The massive development and use of multi omics data permits a better understanding of complex microbial ecosystems at the metabolic level. However, this data and their integration varies depending on the size and characteristics of the considered ecosystems. In this talk, I will discuss of two different approaches applied respectively to a controlled microbial community and natural communities in order to reveal putative bacterium-bacterium interactions. I will first describe how numerical optimisation of metabolic models constrained by genomics and metatranscriptomics data can accurately describe a cheese starter community. I will then detail a discrete method based on only genomic data aiming to characterize the competition and cooperation potentials in natural communities.