MAPPING GENOMIC VARIANTS ASSOCIATED WITH FUNCTIONAL AND HEALTH TRAITS IN DAIRY CATTLE POPULATION

This thesis aimed to integrate high-resolution genomic data with longitudinal phenotypes collected through Automatic Milking Systems (AMS). The objective was to investigate genetic variability, resilience, and udder health in Holstein dairy cattle. By combining genomic tools, sensor-derived time series, and veterinary treatment records, the study provides a multidimensional view of animal functionality under real farming conditions. The first chapter, “Genetic variability in dairy cattle” , studied the genetic diversity of Holsteins bred in seven dairy farms in the Lombardy region through Single Nucleotide Polymorphisms (SNP), Runs of Homozygosity (ROH) and Copy Number Variants (CNVs). The results revealed that intensive selection for milk yield, applied to Holstein cattle during the first decades of the 20th century, strongly shaped the genome of the breed. This process reduced variability in some regions while increasing the frequency of alleles linked to production. CNVs overlapping with genes involved in immunity and fertility suggested a role in adaptation. Within this chapter, these findings were presented in two published articles, providing the genetic background of the animals under study. The second chapter, “Resilience”, addressed resilience using daily AMS data and statistical models to estimate expected lactation curves and calculate resilience indicators. Four indicators were evaluated: variance of milk yield deviations (LnVar), autocorrelation of residuals (rauto), frequency of perturbations (wfPert), and total milk loss (dPert). Analysis of nearly 600,000 daily records showed that older cows displayed greater variability and more frequent perturbations, while the capacity to recover (rauto) remained constant. Genome-wide analyses based on these indicators identified genes related to immune function, metabolism, and tissue integrity, confirming the complex biological nature of resilience. The third chapter, presented as a dedicated article, explores a novel approach that integrates AMS- derived perturbations with veterinary treatment records, thereby incorporating health information into resilience phenotyping. Deviations in milk yield were quantified as perturbation events described by duration, slopes of decline and recovery, and area of loss. Cows were classified into immunological susceptibility groups according to the temporal relationship between veterinary treatments and perturbations. This classification revealed clear differences in resilience indicators across categories. By aligning perturbation profiles with treatment timing (before, during, or after perturbations), the study showed that interventions significantly influenced both the severity of production losses and the trajectories of recovery. The fourth chapter, “Mastitis susceptibility/resistance in Holstein cows”, focused on one of the more expensive diseases in dairy farming. Somatic cell count (SCC) and antibiotic treatments were used as a unique phenotype to map QTL linked to mastitis resistance, highlighting genomic regions associated with immune response and udder integrity. By combining SCC data with antibiotic treatments, the study provided a more comprehensive understanding of susceptibility and resistance to mastitis. Overall, this thesis demonstrates that resilience can be assessed in practice using AMS-derived phenotypes, reflects a partly poligenic genetic basis, and shows a close association with health events. The integration of genomic information with longitudinal data allow to identify animals that not only achieve high productivity but also cope better with stress and recover more quickly, with important implications for reducing antibiotic use, improving animal welfare, and enhancing the sustainability of dairy farming. In summary, the results show that combining genomic tools such as SNP, CNVs, and ROH with dynamic phenotypes derived from AMS provides new opportunities to select cows that are productive, healthy, and resilient.