Figure 1.
Heatmap and hierarchical clustering of kinome microarray profiles from the example experiment.
Samples were taken at four time points from six different subjects, here labeled A-F. The number of the sample from the same subject represents the time point at which the sample was taken; for example, sample C-3 was taken from subject C at time point 3. The distance metric used for clustering was (1 - Pearson correlation), while the linkage method used was average linkage.
Figure 2.
Binary tree representation of the dendrogram shown in Figure 1.
Leaf nodes are shaded in grey and are labeled according to the subject and time point as in Figure 1. Internal nodes are labeled through
, and those internal nodes
for which
is maximized for some group
(where
corresponds to subject A,
corresponds to subject B, and so on; see also Equation 1) are shaded in blue.
Figure 3.
Empirical distribution of random tree scores.
Ten thousand random matrices were created from the matrix used to create the sample dendrogram in Figure 1 by randomly rearranging the peptide intensity values within each sample. For each score
that was given to at least one random tree, the frequency of that score is indicated.
Figure 4.
Heatmap and hierarchical clustering of kinome microarray profiles of samples from the example experiment using 17 peptides chosen according to a local search algorithm.
The same distance metric and linkage method were used as in Figure 1. The sample names are the same as in Figure 1; the peptide names are also indicated on the right side of each row.
Figure 5.
Example of a dendrogram with bootstrap values using PIIKA 2.
The clustering of the samples is the same as in Figure 1. The red numbers represent the approximately unbiased (AU) P-values as determined using the method of Shimodaira [17], [18], while the green numbers represent the standard bootstrap P-value [16]. All calculations and the drawing of the figure were performed using the R package pvclust [19].
Figure 6.
Example of a PCA plot generated in VRML format by PIIKA 2.
In this experiment, samples were taken from subjects labeled A, B, C, D, E, and F. Samples corresponding to subject A are in red, subject B are in yellow, and so on. The label near the top of the figure is the result of hovering the mouse over the leftmost red circle, and shows that the first, second, and third principal components for this sample had the values 2.46, 1.48, and 1.03, respectively. This image is an example of the visualization given using the VRML viewer Instant Player ( http://www.instantreality.org ).
Figure 7.
Example of a volcano plot generated using PIIKA 2.
Points for which FC and P-value
are coloured red, while those with FC
but P-value
are pale red; Similarly, points with FC
and P-value
are green, while those with FC
but P-value
are pale green. All other points are coloured black. The horizontal and vertical blue lines represent the P-value and FC cutoffs, respectively. All coloured points are accompanied by labels showing to which peptide the point corresponds.
Figure 8.
Example of a sample-sample scatterplot generated using PIIKA 2.
Each point represents a peptide, and the and
values of that point represent the normalized intensity values for that peptide for the first sample (A-1) and the second sample (A-2), respectively. The blue line represents the best fit using least squares, whereas the red line simply shows the diagonal (
). The Pearson correlation between the two samples is also indicated.
Figure 9.
Screenshot of the user interface of the PIIKA 2 web server.
Table 1.
Off-the-shelf kinome microarrays that the PIIKA 2 web interface allows the user to select.