Reconstructing Gapless Ancestral Metabolic Networks

We present a method for inferring the structure of ancestral metabolic networks directly from the networks of observed species and their phylogenetic tree. In particular, we aim to minimize the number of mutations on the phylogenetic tree, whilst keeping the ancestral networks structurally feasible, or

gapless

. In gapless metabolic networks all reactions are reachable from external substrates such as nutrients.

To this end, we introduce the gapless minimum mutation problem: finding parsimonious phylogenies of gapless metabolic networks when the topology of the phylogenetic tree is given, but the content of ancestral nodes is unknown. This formulation can be extended also to infer reactions that are missing from the input metabolic networks due to errors in annotation transfer, for example.

The gapless minimum mutation problem is shown to be computationally hard to solve even approximatively. We then propose an efficient dynamic programming based heuristic that combines knowledge on both the metabolic network topology and phylogeny of species. Reconstruction of each ancestral network is guided by the heuristic to minimize the total phylogeny cost. We experiment by reconstructing phylogenies generated under a simple random model and derived from KEGG for a number of fungal species.