Figure 1

The 500 images of 10 fluorescently imaged protein subcellular localisations of Image Set A visualised in iCluster. Each border color represents a different sub-cellular localization. The images are automatically spatially placed in 2D or 3D such that the statistically similar images are close to one another. The spatial placement algorithm only uses the statistics, and is not aware of the subcellular localization categories, these are only used for border coloring. Note the strong clustering of each subcellular localization class, showing that the statistics and algorithm can readily distinguish the localization images. The user may browse, navigate and interact with the image set, show/hide images, show representative images for each class, select subsets of images, detect outliers, reclassify images, and perform tests to give p-values for whether two images classes are different (for instance comparing treated/untreated cells).