Using random walks to identify cancer-associated modules in expression data

Table 3 Comparison of approaches to module-finding in biological networks

Approach	Method	Heuristic for subgraph search	Cluster learning method	Cluster overlap	Tuning parameters	Weighted networks	Platform
Walktrap-BM	Random walk	Differential expression or pairwise similarity	Semi-supervised	No	Modularity, size, score	Yes	R on Unix, Windows, Mac
jActiveModules[20, 21]	Simulated annealing	Differential expression (P-values only)	Semi-supervised	Optional	K-modules, number of paths, iterations	Yes	Cytoscape plugin on Windows, Mac, Linux
Matisse[22]	Seed clustering	Pairwise similarity and significant seed nodes	Semi-supervised	Yes	Module size, seed number	Yes	Linux, Windows
NetWalker[23, 31]	Random walk	Differential expression	Unsupervised	Optional	K-Modules	Yes	Windows, Mac
Affinity Propagation[66]	Seed-based message propagation	Pairwise similarity	Unsupervised	Yes	Preference values, seed number	Yes	Matlab, R on Windows and Linux
MCL[65]	Random walk	Pairwise similarity	Unsupervised	No	Granularity	Yes	UNIX platforms

ISSN: 1756-0381