Effect of population structure corrections on the results of association mapping tests in complex maize diversity panels

Abstract

Association mapping of sequence polymorphisms underlying the phenotypic variability of quantitative agronomical traits is now a widely used method in plant genetics. However, due to the common presence of a complex genetic structure within the plant diversity panels, spurious associations are expected to be highly frequent. Several methods have thus been suggested to control for panel structure. They mainly rely on ad hoc criteria for selecting the number of ancestral groups; which is often not evident for the complex panels that are commonly used in maize. It was thus necessary to evaluate the effect of the selected structure models on the association mapping results. A real maize data set (342 maize inbred lines and 12,000 SNPs) was used for this study. The panel structure was estimated using both Bayesian and dimensional reduction methods, considering an increasing number of ancestral groups. Effect on association tests depends in particular on the number of ancestral groups and on the trait analyzed. The results also show that using a high number of ancestral groups leads to an over-corrected model in which all causal loci vanish. Finally the results of all models tested were combined in a meta-analysis approach. In this way, robust associations were highlighted for each analyzed trait.

Fig. 1

Log-likelihood of the STRUCTURE software outputs as a function of the number of degrees of freedom absorbed by the structure model (filled diamond), and the mean of the log-likelihood of 20 STRUCTURE independent runs as a function of the number of degrees of freedom absorbed by the structure model (solid line). Outlier STRUCTURE runs with very low Log-likelihood values (until −2,356,000) induced a high mean decrease at 4 degrees of freedom (5 groups) and at more than 16 degrees of freedom (17 groups)

Fig. 2

Values of the delta(K) criterion as a function of the number of degrees of freedom absorbed by the structure model

Fig. 3

Neighbor joining tree of the STRUCTURE software outputs. The pairwise Euclidian distances were computed from the predicted allele frequencies from among the 381 STRUCTURE software outputs (no genetic structure model and 20 runs of a number of ancestral groups from 2 to 20). The surrounded part of the tree was zoomed; it corresponds to the outputs that are grouped in accordance with the number of ancestral groups (K)

Fig. 4

BIC values of a male flowering time (MFT) and b thousand-kernel weight (TKW) as a function of the number of degrees of freedom absorbed by the structure model. The blue and red lines represent the most likely (Q ₁) and second most likely (Q ₂) STRUCTURE software models, respectively; the green line represents PCA (P) models and the light-blue line represents the MCA (M) models

Fig. 5

Number of SNPs significantly associated with a male flowering time (MFT) and b thousand-kernel weight (TKW) as a function of the number of degrees of freedom absorbed by the structure model. The blue and red lines represent the most likely (Q ₁) and second most likely (Q ₂) STRUCTURE software models, respectively; the green line represents PCA (P) models and the light-blue line represents the MCA (M) models. The surrounded parts correspond to a zoom of the curves

Fig. 6

Total number of significant SNPs over the four structure models tested (solid line curve) and percentage of SNPs common to all of them (dotted line curve) as a function of the number of degrees of freedom absorbed by the structure model