Genomic prediction in bi-parental tropical maize populations in water-stressed and well-watered environments using low density and GBS SNPs

Authors: ZHANG, XUECAI - International Maize & Wheat Improvement Center (CIMMYT); PEREZ-RODRIQUEZ, PAULINO - Colegio De Postgraduados; KASSA, SEMAGN - International Maize & Wheat Improvement Center (CIMMYT); BEYENE, YOSEPH - International Maize & Wheat Improvement Center (CIMMYT); BABU, RAMAN - International Maize & Wheat Improvement Center (CIMMYT); LOPEZ CRUZ, MARCO - International Maize & Wheat Improvement Center (CIMMYT); SAN VICENTE, FELIX - International Maize & Wheat Improvement Center (CIMMYT); OLSEN, MICHAEL - International Maize & Wheat Improvement Center (CIMMYT); Buckler, Edward - Ed; Jannink, Jean-Luc; PRASANNA, BODDUPALLI - International Maize & Wheat Improvement Center (CIMMYT); CROSSA, JOSE - International Maize & Wheat Improvement Center (CIMMYT)

Submitted to: Heredity
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/17/2014
Publication Date: 11/19/2014
Publication URL: http://www.nature.com/hdy/journal/v114/n3/full/hdy201499a.html
Citation: Zhang, X., Perez-Rodriquez, P., Kassa, S., Beyene, Y., Babu, R., Lopez Cruz, M., San Vicente, F., Olsen, M., Buckler IV, E.S., Jannink, J., Prasanna, B.M., Crossa, J. 2014. Genomic prediction in bi-parental tropical maize populations in water-stressed and well-watered environments using low density and GBS SNPs. Heredity. 114:291-299.

Interpretive Summary: One of the most important applications of genomic selection in maize breeding is to identify the best-untested individuals when the training and validation sets are derived from the same cross. We assessed prediction accuracy in nineteen tropical maize bi-parental populations evaluated in multi-environment trials. We used either low density markers (~200 SNPs) or high density genotyping-by-sequencing (GBS). An extension of the Genomic Best Linear Unbiased Predictor (GBLUP) that incorporates genotype x environment (GE) interaction was used to predict genotypic values. We showed that: (1) low density markers (~200 SNPs) were largely sufficient to get good prediction in bi-parental maize populations for simple traits with moderate-to-high heritability, but GBS outperformed low density SNPs for complex traits and for simple traits evaluated under stress conditions with low-to-moderate heritability. (2) Heritability and genetic architecture of target traits affected genomic prediction performance: prediction accuracy of grain yield, a complex trait, was consistently lower than that of simple traits anthesis date and plant height. Prediction accuracy under stress conditions was consistently lower and more variable than under well-watered conditions due to poor heritability under stress conditions. (3) Models accounting for GE were clearly superior to models that did not for complex traits but only marginally for simple traits.

Technical Abstract: One of the most important applications of genomic selection in maize breeding is to predict and identify the best-untested individuals from bi-parental populations, when the training and validation sets are derived from the same cross. Nineteen tropical maize bi-parental populations evaluated in multi-environment trials were used in this study to assess genomic prediction accuracy of different quantitative traits using low density markers (~200 SNPs) and high density genotyping-by-sequencing (GBS), respectively. An extension of the Genomic Best Linear Unbiased Predictor (GBLUP) that incorporates genotype Å~ environment (GE) interaction was used to predict genotypic values; cross-validation methods were applied to quantify prediction accuracy. Our results showed that: (1) low density markers (~200 SNPs) were largely sufficient to get good prediction in bi-parental maize populations for simple traits with moderate-to-high heritability, but GBS outperformed low density SNPs for complex traits and simple traits evaluated under stress conditions with low-to-moderate heritability; (2) heritability and genetic architecture of target traits affected genomic prediction performance, prediction accuracy of complex traits (grain yield, GY) were consistently lower than those of simple traits (anthesis date, AD and plant height, PH), and prediction accuracy under stress conditions was consistently lower and more variable than under well-watered conditions for all the target traits due to their poor heritability under stress conditions; and (3) superiority of GE models over that of non GE models in prediction accuracy was found for complex traits and marginal for simple traits.