N2SIMBA: from Network topology to SIMulation of interactions and BActerial abundance, using microbial consumer resource model

Microorganisms often coexist and establish complex interaction patterns within their environment. Recent advances in efficient and cost-effective 16S rDNA sequencing have greatly improved our ability to characterize microbial communities and to infer networks of interactions among their members. However, validating microbial interaction inference methods remains challenging, as a true and experimentally accessible reference for microbial interactions is generally unavailable, expecially for large networks. For this reason, in silico approaches are essential to generate realistic synthetic data with known interaction structures that can serve as gold standards. Here, we introduce N2SIMBA, a modular simulation framework for bacterial communities that takes as input weighted and directed interaction networks, interpreted as known ground-truth microbial interactions, and generates realistic 16S rDNA sequencing count tables as output. N2SIMBA models the dynamics of the microbial community using a consumer–resource framework, in which microbial interactions are mediated by metabolites. Each edge in the input network encodes metabolite-mediated interactions in which microbial species affect one another through the consumption and transformation of shared resources, giving rise to both cooperative and competitive effects. These processes are governed by parameters that describe consumer preferences and metabolic transformation rules, allowing the simulation of biologically plausible interaction mechanisms. Finally, to produce synthetic sequencing data, N2SIMBA simulates the experimental sequencing process by introducing sampling variability and compositional effects, thus generating count tables that resemble real 16S rDNA sequencing datasets. We demonstrate that N2SIMBA enables the generation of realistic in silico microbial count tables with known interaction structures, making it possible to systematically evaluate and benchmark microbial network inference methods. As a proof of concept, we compare two widely used interaction inference approaches, showing how N2SIMBA can be used to assess their ability to recover known interactions. In general, N2SIMBA provides a flexible framework for simulating microbial communities and supports the development, testing, and validation of microbial interaction inference methodologies, contributing to more robust interpretations of microbial community data.