Figure 1.
Comparison of mapping results for kernel oil concentration in two different data sets.
(A) Manhattan plots of mixed linear model conducted in data set 1 and 2, respectively (data set 1: n = 368, without imputed genotypic data; data set 2: n = 513, 145 lines with imputed genotypic data). The arrow and red boxes indicate the new loci that were not identified in previous study (Li et al, 2013); (B) Quantile-Quantile plots of p-values of mixed linear model conducted in data sets 1 and 2, respectively.
Table 1.
Additional SNPs and candidate genes significantly associated with oil concentration found using imputed genotype data.
Table 2.
Comparison of significant SNPs identified for 17 traits from MLM using imputed data (dataset 2).
Figure 2.
GWAS of the phenotype of days to heading in Yunnan 2010.
(A) GWAS result by Anderson–Darling test; (B) GWAS result by mixed linear model.
Figure 3.
The nature of statistically significant associations.
The illustration of associations those are highly significant by Anderson–Darling test, with nearly identical trait means for ear height (A) or with an obvious shift of the means for ear leaf width (B).
Figure 4.
Power comparisons in three simulation schemes for four different mapping methods:
A-D, KW, MLM and LM. The “Power” was defined as the detection frequency in 500 repeats for a certain QTN. For the purpose of computing power, a causal SNP was considered to be detected only when the causal SNP was significant at a threshold from 1,000 times permutations. The power and type I error of major QTNs (A–C) and moderate QTNs (D–F) with common allele frequency. The power and type I error of major QTNs (G–I) and moderate QTNs (J–L) with rare allele frequency. A-D test: Anderson–Darling test; LM: linear model; K-W test: Kruskal-Wallis test; MLM: mixed linear model. Scheme 1, phenotypes with normal distribution; Scheme 2, phenotypes with abnormal distribution caused by uncertain effectors; Scheme 3, phenotypes with abnormal distribution caused by a larger effect QTN with rare allele frequency.
Figure 5.
Co-localization of association peaks, QTL and well-annotated candidate genes.
A. Significant association signals on chromosome 7 for kernel width; B. A major kernel width QTL (R2 = 18.7%) was mapped on chromosome 7 from 129 Mb to 149 Mb with BK/Yu8701 RILs and covered the significant association signals; C. The phenotype's frequency distribution histogram and normal distribution curve at the peak SNP of kernel width; D. Significant association signals on chromosome 2 for ear length; E. A major ear length QTL (R2 = 5.7%) was mapped on chromosome 2 in B73/By804 RILs and covered the significant association signals; F. The phenotype's frequency distribution histogram and normal distribution curve at the peak SNP of ear length; G. Significant association signals on chr1 for kernel number per row; H. A major kernel number per row QTL (R2 = 11.78%) was mapped on chr1 in K22/DAN340 RILs and covered the significant association signals; I. The frequency distribution histogram and normal distribution curve at the peak SNP of kernel number per row.