基于不确定数据的功能模块预测

Functional modules play an important role in many molecular processes and functions. Many computational methods have been presented to detect protein functional modules from a protein-protein interaction network. However most of the available methods have failed to take into account the intrinsic biological organization and ignore the negative impact on the algorithms by high false positives. In order to resolve this problem, the protein-protein interaction (PPI) network was modeled as an uncertain graph which every protein-protein interaction endowed with a measure. Through the use of the uncertain data management technology, a novel approach based on the possible-world model was developed to detect functional modules. A subgraph was represented as a functional module with high cohesion between its subunits and low coupling between the subgraph and its neighbor subgraphs. The relationship between a node and a subgraph was described by introducing the concept of expected support degree. A comprehensive comparison among the existing seven algorithms and our algorithm was made. Experimental results indicated that this method performs significantly better the state-of-the-art methods and the discovered functional modules are statistically significant.