OATAO - Open Archive Toulouse Archive Ouverte Open Access Week

Study of the origin and impact of bias in national genetic and genomic evaluations : the case of dairy sheep.

Macedo Fajardo, Fernando. Study of the origin and impact of bias in national genetic and genomic evaluations : the case of dairy sheep. PhD, Pathologie, Toxicologie, Génétique et Nutrition, Institut National Polytechnique de Toulouse, 2021

[img]
Preview
(Document in English)

PDF (Author's version) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
5MB

Abstract

In 2018 a new method called LR method was described to estimate bias in genetic evaluations. The method is based on comparing successive EBVs of a group of individuals, incrementing the information each time. It has estimators for bias, slope, and accuracies. The objective of this thesis were to evaluate bias in French dairy sheep evaluations, considering the effect of selection, using the LR method. Our first study was to check the capability of method LR to detect bias under simulated scenarios. We simulated a dairy species breeding scheme for low (0.10) and moderate (0.30) heritabilities. In both cases, we checked behavior of method LR for 3 scenarios: genetic evaluations (1) with a correct model, (2) when the evaluation model uses an incorrect heritability, and (3) when the data includes an environmental trend. For scenarios in which the evaluation model was correct, the LR method was capable of correctly estimating bias, slope, and accuracies. In cases of incorrect heritabilities in the evaluation model, the bias was correctly estimated in direction but not in magnitude. Similarly, the magnitudes of bias and slope were correctly estimated in scenarios with environmental trends in data, except for cases in which the model was unable to correctly estimate this environmental trend. In general, accuracies were well estimated in all scenarios. In this first study, we demonstrated that the LR method could estimate bias and accuracy in all cases if the evaluation model is correct or robust. The second study was to check the bias over real data. We used the genetic evaluation of milk yield in Manech Tête Rousse. We compared BLUP and single-step genomic BLUP (SSGBLUP) models. To account for missing pedigree, we used unknown parent groups (UPG), UPG with QP transformation in the H matrix (EUPG), and metafounders (MF) to manage the missing pedigree. We applied the LR method using the (G)EBVs of rams born in 2005 until 2014, with their (G)EBVs of the subsequent genetic evaluations. So, we obtained multiple estimators for the same model. We also compared across models: using EBVs of the first evaluation with BLUP and the subsequent GEBVs from SSGBLUP and comparing EBVs at birth with and without genomic information. Within models, in general, bias and overdispersion were small (bias: 0.20 to 0.40 genetic standard deviations; slope: 0.95 to 0.99). The estimates of accuracies confirmed that the addition of genomic information increases the accuracy of EBV in young rams. When we estimated dispersion by comparing a model with no markers to models with markers, SSGBLUPMF showed a value close to 1. In contrast, SSGBLUP-EUPG and SSGBLUP-UPG showed a significant underdispersion. Another important observation was the heterogeneous behavior of the estimates over time, which suggests that a single check could be insufficient to make a good analysis of genetic/genomic evaluations. In the third study, we analyzed the evolution of the genetic variance in Manech Tête Rousse for elite males and for all females (more representative of the population) born in each year from 1981 to 2014. We distinguished the loss of genetic variance due to drift (build-up of coancestry) and selection (mainly due to Bulmer effect). In both groups, the genetic variance decreased from the middle '90s until it reaches a stabilization around 2005. However, due to changes in the breeding objectives, it is impossible to conclude the real cause of the observed stabilization. The Bulmer effect impacted the loss of genetic variance of about 10%, while the build up of coancestry was about 3%.

Item Type:PhD Thesis
Uncontrolled Keywords:
Institution:Université de Toulouse > Institut National Polytechnique de Toulouse - Toulouse INP (FRANCE)
Laboratory name:
Research Director:
Legarra, Andrés
Statistics:download
Deposited On:06 Dec 2021 17:34

Repository Staff Only: item control page