A common analysis in community ecology includes linking variation in system functions to the distribution and abundance of taxa. In inferring processes, functions, and causal taxa, it is common practice to assume a core community can be defined. The core refers to a group of taxa found across samples and is a discretization of continuous occurrence and abundance data. Assuming categorical thresholds in abundance exist, and that a core community exists, has the potential to be misleading. Rather, the existence of a core set of taxa should be treated as a hypothesis that may or may not be supported. An additional challenge in defining a core is that there are no standard criteria for core membership, complicating comparison across studies. In this study, we considered four methods commonly used for defining core membership in studies of microbiomes and applied them to two published microbial data sets and simulations that cover a range of plausible communities. We then evaluated the ability of each method to correctly categorize taxa based on an a priori assignment of core status. Furthermore, we assessed each method’s ability to accurately assign the core while not over or under inflating membership.
Results/Conclusions
Assignment of core taxa varied substantially among methods and datasets, suggesting core membership is not robust to chosen methodology. Distributions of taxon abundance and prevalence offered limited support for a distinguishable set of core taxa. While able to identify known core taxa in select cases, the methods disagreed more often than not and often severely overinflated core membership. Given poor support for the concept of core communities, categorization of taxa into sets corresponding to core and non-core taxa is questionable and requires testing before use in any particular context. Our results do not support applying methods of dimension reduction for core taxa classification but instead provide another rationale to move towards analyses that utilize abundance data in their entirety.