specificity: an R package for analysis of feature specificity to environmental and higher dimensional variables, applied to microbiome species data

• Published in: Environmental Microbiome
• Main Authors: John L. Darcy, Anthony S. Amend, Sean O. I. Swift, Pacifica S. Sommers, Catherine A. Lozupone
• Format: Article in Journal/Newspaper
• Language: English
• Published: BMC 2022
• Subjects: Species distributions; Biogeography; Multi-omic data; Environmental sciences; GE1-350; Microbiology; QR1-502; Antarctic; Antarc*
• Online Access: https://doi.org/10.1186/s40793-022-00426-0; https://doaj.org/article/ce8a3be41d5b4f679637da27ebbe5ce7

Abstract Background Understanding the factors that influence microbes’ environmental distributions is important for determining drivers of microbial community composition. These include environmental variables like temperature and pH, and higher-dimensional variables like geographic distance and host species phylogeny. In microbial ecology, “specificity” is often described in the context of symbiotic or host parasitic interactions, but specificity can be more broadly used to describe the extent to which a species occupies a narrower range of an environmental variable than expected by chance. Using a standardization we describe here, Rao’s (Theor Popul Biol, 1982. https://doi.org/10.1016/0040-5809(82)90004-1, Sankhya A, 2010. https://doi.org/10.1007/s13171-010-0016-3 ) Quadratic Entropy can be conveniently applied to calculate specificity of a feature, such as a species, to many different environmental variables. Results We present our R package specificity for performing the above analyses, and apply it to four real-life microbial data sets to demonstrate its application. We found that many fungi within the leaves of native Hawaiian plants had strong specificity to rainfall and elevation, even though these variables showed minimal importance in a previous analysis of fungal beta-diversity. In Antarctic cryoconite holes, our tool revealed that many bacteria have specificity to co-occurring algal community composition. Similarly, in the human gut microbiome, many bacteria showed specificity to the composition of bile acids. Finally, our analysis of the Earth Microbiome Project data set showed that most bacteria show strong ontological specificity to sample type. Our software performed as expected on synthetic data as well. Conclusions specificity is well-suited to analysis of microbiome data, both in synthetic test cases, and across multiple environment types and experimental designs. The analysis and software we present here can reveal patterns in microbial taxa that may not be evident from a community-level ...