Quantitative genetics aims to bridge the genome to phenome gap. The advent of high-throughput genotyping technologies has accelerated the progress of genome to phenome mapping, but a challenge remains in phenotyping. Various high-throughput phenotyping (HTP) platforms have been developed recently to obtain economically important phenotypes in an automated fashion with less human labor and reduced costs. However, the effective way of designing HTP has not been investigated thoroughly. In addition, high-dimensional HTP data bring up a big challenge for statistical analysis by increasing computational demands. A new strategy for modeling high-dimensional HTP data and elucidating the interrelationships among these phenotypes are needed. Previous studies used pedigree-based connectetdness statistics to study the design of phenotyping. The availability of genetic markers provides a new opportunity to evaluate connectedness based on genomic data, which can serve as a means to design HTP. This dissertation first discusses the utility of connectedness spanning in three studies. In the first study, I introduced genomic connectedness and compared it with traditional pedigree-based connectedness. The relationship between genomic connectedness and prediction accuracy based on cross-validation was investigated in the second study. The third study introduced a user-friendly connectedness R package, which provides a suite of functions to evaluate the extent of connectedness. In the last study, I proposed a new statistical approach to model high-dimensional HTP data by leveraging the combination of confirmatory factor analysis and Bayesian network. Collectively, the results from the first three studies suggested the potential usefulness of applying genomic connectedness to design HTP. The statistical approach I introduced in the last study provides a new avenue to model high-dimensional HTP data holistically to further help us understand the interrelationships among phenotypes derived from HTP. / Doctor of Philosophy / Quantitative genetics aims to bridge the genome to phenome gap. With the advent of genotyping technologies, the genomic information of individuals can be included in a quantitative genetic model. A new challenge is to obtain sufficient and accurate phenotypes in an automated fashion with less human labor and reduced costs. The high-throughput phenotyping (HTP) technologies have emerged recently, opening a new opportunity to address this challenge. However, there is a paucity of research in phenotyping design and modeling high-dimensional HTP data. The main themes of this dissertation are 1) genomic connectedness that could potentially be used as a means to design a phenotyping experiment and 2) a novel statistical approach that aims to handle high-dimensional HTP data. In the first three studies, I first compared genomic connectedness with pedigree-based connectedness. This was followed by investigating the relationship between genomic connectedness and prediction accuracy derived from cross-validation. Additionally, I developed a connectedness R package that implements a variety of connectedness measures. The fourth study investigated a novel statistical approach by leveraging the combination of dimension reduction and graphical models to understand the interrelationships among high-dimensional HTP data.

L'essor des marqueurs moléculaires (SNPs) a révolutionné les méthodes de génétique quantitative en permettant l'identification de régions impliquées dans le déterminisme génétique des caractères (QTLs) via la génétique d'association (GWAS), ou encore la prédiction des performances d'individus sur la base de leur information génomique (GS). La stratification des populations en groupes génétiques est courante en sélection animale et végétale. Cette structure peut impacter les méthodes de GWAS et de GS via des différences de fréquence et d'effets des allèles des QTL, ainsi que par des différences de déséquilibre de liaison (LD) entre SNP et QTL selon les groupes.Pendant cette thèse, deux panels de diversité de maïs ont été utilisés, présentant des niveaux différents de structuration: le panel “Amaizing Dent” représentant les lignées dentées utilisées en Europe et le panel “Flint-Dent” incluant des lignées dentées, cornées européennes, ainsi que des lignées admixées entre ces deux groupes.En GS, l'impact de la structure génétique sur la qualité des prédictions a été évalué au sein du premier panel pour des caractères de productivité et de phénologie. Cette étude a mis en évidence l'intérêt d'une population d'entraînement (TS) dont la constitution en matière de groupes génétiques est similaire à celle de la population à prédire. Assembler les différents groupes au sein d'un TS multi-groupe apparaît comme une solution efficace pour prédire un large spectre de diversité génétique. Des indicateurs a priori de la précision des prédictions génomiques, basés sur le coefficient de détermination, ont également été évalués, mettant en évidence une efficacité variable selon le groupe et le caractère étudié.Une nouvelle méthodologie GWAS a ensuite été développée pour étudier l'hétérogénéité des effets capturés par les SNPs selon les groupes. L'intégration des individus admixés à l'analyse permet de séparer les effets des facteurs responsables de l'hétérogénéité des effets alléliques: différence génomique locale (liée au LD ou à une mutation spécifique d'un groupe) ou interactions épistatiques entre le QTL et le fonds génétique. Cette méthodologie a été appliquée au panel “Flint-Dent” pour la précocité de floraison. Des QTL ont été détéctés comme présentant des effets groupe-spécifiques interagissant ou non avec le fonds génétique. De nombreux QTL présentant un profil original ont pu être mis en évidence, incluant des locus connus tels que Vgt1, Vgt2 ou Vgt3. Une importante épistasie directionnelle a aussi été mise en évidence grâce aux individus admixés, confortant l'existence d'interactions épistatiques avec le fonds génétique pour ce caractère.Sachant l'existence de cette hétérogénéité d’effets alléliques, nous avons développé deux modèles de prédictions génomiques nommées Multi-group Admixed GBLUP (MAGBLUP). Ceux-ci modélisent des effets groupe-spécifiques aux QTLs et sont adaptés à la prédiction d'individus admixés. Le premier permet d'identifier la variance génétique additionnelle créée par l'admixture (variance de ségrégation), alors que le second permet d'évaluer le degré de conservation des effets alléliques entre groupes. Ces deux modèles ont montré un intérêt certain par rapport à des modèles standards pour prédire des caractères simulés, mais plus limité sur des caractères réels.Enfin, l'intérêt des individus admixés dans la constitution de TS multi-groupes a été évalué à l'aide du second panel. Si leur intérêt a clairement été mis en évidence pour des caractères simulés, des résultats plus variables ont été observés avec les caractères réels, pouvant s'expliquer par la présence d'interactions avec le fonds génétique.Les nouvelles méthodes et l'utilisation d'individus admixés ouvrent des pistes de recherches intéressantes pour les études de génétique quantitative en population structurée. / The advent of molecular markers (SNPs) has revolutionized quantitative genetics methods by enabling the identification of regions involved in the genetic determinism of traits (QTLs) thanks to association studies (GWAS), or the prediction of the performance of individuals using genomic information (GS). The stratification of populations into genetic groups is common in animal and plant breeding. This structure can impact GWAS and GS methods through group differences in QTL allele frequencies and effects, as well as in linkage disequilibrium (LD) between SNP and QTL.During this thesis, two maize diversity panels were used, presenting different levels of structuration: the "Amaizing Dent" panel representing the diversity of dent lines used in Europe and the "Flint-Dent" panel including dent, flint and admixed lines between these two groups.In GS, the impact of genetic structure on genomic prediction accuracy was evaluated in the first panel for productivity and phenology traits. This study highlighted the interest of a training population (TS) whose constitution in terms of genetic groups is similar to that of the population to be predicted. Assembling the different groups within a multi-group TS appears as an effective solution to predict a broad spectrum of genetic diversity. A priori indicators of genomic prediction accuracy, based on the coefficient of determination, were also evaluated and highlighted a variable efficiency depending on the group and the trait.A new GWAS methodology was then developed to study the heterogeneity of the allele effects captured by SNPs depending on the group. The integration of admixed individuals to such analyses allows to disentangle the factors causing the heterogeneity of allele effects across groups: local genomic difference (related to LD or group-specific mutation) or epistatic interactions between the QTL and the genetic background. This methodology was applied to the "Flint-Dent" panel for flowering time. QTLs have been detected as presenting group-specific effects interacting or not with the genetic background. QTLs with an original profile have been highlighted, including known loci such as Vgt1, Vgt2 or Vgt3. Significant directional epistasis has also been demonstrated using admixed individuals and supported the existence of epistatic interactions with the genetic background for this trait.Based on the existence of such heterogeneity of allele effects, we have developed two genomic prediction models named Multi-group Admixed GBLUP (MAGBLUP). Both model group-specific QTL effects and are suited to the prediction of admixed individuals. The first allows the identification the additional genetic variance created by the admixture (segregation variance), while the second allows the evaluations of the degree of conservation of SNP allele effects across groups. These two models showed a certain interest compared to standard models to predict simulated traits, but it was more limited on real traits.Finally, the interest of admixed individuals in multi-group TS was evaluated using the second panel. Although their interest has been clearly demonstrated for simulated traits, more variable results have been observed with the real traits, which can be explained by the presence of interactions with the genetic background.The new methods and the use of admixed individuals open interesting lines of research for quantitative genetics studies in structured population.

L'objectif général de cette thèse est d'identifier les variants causaux ou, à défaut, un ensemble de marqueurs prédictifs - qui présentent un déséquilibre de liaison élevé avec les variants causaux - pour la fertilité des vaches laitières. Nous avons abordé cet objectif général dans cinq articles: (i) décrit une approche systématique de cartographie des variants létaux récessifs chez les bovins Normands français basée sur la recherche de déficit en haplotypes homozygotes (HHD). Cette étude montre l’influence de la taille de l’échantillon, de la qualité des génotypes, de la qualité du phasage des génotypes en haplotypes et de l’imputation, de l’âge de l’haplotype et enfin, de la définition des seuils de signification prenant en compte les tests multiples, sur la découverte et la reproductibilité des résultats de HHD. Elle illustre également l’importance de la cartographie fine avec les données de généalogie et de séquence de génome entier (WGS), l’annotation intégrative (entre espèces) pour hiérarchiser les mutations candidates et, enfin, le génotypage à grande échelle de la mutation candidate, pour valider ou invalider les mutations initiales. (ii) décrit une cartographie à haute résolution de grandes délétions chromosomiques de séquences du génome dans une population de 175 animaux appartenant à trois races laitières nordiques. Cette étude utilise trois approches différentes pour valider les résultats de la cartographie. Le chapitre décrit les propriétés génétiques des populations et l’importance fonctionnelle des délétions identifiées. (iii) traite de trois questions liées à l’imputation de variants structuraux, ici de délétions chromosomiques importantes: la disponibilité des génotypes de délétion, la taille du panel de référence d'haplotypes et, enfin, l’imputation elle-même. Pour aborder les deux premières questions, cette étude décrit une approche basée sur un modèle de mélange gaussien dans laquelle les données de profondeur de lecture provenant de fichiers au format VCF (variant call format) sont utilisées pour génotyper un locus de délétion connu, en l’absence d’information sur la séquence brute. Enfin, il présente un pipeline pour l'imputation conjointe de variants WGS et de grandes délétions chromosomiques. (iv) décrit des études d'association pangénomiques de la fertilité femelle dans trois races de bovins laitiers nordiques à l'aide de variants WGS imputés et de grandes délétions chromosomiques. Cette étude concerne huit caractères de fertilité et utilise des analyses d'association mono-marqueur, conditionnelles et conjointes. Cette étude montre qu’une surestimation, ou « inflation », des statistiques de test peut être observée même après correction pour la stratification de la population à l'aide de composantes principales génomiques et pour les structures familiales à l'aide de matrices de relations génomiques. Ce biais était connu pour les caractères très polygéniques. Enfin, cette étude présente plusieurs locus de traits quantitatifs (QTL) nouveaux et confirme plusieurs autres déjà connus. Elle souligne également l’importance d’inclure les grandes délétions (imputées) pour la cartographie par association des caractères de fertilité. (v) décrit la prédiction des valeurs génomiques de fertilité (ou indice de fertilité) à l'aide de génotypes à puces SNP, de QTL sélectionnés et de délétions chromosomiques importantes. En utilisant la méthode de meilleure prédiction linéaire sans biais génomique (GBLUP) avec une ou plusieurs matrices de relations génomiques dérivées d'un ensemble de marqueurs sélectionnés, cette étude rapporte une précision de prédiction améliorée. Cette étude met également en évidence l’influence de la sélection des marqueurs les plus prédictifs, en particulier pour une race ayant une population d’apprentissage réduite, sur la précision des prédictions génomiques. Enfin, les résultats démontrent que les grandes délétions ont en général un pouvoir prédictif élevé. / The overall aim of this PhD thesis is to identify causal variants for recessive lethal mutations and select a set of predictive markers that are in high linkage-disequilibrium with the causal variants for female fertility in dairy cattle. We addressed this broad aim under five articles: (i) describes a systematic approach of mapping recessive lethals in French Normande cattle using homozygous haplotype deficiency (HHD). This study shows the influence of sample size, quality of genotypes, quality of (genotype) phasing and imputation, age of haplotype (of interest), and last but not the least, multiple testing corrections, on discovery and replicability of HHD results. It also illustrates the importance of fine-mapping with pedigree and whole-genome sequence (WGS) data, (cross-species) integrative annotation to prioritize candidate mutation, and finally, large-scale genotyping of the candidate mutation, to validate or invalidate initial results. (ii) describes a high-resolution population-scale mapping of large chromosomal deletions from whole-genome sequences of 175 animals from three Nordic dairy breeds. This study employs three different approaches to validate identified deletions. Next, it describes population genetic properties and functional importance of these deletions. (iii) deals with three main issues related to imputation of structural variants, in this case, large chromosomal deletions, e.g. availability of deletion genotypes, size of haplotype reference panel, and finally, imputation itself. To address the first two issues, this study describes a Gaussian mixture model-based approach where read-depth data from the variant call format (VCF) file is used to genotype a known deletion locus, without the need for raw sequence (BAM) file. Finally, it presents a pipeline for joint imputation of WGS variants along with large chromosomal deletions. (iv) describes genome-wide association studies for female fertility in three Nordic dairy cattle breeds using imputed WGS variants including large chromosomal deletions. This study is based on the analyses of eight fertility related traits using single-marker association, conditional and joint analyses. This study illustrates that inflation in association test-statistics could be seen even after correcting for population stratification using (genomic) principal components, and relatedness among the samples using genomic relationship matrices; however, this was known for traits with strong polygenic effects, among other factors. Finally, mapping of several new quantitative trait loci (QTL), along with the previously known ones, are reported in this study. This study also highlights the importance of including (imputed) large deletions for association mapping of fertility traits. (v) describes prediction of genomic breeding values for fertility using SNP array-chip genotypes, selected QTL and large chromosomal deletion. Using genomic best linear unbiased prediction (GBLUP) method with one or several genomic-relationship matrices derived from a set of selected markers, this study reports higher prediction accuracy compared with previous report. This study also highlights the influence of selecting markers with best predictability, especially for a breed with small training population, in accuracy of genomic prediction. The results demonstrate that large deletions in general have a high predictive performance.

No description available.

In multiple species, genome-wide association (GWA) studies have become an increasingly prevalent method of identifying the quantitative trait loci (QTLs) that underlie complex traits. Despite this, relatively few GWA analyses using high-density genomic markers have been carried out on highly quantitative traits in wheat. We utilized single-nucleotide polymorphism (SNP) data generated via a genotyping-by-sequencing (GBS) protocol to perform GWA on multiple yield-related traits using a panel of 329 soft red winter wheat genotypes grown in four environments. In addition, the SNP data was used to examine linkage disequilibrium and population structure within the testing panel. The results indicated that an alien translocation from the species Triticum timopheevii was responsible for the majority of observed population structure. In addition, a total of 50 significant marker-trait associations were identified. However, a subsequent study cast some doubt upon the reproducibility and reliability of plant QTLs identified via GWA analyses. We used two highly-related panels of different genotypes grown in different sets of environments to attempt to identify highly stable QTLs. No QTLs were shared across panels for any trait, suggesting that QTL-by-environment and QTL-by-genetic background interaction effects are significant, even when testing across many environments. In light of the challenges involved in QTL mapping, prediction of phenotypes using whole-genome marker data is an attractive alternative. However, many evaluations of genomic prediction in crop species have utilized univariate models adapted from animal breeding. These models cannot directly account for genotype-by-environment interaction, and hence are often not suitable for use with lower-heritability traits assessed in multiple environments. We sought to test genomic prediction models capable of more ad-hoc analyses, utilizing highly unbalanced experimental designs consisting of individuals with varying degrees of relatedness. The results suggest that these designs can successfully be used to generate reasonably accurate phenotypic predictions. In addition, multivariate models can dramatically increase predictive accuracy for some traits, though this depends upon the quantity and characteristics of genotype-by-environment interaction. / Ph. D. / Quantitative traits are those traits that can display a wide range of variability within a population of individuals. These traits are influenced by the interaction of many different genes, and are also influenced by the environment to varying degrees. Traditionally, geneticists who studied quantitative traits had to rely on statistical models, while the biological causes of variation in the expression of these traits remained largely unknown. However, the advent of DNA marker technology granted geneticists the ability to identify specific regions of the genome that highly influence quantitative traits. Many studies have since attempted to find these <i>quantitative trait loci</i> (QTLs) across a wide range of traits and species. However, we are faced with something of a paradox when we attempt to find QTLs. Theory tells us that an idealized, truly quantitative trait arises due to the effects of many genes, each with an infinitesimal effect on the trait in question. Therefore, the more quantitative a trait, the fewer QTLs we should expect to find. In addition, QTLs may not be reliable, due to the effects of different environments and different genetic backgrounds within a population. A more recent trend involves using all available marker data simultaneously to predict a particular line’s performance. This method entails ignoring the genomic underpinnings of a trait, and instead focusing solely on its expression, much like classical quantitative genetics. The obvious downside of this method is that it cannot be used to increase our understanding of what is giving rise to the variations in the trait’s expression that we observe. The studies described in this dissertation were designed to 1) test whether we could identify QTLs for highly quantitative yield-related traits in winter wheat, 2) test the reliability of identified QTLs, and 3) use the DNA marker data to instead generate predictions of line performance. The results indicate that while we can identify QTLs for highly quantitative traits in winter wheat, these QTLs may not be very reliable. Therefore, predictive models may be a good alternative to identifying QTLs, and these methods can be readily implemented within breeding programs.

Métodos de predição podem aumentar consideravelmente a eficiência dos programas de melhoramento de milho. O objetivo deste estudo foi predizer a performance de 250 híbridos simples de milho avaliados em múltiplos ambientes utilizando a informação de marcadores moleculares. Para isso, 50 linhagens endogâmicas provenientes de diferentes populações foram cruzadas com cinco linhagens elite, também endogâmicas, para obtenção dos 250 híbridos simples. As matrizes moleculares das linhagens e dos híbridos foram obtidas a partir da genotipagem das 55 linhagens com 614 marcadores AFLP. Os híbridos simples foram avaliados para produção de grãos em 13 ambientes. A predição dos híbridos foi realizada utilizando o modelo misto BLUP considerando diferentes coeficientes de parentesco e similaridade no estado na predição dos efeitos das capacidades geral e específica de combinação dos genitores. As médias preditas dos híbridos a partir de cada coeficiente foram correlacionadas com as médias fenotípicas para obtenção da acurácia de predição. A predição também foi realizada utilizando o modelo de seleção genômica ampla RR-BLUP. Nesse caso, a matriz molecular dos híbridos foi utilizada diretamente no modelo misto de estimação dos efeitos dos marcadores e da contribuição de cada um deles para o valor genético dos híbridos. Foram realizadas validações cruzadas entre e dentro de ambientes e entre e dentro de grupos de híbridos relacionados a fim de verificar os efeitos do tamanho da população de treinamento (N), número de marcas (NM), interação híbridos x ambientes (H x A) e da estrutura da população na estimativa da acurácia de predição. A predição genômica foi comparada com a seleção fenotípica quanto à eficiência em identificar híbridos superiores em um esquema de melhoramento de milho. Todos os coeficientes de parentesco e similaridade no estado apresentaram elevadas estimativas de acurácia, contudo foi possível observar considerável superioridade dos coeficientes Wang e Rogers Modificado tanto na predição quanto na seleção dos híbridos superiores, demonstrando o potencial dessas metodologias como ferramentas a serem utilizadas nos programas de melhoramento de milho. Os resultados da predição utilizando o modelo de seleção genômica ampla indicaram que o aumento de N e NM não alteraram significativamente as estimativas de acurácia. As estimativas da acurácia na validação cruzada dentro de ambientes foram superiores às obtidas entre ambientes, inferindo que o efeito da interação H x A foi expressivo. Também foram observadas estimativas de acurácia expressivamente maiores para populações de treinamento e validação compostas por híbridos relacionados. Em todos os casos, as estimativas de acurácia apresentaram amplos intervalos em função da amostra de híbridos utilizada nas populações de treinamento e validação, indicando que a seleção genômica pode não ser eficiente dependendo da população amostrada. Os resultados deste estudo sugerem que a predição genômica é uma ferramenta para aumentar a eficiência da seleção nos programas de melhoramento se utilizada de forma adequada pelo melhorista, considerando os efeitos de estrutura de população e interação H x A de forma a maximizar a acurácia e, consequentemente, o sucesso da predição. / Prediction using molecular markers information can greatly increase the efficiency of maize breeding programs. This study aimed to predict the performance of maize single-crosses evaluated in multiple environments and using molecular markers information. Five inbred lines used as testers were crossed to 50 inbred lines from multiple populations to obtain 250 maize single-crosses. 614 AFLP markers were used to asses molecular matrices of the inbred lines and single-crosses. The 250 single-crosses were evaluated for grain yield in 13 environments. Genomic prediction was performed using the mixed model BLUP considering different genomic relationship and similarity in state coefficients to predict the effect of general and specific combining abilities of the parents. Predicted means from each coefficient were correlated with phenotypic means for obtaining prediction accuracy. Genomewide prediction was also performed using the linear regression model RR-BLUP in the estimation of markers genotypic values and its contribution to hybrids genetic values. Cross-validations between and within environments and between and within groups of related single-crosses were performed to verify the effects of training population size (N), number of markers (NM), genotype-by-environment interaction (G x E) and population structure in estimating accuracy. Genomic prediction was compared with phenotypic selection in efficiency of selecting better hybrids in a maize breeding program. All relationship coefficients and similarity in state coefficients showed high values of accuracy, however we observed superiority of Wang relationship coefficient and Modified Rogers similarity coefficient both in predicting and in identifying the best single-crosses, showing the potential of these approaches as tools to be used in maize breeding programs. Genomewide prediction results showed that increasing N and NM did not led to higher accuracy estimates. Predicted accuracies of cross validation analysis within environments were higher than between environments, indicating that the effect of G x E interaction was significant. Greater accuracies were achieved when training and validation set were from related single-crosses. In all scenarios, wide intervals of accuracy were found, meaning that genomic prediction may not be effective depending on the sample used. Therefore, the results of this study suggest that genomic prediction is a tool to increase the efficiency of selection in breeding programs if used properly by breeders, considering the population structure and G x E interaction effect so as to reduce sample problems and maximize accuracy and hence the success of prediction.

A predição da variância fenotípica é de grande importância para que os sistemas de produção de bovinos de corte consigam aumentar a rentabilidade otimizando o uso de recursos. Modelos mecanicistas dinâmicos do crescimento bovino vêm sendo utilizados como ferramentas de suporte à tomada de decisão em sistemas de manejo individual do gado. Entretanto, a aplicação desses modelos ainda fundamenta-se em parâmetros populacionais, sem qualquer abordagem para que se consiga capturar a variabilidade entre sujeitos nas simulações. Assumindo que modelos mecanicistas sejam capazes de simular o componente de desvio ambiental da variância fenotípica e considerando que marcadores SNPs possam predizer o componente genético dessa variância, esse projeto objetivou evoluir em direção a um modelo matemático que considere a variabilidade entre animais em seu nível genético. Seguindo conceitos de fisiologia genômica computacional, nós assumimos que a variância genética da característica complexa (i.e. produto do comportamento do modelo) surge de características componentes (i.e. parâmetros dos modelos) em níveis hierárquicos mais baixos do sistema biológico. Esse estudo considerou dois modelos mecanicistas do crescimento de bovinos - Cornell Cattle Value Discovery System (CVDS) e Davis Growth Model (DGM) - e ao questionar se os parâmetros de tais modelos mapeariam regiões genômicas que englobam QTLs já descritos para a característica complexa, verificou as suas interpretações biológicas esperadas. Tal constatação forneceu uma prova de conceito de que os parâmetros do CVDS e do DGM são de fato fenótipos cuja interpretação pode ser confirmada através das regiões genômicas mapeadas. Um método de predição genômica foi então utilizado para computar os parâmetros do CVDS e do DGM. Os efeitos dos marcadores SNPs foram estimados tanto para os parâmetros quanto para os fenótipos observados. Nós buscamos qual o melhor cenário de predição - simulações dos modelos com parâmetros computados a partir das informações genômicas ou predição genômica conduzida diretamente nos fenótipos complexos. Nós encontramos que enquanto a predição genômica dos fenótipos complexos pode ser uma melhor opção em relação aos modelos de crescimento, simulações conduzidas com parâmetros obtidos a partir de dados genômicos estão condizentes com simulações geradas com parâmetros obtidos a partir de métodos regulares. Esse é o principal argumento para chamar atenção da comunidade científica de que a abordagem apresentada nesse projeto representa um caminho para o desenvolvimento de uma nova geração de modelos nutricionais aplicados capazes de capturar a variabilidade genética entre bovinos de corte confinados e produzir simulações com variáveis de entrada específicas de cada genótipo. Esse projeto é a primeira abordagem no Brasil conhecida dos autores a usar genótipos Bos indicus para o estudo da aplicação de genômica integrada à modelos mecanicistas para o manejo e comercialização de animais na pecuária. / The prediction of phenotypic variance is important for beef cattle operations to increase profitability by optimizing resource use. Dynamic mechanistic models of cattle growth have been used as decision support tools for individual cattle management systems. However, the application of such models is still based on population parameters, with no further approach to capture between-subject variability. By assuming that mechanistic models are able to simulate environmental deviations components of phenotypic variance and considering that SNPs markers may predict the genetic component of this variance, this project aimed at evolving towards a mathematical model that takes between-animal variance to its genetic level. Following the concepts of computational physiological genomics, we assumed that genetic variance of the complex trait (i.e. outcome of model behavior) arises from component traits (i.e. model parameters) in lower hierarchical levels of biological systems. This study considered two mechanistic models of cattle growth - Cornell Cattle Value Discovery System (CVDS) and Davis Growth Model (DGM) - and verified their expected biological interpretation by asking whether model parameters would map genomic regions that harbors QTLs already described for the complex trait. This provided a proof of concept that CVDS and DGM parameters are indeed phenotypes whose expected interpretations may be stated by means of their mapped genomic regions. A method of genomic prediction to compute parameters for CVDS and DGM was then used. SNP marker effects were estimated both for their parameters and observed phenotypes. We looked for the best prediction scenario - model simulation with parameters computed from genomic data or genomic prediction on complex phenotypes directly. We found that while genomic prediction on complex phenotypes may still be a better option than predictions from growth models, simulations conducted with genomically computed parameters are in accordance with those performed with parameters obtained from regular methods. This is the main argument to call attention from the research community that this approach may pave the way for the development of a new generation of applied nutritional models capable of representing genetic variability among beef cattle under feedlot conditions and performing simulation with inputs from individual\'s genotypes. To our knowledge, this project is the first of this kind in Brazil and the first using Bos indicus genotypes to study the application of genomics integrated with mechanistic models for the management and marketing of commercial livestock.

Advances in molecular marker technologies have led to the development of high throughput genotyping techniques such as Genotyping by Sequencing (GBS), driving the application of genomics in crop research and breeding. They have also supported the use of novel mapping approaches, including Multi-parent Advanced Generation Inter-Cross (MAGIC) populations which have increased precision in identifying markers to inform plant breeding practices. In the first part of this thesis, a high density physical map derived from GBS was used to identify QTLs controlling key agronomic traits of wheat in a genome-wide association study (GWAS) and to demonstrate the practicability of genomic selection for predicting the trait values. The results from GBS were compared to a previous study conducted on the same association mapping panel using a less dense physical map derived from diversity arrays technology (DArT) markers. GBS detected more QTLs than DArT markers although some of the QTLs were detected by DArT markers alone. Prediction accuracies from the two marker platforms were mostly similar and largely dependent on trait genetic architecture. The second part of this thesis focused on MAGIC populations, which incorporate diversity and novel allelic combinations from several generations of recombination. Pedigrees representing a wild rice MAGIC population were used to model MAGIC populations by simulation to assess the level of recombination and creation of novel haplotypes. The wild rice species are an important reservoir of beneficial genes that have been variously introgressed into rice varieties using bi-parental population approaches. The level of recombination was found to be highly dependent on the number of crosses made and on the resulting population size. Creation of MAGIC populations require adequate planning in order to make sufficient number of crosses that capture optimal haplotype diversity. The third part of the thesis considers models that have been proposed for genomic prediction. The ridge regression best linear unbiased prediction (RR-BLUP) is based on the assumption that all genotyped molecular markers make equal contributions to the variations of a phenotype. Information from underlying candidate molecular markers are however of greater significance and can be used to improve the accuracy of prediction. Here, an existing Differentially Penalized Regression (DiPR) model which uses modifications to a standard RR-BLUP package and allows two or more marker sets from different platforms to be independently weighted was used. The DiPR model performed better than single or combined marker sets for predicting most of the traits both in a MAGIC population and an association mapping panel. Overall the work presented in this thesis shows that while these techniques have great promise, they should be carefully evaluated before introduction into breeding programmes.

Métodos de predição podem aumentar consideravelmente a eficiência dos programas de melhoramento de milho. O objetivo deste estudo foi predizer a performance de 250 híbridos simples de milho avaliados em múltiplos ambientes utilizando a informação de marcadores moleculares. Para isso, 50 linhagens endogâmicas provenientes de diferentes populações foram cruzadas com cinco linhagens elite, também endogâmicas, para obtenção dos 250 híbridos simples. As matrizes moleculares das linhagens e dos híbridos foram obtidas a partir da genotipagem das 55 linhagens com 614 marcadores AFLP. Os híbridos simples foram avaliados para produção de grãos em 13 ambientes. A predição dos híbridos foi realizada utilizando o modelo misto BLUP considerando diferentes coeficientes de parentesco e similaridade no estado na predição dos efeitos das capacidades geral e específica de combinação dos genitores. As médias preditas dos híbridos a partir de cada coeficiente foram correlacionadas com as médias fenotípicas para obtenção da acurácia de predição. A predição também foi realizada utilizando o modelo de seleção genômica ampla RR-BLUP. Nesse caso, a matriz molecular dos híbridos foi utilizada diretamente no modelo misto de estimação dos efeitos dos marcadores e da contribuição de cada um deles para o valor genético dos híbridos. Foram realizadas validações cruzadas entre e dentro de ambientes e entre e dentro de grupos de híbridos relacionados a fim de verificar os efeitos do tamanho da população de treinamento (N), número de marcas (NM), interação híbridos x ambientes (H x A) e da estrutura da população na estimativa da acurácia de predição. A predição genômica foi comparada com a seleção fenotípica quanto à eficiência em identificar híbridos superiores em um esquema de melhoramento de milho. Todos os coeficientes de parentesco e similaridade no estado apresentaram elevadas estimativas de acurácia, contudo foi possível observar considerável superioridade dos coeficientes Wang e Rogers Modificado tanto na predição quanto na seleção dos híbridos superiores, demonstrando o potencial dessas metodologias como ferramentas a serem utilizadas nos programas de melhoramento de milho. Os resultados da predição utilizando o modelo de seleção genômica ampla indicaram que o aumento de N e NM não alteraram significativamente as estimativas de acurácia. As estimativas da acurácia na validação cruzada dentro de ambientes foram superiores às obtidas entre ambientes, inferindo que o efeito da interação H x A foi expressivo. Também foram observadas estimativas de acurácia expressivamente maiores para populações de treinamento e validação compostas por híbridos relacionados. Em todos os casos, as estimativas de acurácia apresentaram amplos intervalos em função da amostra de híbridos utilizada nas populações de treinamento e validação, indicando que a seleção genômica pode não ser eficiente dependendo da população amostrada. Os resultados deste estudo sugerem que a predição genômica é uma ferramenta para aumentar a eficiência da seleção nos programas de melhoramento se utilizada de forma adequada pelo melhorista, considerando os efeitos de estrutura de população e interação H x A de forma a maximizar a acurácia e, consequentemente, o sucesso da predição. / Prediction using molecular markers information can greatly increase the efficiency of maize breeding programs. This study aimed to predict the performance of maize single-crosses evaluated in multiple environments and using molecular markers information. Five inbred lines used as testers were crossed to 50 inbred lines from multiple populations to obtain 250 maize single-crosses. 614 AFLP markers were used to asses molecular matrices of the inbred lines and single-crosses. The 250 single-crosses were evaluated for grain yield in 13 environments. Genomic prediction was performed using the mixed model BLUP considering different genomic relationship and similarity in state coefficients to predict the effect of general and specific combining abilities of the parents. Predicted means from each coefficient were correlated with phenotypic means for obtaining prediction accuracy. Genomewide prediction was also performed using the linear regression model RR-BLUP in the estimation of markers genotypic values and its contribution to hybrids genetic values. Cross-validations between and within environments and between and within groups of related single-crosses were performed to verify the effects of training population size (N), number of markers (NM), genotype-by-environment interaction (G x E) and population structure in estimating accuracy. Genomic prediction was compared with phenotypic selection in efficiency of selecting better hybrids in a maize breeding program. All relationship coefficients and similarity in state coefficients showed high values of accuracy, however we observed superiority of Wang relationship coefficient and Modified Rogers similarity coefficient both in predicting and in identifying the best single-crosses, showing the potential of these approaches as tools to be used in maize breeding programs. Genomewide prediction results showed that increasing N and NM did not led to higher accuracy estimates. Predicted accuracies of cross validation analysis within environments were higher than between environments, indicating that the effect of G x E interaction was significant. Greater accuracies were achieved when training and validation set were from related single-crosses. In all scenarios, wide intervals of accuracy were found, meaning that genomic prediction may not be effective depending on the sample used. Therefore, the results of this study suggest that genomic prediction is a tool to increase the efficiency of selection in breeding programs if used properly by breeders, considering the population structure and G x E interaction effect so as to reduce sample problems and maximize accuracy and hence the success of prediction.

A predição da variância fenotípica é de grande importância para que os sistemas de produção de bovinos de corte consigam aumentar a rentabilidade otimizando o uso de recursos. Modelos mecanicistas dinâmicos do crescimento bovino vêm sendo utilizados como ferramentas de suporte à tomada de decisão em sistemas de manejo individual do gado. Entretanto, a aplicação desses modelos ainda fundamenta-se em parâmetros populacionais, sem qualquer abordagem para que se consiga capturar a variabilidade entre sujeitos nas simulações. Assumindo que modelos mecanicistas sejam capazes de simular o componente de desvio ambiental da variância fenotípica e considerando que marcadores SNPs possam predizer o componente genético dessa variância, esse projeto objetivou evoluir em direção a um modelo matemático que considere a variabilidade entre animais em seu nível genético. Seguindo conceitos de fisiologia genômica computacional, nós assumimos que a variância genética da característica complexa (i.e. produto do comportamento do modelo) surge de características componentes (i.e. parâmetros dos modelos) em níveis hierárquicos mais baixos do sistema biológico. Esse estudo considerou dois modelos mecanicistas do crescimento de bovinos - Cornell Cattle Value Discovery System (CVDS) e Davis Growth Model (DGM) - e ao questionar se os parâmetros de tais modelos mapeariam regiões genômicas que englobam QTLs já descritos para a característica complexa, verificou as suas interpretações biológicas esperadas. Tal constatação forneceu uma prova de conceito de que os parâmetros do CVDS e do DGM são de fato fenótipos cuja interpretação pode ser confirmada através das regiões genômicas mapeadas. Um método de predição genômica foi então utilizado para computar os parâmetros do CVDS e do DGM. Os efeitos dos marcadores SNPs foram estimados tanto para os parâmetros quanto para os fenótipos observados. Nós buscamos qual o melhor cenário de predição - simulações dos modelos com parâmetros computados a partir das informações genômicas ou predição genômica conduzida diretamente nos fenótipos complexos. Nós encontramos que enquanto a predição genômica dos fenótipos complexos pode ser uma melhor opção em relação aos modelos de crescimento, simulações conduzidas com parâmetros obtidos a partir de dados genômicos estão condizentes com simulações geradas com parâmetros obtidos a partir de métodos regulares. Esse é o principal argumento para chamar atenção da comunidade científica de que a abordagem apresentada nesse projeto representa um caminho para o desenvolvimento de uma nova geração de modelos nutricionais aplicados capazes de capturar a variabilidade genética entre bovinos de corte confinados e produzir simulações com variáveis de entrada específicas de cada genótipo. Esse projeto é a primeira abordagem no Brasil conhecida dos autores a usar genótipos Bos indicus para o estudo da aplicação de genômica integrada à modelos mecanicistas para o manejo e comercialização de animais na pecuária. / The prediction of phenotypic variance is important for beef cattle operations to increase profitability by optimizing resource use. Dynamic mechanistic models of cattle growth have been used as decision support tools for individual cattle management systems. However, the application of such models is still based on population parameters, with no further approach to capture between-subject variability. By assuming that mechanistic models are able to simulate environmental deviations components of phenotypic variance and considering that SNPs markers may predict the genetic component of this variance, this project aimed at evolving towards a mathematical model that takes between-animal variance to its genetic level. Following the concepts of computational physiological genomics, we assumed that genetic variance of the complex trait (i.e. outcome of model behavior) arises from component traits (i.e. model parameters) in lower hierarchical levels of biological systems. This study considered two mechanistic models of cattle growth - Cornell Cattle Value Discovery System (CVDS) and Davis Growth Model (DGM) - and verified their expected biological interpretation by asking whether model parameters would map genomic regions that harbors QTLs already described for the complex trait. This provided a proof of concept that CVDS and DGM parameters are indeed phenotypes whose expected interpretations may be stated by means of their mapped genomic regions. A method of genomic prediction to compute parameters for CVDS and DGM was then used. SNP marker effects were estimated both for their parameters and observed phenotypes. We looked for the best prediction scenario - model simulation with parameters computed from genomic data or genomic prediction on complex phenotypes directly. We found that while genomic prediction on complex phenotypes may still be a better option than predictions from growth models, simulations conducted with genomically computed parameters are in accordance with those performed with parameters obtained from regular methods. This is the main argument to call attention from the research community that this approach may pave the way for the development of a new generation of applied nutritional models capable of representing genetic variability among beef cattle under feedlot conditions and performing simulation with inputs from individual\'s genotypes. To our knowledge, this project is the first of this kind in Brazil and the first using Bos indicus genotypes to study the application of genomics integrated with mechanistic models for the management and marketing of commercial livestock.