Development of new machine learning methods and algorithms for the integration of omics and environmental data to quantify the role of the soil microb

A new set of machine learning tools will sift through DNA sequences, soil chemistry, and weather records to work out exactly which microbes lock carbon into the ground and which ones release it. Soils hold more carbon than the atmosphere and all plant life combined, but that carbon is not static. Microbes constantly break down organic matter, releasing carbon dioxide, while also building stable compounds that can stay buried for decades. Current computer models cannot handle the sheer complexity of the data—genetic profiles of thousands of microbial species, plus soil pH, moisture, temperature, and land-use history—all interacting at once. Existing methods either oversimplify the biology or fail to find meaningful patterns in the noise. This project builds fundamentally new algorithms designed specifically for this kind of multi-layered data. If successful, the tools will let researchers predict which farming practices or land-management strategies actually increase long-term carbon storage, rather than just shifting it around. The work is primarily fundamental science—advancing how computers learn from messy biological systems—but the practical payoff could be more accurate carbon accounting for agriculture and better-informed decisions about soil management in climate mitigation efforts.