Abstract
BACKGROUND: Differentiating the microbiome changes associated with diseases is challenging but critically important. Majority of existing efforts have been focused on a community level, but the discerning power of community or holistic metrics such as diversity analysis seems limited. This prompts many researchers to believe that the promise should be downward to species or even strain level-effectively and efficiently identifying unique or enriched species in diseased microbiomes with statistical rigor. Nevertheless, virtually, all species-level approaches such as differential abundance and differential network analysis methods exclusively rely on species abundances without considering species distribution information, while it can be said that distribution is equally, if not more, important than abundance in shaping the spatiotemporal heterogeneity of community compositions. RESULTS: Here, we fill the gap by developing a novel framework-species specificity and specificity diversity (SSD)-that synthesizes both abundance and distribution information to differentiate microbiomes, at both species and community scales, under different environmental gradients such as the healthy and diseased treatments. The proposed SSD framework consists of three essential elements. The first is species specificity (SS), a concept that reincarnates the traditional specialist-generalist continuum and is defined by Mariadassou et al. (Ecol Lett 18:974-82, 2015). The SS synthesizes a species' local prevalence (distribution) and global abundance information and attaches specificity measure to each species in a specific habitat (e.g., healthy or diseased treatment). The second element is a new concept to introduce here, the (species) specificity diversity (SD), which is inspired by traditional species (abundance) diversity in community ecology and measures the diversity of specificity (a proxy for metacommunity heterogeneity, essentially) with Renyi's entropy. The third element is a pair of statistical tests based on the principle of permutation tests. CONCLUSIONS: The SSD framework can (i) identify and catalogue lists of unique species (US), significantly enriched species (ES) in each treatment based on SS and specificity permutation (SP) test and (ii) measure the holistic differences between assemblages (or treatments) based on SD and specificity diversity permutation (SDP) test. Both capacities can be enabling technologies for general comparative microbiome research including risk assessment, diagnosis, and treatment of microbiome-associated diseases.