Fecal microbial biomarkers represent a less invasive alternative for acquiring information on wildlife populations than many traditional sampling methodologies. Our goal was to evaluate linkages between fecal microbiome communities in Rocky Mountain elk (Cervus canadensis) and four host factors including sex, age, population, and physical condition (body-fat). We paired a feature-selection algorithm with an LDA-classifier trained on elk differential bacterial abundance (16S-rRNA amplicon survey) to predict host health factors from 104 elk microbiomes across four elk populations. We validated the accuracy of the various classifier predictions with leave-one-out cross-validation using known measurements. We demonstrate that the elk fecal microbiome can predict the four host factors tested. Our results show that elk microbiomes respond to both the strong extrinsic factor of biogeography and simultaneously occurring, but more subtle, intrinsic forces of individual body-fat, sex, and age-class. Thus, we have developed and described herein a generalizable approach to disentangle microbiome responses attributed to multiple host factors of varying strength from the same bacterial sequence data set. Wildlife conservation and management presents many challenges, but we demonstrate that non-invasive microbiome surveys from scat samples can provide alternative options for wildlife population monitoring. We believe that, with further validation, this method could be broadly applicable in other species and potentially predict other measurements. Our study can help guide the future development of microbiome-based monitoring of wildlife populations and supports hypothetical expectations found in host-microbiome theory.
Keywords: 16S‐rRNA survey; conservation; host‐microbiome; machine‐learning; microbial‐biogeography; non‐invasive; wildlife.
© 2022 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd.