HAMSTER builds a set of minimum spanning trees (MSTs) from the experiments of a microarray data set formatted in GEO SOFT format. Each node in a MST is an experiment and edges between nodes indicate level of dissimilarity through their lengths. This overview looks at two images produced through the tutorial (Available by clicking on Documentation in the left menu.). The MSTs that make up this example are available here.

The data set GDS596 is entitled "Large-scale analysis of the human transcriptome (HG-U133A)". This data set contains 158 experiments that form two complete sets of replicates. The aim of the data set is to examine the expression level of 79 physiologically normal tissues.

We select the first set of replicates and instead of using all 22,283 probes, we focus our attention on the 2,002 probes that are assigned to the Gene Ontology categories:

• Extracellular region (GO:0005576)
• Extracellular region part (GO:0044421)

Applying Euclidean distanceas our dissimilarity measure produces the image below:

Color code: red = cerebral blue = epithelial yellow = immune green = muscle

A coloring scheme has been automatically assigned since all 79 of these experiments have been manually curated beforehand. The text is small, but the overall view is of primary concern (other available options supported by HAMSTER+ include generating an SVG image or a PNG in a higher resolution). The figure shows that the cerebral and immune tissues are closely correlated to each other, while epithelial and muscle are not in this data set.

The most similar experiments in the original data set are recursively merged and the minimum spanning tree algorithm repeated. At the 52nd MST, the following image results:

In this image, most nodes remain colored, indicating that they do not consist of any composite nodes of different colors (manually curated labels). If we expand on the node 51 in the center, we can see that most of the experiments contained in it are immune or cerebral tissue samples:

As a final observation, in both images, note that experiments that are similar to many other experiments are located in the center of the MST and those that are similar with only one experiment are along the outside of the MST.

To determine the steps required to generate this output, select Documentation in the left menu and then click on the Tutorial link.