Abstract
This thesis uses data-driven methods to understand factors influencing the health, welfare, and performance of pullets and laying hens in the poultry industry, with a focus on hatchability, rearing success, and fearfulness. Hatchability was calculated per batch of eggs as hatch of fertile eggs, rearing success was defined as the
... read more
percentage of animals that survived to the laying barn relative to the number of chicks that hatched from a batch while fearfulness referred to how laying hen pullets responded to stressful situations.
In chapter 2, it was shown that ensemble machine learning methods (random forest and gradient boosting machine) could predict hatchability better in terms of performance (root mean square error and coefficient of determination) than a linear regression. Additionally, it was observed that strain, breeder age, egg weight uniformity, length of egg storage and season had a significant effect on hatchability, while egg weight loss had little or no significant effect on hatchability.
In chapters 3 & 4, rearing success traits: clutch size (CS), first week mortality (FWM), rearing abnormalities (RA) and natural death (ND), were included as factors determining rearing success (RS). Chapter 3, identified CS as the most important trait, which connects the incubation and rearing stages of a laying hen’s life cycle. Larger CS was associated with increased rearing success, and CS showed the highest heritability (0.19) compared to FWM, RA and ND, irrespective of the genetic line that was considered. Similarly, in chapter 4, when Bayesian network analysis was applied on the same dataset that was used in chapter 3, CS again was identified as the main connection between the incubation and rearing stages. While CS was associated with 8 SNPs in chapter 3 when Genome Wide Association Study (GWAS) was used to identify significant single nucleotide polymorphisms (SNPs), it was linked to 10 SNPs in chapter 4 when Bayesian network analysis was applied. Consequently, CS was associated with the most SNPs in both chapters.
In chapter 5, five behavioral tests were used to assess fearfulness in pullets, to determine stress response. These tests included novel environment test (NET), voluntary approach test (VAT), manual restraint test (MRT), tonic immobility test (TIT), and open field test (OFT). The following golden standard traits (GSTs) were identified as the key measures of fearfulness in each test: NET (latency to produce isolation calls), VAT (latency to peck at food reward), MRT (latency to struggle), TIT (number of inductions required to induce tonic immobility), and OFT (latency to walk). GSTs denote particular traits from behavioral tests that better explain/measure fearfulness. Additionally, it was shown that a Convolutional Neural Network (CNN) model could predict these GSTs better in terms of performance (accuracy) than a Support Vector Machine (SVM) model, and a Random Forest (RF) model.
Overall, this thesis shows that animal related, environmental, and behavioral factors are essential for understanding the health, welfare, and performance of laying hen embryos and pullets. Therefore, collecting more embryonic and rearing data, and applying advanced analytical techniques on these data, can enable researchers to unlock new opportunities that can improve pullets’ health and welfare, while simultaneously enhancing pullets’ and laying hens’ performance. As a recommendation, animal related factors affecting health, welfare and performance of pullets can be addressed through breeding selection, while environmental factors can be adjusted by management.
show less