Structural equation models for genome-wide association study (SEM-GWAS) of interrelationships among udder health traits in dairy cattle

The quality and safety of milk products depend on animal health and welfare, and particularly on mammary gland health. It is well known that mastitis, either in its acute or sub-clinical form, greatly affects the nutritional and technological quality of milk. Apart from somatic cell count (SCC), milk lactose (LAC) content, pH and casein to protein ratio (CAS:PRT) are sensitive to inflammation of the mammary gland, and these variables could be incorporated together into an udder health indicator. Multi-trait model genome-wide association (MTM-GWAS) can be used to study associations between genomic regions and multiple traits, but it does not con-sider potential causal relationships among phenotypes. Alternatively, structural equation modelling (SEM) represents a powerful tool for modelling causal networks. Recently, SEM in com-bination with GWAS (SEM-GWAS) has been proposed to better understand the genetic basis and the relationships among a set of traits. SEM-GWAS can partition SNP effects into direct and indirect (i.e. mediated by an up-stream trait in the causal network) com-ponents. This study aimed to apply SEM-GWAS on a set of pheno-types related to udder health, i.e. milk yield (MY), somatic cell score (SCS), LAC, pH and CAS:PRT, in a cohort of 1158 Italian Brown Swiss cows. Animals were genotyped with the Illumina BovineSNP50 Bead Chip v.2. A causal phenotypic network was inferred in two stages: (1) a multi-trait model was employed to estimate covariance matrices of additive genetic effects and of residuals, and (2) the causal structure among phenotypes from the covariance matrix between traits, conditionally on additive genetic effects, was inferred by the Hill-Climbing algorithm. The residual (co)variance matrix was inferred using Bayesian Markov-chain Monte Carlo, with samples drawn from the posterior distribution. Results showed positive path coefficients for MY → LACT, LACT → CAS:PRT and SCS → pH, while negative values were obtained for LACT → SCS. Based on the identified phenotypic relationships structures, SEM-GWAS will be run, enabling to identify direct and indirect (i.e. mediated) SNP effects and providing a more complete picture of the genetic basis of these indicators of udder health.