The aim of this study was to estimate (co)variance components for milk coagulation properties (MCP) predicted by mid-infrared spectroscopy (MIRS) during routine milk recording, and to assess their relationships with yield and quality traits. A total of 63 470 milk samples from Holstein-Friesian cows were analyzed for MCP, pH and quality characteristics using MIRS. Casein to protein and protein to fat ratios were calculated from information obtained by MIRS. Records were collected across 1 year on 16 089 cows in 345 herds. The model used for genetic analysis included fixed effects of parity and stage of lactation, and random effects of herd-test-day, cow permanent environmental, animal additive genetic and residual. (Co)variance components were assessed in a Bayesian framework using the Gibbs Sampler. Estimates of heritabilities were consistent with those reported in the literature, being moderate for MCP (0.210 and 0.238 for rennet coagulation time (RCT) and curd firmness (a30), respectively), milk contents (0.213 to 0.333) and pH (0.262), and low for somatic cell score (0.093) and yield traits (0.098 to 0.130). Repeatabilities were 0.391 and 0.434 for RCT and a30, respectively, and genetic correlations were generally low, with estimates greater than 0.30 (in absolute value) only for a30 with fat, protein and casein contents. Overall, results suggest that genetic evaluation for MCP predicted by MIRS is feasible at population level, and several repeated measures per cow during a lactation are required to estimate reliable breeding values for coagulation traits.

Mid-infrared spectroscopy (MIR) is used to predict fatty acid (FA) composition of individual milk samples (n = 267) of Brown Swiss cows. FAs were analyzed by gas chromatography as a reference method. Samples were scanned (4000 to 900 cm−1) by MIR, and predictive models were developed using modified partial least squares regressions with full cross-validation. The methods using a first derivative or multiplicative scatter corrected plus first derivative resulted, on average, in the best predictions. Coefficients of correlation between measured and predicted C8:0, C10:0, C12:0, C14:0, anteiso-C17:0, c9-C18:1, and medium- and long-chain FA, and saturated, monounsaturated and unsaturated FA ranged from 0.71 to 0.77, suggesting that prediction models can be implemented in milk recording schemes to routinely collect information on FA composition from the whole Brown Swiss population for breeding purposes.