New strategies to detect and understand genotype-by-environment interactions and QTL-by-environment interactions

Dissertação para obtenção do Grau de Doutor em Estatística e Gestão do Risco, especialidade em Estatística

Genotype-by-environment interaction (GEI) is frequent in multi-environment trials, and represents differential responses of genotypes across environments. With the development of molecular markers and mapping techniques, researchers can go one step further and analyse the whole genome to detect specific locations of genes which influence a quantitative trait such as yield. These locations are called quantitative trait locus (QTL), and when these QTLs have different expression across environments we talk about QTLby-environment interactions (QEI), which is the base of GEI. Good understandings of these interactions enable researchers to select better genotypes across different environmental conditions and, consequently, to improve crops in developed and developing countries. In this thesis I intend to present new strategies to improve detection and better understanding of QTLs, especially those exhibiting QEI in the context of multi-environment trials, by using and providing open source software. The first part of this thesis presents a comparison between two of the most used methods to analyse and to structure GEI: the joint regression analysis (JRA) and the additive main effects and multiplicative interaction (AMMI) model. This comparison is made in terms of “robustness” with different incidence rates of missing values, and in terms of dominant/winner genotypes. In the following chapters two- and threestages approaches are presented in which the AMMI model is used to gain accuracy in the phenotypic data, and their scores used to order the environments to find ecological or biological patterns. The first approach (two stages) is appropriated when the error variance is constant across environments, whereas the second (three stages) is more general and accounts for differences in the error variances by using the proposed weighted AMMI model (WAMMI). The final part of the thesis illustrates a strategy to simulate and to model GEI and QEI in complex traits, with the example of yield, based on a number of physiological parameters purely genotype dependent. This is done by using an eco-physiological genotype-to-phenotype model with seven parameters defined with a simple QTL basis.

Fundação para a Ciência e Tecnologia - SFRH/BD/35994/2007; project N N310 447838 supported by Ministry of Science and Higher Education, Poland.