Mining Protein Complexes Based On Connected Affinity Extension

Proteomics is one of the very important research points in the post-genome era. It is common sense that proteins are the material foundation for life activities. The research of protein complex and its functions will facilitate the understanding of kinds of life activity patterns and offer theories and solutions for elucidating and overcoming the mechanisms of many diseases. Therefore, it has been the hot research point at home and abroad to mine protein complex with biological significance and predict the function of unknown proteins, according to the known functional proteins and the topological characteristics analyzed from the protein interaction data.This paper designs a series of protein complex and functional module detection algorithms, based on the topological characteristics of the protein interaction network and combining the properties of the protein complex itself and its formation process. The main work is as follows:Unlike the previous mining algorithms mainly concentrated in the unweighted protein interaction network, which ignored the biological characteristics between proteins, this paper proposes a novel protein complex mining algorithm CACE(Connected Affinity Clique Extension) based on the affinity model and clique extension This new algorithm first constructs the weighted yeast protein interaction network(W-PIN) with affinity coefficient, and then mines protein complex based on the affinity density and clique extension model. Experimental results demonstrate that, CACE is able to mine much more protein complexes with biological significance and greatly elevate the detection accuracy compared with traditional algorithms.Traditional algorithms are implemented from the perspective of the network topology properties directly, but overlook the topology properties of the protein complex itself. However, this paper, employing the seed selection, inner clique extension and outer clique extension methods, presents the protein complex mining algorithm CASE based on the affinity model and seed extension model and applies it on the weighted yeast protein interaction network (W-PIN). The results show that, algorithm CASE makes great improvements on the aspects of recall, accuracy, function enrichment and matching degree. Furthermore, it has a better performance on the execution efficiency. Integrating the protein affinity model, essential protein detection, and the fusion technology of multi-conditions, the new protein mining algorithm KCME is proposed and applied on the weighted yeast protein interaction network (W-PIN). Results indicate that, the algorithm can dig out protein complex of high quality. Additionally, with the homo logical similarity principle, KCME can predict the function of the predicted protein complex, as well as the unknown function of the protein node within the complex.