Multiple factors drive imbalance in the global microbial assemblage in soil.

Soil microbial assemblages play a critical role in biogeochemical cycling processes in terrestrial ecosystems. Dynamic global information for these assemblages considering multiple factors is critical for predicting ecological safety concerns but remains unpredictable. Here, we collected microbial data from soil datasets worldwide and used a feature-explicable machine learning (FEML) approach to address this problem. Multiple-factor and factor interaction network analysis based on FEML can be used to visualize the restrictive relationships among multiple factors (e.g., fertilizer application, land use, and changing global climate and natural environments), which are difficult to explore based on limited experimental data and traditional machine learning methods. The FEML approach predicted that areas of bacterial hotspots in South America and Africa will expand by approximately 27% and 83%, respectively, in scenario RCP8.5 in 2100. In contrast, the areas of fungal hotspots in Asia and North America will decline by approximately 34% and 62%, respectively, under RCP8.5. The unbalanced ratios of bacteria to fungi affect the soil ecosystem, and bacterial-dominated communities contribute to the reduction of easily decomposing nutrients, the growth of the bacterivore community and a high proportion of microaggregates in the soil. Therefore, mitigating climate change is critical to reduce the remarkable imbalance between soil bacteria and fungi and predict risks to soil microbial assemblages based on multiple factors.