Up to now, diatom assembly rules are mainly studied through the direct relation between species and environmental parameters. Neither the role of neutral dispersion, nor the influence of interspecific relationships are considered. Here, we studied 115 diatom samples from French reference stations. We predicted each species with environmental parameters, but we obtained different co-occurrence patterns in this predicted dataset compared to the initial dataset. Such results could be explained by the strong influence of interspecific relationships in the community structure. Those relationships were not taken into account by the prediction model. Co-occurrence patterns seem different between most abundant species and scarce species, resulting in complex assembly rules. We investigated which statistical techniques could enable to account for such relationships.