Environmental context shapes microbial co-occurrence

The analysis of microbial co-occurrence patterns promises to uncover the elusive interactions underlying the many important functions microbiomes perform. We test such co-occurrence analysis on a dataset generated by a spatially explicit meta-community generalized Lotka-Volterra model, to investigate how environmental confounders affect co-occurrence patterns. We show that strong species interactions, particularly positive ones, remain detectable despite environmental complexity. However, in this context the number of spurious co-occurrences, with no correspondence to specific ecological drivers are inflated. We also explored two types of homogenization: intrinsic (via dispersal among communities) and extrinsic (via sample aggregation). In our settings both types inflate positive co-occurrences, improving the recon-struction of the environmental signal at the cost of the precision for reconstructing interactions. Negative ones, however, exhibit contrasting responses: intrinsic homogenization obscures them, while extrinsic homogenization increases their number, but at the cost of reduced precision in linking co-occurrences to ecological drivers. Our findings underscore the complex interplay between environmental factors and sampling strategies in shaping microbial co-occurrence patterns and their ecological interpretation.