Overview of the Kiwi workflow. (a) Significant gene-sets are selected based on a user-set cutoff and used as nodes in the visualization network. (b) The shortest path length (SPL) between all node pairs in the gene-set interaction network is calculated. In the example, the SPL between node A and B is 3, and between node C and D is 4. If the SPL between two nodes is below a user-set cutoff (5 in the example), an edge is drawn between those nodes, with a thickness corresponding to the SPL (an SPL = 1 will generate the thickest edge). In the example, the edges between nodes A and B, and C and D, respectively, are marked in red, corresponding to the SPLs shown in the gene-set interaction network. (c) To avoid a cluttered network with too many edges, only the best edges (with lowest SPL) are kept for each node. Note that a node may still have multiple edges of different thickness if, for example, a thinner edge is the best one of a neighbouring node. (This step is optional.) (d) Finally, the visualization network is drawn using a forced-based layout. Nodes are resized according to the gene-set significance and color-coded in order to reflect the general direction of change of the gene-set.