AI guides design of new anti-methanogenic molecules

Ratul Chowdhury, assistant professor of chemical and biological engineering and a Building a World of Difference Faculty Fellow, is using AI to speed the search for molecules that inhibit methane production in dairy cattle digestive systems.

Chowdhury and his research team demonstrated how a graph neural network could quickly and efficiently examine all the molecules in bovine metabolite and milk composition databases – to reveal the “chemical signature” of effective anti-methanogenic molecules.

Researchers can then use the modeling results to design new methane inhibitor molecules.

“The known anti-methanogenic molecules are either efficient but unsafe for humans, or are modestly efficient but safe. To address this gap, we first collated a list of bovine molecules which are likely to be safe for humans.

We then deployed a combination of detailed biomolecular simulations and featurized each molecule in connection to the methanogenesis inhibition enzymatic process. Our simulations account for the chemical environment of the bovine rumen.

This helps us identify a rank-ordered list of molecules which are likely to be effective for inhibiting methane production while ensuring food safety. We are currently testing a large set of these molecules both in vitro and in vivo with our collaborators at Texas A&M and USDA,” says Chowdhury.

Chowdhury’s team AI work is designed to minimize required computing power, while guiding experimental work to the most promising options – and maximizing return on research investment.

“We believe foundational, biologically-aware AI models can enhance understanding of how molecules interact with proteins and the rumen microbiome, potentially uncovering novel compounds.

This predictive modeling is valuable for animal nutritionists, especially in addressing enteric methane emissions. Integrating focused AI models with lab research can accelerate discoveries, offering faster pathways for researchers, nutritionists, and companies to offer solutions.

To this end, we have put forward a blueprint of the analysis pipeline in our recent publication with our collaborators. The study provides a detailed breakdown of both computational and financial costs required to conduct this research on a per-molecule basis. This analysis estimates potential expenses and challenges, offering insights to guide investment decisions for conducting such research entirely in a laboratory setting,” says Chowdhury.

The paper, “Computational approaches for enteric methane mitigation research: From Fermi calculations to artificial intelligence paradigms,” was published in Animal Frontiers. It was co-authored by Chowdhury, Matthew Beck, Jacek Koziel, Nathan Frazier (USDA ARS) and Logan Thompson (Kansas State University).

Iowa State graduate students Randy Aryee and Mohammed Sakib Noor worked with Chowdhury to design this computational framework.