The Classification Compare Window
Graphical representation of classifications

   The Display Panel can display up to nine different classifications of the data currently loaded into ArrayMiner. Each of the classifications is marked on top by the name of the file it was retrieved from (a red point will designate that this is a file based classification), or the number of ArrayMiner's iteration if it comes directly from the current optimization process. The classifications are automatically ordered from the left to the right according to increasing number of clusters.

     The clusters in the classifications are displayed as boxes marked by the number of genes present in each cluster. The contents of the clusters in the different classifications are compared and clusters are joined together by connections mentioning the number of genes in the connection.

    The "noise" clusters (gaussian model only) are drawn in a crossed light red box and their connections are drawn in a thin light red line if there is only one or two swaps, or with a red point with a number if there are more swaps than two.


   Thus in the example in the left figure, the first classification consists of four clusters containing 38, 13, 36 and 25 genes, respectively. The last cluster is a noise cluster. When the same data were classified into five groups, the Classification Compare shows that

  • the first cluster was preserved in its integrity
  • tthe second cluster lost one gene which joined a new cluster with 10 genes coming from the third cluster
  • the third cluster was split into two parts of 10 and 24 genes, respectively, as well as lost two genes to the noise cluster
  • the fourth cluster gives three genes to a new cluster.

   Similar correspondences are found when comparing the evolution of classification into six clusters.

    In order to know exactly what genes are in each cluster and which ones switch clusters among different classifications it is possible to

  • click on a cluster box and keep the mouse button depressed, the genes inside the cluster are then listed on screen
  • click on a connection between two clusters and keep the mouse button depressed, the genes involved in the connection are then listed on screen
  • click on a cluster box or anywhere in the vertical band of a classification, and the classification becomes selected. It can be subsequently saved, updated by the one currently shown in ArrayMiner's main graphic window, or cleared.