Protein Complex Detection Based On Protein-protein Interaction Networks And Multi-objective Evolutionary

The proteins,the basis of life activities,seldomly participate in a life process alone.Actually,they interact together in a complex of interacting proteins.The recognition of protein complexes is conducive to understanding the internal organizational structure of cells and assisting medical research to analyze the principle of diseases.Therefore,how to mine protein complexes accurately and efficiently is one of the hot topics in bioinformatics research at present.Although some achievements have been made in the detection of protein complexes today,the recognition accuracy needs to be improved due to the complexity of Protein-Protein Interaction network(PPIN).Therefore,we hope to propose an effective algorithm to improve the accuracy as much as possible.Based on the static and dynamic PPI networks,We detect protein complexes in this thesis.Based on static PPI network,we first integrate Kernel K-means(KKM),spectral clustering Ratio Cut(RC)and overall complex density as objective functions,and propose a fuzzy fusion density method and multi-objective evolutionary protein complex detection algorithm CWFC.Firstly,we construct a weighted protein-protein interaction network.Then we select the candidate protein key nodes.The similarity of the distance between non-candidate key protein nodes and each candidate key protein is compared to determine the non-overlapping protein complex.Finally,the fuzzy algorithm is used to determine the overlapping complex.To some extent,the experimental results of F-Measure,Acc and Sn indexes show that our algorithm is effective in the Collins dataset and Gavin dataset.The life process is dynamic,actually,while the static PPI network cannot reflect its dynamics.Therefore,based on the static CWFC algorithm,we combine the gene expression data and then propose an algorithm Dy CWFC,which can be applied to the dynamic PPI network.The weighted dynamic PPI sub-network is constructed by gene expression data,protein interactions and Pearson correlation coefficient..Using yeast data in the DIP dataset,our algorithm has advantages over analyzing common statistical indicators,e.g F-Measure.