Non-additive genomic models for the evaluation of dairy cows

Author(s)

Fabricio Pilonetto

• Document type: Doctoral Thesis
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2021-09-29

Examining board members

Gerson Barreto Mourão; Luiz Lehmann Coutinho; Renato Ribeiro de Lima; Renata Veroneze

• Advisor: Gerson Barreto Mourão

Abstract

In dairy farming, the interest in selecting animals with high genetic potential for milk production and quality may cause a decrease in phenotypic and genotypic variability, due to intense selective pressure. Furthermore, genetic evaluations have been limited to additive genetic effects, not including dominance effects in genomic prediction models, for example. The aim of this study was to verify the contribution of non-additive genetic effects in genomic prediction models and to infer molecular heterozygosity and genomic heterosis. Two main lines of research were followed. Two main lines of research were followed. Initially, the ability and bias in predicting genomic breeding values by a purely additive model and an additive-dominant model were tested. The models were compared for ability and bias to genomic estimated breeding values (GEBVs) and the estimates of variance computed and evaluated according to the proposed model. The second phase involved inferring genomic heterozygosity and heterosis, and performing an adaptation of genome-wide association (GWAS) to capture non-additive variances, identify regions of the genome with possible heterozygous advantage, and molecular markers associated with the phenotype. In both phases, simulated data and real phenotypic records of Holstein dairy cattle were used. The proportions of dominance variance in relation to phenotype were 2.9 to 5.2% in simulated data, and from 0.08 to 3.8% for the milk yield and quality traits. The highest accuracy was 0.79 and 0.36 in the simulated additive and additivedominant model. The estimates were similar across models for the real data. Higher variance estimates were found for fat and fatty acid, with a greater advantage for fatty acid when dominance effects were included in the assessment. On average, for the lower heritability scenario, heterozygosity was 35.6%, and similar levels were found for the other scenarios (35.4 and 35.7%). The high heritability scenario presented the highest mean values of genomic heterosis (17.0%), compared to the others (3.02 to 11.0%). Regression coefficients revealed that heterozygosity has a positive impact on genomic heterosis for most simulated populations. The heterozygosity evaluated empirically and through real data revealed optimistic results regarding the possibility of selecting purebred animals to genomic heterosis. Including dominance effects in the adapted GWAS model can contribute to the identification of markers associated with the phenotype in a similar or superior way to the adapted additive model. We conclude that accounting for dominance effects in genomic prediction models should benefit better estimates of genetic variance components, especially for traits with lower heritability and highly influenced by the environment. Furthermore, it is suggested that genomic heterosis in purebred populations can be accessed if molecular heterozygosity factors and their associations with the phenotype of interest are considered. (AU)

• FAPESP's process: 19/03373-9 - Non-additive genomic models for the evaluation of dairy cows
• Grantee: Fabrício Pilonetto
• Support Opportunities: Scholarships in Brazil - Doctorate