Table 1.
Genome-wide association study results for quantitative traits that are dichotomized at the 50th percentile.
Number of statistically significant marker-trait associations and genomic control values from the unified mixed linear model and the logistic mixed model are presented.
Fig 1.
Results for two quantitative traits dichotomized at the 50th percentile.
Manhattan plots summarizing the genome-wide association study (GWAS) results for two quantitative traits dichotomized at the 50th percentile. Each trait was analyzed using both the unified mixed linear model (MLM) and the logistic mixed model (LMM). Quantile quantile (QQ)-plots depicting the observed (Y-axis) and expected (X-axis) −log(10) P-values are inserted into each Manhattan plot. The Manhattan plots on each graph shows the physical bp position of each tested SNP in either the maize B73_RefGen v2 reference genome (for A and B) or the sorghum Btx623 v2.1 reference genome (for C and D) on the X-axis; while the −log(10) P-values from either the unified MLM (A and C) or LMM (B and D) on the Y-axis. The vertical line on each graph indicates the approximate location of the peak marker-trait associations identified in the previously-published GWAS of the original quantitative trait. (A) Results for a GWAS on dichotomized α-tocopherol measured in maize grain using the unified MLM. (B) Results for a GWAS on dichotomized α-tocopherol measured in maize grain using the LMM. (C) Results for a GWAS on dichotomized sorghum plant height using the unified MLM. (D) Results for a GWAS on dichotomized sorghum plant height using the LMM.
Table 2.
Genome-wide association study results for quantitative traits that are dichotomized using the 75th percentile.
Number of statistically significant marker-trait associations and genomic control values from the unified mixed linear model and the logistic mixed model are presented.
Fig 2.
Results for two quantitative traits dichotomized at the 75th percentile.
Manhattan plots summarizing the genome-wide association study (GWAS) results for two quantitative traits dichotomized at the 75th percentile. Each trait was analyzed using both the unified mixed linear model (MLM) and the logistic mixed model (LMM). Quantile quantile (QQ)-plots depicting the observed (Y-axis) and expected (X-axis) −log(10) P-values are inserted into each Manhattan plot. The Manhattan plots on each graph shows the physical bp position of each tested SNP in either the maize B73_RefGen v2 reference genome (for A and B) or the sorghum Btx623 v2.1 reference genome (for C and D) on the X-axis; while the −log(10) P-values from either the unified MLM (A and C) or LMM (B and D) on the Y-axis. The vertical line on each graph indicates the approximate location of the peak marker-trait associations identified in the previously-published GWAS of the original quantitative trait. (A) Results for a GWAS on dichotomized α-tocopherol measured in maize grain using the unified MLM. (B) Results for a GWAS on dichotomized α-tocopherol measured in maize grain using the LMM. (C) Results for a GWAS on dichotomized sorghum plant height using the unified MLM. (D) Results for a GWAS on dichotomized sorghum plant height using the LMM.
Fig 3.
Distribution of -log10(P-values) for dichotomized maize ear height at the 75th percentile.
Quantile-quantile (QQ) plots showing the distribution of −log(10) P-values of 262,191 single nucleotide polymorphisms (SNPs) tested for association with dichotomized maize ear height at the 75th percentile. On each plot the observed −log(10) P-values from the unified mixed linear model (MLM; blue squares) and logistic mixed model (LMM; purple triangles) are plotted against the expected −log(10) P-values. The value of lambda for genomic control for the unified MLM and LMM are presented in the legend of each plot. (A) QQ-plot for all 262,191 SNPs that are non-monomorphic within the tropical and non-tropical subpopulations of the Goodman diversity panel. (B) QQ-plot of the SNPs that were more common in the non-tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was less than 0.80). (C) QQ-plot of SNPs that tended to have similar allele frequencies in both subpopulations (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical lines were between 0.80 and 1.25). (D) QQ-plot of SNPs that were more common in the tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was greater than 1.25).
Table 3.
Summary of observed values of maize ear height dichotomized using the 75th percentile among the tropical and non-tropical subpopulations of 278 lines from Goodman maize diversity panel.
Table 4.
Summary the binary trait simulated on 278 lines from Goodman maize diversity panel where the probability of observing “1” differed in the tropical and subtropical subpopulations.
Table 5.
Summary the binary trait simulated on 278 lines from Goodman maize diversity panel where the probability of observing “1” was the same in the tropical and subtropical subpopulations.
Fig 4.
Distribution of -log10(P-values) for binary trait simulated in maize with unequal proportion of “1’s” in subpopulations.
Quantile-quantile (QQ) plots showing the distribution of −log(10) P-values of 262,191 single nucleotide polymorphisms (SNPs) tested for association with binary trait where the probability of observing “1” differed between the tropical and non-tropical subpopulations. On each plot the observed −log(10) P-values from the unified mixed linear model (MLM; blue squares) and logistic mixed model (LMM; purple triangles) are plotted against the expected −log(10) P-values. The value of lambda for genomic control for the unified MLM and LMM are presented in the legend of each plot. (A) QQ-plot for all 262,191 SNPs that are non-monomorphic within the tropical and non-tropical subpopulations of the Goodman diversity panel. (B) QQ-plot of the SNPs that were more common in the non-tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was less than 0.80). (C) QQ-plot of SNPs that tended to have similar allele frequencies in both subpopulations (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical lines were between 0.80 and 1.25). (D) QQ-plot of SNPs that were more common in the tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was greater than 1.25).
Fig 5.
Distribution of -log10(P-values) for binary trait simulated in maize with equal proportion of “1’s” in subpopulations.
Quantile-quantile (QQ) plots showing the distribution of −log(10) P-values of 262,191 single nucleotide polymorphisms (SNPs) tested for association with binary trait where the probability of observing “1” was the same between the tropical and non-tropical subpopulations. On each plot the observed −log(10) P-values from the unified mixed linear model (MLM; blue squares) and logistic mixed model (LMM; purple triangles) are plotted against the expected −log(10) P-values. The value of lambda for genomic control for the unified MLM and LMM are presented in the legend of each plot. (A) QQ-plot for all 262,191 SNPs that are non-monomorphic within the tropical and non-tropical subpopulations of the Goodman diversity panel. (B) QQ-plot of the SNPs that were more common in the non-tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was less than 0.80). (C) QQ-plot of SNPs that tended to have similar allele frequencies in both subpopulations (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical lines were between 0.80 and 1.25). (D) QQ-plot of SNPs that were more common in the tropical subpopulation (i.e., the SNPs where the ratio of expected variance between tropical and non-tropical subpopulations was greater than 1.25).