Relationships between milk coagulation property traits analyzed with different methodologies.

Milk coagulation properties (MCP) analysis is performed using a wide range of methodologies in different countries and laboratories, using different instruments, coagulant activity in the milk, and type of coagulant. This makes it difficult to compare results and data from different research. The aims of this study were to propose a method for the transformation of values of rennet coagulation time (RCT) and curd firmness (a(30)) and to predict the noncoagulation (NC) probability of milk samples analyzed using different methodologies. Individual milk samples were collected during the morning milking in October 2010 from each of 165 Holstein-Friesian dairy cows in 2 freestall barns in Italy, and sent to 3 laboratories for MCP analysis. For each laboratory, MCP analysis was performed using a different methodology: A, with a computerized renneting meter instrument using 0.051 international milk clotting units (IMCU)/mL of coagulant activity; B, with a Lattodinamografo (Foss-Italia, Padova, Italy) using 0.051 IMCU/mL of coagulant activity; and C, with an Optigraph (Ysebaert, Frépillon, France) using 0.120 IMCU/mL of coagulant activity. The relationships between MCP traits were analyzed with correlation and regression analyses for each pair of methodologies. For each MCP trait, 2 regression models were applied: model 1 was a single regression model, where the dependent and independent variables were the same MCP trait determined by 2 different methodologies; in model 2, both a(30) and RCT were included as independent variables. The NC probabilities for laboratories with the highest number of NC samples were predicted based on the RCT and a(30) values measured in the laboratories with lower number of NC samples using logistic regression and receiver operating characteristic analysis. The percentages of NC samples were 4.2, 11.5, and 0.6% for A, B, and C, respectively. The transformation of MCP traits was more precise with model 1 for RCT (R(2): 0.77-0.82) than for a(30) (R(2): 0.28-0.63). The application of model 2 was needed when the C measurements were transformed into the other scales. The analyses of NC probabilities of milk samples showed that NC samples from one methodology were well distinguishable (with an accuracy of 0.972-0.996) based on the rennet coagulation time measured with the other methodology. A standard definition for MCP traits analysis is needed to enable reliable comparisons between MCP traits recorded in different laboratories and in different animal populations and breeds.