Submitted to: Crop Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: August 3, 2010
Publication Date: November 15, 2010
Citation: Bradbury, P., Parker, T., Hamblin, M.T., Jannink, J. 2010. Assessment of power and false discovery in genome-wide association studies using the BarleyCAP germplasm. Crop Science. 51:52-59. Interpretive Summary: Genome-wide association mapping (GWAS) has the potential to find important connections between genes and agronomic traits in crop plants. Knowledge of those connections will help plant breeders develop improved varieties more quickly and, as a result, have a significant impact on U.S. agriculture. Unfortunately, because few large scale GWAS studies have been conducted in plants, how those studies should be conducted and what methods work best for analyzing the data, need to be better understood. The Barley Coordinated Agricultural Project (CAP) is assembling a large collection of barley varieties and trait data representing most of the barley breeding programs in the U.S. for the purpose of addressing those questions and finding important barley QTL (quantitative trait loci) that can be used by barley breeders. This study has used data for over 3000 genetic markers from 1800 varieties collected over the first two years of the project to simulate trait data and to evaluate methods of analysis. The results show that mixed-model association analysis can be effective in locating moderate to large effect QTL for a variety of traits across the barley genome as long as a sufficiently large number of lines are analyzed. It also provides guidance to geneticists and plant breeders for designing and analyzing trait data collected by the project.
Technical Abstract: Success in genome-wide association mapping studies (GWAS) is dependent on the power to detect QTL with a minimal rate of false discovery. The objective of this study was to determine the potential for GWAS within Hordeum vulgare L. by evaluating several linear models that varied in the way they accounted for population structure (model-based STRUCTURE or principle component analysis) and familial relatedness. Using genotype data from the Barley Coordinated Agricultural Project (BarleyCAP), phenotypic effects were simulated at 1 to 10 QTL with three trait heritability levels. Under each scenario, power and false discovery rate were calculated for sample sizes of 100 or 300 individuals. A mixed-model that accounted for familial relatedness, but not population structure, performed as well as or better than all other models across all heritability levels, QTL numbers, and sample sizes tested. Simulations with 100 lines performed poorly for QTL detection, but simulations with 300 lines performed adequately, suggesting that the Barley CAP data can be used successfully for GWAS if sample sizes are adequate.