When SPADE is run, events are pulled from a selected population across all files in the Cytobank experiment. The events are pooled together and the algorithm operates on all of them. The end result is a single minimum spanning tree that represents all of the data. For analysis, however, the files are split back out and visualized individually in context of the global tree. Each file will fill the tree in a unique way.
Looking at individual SPADE trees is useful, but presents problems when a quantitative comparison between two trees is desired. Consider the situation depicted below. Two annotated SPADE trees are shown with pSTAT5 intensity as the coloring channel. Differences can be observed between the trees, but requires looking back and forth on a node by node basis. This method doesn't scale and comparisons can be hard to make.
(Comparing basal to IL-7 is scientifically interesting, but hard to do well with separate trees Bendall, Simonds, et al, Science, May 2011)
This two tree problem is solved easily in SPADE by overlaying the trees. Each node, instead of displaying a raw value for intensity on a certain channel, will now display a fold intensity. The fold is calculated on a node-by-node basis.
(The two separate SPADE trees can be visualized as one, with fold change intensity indicated as the color. Bendall, Simonds, et al, Science, May 2011)
(two separate SPADE trees are overlaid to view as a single SPADE tree. Coloring is fold intensity on pSTAT5 Bendall, Simonds, et al, Science, May 2011)
How to setup SPADE with fold change
SPADE requires configuration for fold change before the run is executed, and samples cannot be compared dynamically after the fact. To use different baselines, a different SPADE will have to be run. (note: if you are invested in a particular SPADE tree, with effort, any SPADE tree can be manually changed to show whatever values you like).
To get started with configuring fold change, first create a new SPADE run. On the SPADE setup page, click Setup in the Fold-Change Groups box.
(Click the Setup link to navigate to the fold change setup page)
Clicking the setup link will navigate to the fold change setup page. This is visualized below with annotations:
(The SPADE fold change setup page)
Two important concepts to know when setting up SPADE with fold change are Fold Change Groups and Baselines.
A fold change group is a collection of files that will be compared amongst each other for the fold change calculation. In order to compare files with each other, there needs to be a baseline file. Each file will be compared to this baseline in order to calculate fold change. Multiple baseline files can be selected within one group. In this case the statistics for the files will be averaged before serving as the baseline. Again, fold is calculated on a node-by-node basis.
When using multiple fold change groups, the calculations are encapsulated only within those groups. No calculation is done between groups. Each group will have to have its own baseline to compare between files in that group.
Fold change is simply a calculation and visualization and doesn't affect the structure of the SPADE tree.
How to analyze SPADE with fold change
The principle way to characterize a SPADE tree during any analysis is by coloring the tree by a channel. When this is done, each node will report a colorized version of a statistic taken from the events within that node. Besides choosing a channel to display, a metric needs to be chosen for the data in that channel.
There are five options for metric in SPADE, or three if fold change is not enabled. To view the tree in context of fold change, the metric must be changed to Fold or Raw Fold. At this point the nodes will be colored in overlay mode.
(change the SPADE metric to Fold or Raw Fold to view the tree in context of fold change to baseline)
To learn more about SPADE metrics of Medians, Raw Medians, Fold, Raw Fold, and CVs, read the article on SPADE metrics.