MSClustering is a Cytoscape tool for multi-level clustering of complex networks and provides immediate visualization and analyses in the Cytoscape platform. After transforming the N^2-distance matrix into an N-list of shortest connections for a system of N nodes, MSClustering groups the system hierarchically for several characteristic levels of resolution, without the input of parameters. It has a system size dependence in the memory of storage and operations of the algorithm. For a system of 2500 nodes, it takes about 36 seconds to perform a hierarchical clustering of the system. Therefore it is efficient (computation time is linear in N for N < 1500) and has been successfully applied to study various systems in biology, scientometrics, literature analyses, and finance. It is now integrated with Cytoscape and provides immediate visualization and statistical analyses.
Clustering/classification is an important step in understanding the present diversity and past evolutionary history of a complex system. As many important real-world clustering/classification problems are intrinsically hierarchical, it is desired to develop an efficient clustering tool for the automated clustering of complex systems at various characteristic levels. We have developed the MSClustering app in Cytoscape for an automated, efficient, and hierarchical clustering of complex networks without any input parameter. Here, we have demonstrated that the constructed MSC tree provides phylogenetic information for complex systems from a distance-based approach. An example is the clustering of a 46-coronavirus network in three characteristic levels as shown in the screenshots. The constructed MSC tree is consistent with the best phylogenetic models from the character-based approach.
Please cite: "MSClustering: A Cytoscape Tool for Multi-Level Clustering of Biological Networks", BK Ge, GM Hu, R Chen, CM Chen, International Journal of Molecular Sciences 23, 14240 (2022).