Research On Organelle Communication Patterns Based On Frequent Subgraph Mining

Organelles are subcellular structures with relatively independent function andstructure in living cells. Different organelles collaborate together with differentfunctions to finish the complex cellular activities. With the continuousdevelopment of the biological experimental techniques, lots of evidences havebeen found in the communication of different organelles. People desperately wantto reveal whether their communications are under some rules. However, traditionalexperiments can only reflect communication relationship in terms of metabolic orsignal transduction pathways, which can seldom reveal the rule of communicationsystematically. With the development of system biology and network biology,complex network has become a important tool to describe the intermolecularrelationship. Frequent subgraph mining can effectively reveal the communicationpattern through finding the building blocks in the complex network. In addition,with the application of gene microarray and next generation sequencingtechnology, the gene expression data grows explosively. We can deeply recognizethe relationship between genes and organelles with the increasing annotation onfunction and location of genes, and the development of gene ontology, whichmakes it possible for us to study the communication patterns of organelles at thegene level.In this paper, we proposed a new method to study organelle communicationpatterns based on frequent subgraph mining method and gene expression data.First, we constructed an "organelle-gene" interaction network, which can reflectthe communication relationship among organelles, based on gene co-expressionnetwork. In the "organelle-gene" interaction network, an organelle is represent bya set of genes, which construct a function module in the gene co-expressionnetwork. Second, we proposed two kinds of organelle communicationpatterns(broadcast pattern and spreading pattern). To verify our hypothesis, wedesigned an efficient frequent subgraph mining algorithm to find the frequentcommunication patterns from the "organelle-gene" interaction network constructedin last step. Our algorithm used two strategies to decrease the computation spaceof testing whether two graph are isomorphic, so that it improved the systemefficiency. We constructed four "organelle-gene" interaction networks using geneexpression data of arabidopsis under different environmental stresses. And wediscovered the potential organelle communication patterns by mining the commonhighly frequent subgraphs in four networks. At last, we compared our algorithmwith Kavosh. The result shows that our algorithm performs better on the input "organelle-gene" interaction network.