Principal component analysis plot

This will create a principal component plot as shown in figure 23.45.

Image pca_plot
Figure 23.45: A principal component analysis colored by group.

The plot shows the projection of the samples onto the two-dimensional space spanned by the first and second principal component of the covariance matrix. In the bottom part of the side-panel, the 'Projection/Correlation' part, you can change to show the projection onto the correlation matrix rather than the covariance matrix by choosing 'Correlation scatter plot'. Both plots will show how the samples separate along the two directions between which the samples exhibit the largest amount of variation. For the 'projection scatter plot' this variation is measured in absolute terms, and depends on the units in which you have measured your samples. The correlation scatter plot is a normalized version of the projection scatter plot, which makes it possible to compare principal component analysis between experiments, even when these have not been done using the same units (e.g an experiment that uses 'original' scale data and another one that uses 'log-scale' data).

The plot in figure 23.45 is based on a two-group experiment. The group relationships are indicated by color. We expect the samples from within a group to exhibit less variability when compared, than samples from different groups. Thus samples should cluster according to groups and this is what we see. The PCA plot is thus helpful in identifying outlying samples and samples that have been wrongly assigned to a group.

In the Side Panel to the left, there is a number of options to adjust the view. Under Graph preferences, you can adjust the general properties of the plot.

Below the general preferences, you find the Dot properties:

Note that if you wish to use the same settings next time you open a principal component plot, you need to save the settings of the Side Panel.