We show that this method is useful for identifying sub-clusters of reactions in the networks by a simple one step numerical computation.
To determine if this modularity of metabolic networks is also reflected at the level of the constituent biochemical reactions, we carried out analyses to detect sub-clusters of reactions in the giant component by graph spectral analysis.
The 2evc plots constructed for the reaction networks of M. tuberculosis and M. leprae comprised numerous plateau regions which represented sub-clusters of reactions in the giant component of these networks (Figures 3A,B).
Metabolome-based reaction networks of Mycobacterium tuberculosis, Mycobacterium leprae and Escherichia coli have been constructed based on the KEGG LIGAND database, followed by graph spectral analysis of the network to identify hubs as well as the sub-clustering of reactions.
We note that there are two disjoint clusters of fast reactions in the model, corresponding to the outer- and inner parts of the cell membrane.
In this study, we tested adjoined matrices with different number of constituting individual flux-pattern matrices for hierarchical clustering in order to confirm that they produce consistent clusters of metabolic reactions.
Depending on the employed clustering method, there could be clusters exclusively composed of reactions showing flat or fluctuating profiles as well as clusters containing reactions of both profiles.
The within-cluster distance of reaction i is defined as d i = ∑ j =1 k (r ij − R cj ).
Under aerobic conditions (in the presence of molecular oxygen), a cluster of physiological reactions prevails that impacts the geochemical status of waters, sediments, soils, and the atmosphere.
This may affect the clustering of biochemical reactions.
