Bayesian confirmatory factor analysis modeling enables genome-wide discovery of latent feeding behavior processes underlying feed efficiency in beef cattle

Feed efficiency (FE) is linked to sustainable beef production by lowering total feed costs and environmental impacts, but its biological basis remains unclear in tropically adapted beef cattle. Hence, we used feeding-event and FE-related traits from 1,280 young Nellore bulls genotyped for 383,386 markers to derive latent phenotypes and investigate their genomic background. The latent factors represent feeding rhythm (LF1), feeding bout structure (LF2), and overall feed efficiency (LF3). A Bayesian structural equation model identified a hierarchical recursive structure in which LF1 negatively affected LF2 (-0.41) and LF3 (-0.55), whereas LF2 positively affected LF3 (0.29). Heritability estimates were moderate for LF1 (0.36 ± 0.05), LF2 (0.35 ± 0.05), and LF3 (0.30 ± 0.05). LF1 and LF3 were unfavorably genetically correlated (r g = -0.60 ± 0.10), whereas LF2 showed a lower genetic correlation with LF3 (r g = 0.18 ± 0.10). Genome-wide association analyses enabled the identification of 281, 210, and 631 significant markers for LF1, LF2, and LF3, respectively, located across 22 chromosomes. These regions harbored genes involved in hypothalamic appetite and satiety signaling for LF1, circadian-neuroendocrine regulation for LF2 and growth–metabolic signaling for LF3. Shared genes (e.g., LEPR , MAPK1 , PLAG1 , and FTO ) supported an integrated neuroendocrine–metabolic network linking feeding behavior to overall FE. Latent phenotypes derived from feeding-event and efficiency data captured heritable feeding strategies and provided biologically meaningful traits for subsequent genomic research. Their use as auxiliary traits in selection indices should be tested in Nellore cattle populations using multi-trait genomic prediction that accounts for genetic correlations and potential overlap with traditional FE traits.