We have studied that the topological shapes of the subgraphs of MIPS functional categories extracted from the PPI network are arbitrary and the density of them is fairly low. These two unexpected properties of functional categories prohibited other existing approaches from detecting functional modules from PPI networks effectively. A relative excess of emphasis on density and interconnectivity in the existing methods can be preferential for detecting clusters with relatively balanced round shapes and limit performance. The incompleteness of clustering is another distinct drawback of existing algorithms, which produce many clusters with small size and singletons. The preference for strongly connected nodes results in many weakly connected nodes being discarded. Moreover, considering only the topological properties and ignoring the biological characteristics of the network also can damage the effectiveness of clustering.