Development and validation of a model for early prediction of residual feed intake in beef cattle using plasma biomarkers

International audience

Английский. Identification of plasma biomarkers for feed efficiency in growing beef cattle offers a promising opportunity for developing prediction models to improve precision feeding strategies. However, these models must accurately predict feed efficiency at early stages of fattening. Our study aimed to evaluate the reliability of candidate biomarkers previously identified in late-fattening cattle when analysed during early fattening stages and to develop diet-specific prediction equations for residual feed intake (RFI). From a total of 364 Charolais bulls across seven cohorts, we selected 64 animals with extreme RFI values. The animals were fed either a corn- or grass-silage diets. These animals were chosen from four out of the available seven cohorts. Animals from three cohorts (24 high-RFI and 24 low-RFI, having a mean RFI difference of 1.48 kg/d) were used for biomarker confirmation and prediction model training. Animals from a fourth cohort (8 high-RFI and 8 low-RFI, having a mean RFI difference of 0.98 kg/d) were used for model external validation. Blood samples were collected at the beginning of the feed efficiency test (333 +/- 20 days), and plasma underwent targeted metabolomic for 630 metabolites, natural abundance of N-15 (delta N-15), insulin, and IGF-1 analysis. Seven previously identified plasma biomarkers for RFI in late-fattening beef cattle still kept their capability for discriminating low and high RFI animals when analysed during early fattening stages (P < 0.05). Among these confirmed biomarkers, five were common for both grass- and corn-fed animals (creatinine, beta-alanine, triglyceride TG18:0_34:2, symmetric dimethyl-arginine and phosphatidylcholine PC aa C30:2) while two were diet-specific (IGF-1 for grass silage-based diet, and isoleucine for corn silage-based diet. No new plasma biomarkers of RFI were identified at early-fattening stages (false discovery rate > 0.05). Prediction models were developed based on seven confirmed RFI biomarkers analysed during early-fattening. Two logistic regression models incorporating creatinine and either IGF-1 (for grass silage-based diet) or PC aa C30:2 (for corn silage-based diet) effectively distinguished between high- and low-RFI animals with high sensitivity and specificity (area under the curve > 0.80). The biomarkers used in the models showed moderate to high repeatability between early and late fattening stages (0.45 < r < 0.65). The models were successfully externally validated, with more than 85% of animals from the fourth cohort correctly classified. Once validated in larger cohorts and utilising cost-effective and rapid analytical methods, these models could support precision feeding and breeding programmes, aiming to reduce the cost of raising beef cattle.