Predicting soil fungal communities from chemical and physical properties

Summary

Introduction
Biogeography describes spatial patterns of diversity and explains why organisms occur in given conditions. While it is well established that the diversity of soil microbes is largely controlled by edaphic environmental variables, microbiome community prediction from soil properties has received less attention. In this study, we specifically investigated whether it is possible to predict the composition of soil fungal communities based on physicochemical soil data using multivariate ordination.
Materials and Methods
We sampled soil from 59 arable fields in Switzerland and assembled paired data of physicochemical soil properties as well as profiles of soil fungal communities. Fungal communities were characterized using long‐read sequencing of the entire ribosomal internal transcribed spacer. We used redundancy analysis to combine the physical and chemical soil measurements with the fungal community data.
Results
We identified a reduced set of 10 soil properties that explained fungal community composition. Soil properties with the strongest impact on the fungal community included pH, potassium and sand content. Finally, we evaluated the model for its suitability for prediction using leave‐one‐out validation. The prediction of community composition was successful for most soils, and only 3/59 soils could not be well predicted (Pearson correlation coefficients between observed and predicted communities of <0.5). Further, we successfully validated our prediction approach with a publicly available data set. With both data sets, prediction was less successful for soils characterized by very unique properties or diverging fungal communities, while it was successful for soils with similar characteristics and microbiome.
Conclusions
Reliable prediction of microbial communities from chemical soil properties could bypass the complex and laborious sequencing‐based generation of microbiota data, thereby making soil microbiome information available for agricultural purposes such as pathogen monitoring, field inoculation or yield projections.