To navigate this complexity, we developed a method that reveals simple, emergent patterns in natural microbial communities. These patterns, which are often found in microbiomes from soils to oceans, have not been understood in terms of their origins or implications for important climate processes, such as carbon and nitrogen cycling.

 

Our first step was to apply a mathematical approach, unit-invariant Singular Value Decomposition (SVD), to large-scale datasets of the topsoil microbiome. Unit-invariant SVD allowed us to characterize the dominant modes of variation in these complex systems without being biased by changes in the absolute quantities of genes that are present across samples. This technique revealed consistent global patterns in the abundance of key genes involved in nitrogen cycling: nitrate reductase genes (nap and nar). Specifically, we found a striking relationship between soil pH and the relative abundance of these genes: as pH increased, the nap gene became more abundant, while the nar gene decreased. This analysis provided a clear, simplified framework for understanding global microbiome patterns that had previously seemed overwhelmingly complex.

 

Once we identified these statistical patterns using SVD, we sought to understand their biological origins. To do this, we turned to controlled laboratory experiments. We sampled microbial communities from soil and grew them under varying pH conditions to determine whether the patterns we observed in global data could emerge naturally in more controlled settings. Remarkably, we found that the same pH-driven gene abundance patterns—identified mathematically with SVD—emerged in the lab across multiple soil samples.

 

By isolating the relevant microbes, we discovered a pH-modulated interaction between species carrying nap and nar genes. At low pH, the dominance of the nar genotype arose from an interaction between strains harboring nar and nap. Thus, the emergent patterns in gene content across the globe likely reflect ecological interactions in the community – a key insight from our study.

 

In the future, we plan to apply environmental perturbations to intact soil samples and use machine learning to further explore how microbial responses at the species level connect to metabolic processes. By bridging mathematical analysis with experimental biology, we aim to illuminate the microbial mechanisms that shape Earth's nutrient cycles.