With the availability of genome sequences of many organisms and information about gene-protein-reaction (GPR) associations with respect to these organisms genome-scale metabolic networks can be reconstructed. In cellular systems biology these networks are used to model the behavior of metabolism in context of cell growth in terms of fluxes (reaction rates) through reactions in the network. Because the flux through each reaction can generally vary within a range, many flux distributions of the entire network are possible. However, since reactions are connected by common metabolites, reactions that are functionally coherent, are expected to highly correlate in terms of their flux value over different flux distributions. In this paper the genome-scale network of a lactic acid bacterium, named Lactococcus lactis MG1363, is used to generate flux distributions for multiple in silico environmental conditions, mimicking laboratory growth conditions. The flux distributions per condition are used to calculate a correlation matrix for each condition. Subsequently the correlations between the reactions are analyzed in a multivariate approach across the in silico environmental conditions in order to identify correlations that are invariant (i.e. independent of the environment) and correlations that are variant across conditions (i.e. dependent of the environment). The applied multivariate methods are Parallel Factor Analysis (PARAFAC) and Principal Component Analysis (PCA). The discussion of the results of both methods leads to the question whether latent variable models are suitable analyzing this type of data.
- Flux distributions
- Genome-scale network