Typically you want to find marker loci associated with genes controlling the phenotypes for which you have data. For measured traits, QGene uses simple regression to examine all the marker loci in your dataset. The dependent variable is the trait score and the independent one is the number of A-parent alleles at a locus. For categorical traits, contingency analysis is applied in an analogous way.
The contour plot shows for each marker in the dataset the F, RSq, LOD, P value, or other fit statistic from the regression analysis of variance (or chi squared, tau, Wilcoxon rank-sum, or Kruskal-Wallis statistic in case the traits are categorical). This figure shows the contours for eight traits; QGene will allow you to superimpose up to 20 contours color-coded to match their names in the trait list.
What is the meaning of the negative signs on some of the entries, since F is an unsigned statistic? These signs have no statistical meaning. They are put there to indicate the parent whose marker allele at the corresponding locus coincides with a higher mean for the phenotype. At marker Xwg622 in the figure, this is parent Cultivated (genotype aa). "Negative" stats always correspond to a downward peak in the contour.
The histograms at middle left of the dialog show the distribution of the current trait in each of the marker (and missing-marker) classes at the current locus and are updated as you drag the cursor in the plotted contour. Clicking on any of them puts up a line-drawn histogram for exporting. The histograms shown here illustrate what is expected at a locus near a major QTL; the means of the phenotype distributions (shown to right of the histograms) are different and their variances may differ too.
The plot at the middle right is an ordinary trait plot that may be called a uniplot.The Y-axis measures the phenotypic scores for the current master trait PAVY, and the X-axis just holds line numbers. There are as many vertical bars as there are lines that were scored for PAVY, and they're sorted in descending order of the trait. Each bar is given the color keyed to its marker genotype at the current locus, Xwg622. Here it's evident that the lines with highest scores for the trait had mostly the Cultivated marker genotype, another way of seeing the marker association shown by the regression statistics and the histograms.
You may use this dialog to construct and evaluate multiple-regression models for various markers (the Stack contours button reads Multiple regression when only one trait is selected). In a population with both homozygotes and heterozygotes, you can include a dominance term for any marker in the model. If you Command-click on a marker that's in the model, the contour changes to show the two-way interaction F test for this marker with all the other markers in the dataset.
When you view multiple contours and the Multiple regression button turns into the Stack contours button, clicking it brings up a plot like this one, from a different dataset:
QGene has just separated the contours, still plotting them on the scale of the master trait contour (which appears at the top of the window). You may vary this scale via a popup menu.
The single-marker analysis of categorical, in contrast to measured, traits is done with chi square or various nonparametric tests, but QGene's display is quite analogous to that for regression. Here is just the contour plot for three traits that will be familiar to wheat geneticists.
This analysis identifies likely positions for QTLs via a flanking-marker regression method described in Haley and Knott, Heredity 69: 315-324 (1992). At the top of the dialog is the simple-regression contour for the current traits, of which the "master trait" is Rc3_Avg; at the left is the chromosome 7D interval-regression contour for this and several other traits. The contour for the master trait is shown in white, while the other contours are color-coded to match their names in the trait list. As in the Single-point dialog, up to 20 trait contours may be shown at once.
| What QGene does | QGene home |
|---|