Systematizing the generation of missing metabolic knowledge

Abstract

Genome-scale metabolic network reconstructions are built from all of the known metabolic reactions and genes in a target organism. However, since our knowledge of any organism is incomplete, these network reconstructions contain gaps. Reactions may be missing, resulting in dead-ends in pathways, while unknown gene products may catalyze known reactions. New computational methods that analyze data, such as growth phenotypes or gene essentiality, in the context of genome-scale metabolic networks, have been developed to predict these missing reactions or genes likely to fill these knowledge gaps. A growing number of experimental studies are appearing that address these computational predictions, leading to discovery of new metabolic capabilities in the target organism. Gap-filling methods can thus be used to improve metabolic network models while simultaneously leading to discovery of new metabolic gene functions.

Figure 1

Overview of the different algorithms used for predicting gap-filling reactions and orphan-filling genes.

Figure 2

Examples of a gap (A) and an orphan reaction (B) in metabolic networks. Gaps occur when the reaction that consumes or produces a particular metabolite is completely unknown. Orphans occur when a particular reaction is known to occur, but it is not known which gene encodes the catalyzing enzyme. In this figure, metabolites are represented by orange circles and reactions by blue arrows connecting them. Gene-protein-reaction interactions are shown for each reaction.

Figure 3

An example of how a cycle in a metabolic network can lead to a non-obvious gap. Reactions 2–5 form a cycle with reaction 1 as the only input or output. At steady-state, reaction 1 is unable to carry flux without violating mass balance constraints because of this cycle. The four reactions in the cycle are free to carry any flux (as long as all reactions carry the same flux), but metabolite A is blocked even though it has both producing and consuming reactions. Algorithms such as GapFind are useful for identifying these types of gaps.