Statistical genetics tools for empowered data-driven decisions

The pressure to accelerate results in plant breeding programs is intensifying. Conversely, there is a concerning decline in the genetic diversity of staple crops, making it increasingly difficult to achieve genetic gains. Consequently, efficient resource allocation within breeding programs requires the strategic implementation of statistical genetics tools. This shift necessitates data-driven decision-making, placing professionals proficient in this toolkit at a significant advantage for addressing both traditional and emerging challenges. This thesis serves as a practical demonstration of utilizing statistical genetics in various plant breeding endeavours. Divided into six chapters, each with distinct objectives, the work showcases a range of applications. In Chapter 1, we determined the optimal number of harvests for selection in cacao breeding, considering both recommendation and recombination. Chapter 2 explores the application of covariance structure modelling in two common scenarios of perennial plant breeding: multi-harvest and multi-site data analysis. Chapter 3 demonstrates the use of factor analytic mixed models in maize breeding, including the incorporation of selection tools for streamlined decision-making. Notably, this chapter highlights the advantage of seasonal selection for achieving greater genetic gains compared to a combined approach. In Chapter 4, we evaluated the efficacy of the reciprocal recurrent selection (RRS) scheme within a eucalyptus breeding program. This chapter acknowledges the extended timeframe associated with RRS but also demonstrates its success in enhancing the hybrid population. Additionally, the chapter emphasizes the importance of considering dominance effects during the selection process. Chapter 5 offers a comprehensive tutorial on conducting linear mixed model analyses in perennial plant breeding. The chapter covers various analyses, including individual trials, multi- environment trials, spatial analysis, and competition analysis. Finally, Chapter 6 introduces the R package ProbBreed, which utilizes Bayesian principles and probabilistic concepts to support selection in multi-environment trials. ProbBreed estimates the risk associated with selecting candidates, empowering more informed decision-making. This chapter also introduces a novel multi-location-year model and compares the outcomes of ProbBreed and ASReml-R using simulated data. By showcasing the applications of statistical genetics tools and facilitating knowledge sharing through open-source code and reproducible examples, this thesis emphasizes the versatility and importance of this field in tackling diverse challenges within the dynamic field of plant breeding. Keywords: Data Analysis. Linear Mixed Models. Bayesian models. Genotype-by-Environment interaction. Spatial Analysis. Reciprocal Recurrent Selection

Grantor dc:publisher

Universidade Federal de Viçosa

Year dc:date.issued

2024

Author dc:creator

• Chaves, Saulo Fabrício da Silva

Advisor dc:contributor.advisor

• Dias, Luiz Antônio dos Santos

Contributors dc:contributor

• Dias, Kaio Olimpio das Graças
• Alves, Rodrigo Silva

• Plantas - Melhoramento genético
• Genética quantitativa

Statement dc:rights

• Acesso Aberto

Language dc:language.iso

eng

DOI

https://doi.org/10.47328/ufvbbt.2024.122

OAI identifier oai:identifier

oai:locus.ufv.br:123456789/32302