Exploring Cytoscape Clusters
UCSF clusterExplorer is a Cytoscape plugin that provides a number of tools for exploring clusters created by Cytoscape clustering tools such as clusterMaker. clusterExplorer provides a number of tools to view information about the cluster sizes, inter and intra cluster edge weights, and five comparative methods to delve into specific features of nodes and clusters with respect to other nodes and clusters. clusterExplorer currently has five methods that can be applied to the graph:
- Cluster: central element: determines the "central" node in a cluster. This can be of use to suggest a representative cluster member for further research.
- Cluster: similarity to other clusters: given a node in a cluster, computes the weighted distance between this cluster and all other clusters. This can be useful to explore potential functional relationships between clusters
- Cluster: similarity to other elements: given a node in a cluster, computes the weighted distance between this cluster and all connected nodes not in this cluster. This could be useful to explore potential intermediate proteins.
- Element: similarity to other clusters: looks at the weighted distance between a selected node and any clusters that this node is connected to, excluding the cluster the node is a member of. This can be useful for looking at how close boundary nodes are to other clusters.
- Element: similarity to other elements: looks at the weighted distance between a selected node and any elements that this node is connected to. This can be useful for looking at how well this node fits within a cluster, and how close it is to other clusters.
clusterMaker also provides a prototype tool for comparing clusters to a "gold standard". In the future this tool will be significantly expanded to provide statistics about the cluster "fit" and expanded possible value types to be used for the gold standard attribute.
Figure 1. clusterExplorer in action. In this screenshot the Amidohydrlase superfamily from the Structure-Function Linkage Database has been clustered using the MCL algorithm from clusterMaker. The clusters are depicted as attributes on the nodes and as groups in the Groups panel. clusterExplorer has been used to view the distribution of intercluster and intracluster edge weights. In addition, clusterExplorer was used to determine the nearest cluster to the node representing the protein gi5805954, which is a member of the unkown119 subgroup (the phosphotriesterase-homology proteins).