MIM: Using Machine Learning and a Model Watershed to Understand how Microbes Govern Food Web Architecture and Efficiency

Rules that govern food web dynamics, which describe how energy is transferred among different living organisms, are among the most universal laws of nature. Consumption up a food-chain is an inherently inefficient process that leads to significant and predictable losses through waste and respiration. This rule of life can be leveraged to model how biological diversity will respond to phenomena such as sudden changes in the environment or species extinctions, and is an important constraint in food production.

Until now, food web research has largely focused on the interactions among plants and animals, however, microbes living in and on larger organisms play essential roles in their health, rates of reproduction, and ability to digest food. This project will examine how symbiotic microbes govern the efficiency of food webs, and how aspects of food webs, in turn, determine the composition of symbiotic microbes.

The predictive insight gained from this research may make it possible to manipulate the composition of microbes to create more efficient food webs that can potentially guide restoration of degraded habitats, capture carbon, and increase yield in agriculture, aquaculture and biofuels systems. In addition, workforce development and outreach to under-represented groups including native Hawaiians and Pacific Islanders, will be performed.

Postdoctoral researchers, graduate students and undergraduates will be trained in microbiome science through research experiences and class modules.

This proposal addresses the hypothesis that canonical laws governing the transfer of energy among trophic levels of food webs both constrain, and are constrained by the composition and function of microbiomes. Leveraging a model Hawaiian watershed system, this project aims to understand how host-associated microbiomes govern food chain efficiency and how, in turn, trophic position within a food web affects the microbiome.

The project will develop transfer learning approaches based on machine-learning tools trained on higher-feature datasets (such as the Earth Microbiome Project) to enable robust predictions of the interaction between food chain length, trophic position and microbiome diversity. Two tractable experimental systems will be used to explore these predictions.

The first is a simple four-tiered bromeliad food web mesocosm where the number and of trophic levels is controlled. The second consists of a three-tiered mosquito microcosm in which all microbial symbionts are isolated and manipulated. Associated genomic data will enable a mechanistic understanding of how microbiomes influence food web efficiency and function by altering metabolic capacity of hosts.

In summary, this project will employ food web theory to explain and predict the interactions between the microbiome, the host, and the environment.

This project is jointly funded by the Understanding Rules of Life: Microbiome Interactions and Mechanisms Program and the Established Program to Stimulate Competitive Research (EPSCoR).

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.