Breeding objectives in the turkey industry are heavily weighted towards improving growth traits. This thesis focused on methods to efficiently select for other important production traits such as reproduction, feed efficiency, and meat yield. Based on bivariate and random regression modeling it was determined that egg production, fertility, and hatchability were influenced by genotype by environment interactions and, as a result, the regulation of reproductive traits is by some unique genes in the summer and winter. This may be due to changes in day length and temperature. Feed efficiency is another important consideration in a breeding objective and feed conversion ratio and residual feed intake were both moderately heritable. Residual feed intake was also more independent of production traits than feed conversion. Feed intake, body weight, and weight gain were moderately heritable and progress can be made in feed efficiency by appropriately weighting these traits in an index. Infrared measures of surface temperature were then investigated to determine if they can be used to select for feed efficiency. Temperatures of the distal metatarsus, eye, neck, and head did not show a strong relationship to feed efficiency and therefore offer limited advantages to a breeding program. Selection for breast meat yield (BMY) is important and it was determined that breast muscle depth, measured with ultrasound technology, is heritable and highly correlated to the carcass trait. As a result, ultrasound traits can compliment conformation scoring and sibling testing in a breeding program to increase the accuracy of selection for BMY and increase response to selection. A deterministic model was also developed and could be used to determine optimum slaughter weight. This would optimise profits in an integrated system, enabling the industry to account for and capitalize on genetic gains. Overall, the population parameters and selection criteria identified for reproduction, efficiency, and meat yield traits identified in the present thesis could be used to increase selection efficiency in turkey breeding programs. Further, the developed production model can be used by the industry to slaughter turkeys at a time that maximizes profits, based on performance levels.

La filière « poule pondeuse » représente un marché en pleine expansion. L’amélioration des caractères est essentielle pour satisfaire les attentes des consommateurs et des industriels. Cette amélioration est réalisée via la sélection. Actuellement, les candidats à la sélection sont évalués à partir de leur valeur génétique estimée (Estimated Breeding Value, EBV) en appliquant un modèle statistique prenant en compte l’ensemble des phénotypes disponibles sur leurs apparentés (BLUP). L’essor de nouvelle technologie permettant le génotypage à moindre coût de nombreux individus, permet d’envisager la mise en place d’une sélection génomique dans cette filière. La valeur génomique estimée (Genomic Estimated Breeding Value, GEBV) serait potentiellement plus précise que l’EBV, disponible dès la naissance de l’individu et sûrement pour un plus grand nombre de candidats, engendrant ainsi un gain de progrès génétique.Par ailleurs, un même type génétique de poule pondeuse étant largement diffusé à travers le monde, les animaux produisent dans des environnements différents (alimentation, température…). Des interactions génotype – environnement pourraient donc affecter l’estimation des valeurs génétiques des candidats à la sélection. L’objectif premier de ce travail est de préciser l’impact de celles-ci sur un panel large de caractères de production et de qualité des œufs aussi bien au niveau de leur architecture génétique qu’au niveau des évaluations. De plus, les conséquences de l’architecture génétique des caractères sur l’estimation des valeurs génétiques également ont été étudiées. / The laying hens farming represents a growing market. The improvement of traits is needed to satisfy the willing of customers and industrials. This improvement is done with selection. Actually, (candidates for selection are evaluated according to their Estimated Breeding Value (EBV), which is estimated, using a statistic model which considers all the available phenotypes of their relative BLUP). The development of new technologies which allow the genotyping at a lower cost of numerous individuals, could allow the development of genomic selection in this farming. The Genomic Estimated Breeding Value (GEBV) could be potentially more accurate than the EBV, available at the birth of the individual and for probably a larger number of candidates, increasing the rates of genetic progress.Besides, a same genetic type of laying hens is widely distributed around the world, so animals produce in various environments ((alimentation, temperature, hygiene standards…). So, genotype – environment interactions could affect the estimation of breeding values of the candidates for selection. The first objective of this work is to determine of the impact of these interactions on a large panel of egg production and egg quality traits, as well at the genetic architecture level than at the evaluations level. Moreover, the consequences of genetic architecture of these traits on breeding value estimation have been studied.

Les microARN sont des petits ARN non-codant impliqués dans la régulation de multiples fonctions biologiques dont la réponse immunitaire. L'infection par un pathogène induit un changement transcriptomique fort chez l'hôte. Cependant, la variabilité de ces dérégulations reste encore mal décrite. Cette thèse avait pour principaux objectifs de mieux comprendre la spécificité et la variabilité de la réponse des microARN chez l'homme ainsi que mettre en évidence les bases génétiques de cette diversité en utilisant comme modèle l'infection des cellules dendritiques par Mycobacterium tuberculosis (MTB). Nous avons utilisé une approche ex vivo et des techniques à haut débit dans le but de décrire la réponse des microARN suite à l'infection par MTB dans la population générale, et de la comparer à celle induite par d'autres mycobactéries et bactéries intracellulaires. Nous montons que l'infection modifie profondément l'expression des microARN ainsi que la diversité de leurs isoformes, dont un certain nombre de microARN sont impliqués dans une réponse très conservée. Nos résultats soulignent aussi l'effet de l'infection sur les réseaux de régulation de l'expression des gènes impliquant les microARN et montrent que l'expression de 3% de ces transcrits peut-être corrélée à un marqueur génétique. Grâce à l'intégration de ces différentes analyses, nous proposons certains microARN candidats qui pourraient jouer un rôle dans la variabilité de la réponse immunitaire. L'ensemble de nos résultats constitue la première tentative de compréhension de l'architecture génétique de la réponse des microARN et apporte un nouvel éclairage sur le rôle de ces transcrits dans la réponse antibactérienne. / MicroRNAs (miRNAs) are important epigenetic regulators of gene expression that play a key role in many biological processes, including the immune response. Although infection is accompanied by marked changes in the transcriptional profiles of host cells, little is known about the variability of host miRNA responses to infection. In this thesis, we aimed to define the extent and specificity of pathogen-induced miRNA transcriptional responses of host cells, and to characterise the genetic basis of miRNA variability upon infection, using the model of Mycobacterium tuberculosis (MTB) infection of human dendritic cells. To this end, we have combined ex vivo approaches with a range of high-throughput genomic techniques to profile miRNA responses to MTB at the population-level and to compare this response with other mycobacterial and non-mycobacterial infections. We show that miRNAs display marked changes in expression and in isomiR distribution upon infection that are highly consistent across diverse bacteria, demonstrating the presence of a strong core miRNA response to bacterial infection. Our results highlight the impact of infection on miRNA-mediated gene regulatory networks and show that the expression of 3% of miRNAs are controlled by proximate expression quantitative trait loci (eQTLs) and identify a number of candidate miRNAs that may play a role in variability in the immune response to infection. Together, these results provide the first assessment of the impact of genotype-environment interactions on the regulation of miRNA expression, as well as offering novel insights into the specificity of these miRNAs in the response to mycobacterial infections.

Due to increased environmental stewardship and fertilizer prices, there is increased interest in using legume mixes in perennial croplands. The objective of this study was to compare quantitative genetic parameters and quantitative trait loci (QTLs) associated with intermediate wheatgrass (Thinopyrum intermedium) when grown in 1) a non-competitive spaced environment, 2) a polyculture with alfalfa (Medicago sativa), and 3) a monoculture with crested wheatgrass (Agropyron desertorum). Traits evaluated include plant growth characteristics (Zadok’s maturity, height, and tiller count), biomass, and forage nutritive value (CP, NDF, ADF, ADL, IVTD, NDFD, NFC, ME, RFQ). A linkage map comprised of 3568 single nucleotide polymorphisms in 21 linkage groups corresponding to 21 homologous chromosome pairs of both parents was used to identify QTLs and QTL x environment interactions (QxE) based on trait averages for each genotype in each environment. Significant genotype x environment interactions were detected for biomass, NDF etc. A total of 26 QTLs were identified, including 6 MASS, 2 TILE, 2 TICR, 1 ZAMA, 1 CP, 3 NDF, 2 ADF, 3 IVTD, and 5 NDFD. A subset of 7 QTLs showed significant QxE interaction. These results indicate that breeders need to evaluate plants in polyculture or swards if these are the intended crop management systems.

No description available.

A series of studies are presented on the genetic evaluation of cultivated potato (Solanum tuberosum L.) to improve the accuracy and efficiency of selection at various stages of a breeding programme. The central theme was the use of correlated data, such as relationship information and spatial and across-trial correlations, within a linear mixed modelling framework to enhance the evaluation of candidate genotypes and to improve the genetic response to selection. Analyses focused on several social and economically-important traits for the enhancement of the nutritional value, disease resistance and yield of potato tubers. At the formative stages of a breeding scheme, devising a breeding strategy requires an improved understanding of the genetic control of target traits for selection. To guide a strategy that aims to enhance the micronutrient content of potato tubers (biofortification), univariate and multivariate Bayesian models were developed to estimate genetic parameters for micronutrient tuber content from a breeding population generated from crosses between Andean landrace cultivars. The importance of the additive genetic components and extent of the narrow-sense heritability estimates indicated that genotypic 'individual' recurrent selection based on empirical breeding values rather than family-based selection is likely to be the most effective strategy in this breeding population. The magnitude of genetic correlations also indicated that simultaneous increases in important tuber minerals, iron and zinc, could be achieved. Optimising selection efficiency is an important ambition of plant breeding programmes. Reducing the level of candidate replication in field trials may, under certain circumstances, contribute to this aim. Empirical field data and computer simulations inferred that improved rates of genetic gain with p-rep (partially replicated) testing could be obtained compared with testing in fully replicated trials at the early selection stages, particularly when testing over two locations. P-rep testing was able to increase the intensity of selection and the distribution of candidate entries across locations to account for G×E effects was possible at an earlier stage than is currently practised. On the basis of these results, it was recommended that the full replication of trials (at the first opportunity, when enough planting material is available) at a single location in the early stages of selection should be replaced with the partial replication of selection candidates that are distributed over two locations. Genetic evaluation aims to identify genotypes with high empirical breeding values (EBVs) for selection as parents. Using mixed models, spatial parameters to target greater control of localised field heterogeneity were estimated and variance models to account for across-trial genetic heterogeneity were tested for the evaluation of soil-borne powdery scab disease and tuber yield traits at the early stages of a selection programme. When spatial effects improved model fit, spatial correlations for rows and columns were mostly small for powdery scab, and often small and negative for marketable and total tuber yield suggesting the presence of interplot competition in some years for tuber yield traits. For the evaluation of powdery scab, genetic variance structures were tested using data from 12 years of long-term potato breeding METs (multi-environment trials). A simple homogeneous correlation model for the genetic effects was preferred over a more complex factor analytic (FA) model. Similarly, for the MET evaluation of tuber yield at the early stages, there was little benefit in using more complex FA models, with simple correlation structures generally the most favourable models fitted. The use of less complex models will be more straightforward for routine implementation of potato genetic evaluations in breeding programmes. Evaluations for (marketable) tuber yield were extended to multi-location MET data to characterise both genotypes and environments, allowing a re-evaluation of New Zealand MET selection strategies aimed at broad adaptation. Using a factor analytic mixed model, results indicated that the programme’s two main trial locations in the North and the South Islands optimised differentiation between genotypes in terms of G×E effects. There was reasonable performance stability of genotypes across test locations and evidence was presented for some, but limited, genetic progress of cultivars and advanced clonal selections for tuber marketable yield in New Zealand over recent years. The models and selection strategies investigated and developed in this thesis will allow an improved and more systematic application of genetic evaluations in potato selection schemes. This will provide the basis for well informed decisions to be made on selection candidates for the genetic improvement of potato in breeding programmes.

In addition to habitat loss and fragmentation, habitat degradation can have important consequences for biodiversity and population persistence, including effects on ecological and genetic processes beyond decreased demographic viability and the loss of genetic variation. Particularly interesting is the potential for evolutionary changes and adaptation to degraded habitats, that can affect population viability even in the short-term. Here, I explore how environmental changes after habitat degradation affect the evolutionary dynamics of populations of the rainforest cycad Zamia fairchildiana, specifically how habitat degradation affects gene dispersal, inbreeding, directional selection, and genotype-by-environment interactions, and the potential for genetic differentiation between populations. Colonies of Z. fairchildiana showed little genetic differentiation in neutral molecular markers across study sites, thus can be considered as subpopulations. Subpopulations in the disturbed habitat are experiencing different environmental conditions when compared to subpopulation in their native habitat. Disturbed-habitat subpopulations showed a faster life-history. This faster life history is associated with a weaker spatial genetic structure and higher levels of inbreeding in the disturbed-habitat subpopulations. In addition, higher light availability in the disturbed habitat seems to be a major agent of selection on traits like leaf production that have the potential to respond to selection in these subpopulations. Different traits were under selection in the native-habitat subpopulations, suggesting the potential for genetic differentiation between native and disturbed-habitat subpopulations. Genotype by environment interactions in seed germination and seedling survival, in response to light and water availability, further suggested that subpopulations can adaptively diverge between habitats, but the relative role of genetic and environmental factors, particularly maternal effects, on the magnitude and rate of genetic differentiation between subpopulations remains to be evaluated. These results suggest that habitat degradation can have important consequences for the evolutionary dynamics of populations of this cycad, not necessarily typical of habitat loss and fragmentation. This study identified factors and processes important for population persistence in degraded habitats, but population responses to habitat degradation are complex. Thus further studies and long-term experiments are required for better understanding the effects of habitat degradation on population viability.

Objetivou-se com o presente estudo foram caracterizar e definir ambientes homogêneos de produção de bovinos de corte compostos no Brasil com relação às variáveis climáticas e geográficas, utilizando técnicas exploratórias multivariadas. Verificar a presença de interação genótipo-ambiente (GxE) nas características peso ao nascimento (PN), peso a desmama (PD), ganho de peso da desmama ao sobreano (GP), perímetro escrotal (PE) e musculosidade (MUS). Pela análise de agrupamento não-hierárquico foram agrupadas as regiões similares com relação às variáveis ambientais. Foram formados seis grupos de fazendas. A inclusão do efeito de interação touro-grupo foi avaliada em análises uni-característica. Comparou-se um modelo com o efeito de interação touro-grupo com outro sem esse efeito. Incluir o efeito de interação touro-GEO no modelo de avaliação genética do PN, PD e PE não resultou melhor ajuste aos dados, no entanto não deve ser descartada a hipótese de se considerar outros tipos de efeitos de GxE. Foram estimados parâmetros genéticos por meio de análises multi-característica, considerando-se a mesma característica como diferente em cada grupo de fazendas. Foi verificada heterogeneidade de variância para todas as características. Os coeficientes de herdabilidade nos grupos de fazendas para PN, PD, GP, PE e MUS variaram de 0,15 a 0,25; 0,16 a 0,25; 0,10 a 0,20; 0,17 a 0,31 e 0,17 a 0,24, respectivamente. As correlações genéticas variaram de 0,19 a 0,90 para PN, -0,02 a 0,92 para PD, 0,31 a 0,93 para GP, 0,64 a 0,89 para PE e de 0,18 a 0,80 para MUS nos grupos fazendas. As diferentes estimativas de herdabilidade obtidas entre grupos de fazendas implicam resposta à seleção diferenciada conforme o ambiente em que os animais são criados e selecionados. Pelas correlações genéticas entre as características nas diversas regiões, constatou-se GxE, indicando que os melhores reprodutores para uma determinada região não são sempre os mesmos para as demais. Um modelo hierárquico de norma de reação sob abordagem Bayesiana também foi utilizado para estimação dos componentes de variância, parâmetros genéticos e verificação da existência de GxE. Os gradientes ambientais baseados nas soluções para o efeito de grupo de contemporâneos para PN, PD, GP e PE foram -6,45 a +4,75 kg, -65 a +65 kg, -72 a +112 kg e -6.5 a +5.5 cm, respectivamente. As estimativas de herdabilidade foram crescentes no gradiente ambiental, PN (0,04 a 0,55), PD (0,39 a 0,47), GP (0,01 a 0,43) e PE (0,21 a 0,23). A correlação entre o nível e a inclinação da norma de reação para PN e GP foi de alta magnitude, indicando que os animais de maior valor genético médio foram os que apresentaram maior resposta à melhoria das condições ambientais, caracterizando o efeito de escala da GxE. Para PD e PE, a correlação entre intercepto e inclinação foi baixa implicando reclassificação dos animais em ambientes diferentes. O modelo hierárquico de normas de reação foi útil para descrever alterações nos componentes de variância decorrentes do ambiente e para descrever a presença de GxE nas características estudadas de bovinos compostos. Existe variação genética com respeito à sensibilidade dos animais, o que possibilita a seleção de genótipos mais plásticos ou mais robustos. / The objectives of this study were to characterize and define homogenous production environments of composite beef cattle in Brazil in terms of climatic and geographic variables using multivariate exploratory techniques; to evaluate the presence of genotype by environment interaction (GxE) for birth weight (BW), weaning weight (WW), postweaning gain (PWG), scrotal circumference (SC) and muscling. Nonhierarchical cluster analysis was used to group farms located in regions with similar environmental variables into clusters. Six clusters of farms were formed. The effect of sire-cluster interaction was tested by single-trait analysis. The inclusion of sire-cluster interaction in the genetic evaluation model may not result in better fit to the data for BW, WW and SC. Genetic parameters were estimated by multiple-trait analysis considering the same trait to be different in each cluster. The heritability coefficient in the clusters for BW, WW, PWG, SC and muscling ranged from 0.15 to 0.25; 0.16 to 0.25; 0.10 to 0.20; 0.17 to 0.31 and 0.17 to 0.24, respectively. The genetic correlations ranged from 0.19 to 0.90 for BW, -0.02 to 0.92 for WW, 0.31 to 0.93 for PWG, 0.64 a 0.89 for SC and 0.18 to 0.80 for muscling in the clusters of farms. The different heritability estimates between groups of farms indicates that the response to selection varies with the environment in which animals are selected. The low genetic correlations between traits in the different regions demonstrated the presence of GxE, indicating that the best sires in a certain region are not the same for the other regions. A reaction norm hierarchical model using Bayesian approach was also used for estimation of variance components, genetic parameters and to verify the existence of GxE. Environmental gradients based in solutions for the effect of contemporary groups for BW, WW, PWG and SC were -6.45 to +4.75 kg, -65 kg to +65, -72 to +112 kg and -6.5 to +5.5 cm, respectively. Heritability estimates were increasing in the environmental gradient, BW (0.04 to 0.55), WW (0.39 to 0.47), PWG (0.01 to 0.43) and SC (0.21 to 0.23). The correlation between the level and slope of reaction norm for BW and PWG was of high magnitude, indicating that animals of higher average breeding value were the ones which presented a best response to environmental improvement, characterizing a scale effect on GxE. For WW and SC, the correlation between intercept and slope was low implying reranking of animals in different environments. The reaction norm hierarchical model has been useful to describe changes in the variance components due to the environment and to describe the presence of GxE traits in composite beef cattle. There is genetic variation with respect to the sensitivity of the animals, which enables the selection of genotypes most plastics or more robust.

Com a finalidade de avaliar o programa de melhoramento da cana-de-açúcar, este trabalho teve como objetivos: determinar o grau da interação famílias x ambientes (FA); comparar métodos de seleção na etapa inicial em termos de resposta esperada à seleção; e propor o método de seleção sequencial modificado. Para isso, foram avaliados 4752 'seedlings' pertencentes a 33 famílias (irmãs germanas e meia irmãs), em dois locais do Estado de São Paulo, Piracicaba e Jaú. Os caracteres estimados foram altura e diâmetro dos colmos, número de perfilhos, Brix % caldo da cana, toneladas de cana por hectare (TCH) e toneladas de Brix por hectare (TBH). Para validar a resposta esperada à seleção sequencial modificada, 40 famílias foram selecionadas por este método e os clones foram multiplicados até a etapa III do programa de melhoramento, a fim de fornecer a resposta realizada à seleção. Os resultados mostraram interação FA significativa para todos os caracteres avaliados. A opção por uma seleção específica implicou em aumentos de ganhos de 4,1% e 5,8% para TBH, para Piracicaba e Jaú, respectivamente, em relação à seleção de famílias generalistas, ou seja, baseada na média dos dois locais. A seleção sequencial modificada, proposta neste trabalho e a seleção combinada através de índices apresentaram pouca vantagem em relação à seleção massal, na condição de seleção direta (sem avaliação visual). Nessas mesmas condições, a seleção sequencial tradicional e a australiana foram inferiores à seleção massal, em termos de progresso esperado, para todos os caracteres. Incluindo a seleção indireta (com avaliação visual) o método sequencial modificado foi sempre superior à seleção massal, para os principais caracteres avaliados (TCH e TBH). Para os métodos australiano e tradicional, houve vantagem apenas quando a correlação genética entre a avaliação visual e o caráter principal foi igual ou inferior a 0,7. Dessa forma, e devido à dificuldade prática de utilizar índices (seleção combinada), concluiu-se que o método proposto de seleção sequencial modificada é perfeitamente viável, sendo, portanto, recomendado. Na comparação entre as respostas estimada e a realizada com a seleção sequencial modificada, houve concordância dos resultados, na condição de correlação genética igual a 0,62. Isso veio reforçar a confiabilidade das estimativas obtidas no trabalho, bem como assegurar as vantagens do método proposto sobre os demais aqui apresentados. Se considerada uma correlação genética entre o critério de avaliação visual e o caráter principal variando entre 0,5 a 0,7, a superioridade do método proposto em relação ao massal é da ordem de 3% a 5%, em termos de ganho esperado para TCH, em um ciclo de seleção. Para alcançar estes mesmos ganhos com a seleção massal, ter-se-ia que aumentar a população inicial entre 36% a 100%. Alternativamente, mantendo os mesmos tamanhos populacionais, o número provável de clones no final do processo seletivo seria aumentado em 85% a 150%, apenas com a substituição do método de seleção massal pelo sequencial modificado na etapa inicial. / Aiming at the evaluation of the sugarcane breeding program, this study had the following objectives: to determine the degree of the family x environments interaction (FE); to compare selection methods in the initial phase, in terms of expected response to selection, and to propose the modified sequential selection method. For this, 4752 seedlings, originated from 33 families (full-sibs and half-sibs), were evaluated in two sites of the State of São Paulo: Piracicaba and Jaú. The traits evaluated were stalk height and diameter, number of stalks, Brix % cane juice, tons of cane per hectare (TCH) and tons of brix per hectare (TBH). To validate the expected response to the modified sequential selection, 40 families were selected by this method and the clones were multiplied until phase III of the breeding program and the realized gains in selection measured. The results showed significant FE interaction for all evaluated traits. The option for a site specific selection would increase gains by 4.1% and 5.8% for TBH in Piracicaba and Jau, respectively, in relation to the generalized family selections, that is, based on the average of the two sites. The proposed modified sequential selection and the combined selection using indices showed little advantage over the mass selection using direct selection (without visual selection). In these conditions, the traditional sequential selection and the Australian one were inferior to the mass selection, in terms of expected progress, for all traits. Allowing for the indirect selection (with visual selection) the modified sequential method was always superior to the mass for the main traits evaluated (TCH and TBH). The Australian method and the Traditional one, were advantageous in relation to the mass selection only when the genetic correlation between visual selection and the main trait was equal or inferior to 0.7. For this reason, and given the practical difficulty of applying indices (combined selection), it was concluded that the proposed method of modified sequential is perfectly viable and was reccommended. Comparing the estimated and realized response of the modified sequential selection, there was correspondence of results, given rG equals to 0.62. This result confirmed the reliance of the estimates obtained in this study, and assured the advantages of the proposed method over those compared. Considering a genetic correlation (rG), between the visual selection criterion and the main character, varying from 0.5 to 0.7, the superiority of the proposed method over the mass one, varies from 3 to 5%, in terms of expected gain, for TCH in one selection cycle. To obtain these same gains with mass selection, one would have to increase the initial population between 36% and 100%. Alternatively, keeping the same population sizes, the likely number in the end of the selection process would be increased in 85% to 150%, just with the substitution of the method of mass selection by the modified sequential one, in the initial phase.

A interação QTL por ambientes (QE) têm sido relatada como uma das principais causas de insucesso da seleção assistida por marcadores moleculares (SAM). Estudos que visam o melhor entendimento da interação QE podem contribuir para o aumento da eficiência dos programas de SAM. O objetivo deste trabalho foi mapear QTL para produção de grãos (PG), prolificidade (PROL), peso de 500 grãos (P500), comprimento (CE) e diâmetro de espiga (DE), profundidade de grão (PROF), número de fileiras (NFil) e de grãos por fileira (NGFil) em uma população de milho tropical, verificar a importância da interação QE para estes caracteres e avaliar a estabilidade dos efeitos genéticos dos QTL mapeados. Uma população de 256 progênies F2:3 obtida do cruzamento entre duas linhagens de grupos heteróticos distintos e contrastantes para diversos caracteres foi avaliada em 13 ambientes. Os ambientes foram alocados em grupo de ambientes utilizando um método de agrupamento e o modelo AMMI, sendo que ambos os métodos levaram a identificação de três grupos de ambientes. O mapeamento de QTL foi realizado considerando um mapa genético com 177 marcadores microssatélites e mapeamento por intervalo composto expandido para múltiplos ambientes (mCIM). As médias de grupo de ambientes para cada caráter foram utilizadas nas análises. Foram mapeados 87 QTL, sendo 9 para PG, 9 para PROL, 14 para P500, 7 para CE, 9 para DE, 14 para PROF, 17 para NFil e 8 para NGFil. A maioria dos QTL mapeados localizou-se em regiões genômicas que ainda não foram reportados QTL, tanto para germoplasma temperado quanto tropical. A interação QTL por grupo de ambientes foi significativa para PG e não significativa para os componentes de produção. Para PG, as metodologias QQE biplot e AMMI foram utilizadas para estudar a interação QE dos efeitos genéticos dos QTL. As estimativas dos efeitos aditivos e de dominância dos QTL foram influenciadas pela interação QTL por grupo de ambientes, sendo que o padrão de interação foi específico para cada efeito genético. A expressiva interação QTL por grupo de ambientes para PG e o padrão específico de interação dos efeitos genéticos dos QTL impõe desafios adicionais à incorporação da SAM nos programas de melhoramento que visam o desenvolvimento de genótipos produtivos e com estabilidade de produção. / QTL by environment (QE) interaction has been reported as one of the main reasons for the unsuccessful of marker-assisted selection (MAS). Studies aimed to a better understanding of QE interaction could contribute to increase the efficiency of MAS programs. The objectives of this study were to map QTL for grain yield (GY), prolificacy (PROL), 500 Kernels weight (W500), ear length (EL), ear diameter (ED), kernel depth (KD), row number per ear (RN) and kernels per row number (KRN) in a tropical maize population, to assess the importance of QE interaction for these traits and to evaluate the stability of the genetic effects of mapped QTL. A population of two-hundred and fifty-six progenies obtained from the cross between two inbred lines, which belong to different heterotic groups and divergent for different traits, was evaluated in 13 environments. The environments were jointed into groups using a cluster method and an AMMI model. Both methods led to the identification of three groups of environments. The QTL mapping was performed considering a genetic map with 177 microsatellites markers and the multiple-environment composite interval mapping analysis (mCIM). The means from each group of environments of each trait were used in the analyses. Eighty seven QTL were mapped: 9 for GY, 9 for PROL, 14 for W500, 7 for EL, 9 for ED, 14 for KD, 17 for RN and 8 for KRN. Most of the mapped QTL was located in genomic regions that have not been reported QTL in both temperate and tropical germplasm. The QTL by group of environments interaction was significant for GY and not significant for yield components. For GY, QQE biplot and AMMI methodologies were used to study the QE interaction of the genetic effects of the QTL. The estimates of additive and dominance effects of QTL were affected by QTL by group of environments interaction and the interaction pattern was specific for each genetic effect. The large QTL by group of environments interaction for GY and the specific interaction pattern of the genetic effects of QTL impose additional challenges for the incorporation of MAS in breeding programs that aim to develop high yielding genotypes with grain yield stability.