A recent study indicates that generative Artificial Intelligence (AI) can significantly speed up the process of finding ways to reduce methane emissions produced by cows in animal farming. This farming sector makes up roughly 33% of U.S. agriculture and contributes about 3% to the nation’s overall greenhouse gas emissions.
A recent study from USDA’s Agricultural Research Service (ARS) and Iowa State University (ISU) indicates that generative Artificial Intelligence (AI) can significantly speed up the process of finding ways to reduce methane emissions produced by cows in animal farming. This farming sector makes up roughly 33% of U.S. agriculture and contributes about 3% to the nation’s overall greenhouse gas emissions.
“Finding ways to tackle methane emissions from livestock is a crucial issue. Our researchers are applying innovative, data-driven methods to assist cattle producers in achieving emission reduction targets that protect the environment and foster a more sustainable agricultural future,” stated ARS Administrator Simon Liu.
One promising method starts in the cow’s stomach, where microorganisms play a role in enteric fermentation, causing cows to release methane through normal digestion. Researchers discovered a series of compounds that could inhibit methane production in the cow’s largest stomach chamber, the rumen, which may help in cutting methane emissions.
Among these compounds, bromoform, found naturally in seaweed, can reduce enteric methane production in cattle by 80-98%. However, bromoform is recognized as a carcinogen, limiting its use in cattle due to food safety concerns. Consequently, researchers are on the lookout for other compounds that share methane-inhibiting capabilities without the associated toxicity. Yet, this search is often time-consuming and expensive.
To tackle these challenges, a group of scientists from the ARS Livestock Nutrient Management Research Unit and ISU’s Department of Chemical and Biological Engineering utilized generative AI alongside large computational models to accelerate the search for alternatives to bromoform that are non-toxic.
“We are employing advanced molecular simulations and AI to discover new methane inhibitors by studying the properties of previously tested compounds [like bromoform], ensuring they are safe, scalable, and have significant potential for methane emission reduction,” explained Matthew Beck, a research animal scientist involved in the study, now with Texas A&M University’s Department of Animal Science. “Iowa State University is spearheading the computer simulation and AI aspects, while ARS focuses on identifying and testing compounds through both in vitro [laboratory] and in vivo [live cattle] studies.”
The researchers utilized publicly accessible databases containing scientific observations from earlier studies about cows’ rumen to create expansive computational models. Using AI with these models enabled them to anticipate molecular behaviors and pinpoint those that warrant further laboratory testing. Results from these lab tests were then integrated back into the computer models, enhancing AI predictions in a feedback loop process known as a graph neural network.
“Our graph neural network, a type of machine learning model, helps map out the properties of molecules, including atomic details and chemical bonds, while preserving crucial information to understand how these molecules behave in the cow’s stomach,” noted ISU Assistant Professor Ratul Chowdhury. “We analyzed their biochemical signatures to determine what enables them to inhibit methane effectively, unlike the fifty thousand other molecules present in the cow’s rumen that do not impede methane production.”
“This study successfully identified fifteen molecules that are closely related in what we refer to as a ‘functional methanogenesis inhibition space,’ indicating they share similar methane inhibition potential, chemical structure, and cell permeability as bromoform,” Chowdhury added.
Scientists believe AI can be essential in exploring how known molecules interact with proteins and the microbial community in the rumen, potentially leading to the discovery of new molecules and vital interactions within the rumen microbiome. This kind of predictive modeling could prove particularly beneficial for animal nutritionists.
“Currently, there are other promising strategies available to reduce enteric methane emissions, but they remain limited,” remarked USDA-ARS Research Leader Jacek Koziel. “That’s why integrating AI with laboratory studies, through continuous refinement, is a powerful scientific approach. AI can accelerate research and enhance the various strategies that animal nutritionists, researchers, and companies can pursue to make strides toward the ambitious goal of reducing greenhouse gas emissions and helping mitigate climate change.”
The study also outlines the overall computational and financial expenses associated with researching each molecule, providing an estimate of potential costs and foreseeable challenges. This information can help guide investment decisions for this type of laboratory research.
Chowdhury, Beck, and Koziel co-authored the paper published in Animal Frontiers, along with Nathan Frazier from ARS and Logan Thompson from Kansas State University. Mohammed Sakib Noor, a graduate student at ISU, is collaborating with Chowdhury to develop the graph neural networks.
The Agricultural Research Service is the principal scientific research agency of the U.S. Department of Agriculture. Daily, ARS works on solutions to agricultural issues affecting America. For every dollar invested in U.S. agricultural research, there is a $20 economic return.
