Fuzzy Temporal Association Rule Mining fromDynamic Protein-Protein Interaction Network

One of the basic concepts in the functioning ofcells is the dynamics of PPI networks, the formation, mergingand finding the temporal relationship between proteincomplexes. Converting PPI networks from static form to timegraph form helps to understand their dynamics. To convertthe PPI static graph into a dynamic graph, gene expressiondata is used according to the expression level. Gene expressionsamples were collected at different time points according tovarious conditions such as stress, adding the oxidant or addinggalactose to the cell cultures. Therefore, biological knowledgeextracted from dynamic networks will be affected by thistemporal heterogeneity. In this article, a method forextracting fuzzy temporal association rules of referenceprotein complexes in dynamic PPI networks is presented. Thegeneralized MCODE graph clustering algorithm is applied todynamic networks to handle the volume of calculations and toreduce the number of data items. The identified clusters areconsidered reference protein complexes. Extracted clusters, ateach time point, are fuzzified according to the referenceprotein complexes with different membership degrees. Ruleswill be extracted based on fuzzified reference proteincomplexes. By clustering and cluster fuzzifying, the volume ofitems in each transaction is reduced from the graph size to thenumber of protein complexes. The numbers of extracted rulesfrom YeastNet datasets is, ۹۵۷, with support ۰.۱ andconfidence ۰.۱۵. Also, the evaluation of multidimensionalrules at different times, extracted with the proposedalgorithm, using EBI data emerged informative results.

Keywords:

Inter-transaction rules , Protein-ProteinInteraction Network , fuzzy temporal association rule , graphmining.