COMBat: Visualizing co-occurrence of annotation terms

Remko B. J. van Brakel, Mark W. E. J. Fiers, Christof Francke, Michel A. Westenberg, Huub van de Wetering
2013 2013 IEEE Symposium on Biological Data Visualization (BioVis)  
Figure 1 : COMBat applied to a dataset with 268 genes-annotated with in total 634 gene ontology (GO) identifiers [1]-on both the rows and the columns. Three property matrices are shown, in which each cell contains the set of GO identifiers that the corresponding genes have in common. Each cell is colored by the number of elements in the set as given by the colormap (see legend in the lower-left corner). The cell is gray if no data is associated with it. In the top-left corner, the initial,
more » ... rted property matrix is shown. The image just below it shows the same data, but now sorted using a similarity sort. Typically, the resulting blocks of constant color correspond to genes with the same set of GO identifiers. At the right, more details are visible after zooming in on a region somewhat to the right and below the center of the matrix; in the top-left corner of the property matrix, a minimap provides information about the location and size of this zoomed region relative to the whole matrix. The cell corresponding to gene G71764 and G56281 is highlighted, and the GO identifiers that they have in common are shown in a separate inset. ABSTRACT We propose a visual analysis approach that employs a matrixbased visualization technique to explore relations between annotation terms in biological data sets. Our flexible framework provides various ways to form combinations of data elements, which results in a co-occurrence matrix. Each cell in this matrix stores a list of items associated with the combination of the corresponding row and column element. By re-arranging the rows and columns of this matrix, and color-coding the cell contents, patterns become visible. Our prototype tool COMBat allows users to construct a new matrix on the fly by selecting subsets of items of interest, or filtering out uninteresting ones, and it provides various additional * interaction techniques. We illustrate our approach with a few case studies concerning the identification of functional links between the presence of particular genes or genomic sequences and particular cellular processes.
doi:10.1109/biovis.2013.6664342 dblp:conf/biovis/BrakelFFWW13 fatcat:3mijirvjebawlfm4rdwmh3c36a