CONJUGATE VARIABILITY OF LINEAR TRAITS AND THEIR COMBINATIONS WITH MILK PROTEIN CONTENT IN HOLSTEIN FIRST-CALF HEIFERS

<jats:p>The objective of the study is to determine the predictive ability of linear traits and their combinations for assessing milk protein yield in first-calf heifers. The study subjects were Holstein first-calf heifers with conformation and milk protein content assessed using 18 linear traits over 305 days of lactation. The subject of the study was the phenotypic correlation of linear traits with milk protein yield. The relationship between the studied traits was assessed in 1,243 first-calf heifers from a high-yielding Holstein herd in the Novosibirsk Region from 2017 to 2023, lactating for 305 days or more, using the Spearman rank correlation coefficient. Preprocessing of the primary data consisted of transforming the linear trait estimates to correctly calculate the correlation coefficient. Grouping of conformation parameters to determine the complex most closely associated with the productive trait was performed using two algorithms. Milk protein content over 305 days of lactation ranged from 156 to 458 kg. A weak correlation was found between conformation and productivity indicators for five udder traits and one body trait, ranging from +0.073 to +0.147 (α &lt; 0.01, α &lt; 0.001). The correlation coefficient between the composite score calculated using the first algorithm and milk protein yield was +0.275 (α &lt; 0.001), while for the second algorithm it was +0.292 (α &lt; 0.001). The approximation coefficients (R2) of the trend functions plotted against the average milk protein level of animal groups, aggregated by the composite score, indicated satisfactory model accuracy in the first case (≈ 0.95) and high accuracy in the second (≈ 0.98). The comparison results suggest that the second method is the preferred algorithm for identifying the best combination of linear features for indirectly assessing and predicting productivity in groups of first-calf heifers. The first method is recommended as a rapid test to determine the feasibility of further application of the second, which is more computationally expensive but more effective.</jats:p>