Association mapping in an elite maize breeding population

Abstract

Association mapping (AM) is a powerful approach to dissect the genetic architecture of quantitative traits. The main goal of our study was to empirically compare several statistical methods of AM using data of an elite maize breeding program with respect to QTL detection power and possibility to correct for population stratification. These models were based on the inclusion of cofactors (Model A), cofactors and population effect (Model B), and SNP effects nested within populations (Model C). A total of 930 testcross progenies of an elite maize breeding population were field-evaluated for grain yield and grain moisture in multi-location trials and fingerprinted with 425 SNP markers. For grain yield, population stratification was effectively controlled by Model A. For grain moisture with a high ratio of variance among versus within populations, Model B should be applied in order to avoid potential false positives. Model C revealed large differences among allele substitution effects for trait-associated SNPs across multiple plant breeding populations. This heterogeneous SNP allele substitution effects have a severe impact for genomic selection studies, where SNP effects are often assumed to be independent of the genetic background.

Comparison of biometrical models for joint linkage association mapping

“…Model-A generally detected more QTL than Model-B, which are likely QTL due to exploitation of variance among populations (Figure 3; Liu et al, 2011). This corroborates findings of simulation studies (Yu et al, 2008) reporting that QTL detection power was higher for a model including only cofactors (Model-A) compared with a model, which in addition included a population effect (Model-B).…”

“…The difference between Model-A and Model-B with regard to the explained genotypic variance was much more pronounced for beet yield, sugar content and potassium content than for the other three traits (Table 2). A recent publication indicated that this might be due to higher associations between the phenotype and the population structure for some traits, or by higher R 2 values of the population effect (Liu et al, 2011). None of these parameters, however, explained the differences in this data set (Supplementary Table 3).…”

“…Several statistical approaches for JLAM have been suggested based on linear (Yu et al, 2008;Liu et al, 2011) or linear mixed models (Stich et al, 2008). In addition, the QIPDT approach has been suggested (Stich et al, 2006), but the relationship between parents can better be exploited by using linear mixed models including a kinship matrix.…”

“…A lower QTL detection power was observed for the latter model, indicating that incorporating cofactors might be advantageous as they control both population structure and genetic background within populations. Another conceptually different approach is to model the marker effect as nested within population (Liu et al, 2011). These three approaches based on linear models have been compared in a recent analysis based on experimental data and were shown to result in different QTL being detected (Reif et al, 2010;Liu et al, 2011).…”

“…Another conceptually different approach is to model the marker effect as nested within population (Liu et al, 2011). These three approaches based on linear models have been compared in a recent analysis based on experimental data and were shown to result in different QTL being detected (Reif et al, 2010;Liu et al, 2011). Although linear mixed models have been used to analyze JLAM studies in elite breeding populations (for example, Stich et al, 2008), they have to our knowledge not yet been compared with other biometrical models for JLAM.…”