Democratizing precision farming through prediction of the soil microbiome from environmental features

Microbial communities influence their environment by modifying the availability of compounds, such as nutrients or chemical elicitors. Knowing the microbial composition of a site is therefore relevant to improve productivity or health. However, sequencing facilities are not always available, or may be prohibitively expensive in some cases. Thus, it would be desirable to computationally predict the microbial composition from more accessible, inexpensive, and easily-measured features.

To address this goal we have developed a computational system, based on deep learning architectures. The model was built with experimental data from crop soil, in particular, maize rhizosphere microbiome. Maize is one of the three most important crops in the world, together with wheat and rice, then a model for the development of strategies to increase the soil productivity. Our system reconstructs the microbial composition (more than 700 taxa) with high fidelity (>0.9 correlation), and successfully predicts microbial composition from a few environmental variables (0.73 correlation).

Our system is able to predict the microbial composition under hypothetical scenarios, such as climate change conditions, that could be useful to anticipate suitable crop strategies. In addition, the microbiome knowledge stored in the AI model can be reused, for example by farmers in emerging economies, where only phenotypic data are available. Thus, this technology represents a low-cost tool that contributes to the design of precision agriculture strategies.

For demonstration purposes, we have provided a user-friendly interface, where you can predict maize soil microbiome from your selection of temperature, rain and plant age parameters. It runs in your browsers, so anyone can use it.