Inferring microbial interaction networks from abundance data

In large and complex ecosystems, microorganisms such as bacteria, archaea, and other eukaryotic cells coexist. Microbial communities, in fact, represent the world's largest and most diversified ecosystems. Individuals interact in a variety of ways, including predation, mutualism, comensalism, amensalism, and competitiveness. Measuring these interactions in terms of direction and strength on a broad scale is a difficult task that necessitates a combination of data analysis and modeling. Furthermore, the dynamic nature of the abundances of various microbe species cannot be overlooked in order to give a valid theory on microbial interactions.

We start discussing a method based on cooccurrences for later fitting a more sophisticated dynamical model (generalized Lotka-Volterrra) to longitudinal data. The fitted model displays the intrinsic growth rates as well as the interaction network between OTUs. The resulting interaction networks are mostly sparser than the ones derived from randomizations of the data. The dominant interaction types are also non-reciprocal like ammensalism and commensalism. These kinds of interactions have largely gone unnoticed in the modeling literature, but they demand more study for a complete picture of microbial ecosystems. We also find that mutualism is avoided and competition and predation slightly favored. 

https://uibuniversitat.zoom.us/j/97051592159