Heterogeneous Graph Mining for Biological Pattern Discovery in Metabolic Pathways

Alexandra Zaharia (1) , Bernard Labedan (1) , Christine Froidevaux (1) , Alain Denise (1, 2)

1 BioInfo - LRI - Bioinformatique (LRI)
2 I2BC - Institut de Biologie Intégrative de la Cellule

Systems biology studies biological networks and the relations between them. Among the various types of biological networks, we focus on metabolic pathways and gene neighboring networks, respectively modeled by directed and undirected graphs. We attempt to identify maximal sets of consecutive metabolic reactions catalyzed by products of neighboring genes. The approach proposed here is HNet, a non-exhaustive exact method that is capable to take into account (i) skipped genes and/or reactions and (ii) metabolic pathways containing cycles. HNet relies on a previously described graph reduction method and on trail finding in a directed graph by performing path finding in its line graph. A trail is a path that can contain repeated vertices, but not repeated arcs. HNet is used to analyze the genomes and metabolic networks of 50 prokaryotic species in order to gain insight into metabolic pathway evolution.

Domaines: Bio-informatique [q-bio.QM]; Bio-Informatique, Biologie Systémique [q-bio.QM]

Alexandra Zaharia, Bernard Labedan, Christine Froidevaux, Alain Denise. Heterogeneous Graph Mining for Biological Pattern Discovery in Metabolic Pathways. SeqBio 2016, Nov 2016, Nantes, France. ⟨hal-01745390⟩