Figure 3

Functional association mining results by M-Finder. We weighted the yeast PPI network from BioGRID by simICND, and implemented the information flow-based algorithm using 1,000 random proteins in a query. For each output sub-network, we found the best match to functional modules in terms of precision (the ratio of the proteins in the output sub-network, which occur in the functional module). We finally calculated the average precision of all outputs and plotted it by altering the minimum association threshold. The reference functional modules on three different levels of functional hierarchy were obtained from MIPS. The modules with the most general descriptions are located on level-1, whereas those with the most specific functional descriptions are on level-3.