Files
Abstract
ABSTRACTRatio traits such as fertility, hatchability, and egg production are critical fitness characteristics in poultry breeding programs, but their genetic evaluation and improvement present significant challenges. Firstly, the interpretation and selection on ratio values themselves can be problematic, as the same ratio can arise from vastly different numerator and denominator combinations, and small changes in the denominator can drastically alter the ratio. Secondly, these fitness-related ratio traits frequently exhibit non-normal distributions with outliers, asymmetry, violating the assumptions of traditional statistical methods and potentially leading to biased estimates if not properly transformed. Thirdly, limited sampling or incomplete data collection for these ratio traits, such as when only a subset of laid eggs is evaluated, can fail to accurately capture the true trait distribution, resulting in biased estimates of genetic parameters like heritability and breeding values, as well as potential reranking of selection candidates. Fourthly, the extreme skewness and prevalence of outliers associated with limited sampling exacerbates the challenges in obtaining unbiased estimates. Finally, the low heritability and substantial environmental influence on fitness traits render an animal's phenotypic performance an unreliable predictor of its underlying genetic merit, further increasing the likelihood of reranking across different environments or generations.
Thus, the first objective of is to see the effects of seven common data transformations (log, square root, probit, arcsine, logit, Box-Cox, and Yeo-Johnson) on genetic parameter estimates for the non-normal fitness traits FERT, PEP, and HOF were evaluated. Transformations significantly impacted heritability estimates and animal re-ranking, with effects varying by trait and transformation method. The Box-Cox transformation performed best based on the Akaike Information Criterion, resulting in higher heritabilities and more re-ranking compared to other methods. These findings highlight the importance of carefully considering data transformation approaches prior to genetic analysis of non-normal fitness traits in turkeys and other species.
Second objective is to address the situation where hens experience limited egg laying opportunities, the ratio traits used to measure these reproductive characteristics can become severely biased, inflating residual variances and reducing heritability estimates by up to 50%. This study explored imputation methods and efficient estimators to improve genetic parameter estimation for ratio reproductive traits under limited sampling in turkeys. Simulations mimicking limited egg production sampling showed that imputation alone provided modest 1-2% improvements in heritability estimates, while Hanushek's method accounting for sampling variance yielded more substantial 15-130% improvements. Combining imputation with Hanushek's efficient estimators further enhanced heritability estimates by 23-270%. Rank correlations of breeding values were generally improved for the highest ranked animals using the integrated approach. Across 30% limited sampling, this combined method maintained heritability around 0.11 for fertility and 0.22 for hatchability. By leveraging imputation and efficient estimators, the integrated approach shows potential to enhance genetic progress for lowly heritable reproductive ratio traits in turkey breeding programs impacted by limited phenotypic sampling. The results provide a framework for addressing challenges with ratio trait analysis to improve selection accuracy.
One way to enhance the prediction and effectiveness of genomic selection (GS) is to optimize the genetic relationship matrix (G matrix) used to model the association between phenotypic traits and genotypic data. Specifically, the study aims to improve upon the current methodological approaches by incorporating SNP prioritization based on the fixation index (FST) statistic. Genomic selection (GS) has the potential to improve the proportion of genetic variance explained over traditional pedigree-based approaches, but its effectiveness is limited by the sparse distribution of markers and incomplete understanding of the specific loci influencing quantitative traits. This section of study evaluated the use of the fixation index (FST) to prioritize and select informative single nucleotide polymorphisms (SNPs) for incorporation into GS models for reproductive traits in turkeys. Phenotypic data on percent egg production (PEP), rate of fertile eggs (FERT), and hatch of fertile eggs (HOF) from 2009-2018 were combined with a 48,008 SNP panel. The FST statistic was used to identify loci showing high genetic differentiation between subpopulations representing the phenotypic extremes for each trait. Heritability estimates were compared across pedigree-based BLUP, standard genomic BLUP using all SNPs, and hybrid models including prioritized SNP subsets determined by FST alongside polygenic effects. Including prioritized SNPs based on a 90th percentile FST threshold increased heritability estimates by 16-48% compared to using all SNPs. The FST approach enabled capturing major QTLs while substantially reducing the number of SNPs required. This prioritization strategy leveraging FST could be combined with other methods to further enhance the precision of genomic selection models in turkeys and other livestock species.
INDEX WORDS: Genetic evaluation, transformation, Simulation, fitness, ratio traits, FST