Prediction accuracy for feed intake and body weight gain using host genomic and rumen metagenomic data in beef cattle

Authors: LAKAMP, ANDRES - University Of Nebraska; ADAMS, SEIDU - University Of Nebraska; Kuehn, Larry; Snelling, Warren; Wells, James; Hales Paxton, Kristin; Neville, Bryan; FERNANDO, SAMODHA - University Of Nebraska; SPANGLER, MATTHEW - University Of Nebraska

Submitted to: Genetics Selection Evolution
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/30/2025
Publication Date: N/A
Citation: N/A

Interpretive Summary: The cattle rumen enables the breakdown of plant fibers so that the animal can use the resulting products for energy and as a protein source. Thus, the genetic makeup of both the animal and the bacteria in the rumen can have important implications for cattle feed efficiency. This study used cattle genotypes, bacterial genomic data (metagenome), and their combination to predict cattle intake and gain using several different models. Results suggested that feed intake and gain could be predicted accurately especially when both sources of genomic information were used. These results suggest that knowledge of cattle genotypes and rumen composition could dramatically improve management of cattle to peak efficiencies.

Technical Abstract: Background Host genomic and rumen microbiome data can predict feed efficiency traits, supporting management decisions and increasing profitability. This study estimated the proportion of variation of average daily dry matter intake and average daily gain explained by the rumen metagenome in beef cattle, evaluated prediction accuracy using genomic data, metagenomic data, or their combination, and explored methods for modelling the rumen metagenome to improve phenotypic prediction accuracy. Data from 717 animals on four diets (two concentrate-based, two forage-based) were analyzed. Animal genotypes consisted of 749,922 imputed sequence variants, while metagenomic data comprised 16,583 open reading frames from ruminal microbiota. The metagenome was modelled using six (co)variance matrices, based on combinations of two creation methods and three modifications. Nineteen mixed linear models were used per trait: one with genomic effects only, six with metagenomic effects, six combining genomic and metagenomic effects, and six adding interaction effects. Two cross-validation schemes were applied to evaluate prediction accuracy: 4-fold cross-validation balanced for diet type and leave-one-out cross-validation, where three diets served as training and the fourth as testing. Prediction accuracy was measured as the correlation between an animal’s summed random effects and its adjusted phenotype. Results Although minimal, differences existed in parameter estimates and validation accuracy depending on how the metagenome effect was modelled. Phenotype prediction accuracy ranged from 0.01 – 0.30. No specific set of model characteristics consistently lead to the highest accuracies. Models which combined genome and metagenome data outperformed those using either data source alone. Models where the rumen metagenome (co)variances matrix was scaled within each diet composition generally led to lower prediction accuracies in this study. Conclusions The rumen metagenome can explain a significant proportion of variation in beef cattle feed efficiency traits. Those traits can also be predicted using either host genome or rumen metagenome, though using both sources of information proved more accurate. Multiple methods of forming the metagenome (co)variance matrix can lead to similar prediction accuracies.