Predicting feed efficiency traits in beef cattle using genomics and rumen microbiomics

Authors: LAKAMP, ANDREW - University Of Nebraska; ADAMS, SEIDU - University Of Nebraska; Wells, James; Snelling, Warren; Kuehn, Larry; FERNANDO, SAMODHA - University Of Nebraska; SPANGLER, MATTHEW - University Of Nebraska

Submitted to: Midwestern Section of the American Society of Animal Science
Publication Type: Abstract Only
Publication Acceptance Date: 12/30/2024
Publication Date: 5/20/2025
Citation: Lakamp, A.D., Adams, S., Wells, J., Snelling, W.M., Kuehn, L.A., Fernando, S.C., Spangler, M.L. 2025. Predicting feed efficiency traits in beef cattle using genomics and rumen microbiomics [abstract]. Journal of Animal Science. 103(Supplement 1):18-19. https://doi.org/10.1093/jas/skaf102.020.
DOI: https://doi.org/10.1093/jas/skaf102.020

Interpretive Summary:

Technical Abstract: In the beef industry, average daily dry matter intake (ADDMI) and average daily gain (ADG) are economically relevant traits. Both ADDMI and ADG are influenced by host genomics and rumen microbiomic composition. Therefore, these data can be used to predict the future performance of animals for better management decisions. This study used 417 steers and 300 heifers with phenotypic, genomic, and microbiomic data. There were four diet groups, two for the steers (high concentrate) and two for the heifers (high forage). Two methods of creating the microbial (co) variance matrices were tested, the Van Raden and Ross method, wherein the scaling factors differed. Three versions of the microbiome (co)variance matrix (M) were created: no adjustment (naïve), microbial features were weighted by 1-h2 (weighted), and scaling microbial features in M differently for different clusters (data-driven), where clusters were determined by microbiome composition similarity. Three levels of model complexity were introduced: a single random effect including either a genomic or microbiome effect, both genomic and microbiome effects (additive), and the additive model including the interaction among the two random effects (interactive). In total, there were 38 models: the host genomic model, 6 single effect microbial models (one for each M method-adjustment combination), 6 additive models, and 6 interaction models for each of the two phenotypes of interest. Two forms of cross-validation were used. A four-fold cross-validation where data were evenly divided at random and a leave-one-out approach where one diet was predicted by the others. Accuracy was defined as the Pearson correlation between the adjusted phenotype and the sum of the random effects of a given animal. Accuracy estimates ranged from 0.01 – 0.30 depending on model attributes, cross-validation strategy, and trait of interest. In general, the construction method of M, Van Raden or Ross, did not have a large effect. Models that included a naïve or weighted M performed similarly and out-performed models with a data-driven M. Models with multiple random effects were always more accurate than models with a single random effect. The inclusion of an interaction term never decreased accuracy to a significant degree, but often slightly increased accuracy. The accuracies from the leave-one-out cross-validation were often lower than those from the four-fold strategy. Therefore, it is possible to utilize both host genomic and rumen microbiomic data to make accurate phenotypic predictions of feed efficiency traits in beef cattle.