TopEVM: Using Co-occurrence and Topology Patterns of Enzymes in Metabolic Networks to Construct Phylogenetic Trees

Network-based phylogenetic analysis typically involves representing metabolic networks as graphs and analyzing the characteristics of vertex sets using set theoretic measures. Such approaches, however, fail to take into account the structural characteristics of graphs. In this paper we propose a new pattern recognition technique, TopEVM, to help representing metabolic networks as weighted vectors. We assign weights according to co-occurrence patterns and topology patterns of enzymes, where the former are determined in a manner similar to the Tf-Idf approach used in document clustering, and the latter are determined using the degree centrality of enzymes. By comparing the weighted vectors of organisms, we determine the evolutionary distances and construct the phylogenetic trees. The resulting TopEVM trees are compared to the previous NCE trees with the NCBI Taxonomy trees as reference. It shows that TopEVM can construct trees much closer to the NCBI Taxonomy trees than the previous NCE methods.