Principal Component Analysis

In a principal component analysis (PCA), the total number of variables in the dataset is reduced to a smaller number of variables – the principle components of the dataset.

Principal component variables are calculated from correlated variables in the total dataset. In other words, the principal component analysis is a workflow used to identify variance in a dataset. The analysis can be run on an entire microarray chip, or on a pathway.

To perform a principal component analysis:

1.  Run tranSMART, then click the Dataset Explorer tab.

2.  Define the cohorts you wish to analyze by dragging one or more concepts from a study into empty subset definition boxes. For more information, see Populating the Study Groups.

Description: C:\Users\bkingsbury\Dropbox\Stuff\Recombinant\note.png

Only one subset may be specified in this analysis. Information in Subset 2 will be ignored.

3.  Click the Advanced Workflow tab:

Description: C:\Users\ctucker\AppData\Local\Temp\SNAGHTML27dca9d6.PNG

4.  Select PCA from the Analysis dropdown menu:

The Variable Selection section appears.

5.  Define the heatmap variable by selecting a high dimensional data node from the Dataset Explorer tree and dragging it into the PCA Variable definition box:

Description: C:\Users\bkingsbury\Dropbox\Stuff\Recombinant\note.png

High dimensional data nodes are indicated by the (  ) icon to the left of study data.

6.  Click the High Dimensional Data button.

The Compare Subsets-Pathway Selection dialog appears.

7.  Specify the platform and other factors of interest.

For more information, see High Dimensional Data.

8.  Click Apply Selections.

9.  Click Run.

Your analysis appears below:

Description: C:\Users\ctucker\AppData\Local\Temp\SNAGHTML52b248b.PNG

Description: C:\Users\bkingsbury\Dropbox\Stuff\Recombinant\note.png

For more information regarding PCAs, see: http://psb.stanford.edu/psb-online/proceedings/psb00/raychaudhuri.pdf.