Subsections

Viewing the result of k-medoids clustering

The result of the clustering is a Clustering Collection (Image clustering_collection_16_h_p). When opened, this shows a Sankey plot view with one column for each condition chosen in Perform a separate clustering for each and one row for each cluster, see the top part of figure 33.67. A thumbnail in each cluster shows the average expression for the features in that cluster.

Image k-medoids_sankey
Figure 33.67: Sankey plot example. The data set contains four mouse brain tissues and 6 time points, from virgin to postpartum (GEO accession GSE70732). Top: Features in each brain tissue has been divided into the same number of clusters and the flows indicate how the features change the clusters they belong to in the different tissues. Two clusters are selected, as indicated by the red border. Bottom: The line graph of the clusters to be compared shows the feature expression across the time points for the features found in both selected clusters. Here, cluster 3 in Hippocampus and cluster 2 in Neocortex are compared.

The Sankey plot view is linked to a graph view. Clicking on a thumbnail shows the expression of all the features in the cluster in the linked graph view. Clicking two thumbnails while holding down the Ctrl key (Image command_key_web on Mac) shows the features in common to the two selected clusters, see the bottom part of figure 33.67.

Thumbnails and flows have a right-click menu which, for example, allows selecting the corresponding features in other views, such as an expression browser, heat map, or the volcano plot of a statistical comparison track.

If you close one of the two views, you can re-open it by holding down the Ctrl key (Image command_key_web on Mac) and clicking on the icon for the view you wish to re-open.

Side panel settings

Adjustments to the Sankey plot and line graph can be done from the side panel.

It is possible to re-order or remove the columns in the Sankey plot, to remove clusters from a column, to select where flows start (defaults to the first column), and to remove all flows except for those originating in specified clusters of the start column. Furthermore, selected features can be highlighted.

The line graph allows two types of adjustments from the side panel, selection of "Cluster" and "Genes". Only features present in the selected clusters are available in the features pick list and they can be added and removed from the line graph using the tick boxes.

Line graph representation

The line representations are dependent on selections: