In silico analysis of microbial communities through constraint-based metabolic modelling
Microbial communities are involved in many vital biological processes from elemental cycles to sustaining human health. The bacterial assemblages are remarkably under-studied as they are reluctant to grow in the laboratory conditions. Therefore, alternative omics-based approaches and computational modelling methods have been an active area of research to investigate microbial communities physiologically, ecologically and biochemically. In this thesis different microbial consortia involved in food production and also the human gut microbiota have been modelled and investigated. In the case of the human gut microbiota, the effects of malnutrition on the overall health of children from three different countries, namely, Malawi and Bangladesh, and Sweden have been studied. In each of the first two countries, a group of malnourished children going through food therapy as well as a healthy cohort were monitored to investigate the effect of food intervention on malnutrition, with their gut microbiota being the focal point. In this project, using metagenomics data we identified the dominant strains in each cohort, reconstructed genome-scale metabolic models (GEMs) for the most abundant ones and used our models to predict diet-microbe, microbe-microbe, and microbe- host interactions. Based on our results in this project, in addition to being less diverse, the gut microbiota of malnourished children showed a lower potency regarding the production of valuable metabolites. The second investigated microbial consortia were the ones used in fermented milk products. Based on the genome sequence and also experimental data for five selected strains, we reconstructed GEMs, curated the models and performed community modelling to predict their metabolic interactions. Using the simulation outcomes, we could predict a ratio for bacterial strains used in yogurt starter culture to maximise the production of acetaldehyde which is a key contributor to yogurt’s unique taste and aroma.
GEMs are powerful tools to model an organism’s metabolic capabilities, and although numerous GEMs have been reconstructed, their quality control has not gained enough attention. Evaluation of a repository of semi-automatically reconstructed GEMs related to the human gut microbiota and another repository of manually curated ones was performed comparatively. Assessing these models from topological and functional aspects, it was shown that semi-automatically reconstructed models required extensive manual curation before they could be used for target-specific simulations.
In constraint-based modelling, an objective function is usually optimised under particular environmental conditions, however, in case of the microbial communities, there is no distinct and relevant objective function. Therefore, an unbiased uniform randomised sampling algorithm was implemented for microbial communities. The samples acquired from the solution space were analysed statistically to see clustering patterns of the reactions and commensalistic relationships between the community members were identified. Overall, computational modelling paves the way towards gaining a mechanistic understanding of microbial communities and provides us with testable hypotheses and insight.
fermented food production
genome-scale metabolic model
lactic acid bacteria
uniform randomised sampling