Inferring causal relationships among milk protein fractions in dairy cattle

The quantities and proportion of milk protein fractions have a fundamental role in its nutritional and technological properties. A deeper knowledge of the possible casual relationships among these traits would be useful not only to improve the understanding of their biology but also for setting up management and selection strategies. Aims of this study were: i) to estimate genomic relationships among protein fractions and ii) to infer a Bayesian network structure among them. To achieve these aims, we first fitted a Bayesian multi-trait genomic best linear unbiased prediction (GBLUP) model to infer the genomic and residual correlations among six milk nitrogen fractions (four caseins, namely k-, β-, αs1- and αs2-casein (CN), and two whey proteins, namely β-lactoglobulin and α-lactalbumin), in a population of 1,011 Italian Brown Swiss cows. Animals were genotyped with the Illumina BovineSNP50 Bead Chip v.2. The posterior means of breeding values were used as input to infer putative casual structures among traits by using the Bayesian Max-Min Hill-Climbing (MMHC) algorithm. Strong genomic correlations were found between β-CN and αs1-CN and, between κ-CN and β-CN. The application of the MMHC algorithm pointed out that κ-CN seemed to directly or indirectly influence all the other milk protein fractions (with the exception of β-lactoglobulin), suggesting its possible role of leading trait in the control of milk protein composition. Further and thorough studies are however needed for the validation of the putative causal links among milk protein fractions identified in this study.