INVESTIGATION OF RESILIENCE AND ITS GENETICSTHROUGH THE USE OF LONGITUDINAL DATA

This thesis investigates resilience in dairy cattle using an integrated approach that combines phenotypic characterization, simulation of perturbations, and genomic analysis. The research was motivated by the need to breed animals capable of maintaining productivity amid the environmental and economic challenges facing modern agriculture. Thanks to close collaboration with a medium-large dairy farm in Northern Italy, we had access to a comprehensive dataset of daily milk records and genotype information, which formed the basis for the analysis. Over the past twenty years, resilience assessment in livestock has become feasible thanks to new technologies that enable the collection of high-frequency longitudinal data. Resilience indicators are recently defined proxies that combine routinely collected data, such as milk yield records, to capture an animal’s response to environmental challenges. Although it has been demonstrated that resilience indicators possess some genetic merit to some extent, their true significance and methodological foundations require further investigation to enhance the understanding of these phenotypes and their application in breeding programs. Lactation curve models aiming to represent the unperturbed state of each individual cow are essential for deriving meaningful resilience indicators; however, their impact on these indicators has not been systematically evaluated. This thesis addresses these gaps through three main studies: (1) characterization of resilience indicators derived from various lactation curve models; (2) a simulation study evaluating the ability of resilience indicators to capture true genetic resilience; (3) a genome-wide association study (GWAS) identifying genomic regions associated with resilience indicators. Additionally, one further contribution is included: (4) construction of a copy number variation map for the population studied. The characterization study demonstrated that methodological choices in lactation curve modelling significantly influence phenotypic resilience indicator values and cow rankings. The simulation study revealed that current resilience indicators capture only a fraction of true genetic resilience, with the best model-indicator combinations achieving correlations of 0.15 to 0.25 with simulated resilience breeding values. The genome-wide association study identified genomic regions associated with resilience indicators, confirming the polygenic nature of the trait and highlighting genes involved in immune response, energy metabolism, and tissue integrity. Additionally, the copy number variation map provided a valuable genomic resource for future integrated analyses. These findings advance our understanding of resilience as a complex, multifaceted trait while providing practical tools for its genetic improvement. The practical implications are clear: breeding programs should incorporate resilience indicators derived from meaningfully selected lactation curve models, set realistic expectations for genetic progress given the modest correlations with true resilience, and balance selection for resilience alongside production and other traits through appropriately weighted selection indices. Looking ahead, continued refinement of resilience phenotypes, functional characterization of candidate genes and economic analyses will advance breeding for resilience.