Assessing the methane-mitigating effects of feed additives on dairy cows: Validation of an AI-Based predictive model using a monensin feed additive

In response to the growing demand for sustainable agriculture and process optimization, our study introduces an innovative approach that meets these needs; utilizing rumen microbiome samples from Israeli Holstein cows across 14 commercial dairy farms, we constructed an AI-driven model that predicts the effect of feed additive on enteric methane emissions. The model extracts patterns from microbiome datasets using a network-oriented approach to process raw sequencing data and identifies statistically significant DNA patterns. The identified patterns serve as biomarkers to confirm their significant correlation with the efficacy of the feed additive. For the model validation, cows were given a monensin feed additive; in addition, enteric methane emissions were obtained before and during the 12 weeks of the trial, then in each farm the methane measurements were compared with the respective control group to estimate the accuracy of the AI-model predictions. The validation was performed on independent cohorts to ensure robustness. The results obtained indicated a high accuracy in the model predictions, achieving an average reduction of 20 % on enteric methane emissions across 14 dairy farms, with peaks of 30 % of reduction.