Understanding Rumen Methanogen Interactions in Sheep Using Machine Learning

Abstract

Methane emissions from livestock pose a significant challenge globally, particularly in countries with a strong farming industry dominated by sheep farming, such as Aotearoa, New Zealand (NZ). Chemical inhibitors such as feed additives or vaccines help to decrease methane emissions. However, their successful development has been hindered by a limited understanding of the complex interactions among the microorganisms in the rumen (forestomach). This study serves as a proof-of-concept to explore the potential of using metatranscriptome data to understand the genetic basis of microbial interactions in the rumen and identify potential inhibitor targets. We analyzed a small but carefully curated dataset of 10 sheep emitting different levels of methane. We employed various statistical and machine learning techniques to uncover new contigs (continuous sequences of DNA) linked to high levels of methane output. Despite the limited sample size, our findings revealed new insights into microbial mechanisms, validated by domain experts. These preliminary results suggest that expanding the dataset and integrating machine learning can enhance our understanding of the complex microbial interactions in the rumen, ultimately contributing to the development of effective strategies to reduce methane emissions in livestock.