Figure 1.
DyCoNet architecture and example analysis.
a) Diagram depicting the analysis work-flow and plugin execution. b) Sub-community model applied on the TC-PIN from [14]. Top-left: An overview of the network, consisting of 3901 nodes and 16891 edges. Top-right to bottom-right: Snapshots of the network at time points 1–5. For clarity only the k-core of the network with is shown. Node colours correspond to the community membership of each node. Layout of the network is done using the ForceAtlas2 algorithm. Parts of the network that change between time-points are indicated in circles to aid visualisation.
Figure 2.
Communities found in the Southern Women dataset with Jaccard index group matching.
For a detailed explanation of the output, please see text.
Figure 3.
Community structure in the first 6 time steps on TC-PIN network from [14].
a) Identified communities using the social cost model, b) Assignment of proteins to the communities to which their groups are associated at each time point (i.e. no cost model), c) Identifying sub-communities of nodes with the same dynamic pattern. Colour and number of the first column indicate the sub-community that the protein has been assigned to, the second column the original community membership is shown. The number at the bottom of each column shows the number of members of the corresponding sub-community, or “-” if no sub-community has been assigned.
Figure 4.
Sample sub-community for all 6 time steps.
a-f) Cyan coloured nodes correspond to the sub-community discovered from the set of nodes with the same community change pattern. Rest of coloured nodes are the ones belonging in the original communities, with the original colour shown. g) A sample protein belonging to the depicted community (left column) and its corresponding community pattern change (right column).
Figure 5.
TC-PIN for all 6 time steps from [14].
Node colours correspond to sub-communities identified by DyCoNet, node size corresponds to node promiscuity i.e. frequency with which nodes change communities. This visualisation is produced in Gephi by setting the ranking feature of each node such that ‘Node Promiscuity’ corresponds to node size.