Predictive protein module based on PPI network and double clustering algorithm

Protein-protein interaction (PPI) is a kind of biomolecular network which plays an important role in biological activities. In order to improve the accuracy of protein function module prediction, obtain protein function module and run timely, this paper proposes predictive protein module based on PPI network and double clustering algorithm (ISCC), which considers the characteristics of PPI network and considers nodes as two-dimensional data points. First of all, improved density-based spatial clustering of applications with noise (IDBSCAN) determines the central cluster, and then uses spectral clustering (SC) to redivide the weight; secondly, CFSFDP and Chameleon algorithm are used to filter the similarity of the central cluster, and use support vector machine (SVM) to get the final clustering result. Finally, the experiments are compared with CDUN, EA, MCL and MCODE in terms of accuracy, sensitivity, F value and the number of protein functional modules. The experimental results show that the F value of ISCCD is 70% higher than that of EA, the number of recognition modules is 257 higher than that of CDUN, and the running time is 494s faster than that of MCODE.