Submitted by HERMENEGILDO LUCAS JUSTINO CHIAIA null (chiaia1@yahoo.com.br) on 2017-10-26T12:35:23Z No. of bitstreams: 1 Tese_Hermenegildo_Lucas_Justino_Chiaia.pdf: 1161047 bytes, checksum: 270c67e433de8ecdefbd997478489876 (MD5) / Approved for entry into archive by Monique Sasaki (sayumi_sasaki@hotmail.com) on 2017-10-31T17:44:43Z (GMT) No. of bitstreams: 1 chiaia_hlj_dr_jabo.pdf: 1161047 bytes, checksum: 270c67e433de8ecdefbd997478489876 (MD5) / Made available in DSpace on 2017-10-31T17:44:43Z (GMT). No. of bitstreams: 1 chiaia_hlj_dr_jabo.pdf: 1161047 bytes, checksum: 270c67e433de8ecdefbd997478489876 (MD5) Previous issue date: 2017-10-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Existem diferentes métodos e pseudo-fenótipos utilizados em predições genômicas, sendo necessário determinar o ideal para a característica de interesse. Os ácidos graxos da carne que contribuem de forma prejudicial para a saúde humana têm recebido considerável atenção nos últimos anos. O objetivo desse estudo foi avaliar a acurácia de predição genômica de diferentes métodos (SNP-BLUP, BayesC, BayesCπ e Bayesian Lasso) e pseudo-fenótipos (fenótipo ajustado para os efeitos fixos, valor genético estimado via pedigree e valor genético genômico obtido pelo método single step GBLUP) em dados simulados e reais de perfil de ácidos graxos no músculo Longissimus thoracis da raça Nelore terminados em confinamento. A proposta de inclusão do GEBV obtido pelo método passo único genômico BLUP (ssGBLUP) como pseudofenótipo nesta pesquisa é inédita em predição genômica, para responder um dos problemas frequentes ao utilizar a matriz advinda da informação genealógica, pela maioria dos produtores de gado de corte utilizarem o sistema de acasalamento que utiliza reprodutores multiplos. Foram utilizados cerca de 963 bovinos machos inteiros da raça Nelore terminados em confinamento e genotipados com um painel de 777.962 SNPs do Ilumina BovineSNP BeadChip. A informação genômica pode servir para melhorar o perfil de ácidos graxos em animais do grupo Zebu, por ter-se observado valores genômicos preditos com com acurácia de baixa a moderada magnitude. Nenhum dos métodos foi melhor para todos os ácidos graxos em termos de acurácia, no entanto, o método SNP-BLUP permitiu realizar avaliação menos viesada. O método ssGBLUP apresentou-se como ferramenta alternativa para obter GBVs como pseudo-fenôtipo mais acurado em situações de pedigree incompleto, pela alta proporção de de reprodutores múltiplos, sendo mais apropriado que o EBV e o fenótipo ajustado aos efeitos fixos para predizer o valor genômico dos animais. / There are different methods and pseudo-phenotypes used in genomic predictions, being necessary to determine the ideal for each trait of interest. Beef fatty acids that contribute to human health have received considerable attention in the last years. Thus, the objective of this study was to evaluate genomic predition accuracy of different methodologies (SNP-BLUP, BayesCπ, BayesC and Bayesian Lasso) pseudo-phenotypes (adjusted phenotype, estimated breeding value and genomic estimated breeding value by ssGBLUP) in simulated and real data of fatty acid profile in the Longissimus thoracis muscle of Nelore cattle finished in feedlot. The inclusion of the GEBV obtained by single step genomic BLUP (ssGBLUP) method as pseudo-phenotype in this research is innovator in genomic prediction, to answer one of the frequent problems when using the matrix derived from pedigree, by the mating system that uses multiple sires for many cattle breeders. A total of 963 Nelore bulls with phenotype for fatty acid profiles, were genotyped using the Illumina Bovine HD Bead Chip with 777,962 SNPs. Genomic information can assist in improving fatty acid profile in Zebu animals, since the use of genomic information yielded genomic values for fatty acid profile with accuracies ranging from low to moderate. None of the methods excelled in terms of accuracy, however the SNP-BLUP method allows obtaining less biased genomic evaluations. The ssGBLUP model is an appropriate alternative to obtaining more reliable and less biased GEBVs as pseudo-phenotypes in situations of missing pedigree, due to high proportion of multiple sires, being more appropriate than the EBVs or adjusted phenotypes for fixed effect to predict direct genomic values.

O alcance de altas produtividades em milho (Zea mays L.) depende do desenvolvimento de híbridos, o principal produto explorado nos programas de melhoramento. O sucesso na obtenção deste tipo de cultivar é conseguido com extensivo cruzamento de linhagens, seguido de avaliações para identificação das combinações de maior potencial. Geralmente, o melhorista tem à sua disponibilidade grande número de linhagens, possibilitando a realização de centenas a milhares de cruzamentos distintos, dos quais apenas uma pequena quantidade pode ser avaliada experimentalmente devido a limitação de tempo e recursos. Com o advento da Seleção Genômica (GS) tornou-se possível predizer o comportamento destes indivíduos não avaliados com base em seu genoma. No decorrer do processo de consolidação da GS várias metodologias foram propostas. A aptidão destas em predizer desempenhos fenotípicos é dependente da sua capacidade de acomodar a arquitetura genética das características e lidar com a multicolinearidade das matrizes genômicas. Neste sentido, métodos baseados em modelos mistos podem apresentar menor eficiência na predição de características oligogênicas devido à não capacidade de representar a distribuição real do efeito dos QTL. Além disso, a regularização das predições na presença de multicolinearidade é realizada por meio de um parâmetro de penalização (λ), o qual pode ser estimado de várias formas e consequentemente modificar a acurácia dos modelos. Além do aprimoramento dos métodos, outro aspecto importante é o procedimento de correção dos dados fenotípicos previamente à GS, o qual não é consenso na comunidade científica. Diante do exposto, este trabalho objetivou: verificar o efeito das formas de obtenção do λ (via REML na GS e pela herdabilidade da característica) e da correção do fenótipo (valor genotípico e média ajustada) na GS e avaliar a eficiência da modelagem diferencial de QTL de maior efeito na capacidade preditiva da metodologia G-BLUP, comparando-a ao LASSO Bayesiano, BayesB e G-BLUP convencional. Para isso foram utilizadas informações de híbridos simples de milho tropical avaliados em cinco locais para produtividade de grãos, altura de planta e espiga no ano de 2015. Os dados genômicos foram obtidos com a plataforma Affymetrix® Axiom® Maize Genotyping Array de 616.201 SNPs. Foram estudados diferentes cenários de GS considerando os fatores supracitados, sendo estes comparados entre si por suas capacidades preditivas e seletivas. Os resultados obtidos indicam que a correção do fenótipo e a forma de estimação de λ afetam a capacidade preditiva. O uso de valores genotípicos como correção dos fenótipos e estimação de λ via REML apresentaram os melhores resultados. Foi também observado que a modelagem de SNPs de maior efeito como fator fixo aumenta discretamente a capacidade preditiva da metodologia G-BLUP para as características oligogênicas avaliadas (altura de planta e espiga), sendo indicado o uso do G-BLUP convencional. Complementarmente, observou-se que a GS apresentou modesta eficiência na seleção de híbridos superiores sob intensidades moderadas. Entretanto, a sua alta capacidade de selecionar sob baixa intensidade pode ser amplamente explorada nos programas de melhoramento de milho visando a seleção precoce direta. / The achievement of high yield in maize (Zea mays L.) relies on the development of hybrids, which is the main product of breeding programs. The success in obtaining this kind of cultivar is achieved through extensive crossing of inbred lines followed by field trials to identify the combinations with greatest potential. Generally, breeders have a large number of inbred lines on their hands, being able to perform hundreds to thousands of different crosses, of which only a small portion can be experimentally evaluated due to time and resource limitations. Genomic Selection (GS) has made it possible to predict phenotypes of unevaluated individuals based on their genome. Throughout the establishment process of GS many approaches have been proposed. The ability of these approaches at predicting phenotypic performance depends on their capacity of accommodating the genetic architecture of the traits and dealing with the multicollinearity of the genomic matrices. Hence, methods based on mixed model equations may present lower prediction efficiency for oligogenic traits due to their inability of depicting the real distribution of the QTL effects. Moreover, the prediction regularization in the presence of multicollinearity is done by a shrinkage factor (λ), which can be estimated in a number of ways and may affect the accuracy of the models. In addition to the improvement of the models, the correction of the phenotype utilized in the predictions is also important, which is not a consensus among researchers. Based on these facts, this study aimed to assess the effect of estimation of λ (by REML in the GS model and by the heritability of the traits) and the correction of the phenotype (genotypic value and adjusted mean) on the GS. It also targeted to evaluate the effect of differential modeling of major makers on the prediction accuracy of G-BLUP, comparing it to Bayesian LASSO, BayesB and ordinary G-BLUP. To those ends, tropical maize single-crosses evaluated at five sites for grain yield, plant and ear height in 2015 were utilized. The genomic data was obtained with the Affymetrix® Axiom® Maize Genotyping Array of 616,201 SNPs. Distinct GS scenarios were studied considering the aforementioned factors which were compared by their prediction and selection accuracy. The results suggest that the correction of the phenotype and the way of estimation of λ do affect prediction accuracies. The use of genotypic values as the correction of phenotypes and the estimation of λ by REML showed best results. It was also observed that modeling major SNPs as fixed effect factors had little improvement on the prediction accuracy of G-BLUP for the oligogenic traits evaluated (plant and ear height). Thereby, ordinary G-BLUP should be the method of choice to predict these traits. Additionally, it was observed that GS presented modest efficiency for selecting superior hybrids under moderate intensities. However, its high effectiveness at selecting under low intensities might be exploited on maize breeding programs for early direct selection.

Convergence of quantitative genetics and genomics is becoming the way that fundamental genetics and applied breeding will be carried out in the next decades. This study bridges the quantitative genetics of complex growth and wood properties traits with genomic technologies towards a more innovative approach to tree breeding. Planted forests play a major role to fulfill the growing world demand for wood products and energy. Eucalypts stand out for their high productivity and versatile wood resulting from the advanced breeding programs associated to clonal propagation and modern silviculture. Despite their fast growth, breeding cycles still take several years and wood properties assessment is limited to a sample of trees in the late stages of selection due to the costs involved in wood phenotyping, not exploitingthe range of genetic variation in wood properties. In this study, we examined fifteen traits including growth and wood chemical and physical properties in 1,000 individuals sampled from an elite Eucalyptus breeding population. Near-infrared spectroscopy (NIRS) models were developed and used for high-throughput phenotyping of wood traits.Highdensity data for 29,090 SNPs was used to obtain accurate pedigree-record-free estimates of trait variance components, heritabilities, genetic and phenotypic correlations, based on a realized relationship matrix, comparing them to pedigree-based estimates. To the best of our knowledge, this is the first study to do this in plants. NIRS predictions were accurate for wood chemical traits and wood density, and variably successful for physical traits. Heritabilities were medium for growth (0.34 to 0.44), high for wood chemical traits (0.56 to 0.85) and variable for wood physical traits (0.11 to 0.63). High positive correlations among growth traits and negative between cellulose and lignin content were observed, while correlations between wood chemical and physical traits and between growth and wood quality traits were low although significant. Phenotypes and SNP markers were then used to build genomic predictive models using a marker density higher than any previous genomic selection study in trees (1 SNP/21 kbp). Two models (RR-BLUP and Bayesian LASSO) that differ regarding the assumed distribution of marker effects were used for genomic predictions. Predictions were compared to those obtained by phenotypic BLUP. Predictive abilities very similar by the two models and strongly correlated to the heritabilities. Accurate genomic-enabled predictions were obtained for wood chemical traits related to lignin, wood density and growth, although generally 15 to 25% lower than those achieved by phenotypic BLUP prediction. Nevertheless, genomic predictions yielded a coincidence above 70% in selecting the top 30 trees ranked by phenotypic selection for growth, wood density and S:G ratio, and 60% when tandem selection was applied. The results of this study open opportunities for an increased use of highthroughput NIRS phenotyping and genome-wide SNP genotyping in Eucalyptus breeding, allowing accurate pedigree-record-free estimation of genetic parameters and prediction of genomic breeding values for yet to be phenotyped trees. These applications should become routine in tree breeding programs for the years to come, significantly reducing the length of breeding cycles while optimizing resource allocation and sustainability of the breeding endeavor. / A convergência da genética quantitativa com a genômica está se tornando a maneira pela qual a genética fundamental e aplicada serão conduzidas nas próximas décadas. Este estudo buscou conectar a genética de fenótipos complexos de crescimento e propriedades de madeira às tecnologias genômicas, em uma abordagem inovadora para o melhoramento florestal. Florestas plantadas têm papel fundamental para satisfazer a crescente demanda mundial por produtos madeireiros e energia. O eucalipto,com sua alta produtividade e madeira versátil, é resultado de programas avançados de melhoramento associados à propagação clonal e silvicultura moderna. Apesar de seu rápido crescimento, ciclos de melhoramento ainda levam muitos anos e a avaliação detalhada de propriedades da madeira é limitada a apenas uma amostra das árvores em estágios avançados de seleção, devido aos altos custos de fenotipagem, não explorando assim toda a variação genética disponível. Neste estudo, examinamos quinze caracteres, incluindo crescimento e propriedades químicas e físicas da madeira, em 1000 indivíduos amostrados de uma população elite de melhoramento. Modelos de espectroscopia de infravermelho próximo (NIRS) foram desenvolvidos e utilizados para fenotipagem de alto desempenho de propriedades de madeira. Genotipagem de alta densidade com 29.090 SNPs foi utilizada para obter estimativas acuradas de componentes de variância, herdabilidades e correlações genéticas baseadas em uma matriz de parentesco realizado, ou seja,sem o uso de pedigree. Este é o primeiro estudo de que temos conhecimento a fazer isso em plantas. Predições NIRS foram precisas para caracteres químicos da madeira e densidade, e apresentaram sucesso variável para caracteres físicos. As herdabilidades foram médias para crescimento (0,34 a 0,44), altas para caracteres químicos de madeira (0,56 a 0,85) e variáveis para caracteres físicos da madeira (0,11 a 0,63). Altas correlações positivas entre caracteres de crescimento e negativas entre celulose e lignina foram observadas, enquanto correlações entre caracteres químicos e físicos da madeira foram baixas, porém significativas. Fenótipos e marcadores SNP foram em seguida utilizados na construção de modelos preditivos com a maior densidade de marcadores já utilizada em estudos de seleção genômica em espécies florestais (1 SNP/21 kpb). Dois modelos de predição (RR-BLUP e LASSO Bayesiano)foram usados nas predições genômicas e comparados ao BLUP fenotípico. Os modelos apresentaram capacidades preditivas similares, fortemente correlacionadas às herdabilidades. Predições genômicas precisas foram obtidas para caracteres relacionados à lignina, densidade e crescimento, embora geralmente 15 a 25% menores do que as predições obtidas por BLUP fenotípico. Contudo, predições genômicas alcançaram coincidências acima de 70% na seleção das melhores 30 árvores ranqueadas pela seleção fenotípica para crescimento, densidade e relação S:G, e de 60% quando seleção em tandem foi aplicada. Os resultados deste estudo abrem enormes oportunidades para o uso combinado de fenotipagem NIRS e genotipagem com SNPs no melhoramento do eucalipto, permitindo estimativas acuradas de parâmetros genéticos e a predição de valores genéticos genômicos para plantas jovens ainda não fenotipadas. Estas aplicações deverão se tornar rotineiras nos programas de melhoramento florestal nos próximos anos, reduzindo significativamente a duração dos ciclos de seleção e, consequentemente, otimizando a alocação de recursos e a sustentabilidade do melhoramento.

La sélection génomique, qui a révolutionné la sélection génétique des bovins laitiers notamment, est désormais envisagée dans d’autres espèces comme l’espèce caprine. La clé du succès de la sélection génomique réside dans la précision des évaluations génomiques. Chez les caprins laitiers français, le gain de précision attendu avec la sélection génomique était un des questionnements de la filière en raison de la petite taille de la population de référence disponible (825 mâles et 1945 femelles génotypés sur une puce SNP 50K). Le but de cette étude est d’évaluer comment augmenter la précision des évaluations génomiques dans l’espèce caprine. Une étude de la structure génétique de la population de référence caprine constituée d’animaux de races Saanen et Alpine, a permis de montrer que la population de référence choisie est représentative de la population élevée sur le territoire français. En revanche, les faibles niveaux de déséquilibre de liaison (0,17 entre deux SNP consécutifs) de consanguinité et de parenté au sein de la population, similaires à ceux trouvés en ovins Lacaune, ne sont pas idéaux pour obtenir une bonne précision des évaluations génomiques. De plus, malgré l’origine commune des races Alpine et Saanen, leurs structures génétiques suggèrent qu’elles se distinguent clairement d’un point de vue génétique. Les méthodes génomiques (GBLUP ou Bayésiennes) « two-step », basées sur des performances pré-corrigées (DYD, EBV dérégressées) n’ont pas permis une amélioration significative des précisions des évaluations génomiques pour les caractères évalués en routine (caractères de production, de morphologie et de comptages de cellules somatiques) chez les caprins laitiers. La prise en compte des phénotypes des mâles non génotypés permet d’augmenter les précisions des évaluations de 3 à 47% selon le caractère. L’ajout des génotypes de femelles issues d’un dispositif de détection de QTL améliore également les précisions (de 2 à 14%) que ce soit pour les évaluations two steps ou les évaluations basées sur les performances propres des femelles (single step). Les précisions sont augmentées de 10 à 74% avec les évaluations single step comparées aux évaluations two steps, ce qui permet d’atteindre des précisions supérieures à celles obtenues sur ascendance. Les précisions obtenues avec les évaluations génomiques multiraciales, bicaractères et uniraciales sont similaires même si la précision des valeurs génomiques estimées des candidats à la sélection est plus élevée avec les évaluations multiraciales. La sélection génomique est donc envisageable chez les caprins laitiers français à l’aide d’un modèle génomique multiracial single step. Les précisions peuvent être légèrement augmentées par l’inclusion de gènes majeurs tels que celui de la caséine αs1 notamment à l’aide d’un modèle « gene content » pour prédire le génotype des animaux non génotypés. / Genomic selection which is revolutionizing genetic selection in dairy cattle has been tested in several species like dairy goat. Key point in genomic selection is accuracy of genomic evaluation. In French dairy goats, gain in accuracy using genomic selection was questioning due to the small size of the reference population (825 males and 1 945 females genotyped). The aim of this study was to investigate how to reach adequate genomic evaluation accuracy in French dairy goat population. The study of reference population structure (Alpine and Saanen breeds) showed that reference population is similar to the whole population of French dairy goats. But the weak level of linkage disequilibrium (0.17 between two consecutive SNP), inbreeding and relationship between reference and candidate population were not ideal to maximize genomic evaluation accuracy. Despite their common origin, genetic structure of Alpine and Saanen breeds suggested that they were genetically distant. Two steps genomic evaluation (GBLUP, Bayesian) based on performances corrected for fixed effect (DYD, deregressed EBV) did not improve genetic evaluation accuracy compared to classical evaluations for milk production traits, udder type traits and somatic cells score classically selected in French dairy goat. Taking into account phenotypes of ungenotyped sires increased genomic evaluation from 3 to 47% depending on the trait considered. Adding female genotypes also improved genomic evaluation accuracies from 2 to 4% depending on the method (two steps or single step) and on the trait. When using gemomic evaluation directly based on female performances (single step), accuracy of genomic evaluation reach the level obtained from ascendance in classic evaluation which was not the case using two steps evaluations. Genomic evaluation accuracies were similar when using multiple-trait model, multi-breed or single breed evaluation. But accuracies derived from prediction error variances were better when using multi-breed genomic evaluations. Genomic selection is feasible in French dairy goats using single step multi-breed genomic evaluations. Accuracies could be slightly improved integrating major gene as αs1 casein especially when using « gene content » approach to predict genotypes of ungenotyped animals.

La selección genómica está cambiando profundamente el mercado del vacuno de leche. En la actualidad, es posible obtener una alta precisión en las valoraciones genéticas de animales muy jóvenes sin la necesidad del fenotipo propio o el de sus hijas. Por tanto, la respuesta genética de un programa genómico bien diseñado supera netamente a la selección tradicional. Esta mejora está modificando uno de los principios tradicionales del mercado de vacuno de leche como era la preferencia de uso de toros con altas fiabilidades frente a otros animales con valores genéticos a priori superiores. Esta tesis contiene seis capítulos en los cuales se estudian de las bases para la implementación del programa de selección genómica en el vacuno de leche español. Para ello se realizaron estudios de simulación y valoraciones genómicas con datos reales de la primera población nacional de referencia. El objetivo principal de esta tesis es contribuir a la implementación de la selección genómica en el vacuno de leche español. Los objetivos específicos son: (1) Estudiar alternativas de genotipado en poblaciones reducidas de vacuno lechero. (2) Desarrollar y validar metodología para la evaluación de grandes cantidades de genotipos. (3) Estudiar el efecto de los procesos de imputación de genotipos en la capacidad predictiva de los genotipos resultantes. Las principales cuestiones relacionadas con la selección genómica en vacuno lechero fueron discutidas en el capítulo 1 incluyendo: aspectos estadísticos y genéticos en los que se basa la selección genómica, diseño de poblaciones de referencia, revisión del estado del arte en cuanto a la metodología desarrollada para evaluación genómica, diseño y métodos de los algoritmos de imputación, e implementación de la selección genómica en vacuno de leche a nivel de programa de selección, centro de inseminación y de granja comercial. En el capítulo 2 se realizó un estudio de simulación comparando estrategias de genotipado selectivo en poblaciones de hembras frente al uso de selección tradicional o selección genómica con una población de referencia de machos. La población de referencia española estaba formada en principio por algo más de 1,600 toros con prueba de progenie. Este tamaño no es, en principio, suficiente para obtener predicciones genómicas de alta fiabilidad. Por tanto, debían evaluarse diferentes alternativas para incrementar la habilidad predictiva de las evaluaciones. Las estrategias que consisten en usar como población de referencia los animales en los extremos de la distribución fenotípica permitían mejorar la precisión de la evaluación. Los resultados usando 1,000 genotipos fueron 0.50 para el carácter de baja heredabilidad y 0.63 para el de heredabilidad media cuando la variable dependiente fue el fenotipo ajustado. Cuando se usaron valores genéticos como variable dependiente las correlaciones fueron 0.48 y 0.63 respectivamente. Para los mismos caracteres, una población de 996 machos obtuvo correlaciones de 0.48 y 0.55 en las predicciones posteriores. El estudio concluye que la estrategia de genotipado que proporciona la mayor correlación es la que incluye las hembras de ambas colas de la distribución de fenotipos. Por otro lado se pone de manifiesto que la mera inclusión de las hembras élite que son las habitualmente genotipadas en las poblaciones reales produce resultados no satisfactorios en la predicción de valores genómicos. En el capítulo 3, el Random Boosting (R-Boost) es comparado con otros métodos de evaluación genómica como Bayes-A, LASSO Bayesiano y G-BLUP. La población de referencia española y caracteres incluidos en las evaluaciones genéticas tradicionales de vacuno lechero fueron usados para comparar estos métodos en términos de precisión y sesgo. Las predicciones genómicas fueron más precisas que el índice de pedigrí tradicional a la hora de predecir los resultados de futuros test de progenie como era de esperar. Las ganancias en precisión debidas al empleo de la selección genómica dependen del carácter evaluado y variaron entre 0.04 (Profundidad de ubre) y 0.42 (Porcentaje de grasa) unidades de correlación de Pearson. Los resultados promediados entre caracteres mostraron que el LASSO Bayesiano obtuvo mayores correlaciones superando al R-Boost, Bayes-A y G-BLUP en 0.01, 0.03 y 0.03 unidades respectivamente. Las predicciones obtenidas con el LASSO Bayesiano también mostraron menos desviaciones en la media, 0.02, 0.03 y 0.10 menos que Bayes-A, R-Boost y G-BLUP, respectivamente. Las predicciones usando R-Boost obtuvieron coeficientes de regresión más próximos a la unidad que el resto de métodos y los errores medios cuadráticos fueron un 2%, 10% y 12% inferiores a los obtenidos a partir del B-LASSO, Bayes-A y G-BLUP, respectivamente. El estudio concluye que R- Boost es una metodología aplicable a selección genómica y competitiva en términos de capacidad predictiva. En el capítulo 4, el algoritmo de machine learning R-Boost evaluado en el capítulo 3 es descrito e implementado para selección genómica adaptado a la evaluación de grandes bases de datos de una forma eficiente. Tras la incorporación en el consorcio Eurogenomics, el programa genómico español pasó a disponer de más de 22,000 toros probados como población de referencia, por tanto era necesario implementar un método capaz de evaluar éste gran conjunto de datos en un tiempo razonable. El nuevo algoritmo denominado R-Boost realiza de forma secuencial un muestreo aleatorio de SNPs en cada iteración sobre los cuales se aplica un predictor débil. El algoritmo fue evaluado sobre datos reales de vacuno de leche empleados en el capítulo 3 estudiando más en profundidad el comportamiento de los parámetros de sintonización. Esta propuesta de modificación del Boosting puede obtener predicciones sin perdida de precisión o incrementos de sesgo empleando tan solo un 1% del tiempo de computación original. En el capítulo 5 se evalúa el efecto de usar genotipos de baja densidad imputados con el software Beagle en cuanto a su posterior habilidad predictiva cuando son incorporados a la población de referencia. Para ello se emplearon dos métodos de evaluación R-Boost y un BLUP con matriz genómica. Animales de los que se conocían los SNPs incluidos en los chips GoldenGate Bovine 3K y BovineLD BeadChip, fueron imputados hasta conocer los SNPs incluidos en el BovineSNP50v2 BeadChip. Posteriormente, un segundo proceso de imputación obtuvo los SNPs incluidos en el BovineHD BeadChip. Tras imputatar desde dos genotipados a baja densidad, se obtuvo similar capacidad predictiva a la obtenida empleando los originales en densidad 50K. Sin embargo, sólo se obtuvo una pequeña mejora (0.002 unidades de Pearson) al imputar a HD. El mayor incremento se obtuvo para el carácter días abiertos donde las correlaciones en el grupo de validación aumentaron en 0.06 unidades de Pearson las correlaciones en el grupo de validación cuando se emplearon los genotipos imputados a HD. En función de la densidad de genotipado, el algoritmo R-Boost mostró mayores diferencias que el G-BLUP. Ambos métodos obtuvieron resultados similares salvo en el caso de porcentaje de grasa, donde las predicciones obtenidas con el R-Boost fueron superiores a las del G-BLUP en 0.20 unidades de correlación de Pearson. El estudio concluye que la capacidad predictiva para algunos caracteres puede mejorar imputando la población de referencia a HD así como empleando métodos de evaluación capaces de adaptarse a las distintas arquitecturas genéticas posibles. Finalmente en el capitulo 6 se desarrolla una discusión general de los estudios presentados en los capítulos anteriores y se enlazan con la implementación de la selección genómica en el vacuno lechero español, que se ha desarrollado en paralelo a esta tesis doctoral. La primera población de referencia con unos 1.600 toros fue evaluada en el capítulo 4 y fue usada para comparar los distintos métodos y escenarios propuestos en los capítulos 3, 4 y 5. La primera evaluación genómica obtenida para los caracteres incluidos en el capítulo 4 de esta tesis estuvo disponible para los centros de inseminación incluidos en el programa en septiembre de 2011. La población de Eurogenomics se incorporó en Noviembre de dicho año, completándose la primera evaluación para los caracteres incluidos en el índice de selección ICO en Febrero de 2012 empleando el R-Boost descrito en el capítulo 3. En mayo de 2012 las evaluaciones del carácter proteína fueron validadas por Interbull y finalmente el 30 de Noviembre del 2012 las primeras evaluaciones genómicas oficiales fueron publicadas on-line por la federación de ganaderos CONAFE (http://www.conafe.com/noticias/20121130a.htm). / Jiménez Montero, JA. (2013). Selección genómica en poblaciones reducidas de vacuno de leche [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/27649 / TESIS

La sélection génomique offre de nouvelles perspectives en amélioration des plantes pour la sélection de caractères complexes. Le travail proposé porte sur l’évaluation de son intérêt dans le cadre d’un programme de sélection réciproque pour la valeur d’hybrides entre deux groupes génétiques de maïs complémentaires. Il s’appuie sur un dispositif expérimental original constitué de 900 hybrides produits dans un plan factoriel entre deux dispositifs multiparentaux connectés. L’objectif de la sélection est d’améliorer le rendement ensilage des hybrides tout en améliorant leur digestibilité. Une réflexion sur les modèles permettant de prédire la valeur hybride sera conduite et testée sur les données expérimentales et par simulations. Ce travail, conduit en collaboration avec sept sociétés de sélection (au sein de PROMAÏS) devrait permettre d’améliorer les dispositifs de sélection classiques et de produire des hybrides d’intérêt agronomique. Il s’inscrit dans le cadre plus général de l’amélioration pour la valeur en croisement commune à de nombreuses espèces végétales allogames et à certaines espèces animales. / Genomic selection opens new prospects in plant breeding for the selection of complex traits. The proposed study aims to evaluate its efficiency in the context of a reciprocal selection schemes for the hybrid value between two complementary maize groups. The work will rely on an original experimental design including 900 hybrids produced from a factorial between two multiparental connected designs. The selection objective is to increase the hybrids silage yield as well as their digestibility. Several models for the hybrid value prediction will be proposed and tested on the experimental data and by simulations. This study, carried out in close connection with seven plant breeding companies (members of PROMAÏS) will contribute to the improvement of breeding designs and will produce new interesting hybrids. It falls within the general context of the selection for hybrid value which is common to numerous plant allogamous species and animal species.

Le maïs (Zea mays L.) est la plante la plus cultivée au monde. Pour valoriser le fort effet d’hétérosis pour les caractères liés à la biomasse, la diversité génétique du maïs est structurée en groupes hétérotiques et les variétés cultivées sont majoritairement des hybrides F1 entre lignées de groupes complémentaires. La valeur hybride se décompose comme la somme de l’Aptitude Générale à la Combinaison (AGC) de chacune des lignées parentales et de l’Aptitude Spécifique à la Combinaison (ASC) du couple. En Europe du Nord, le maïs est souvent utilisé en ensilage destiné à l'alimentation animale. L’objectif de sélection est d’améliorer la productivité et d'assurer une bonne digestibilité du maïs ensilage. Les objectifs de cette thèse étaient : (i) d’estimer l’importance de l’AGC et de l’ASC dans la variance génétique hybride pour les caractères de qualité de l’ensilage, (ii) d’identifier les locus (QTL) impliqués dans ces caractères et d’étudier leur colocalisation avec des QTL de productivité, (iii) d’évaluer l’intérêt de la sélection génomique pour la prédiction des performances hybrides et (iv) de comparer l’efficacité de deux dispositifs de calibration des prédictions basés sur un factoriel ou l’utilisation classique de testeurs du groupe complémentaire. Dans le cadre du projet SAM-MCR, 6 familles biparentales connectées ont été créés dans les groupes « corné » et « denté » à partir de 4 lignées fondatrices. Dans une première phase, 822 lignées cornées et 802 dentées ont été génotypées pour 20k SNP et croisées selon un factoriel incomplet pour produire 951 hybrides, phénotypés pour des caractères de qualité et de productivité (étudiés par H. Giraud pendant sa thèse). L’analyse des caractères de qualité a montré une prédominance de l’AGC par rapport à l’ASC ainsi qu’une corrélation négative entre les caractères de qualité et le rendement. De nombreux QTLs multi-alléliques ont été détectés, la plupart spécifiques d’un groupe et dont certains colocalisent avec des QTL de rendement. Par validation croisée, la qualité de prédictions basées sur les QTL détectés s’est avérée plus faible que celle obtenue par prédiction génomique. La prise en compte de l’ASC n‘a pas permis d’améliorer sensiblement la qualité de prédictions des modèles. Dans une seconde phase, 90 lignées ont été choisies par groupe : 30 sélectionnées sur la base de leurs prédictions génomiques pour la productivité et la valeur énergétique et 60 choisies aléatoirement parmi les 6 familles. Ces lignées ont été croisées selon un factoriel incomplet pour produire 360 nouveaux hybrides : 120 issus des lignées sélectionnées et 240 issus des lignées choisies au hasard. Les 90 lignées de chaque groupe ont aussi été croisées à deux lignées fondatrices du groupe complémentaire (testeurs). Les hybrides issus des lignées sélectionnées se sont avérés plus productifs mais de moins bonne qualité. Nous avons confirmé la bonne qualité des prédictions génomiques obtenus dans le factoriel initial sur les nouveaux hybrides évalués dans d’autres environnements et après sélection et observé une bonne corrélation entre les AGC estimées dans le factoriel et dans le dispositif testeurs. Des dispositifs factoriels et testeurs ont été simulés en faisant varier la part d’ASC, le nombre d’hybrides et la contribution de chaque lignée dans le jeu de calibration. A moyens expérimentaux égaux, le dispositif factoriel s’est avéré plus efficace en termes de capacité prédictive et de gain génétique cumulé que le dispositif testeur (jusqu’à +50%) pour un caractère présentant de l’ASC et équivalent pour un caractère purement additif. Les résultats de cette thèse ouvrent de nouvelles perspectives pour revisiter les schémas de sélection hybrides en remplaçant l’évaluation des lignées candidates, classiquement faite sur testeur, par l’évaluation directe d’hybrides issus d’un factoriel incomplet. La mise en œuvre de tels dispositifs nécessitera de réorganiser la logistique des programmes de sélection. / Maize (Zea mays L.) is the most cultivated crop in the world. To exploit the strong heterosis for traits related to biomass, the genetic diversity of maize is structured into heterotic groups and cultivated varieties are mainly F1 hybrids obtained by crossing lines from complementary groups. The hybrid value can be decomposed as the sum of the General Combining Ability (GCA) of each parental line and the Specific Combining Ability (ASC) of the cross. In northern Europe, maize is often used as silage for animal feed and the breeding objective is to improve productivity while ensuring a good energetic value and digestibility of the silage. The objectives of this thesis were: (i) to estimate the importance of GCA and SCA in hybrid genetic variance for silage quality traits, (ii) to identify loci (QTL) involved in these traits and to study their colocalization with QTL for productivity traits, (iii) to evaluate the interest of genomic selection for the prediction of hybrid performances and (iv) to compare the prediction accuracies of two calibration designs either based on a factorial or on the conventional use of testers from the complementary group. As part of the SAM-MCR project, 6 biparental connected families were created in the "flint" and "dent" groups from 4 founder lines. In a first phase, 822 flint and 802 dent lines were genotyped for 20k SNPs and crossed according to an incomplete factorial to produce 951 hybrids which were phenotyped for quality traits and for productivity traits (studied by H. Giraud during her phD). Quality trait analysis showed a predominance of GCA over SCA and a negative correlation between digestibility traits and silage yield. Several multi-allelic QTLs were detected, most of them being specific to one group. Several colocalizations were found with yield QTL. Using cross-validation, we observed that the predictive ability of models based on detected QTLs was lower than that obtained by genomic predictions. Considering the SCA did not improve model predictive abilities for most of the traits. In a second phase, 90 lines were chosen per group: 30 were selected based on their genomic predictions for productivity and the energetic value and 60 were randomly sampled from the 6 families. These lines were crossed according to an incomplete factorial to produce 360 new hybrids: 120 from selected lines and 240 from randomly chosen lines. The 90 lines of each group were also crossed to two lines of the complementary group (testers). Hybrids from the selected lines were more productive but had a lower silage quality. We confirmed the good accuracy of the genomic predictions obtained in the initial factorial on the new hybrids evaluated in other environments and after selection. We also observed good correlations between GCA estimated in the factorial and in the testcross design. Different factorial and testcross designs were simulated by varying the proportion of dominance/SCA, the number of hybrids and the contribution of each line to the calibration set. Considering the same number of hybrids in the calibration set, the factorial was more efficient in terms of predictive ability and cumulative genetic gain (up to + 50%) than the testcross design for traits showing SCA and was similar for purely additive traits. The results of this thesis open new perspectives to revisit hybrid breeding schemes by replacing the evaluation of candidate lines, classically made on testcross, by the direct evaluation of hybrids resulting from an incomplete factorial. The implementation of such designs will require reorganizing the logistics of selection programs.

Les espèces forestières sont particulières à bien des égards par rapport aux autres espècesdomestiquées. Les arbres forestiers ont de longues phases juvéniles, entrainant de long et couteuxcycles de sélection et nécessitant une sélection en plusieurs étapes indépendantes. Bien que cetteméthode soit efficace du point de vue opérationnel, elle reste couteuse en temps et en ressources,entrainant une dilution de l’intensité et de la précision de sélection. Au vu de ces contraintes,les arbres sont de bons candidats pour la mise en oeuvre de l’évaluation génomique. La sélectiongénomique (SG) repose sur le classement et la sélection d’individus à partir de l’informationcontenu dans leur génome sans utilisé une étape d’évaluation phénotypique et ainsi accélérerle processus de sélection.Ce travail visait à identifier les situations, les critères et les facteursdans lesquelles la SG pourrait être une option réalisable pour le peuplier. Notre étude a montréque les avantages de l’évaluation génomique dépendent du contexte. C’est dans des situationsles moins avantageuse que l’évaluation génomique se montre la plus performante, elle profiteégalement de la densification de l’information génétique de faible à moyenne suite à une étaped’imputation de haute qualité. La sélection génomique pourrait être une option intéressante àstade précoce, où la précision de la sélection est généralement faible et la variabilité génétiqueabondante. Notre travail a également montré qu’il est important d’évaluer les performancesavec des critères alternatifs, comme ceux liés au classement, notamment lorsque ces critèresrépondent au contexte opérationnel du programme d’élevage étudié. / Forest species are unique in many ways compared to other domesticated species. Forest trees have long juvenile phases, leading to long and costly selection cycles and requiring selection in several independent stages. Even if this method is operationally effective, it remains costly in terms of time and resources, resulting in a diluted intensity and accuracy of selection.In view of these constraints, trees are good candidates for the implementation of genomic evaluation. Genomic selection (SG) is based on the classification and selection of individuals from the information contained in their genome without using a phenotypic evaluation step and thus accelerating the selection process, in order to identify the situations, criteria and factors in which SG could be a feasible option for poplar. Our study showed that the benefits of genomic evaluation are context-dependent. Genomic evaluation is most effective in theless-advantageous situations, it also benefits from low to medium density genetic information following a high-quality imputation step. Genomic selection could be an interesting option at an early stage, when the accuracy of selection is generally low and genetic variability is abundant.Our work has also shown that it is important to evaluate performance with alternative criteria,such as those related to ranking, especially when these criteria fit the operational context of the breeding programme under study.

Fusarium head blight (FHB) is a devastating wheat (Triticum aestivum L.) disease worldwide. Presently, there is insufficient FHB resistance in the Canadian wheat germplasm. Genome-wide association study (GWAS) and genomic selection (GS) can be utilized to identify sources of resistance that could benefit wheat breeding. To define the genetic architecture of FHB resistance, association panels from a spring and a winter collection were evaluated using the Wheat Illumina Infinium 90K array. A total of 206 accessions from the spring panel and 73 from the winter panel were evaluated in field trials for 3-4 years at two locations, namely Morden (Manitoba) and Ottawa (Ontario). These accessions were phenotyped for FHB incidence (INC), severity (SEV), visual rating index (VRI), and deoxynivalenol (DON) content. Significant (p < 0.05) differences among genotypes for all traits were found. Genetic characterization using the wheat 90K array identified a set of 20,501 single nucleotide polymorphisms (SNPs). The probe sequences (~100 bp) of these SNPs were mapped to the Chinese Spring reference genome v2.0 to identify 13,760 SNPs in the spring panel, and 10,421 SNPs in the winter panel covering all 21 wheat chromosomes. GWAS was performed to identify novel FHB resistance loci for INC, SEV, VRI and DON content for the spring and the combined panels separately using these 13,760 SNPs and for the winter panel using 10,421 SNPs. A total of 107, 157, 174 unique quantitative trait loci (QTNs) were identified for the four traits using two single-locus and seven multi-locus GWAS models for the spring, winter, and combined panels, respectively. These QTNs represent a valuable genetic resource for the improvement of FHB resistance in commercially grown wheat cultivars. In addition, these GWAS-defined QTNs were further used for GS to determine the breeding value (BV) of individuals as outlined below. In order to understand the role of the model and that of the marker type and density in trait prediction modelling, a GS study was conducted. GS is considered as an important tool for increasing genetic gain for economically important traits such as FHB resistance. GS uses genome-wide molecular markers to develop statistical models that predict genomic estimated breeding values (GEBVs) of an individual. Our results support genomic prediction (GP) as an alternative to phenotypic selection to predict the BVs of individuals for this trait. GS accounts for minor effect QTNs, which is beneficial when breeding for quantitative traits. Moderate to high GP accuracies can be achieved for FHB resistance-related traits when implemented in a breeding program. The correlation between the estimate of the missing phenotypic value and the observed phenotype is known as predictive ability (r). Overall, the predictive ability increased significantly using a QTN-based GP approach for FHB traits in wheat and its wild relatives. DON content had the highest predictive ability among all FHB traits, and that was in the winter panel, highlighting the importance of objectively measured traits in breeding for disease resistant genotypes. Interestingly, the winter panel contained several wild relative species that may harbor genes of interest to prevent the accumulation of mycotoxins in the grain. This study showed the usability of genomic prediction by improving the predictive ability of the FHB traits, which can be applied in early generation selection to accelerate the improvement of FHB resistance in wheat. The results show that GS can be successfully implemented in wheat breeding programs over multiple breeding cycles and can be effective for economically important traits. It is anticipated that GS will play a substantial role in the future of wheat breeding.

No description available.