Figure 1

Schematic overview of the method. For each reaction, features were calculated describing 1) its local topology in the metabolic network, 2) its genomic and transcriptomic relations, and 3) biomass production using flux balance analyses when discarding the reaction. E.g. if a given reaction i had three educts, one product and 4 neighbors, the feature NS (number of substrates), NP (number of products), and NNR (number of neighboring reactions) were set to 3, 1 and 4, respectively. A table for the reaction profiles was created and used for training the classifier to distinguish between essential and non-essential reactions. A Support Vector Machine (SVM) was applied as a classifier to find the optimal separating hyperplane.