Effect of population stratification
WGAViewer offers an easy graphical way to inspect the possible population stratification effects by comparing the distribution of observed P values with expected distribution through a Q-Q plot. WGAViewer also calculates a lambda value for quantifying the population stratification effects assuming a 2 df chi distribution of the -2log(e)P measures (Clayton et al. 2005) (Weale 2007) . This feature can be especially useful for confirming that methods for controlling stratification have been properly implemented.
To perform this function, click on menu “ Tools->Q-Q plot ”. This will activate a dialog for plotting parameters (Figure 3.7-1). One has the option to plot percentage line and lifting line to help inspect the P value distribution. Figure 3.7-2 shows the plot result. The lower red line denotes the 90 th percentile, while the upper one indicates the point where the P values lift -up line from the expected line. As discussed in (Fellay et al. 2007) , we applied a principal component method (Price et al. 2006) to control for population stratification effects. The P value distribution here therefore takes account of the correction. As from Figure 3.7-2, it is clearly shown that over 90% of the P values distribute in accordance with random expectation, only 323 P value data points lift from the expectation distribution. The lambda value of 1.0053 also indicates the residual population stratification effect (after correction) is minimal.
nother usage of this plot is to easily inspect how the top hit P values depart from the random distribution, in addition to any statistical method used for correcting for multiple testing. As from Figure 3.7-2, the top two data points (rs2395029, HLA-B/HCP5 and rs9264942, HLA-C) can be clearly distinguished from any other points plotted. These two are the two genome-wide significant hits that we reported for setpoint phenotype.
Figure 3.7-1. Parameters for Q-Q plot. (Click to enlarge)
Figure 3.7-2. Q-Q plot. (Click to enlarge)