Molecular Codes in Large Metabolic Networks

We provide an approach for identifying molecular codes in large reaction networks. The method exploits particular algebraic properties of closed sets of species forming an algebraic lattice. In a first step the network is reduced by unconnected subnetwork removal and pair merging both preserving molecular code properties. Then connected and closed sub-networks are sampled, each being subsequently analyzed for molecular codes separately by a deterministic algorithm improved by memoization. Thus a parallel computing environment can be easily exploited. We apply our method to a large-scale metabolic network model of Helicobacter pylori encompassing 485 species and 554 reactions. 421 unique molecular codes have been found. The vast majority of these codes contain at least one ubiquitous species like protons or water. Filtering for molecular codes without these species in key positions like signs, meaning, or intermediate species, reduces the number of identified codes to 22 only. Whether these codes are utilized by the cell in processing "meaningful" information is yet unknown. All presented data and source code of the new algorithms is available for download.(1)