IC4R006-GWAS-2015-25689273

From RiceWiki
Jump to: navigation, search

Project Title

Genomic Selection and Association Mapping in Rice (Oryza sativa): Effect of Trait Genetic Architecture, Training Population Composition, Marker Number and Statistical Model on Accuracy of Rice Genomic Selection in Elite, Tropical Rice Breeding Lines

The Background of This Project

  • Genomic Selection (GS) is a new breeding method in which genome-wide markers are used to predict the breeding value of individuals in a breeding population. GS has been shown to improve breeding efficiency in dairy cattle and several crop plant species, and here we evaluate for the first time its efficacy for breeding inbred lines of rice.
  • Genomic selection is a promising breeding technique that aims to improve the efficiency and speed of the breeding process. While it has been shown to be effective in crops such as wheat and corn, it has not yet been applied to rice breeding. Genome-wide association studies (GWAS), by contrast, are used to identify genes or QTLs that underlie traits of importance to breeding such as yield, flowering time, or plant height, and has been performed successfully in rice. Here, we experiment with applying genomic selection in conjunction with GWAS to a rice breeding program at the International Rice Research In- stitute in the Philippines and show that genomic selection can result in more accurate predictions of breeding line performance than pedigree data alone and that GWAS results can inform the results of GS. Our results suggest that GS could be an effective tool for increasing the efficiency of rice breeding.
  • Over the next 30 years, the production of staple cereal grains including wheat, maize, and rice must to be doubled to keep pace with global population and income growth. At the same time, agriculture, in general, is imperiled by human-induced climate change, and plant breeders and farmers together must contend with increased biotic and abiotic stresses that are the direct result of climate unpredictability. Breeding rice varieties adapted to the Asian tropics is already a challenging and resource-intesive endeavor. The number of bacterial, fungal, viral and insect pests for tropical irrigated rice outnumber those for other major cereals. For non irrigated rice, abiotic stresses such as flooding and drought also negatively affect production [1,2,3]. Rice breeders must therefore consider a large number of simple and quantitative traits in combination when developing new lines while, at the same time, maintaining and improving quality and ensuring yield improvements over existing varieties. Using coventional breeding methods, this process is extremely time consuming—on average, it takes up to ten years for elite varieties to be developed and identified.
Figure 1 Population structure of current association panel which consisted mostly of the indica accessions. (A) Scree plot from GAPIT showing the selection of PCs for association study. (B) PCA plot of first two components. (C) Bayesian clustering of 220 rice accessions using STRUCTURE program.

Plant Culture & Treatment

  • 369 elite breeding lines were selected for genotyping from the International Rice Research Institute (IRRI) irrigated rice breeding program based on the planned inclusion of the lines in the 2011 Multi-Environment Testing Program and presence in the 2011 and 2012 Replicated Yield Trials (RYT) at IRRI (Los Baños). Approximately half of the lines were also included in the 2009–2010 RYTs at IRRI. The other lines were promoted from the observational yield trial (OYT) to the RYT in 2011.
  • Phenotypes for the replicated yield trials (RYT) were used for all the experiments and curated from the IRRI database for years 2009–2012, including wet and dry seasons each year. All of the RYT breeding lines, of which our selected 369 lines are a subset, were grown in a randomized complete block design with three replicates in the same field location at IRRI every season and year. The following data were curated for each year, with the exception that plant height data was not available for the 2009 wet season.

Research Findings

  • Narrow-sense heritabilities were calculated on a per line basis for each trait for both validation seasons—the 2012 dry season (2012 DS) and the 2012 wet season (2012 WS) and ranged from 0.31–0.32 for grain yield, 0.30–0.35 for plant height, and 0.32–0.44 for flowering time (Table 1). Heritabilities for all three traits were slightly higher in the dry season than the wet season.
  • For all three traits and both validation seasons, it is clear from the marker subset results that 73,147 markers is more than is necessary to capture the QTL segregating in this population. For almost all traits, there was no significant difference in the best-performing GS method for a given trait or validation season when 7,142 SNPs (approximately 1 SNP for every 0.2 cM) were used versus when 13,101 SNPs (1 SNP for every 0.1 cM) or the full 73,147 SNPs were used. This was true for the randomly chosen SNPs as well as for the evenly distributed SNPs, however the accuracy variances were higher for the randomly chosen SNPs, so it is our recommendation that SNPs be evenly distributed across the genome when possible (Figure. 1). Although it is possible that the variation in the call rates and minor allele frequencies of the randomly selected SNPs also contributed to the larger variations in accuracy in the random SNP subsets, it is still thought that the position of the SNPs was the most important contributor to prediction accuracy.
Table 1 Narrow-sense heritabilities (h 2 ) for the two validation season, 2012 dry season (DS 2012) and the 2012 wet season (WS 2012). YLD = grain yield, FL = days to 50% flowering, PH = plant height.
  • The matrix of genotypes and phenotypes on a breeding population provides the opportunity to perform GWAS in addition to testing any GS models that are available. This paper describes the GS-side of a joint GS-GWAS project on a single rice breeding population, and is the first study to suggest that GWAS on a set of breeding lines might provide information about both the genetic architecture of the traits-of-interest and the population structure of the breeding materials. Specifically, our results on performing GS for grain yield, plant height, and flowering time demonstrate that performing GWAS using the inputs to GS can reveal the presence of large-effect QTL segregating in a breeding population, which can then be modeled accurately using GS.

Labs working on this Project

  • Department of Plant Breeding and Genetics, Cornell University, Ithaca, New York, United States of America
  • International Rice Research Institute, Los Baños, Philippines
  • International Center for Tropical Agriculture, Cali, Colombia
  • Bill & Melinda Gates Foundation, Seattle, Washington, United States of America
  • US Department of Agriculture—Agricultural Research Service (USDA-ARS), Ithaca, New York, United States of America

Corresponding Author

  • Jean-Luc Jannink(jj332@cornell.edu)
  • Susan R. McCouch(srm4@cornell.edu)