm collections.