Resumo -O objetivo deste trabalho foi alterar o método centroide de avaliação da adaptabilidade e estabilidade fenotípica de genótipos, para deixá-lo com maior sentido biológico e melhorar aspectos quantitativos e qualitativos de sua análise. A alteração se deu pela adição de mais três ideótipos, definidos de acordo com valores médios dos genótipos nos ambientes. Foram utilizados dados provenientes de um experimento sobre produção de matéria seca de 92 genótipos de alfafa (Medicago sativa) realizado em blocos ao acaso, com duas repetições. Os genótipos foram submetidos a 20 cortes, no período de novembro de 2004 a junho de 2006. Cada corte foi considerado um ambiente. A inclusão dos ideótipos de maior sentido biológico (valores médios nos ambientes) resultou em uma dispersão gráfica em forma de uma seta voltada para a direita, na qual os genótipos mais produtivos ficaram próximos à ponta da seta. Com a alteração, apenas cinco genótipos foram classificados nas mesmas classes do método centroide original. A figura em forma de seta proporciona uma comparação direta dos genótipos, por meio da formação de um gradiente de produtividade. A alteração no método mantém a facilidade de interpretação dos resultados para a recomendação dos genótipos presente no método original e não permite duplicidade de interpretação dos resultados.Termos para indexação: Medicago sativa, análise gráfica, componentes principais, interação genótipos x ambientes. Alteration of the centroid method to evaluate genotypic adaptabilityAbstract -The objective of this work was to modify the centroid method of evaluation of phenotypic adaptability and the phenotype stability of genotypes in order for the method to make greater biological sense and improve its quantitative and qualitative performance. The method was modified by means of the inclusion of three additional ideotypes defined in accordance with the genotypes' average yield in the environments tested. The alfalfa (Medicago sativa L.) forage yield of 92 genotypes was used. The trial had a randomized block design, with two replicates, and the data were used to test the method. The genotypes underwent 20 cuts, from November 2004 to June 2006. Each cut was considered an environment. The inclusion of ideotypes of greater biological average production in the environments produced an arrow-shaped graphical dispersion directed to the right in which the most productive genotypes were placed near the tip of the arrow. With the alteration only five genotypes were classified into the former classes of the original centroid method. The arrow-shaped figure allowed a direct comparison of genotypes throughout the productivity gradient. The alteration performed in the method preserved the easy interpretation of results for genotype recommendations of the original method, and does leaves no room for ambiguity in interpretation of the results.

Analysis using Artificial Neural Networks has been described as an approach in the decision-making process that, although incipient, has been reported as presenting high potential for use in animal and plant breeding. In this study, we introduce the procedure of using the expanded data set for training the network. Wealso proposed using statistical parameters to estimate the breeding value of genotypes in simulated scenarios, in addition to the mean phenotypic value in a feed-forward back propagation multilayer perceptron network. After evaluating artificial neural network configurations, our results showed its superiority to estimates based on linear models, as well as its applicability in the genetic value prediction process. The results further indicated the good generalization performance of the neural network model in several additional validation experiments.

ABSTRACT. The correct classification of individuals is extremely important for the preservation of genetic variability and for maximization of yield in breeding programs using phenotypic traits and genetic markers. The Fisher and Anderson discriminant functions are commonly used multivariate statistical techniques for these situations, which allow for the allocation of an initially unknown individual to predefined groups. However, for higher levels of similarity, such as those found in backcrossed populations, these methods have proven to be inefficient. Recently, much research has been devoted to developing a new paradigm of computing known as artificial neural networks (ANNs), which can be used to solve many statistical problems, including classification problems. The aim of this study was to evaluate the feasibility of ANNs as an evaluation technique of genetic diversity by comparing their performance with that of traditional methods. The discriminant functions were equally ineffective in discriminating the populations, with error rates of 23-82%, thereby preventing the correct discrimination of individuals between populations. The ANN was effective in classifying populations with low and high differentiation, such as those derived from a genetic design established from backcrosses, even in cases of low differentiation of the data sets. The ANN appears to be a promising technique to solve classification problems, since the number of individuals classified incorrectly by the ANN was always lower than that of the discriminant functions. We envisage the potential relevant application of this improved procedure in the genomic classification of markers to distinguish between breeds and accessions.

No abstract

Based upon qualitative parameters experiments, this study aims to investigate how the elements of the environment, where the coffee is produced, contribute to the final quality of the product. For the analyses, it was used approximately one kilogram of coffee cherry samples collected in 14 municipalities previously chosen on the East side of the Minas Gerais State, Brazil. The coffee cherry samples were collected and analyzed for each of the two varieties in four levels of altitude for each exposure side of the mountain in relation to the Sun. The quality of the coffee was evaluated through the analysis of its physical characteristics and sensory analysis, popularly known as "Test of drink or Cupping" carried out by three tasters that belonging to the group of Q-Graders, according to the rules of national and international competitions of the Brazilian Association of Special Coffees (BSCA). Were performed analysis by means descriptive statistics, analysis of variance and multivariate analysis, all of them aiming to study the individual sensory characteristics of quality of the coffee beverage from the "Matas de Minas" region. Path coefficient analysis also was carried out for the partition of the phenotypic correlation coefficients into measures of direct * Corresponding author. W. P. M. Ferreira et al. 1292and indirect effects, in order to determine the individual sensory characteristics that played a major role in the beverage final score. The results demonstrate that it is not possible to conclusively establish the differences among coffees evaluated with basis on varieties and environmental factors previously cited. It can be concluded that it is not recommended to associate the quality of coffee only to a specific factor whether from the environment or being it a specific of the culture of coffee. However, the cafes that were evaluated had intrinsic quality, which were derived from the specific characteristics of the "Matas de Minas" region where they were grown.

Artificial neural networks (ANN) are computational models inspired by the neural systems of living beings capable of learning from examples and using them to solve problems such as non-linear prediction, and pattern recognition, in addition to several other applications.In this study, ANN were used to predict the value of the area under the disease progress curve (AUDPC) for the tomato late blight pathosystem. The AUDPC is widely used by epidemiologic studies of polycyclic diseases, especially those regarding quantitative resistance of genotypes.However, a series of six evaluations over time is necessary to obtain the final area value for this pathosystem. This study aimed to investigate the utilization of ANN to construct an AUDPC in the tomato late blight pathosystem, using a reduced number of severity evaluations. For this, four independent experiments were performed giving a total of 1836 plants infected with Phytophthora infestans pathogen. They were assessed every three days, comprised six opportunities and AUDPC calculations were performed by the conventional method. After the ANN were created it was possible to predict the AUDPC with correlations of 0.97 and 0.84 when compared to conventional methods, using 50 % and 67 % of the genotype evaluations, respectively. When using the ANN created in an experiment to predict the AUDPC of the other experiments the average correlation was 0.94, with two evaluations, 0.96, with three evaluations, between the predicted values of the ANN and they were observed in six evaluations. We present in this study a new paradigm for the use of AUDPC information in tomato experiments faced with P. infestans. This new proposed paradigm might be adapted to different pathosystems.

-In Brazil the rust caused by Puccinia psidii Winter stands out as the most important disease of eucalyptus. The use of resistant genotypes is the main control method, which makes the detection of markers linked to rust resistance essential to the selection of resistant genotypes. In this study, an F 1 progeny of 131 plants from interspecific crossings of Eucalyptus was used to identify markers linked to resistance genes for this pathogen. An integrated map was constructed for linkage group three based on microsatellite markers. For QTL mapping two methodologies based on alleles identical-by-descent (IBD) were used: single marker analysis of Haseman and Elston and the interval mapping procedure of Fulker and

ABSTRACT. Artificial neural networks have shown great potential when applied to breeding programs. In this study, we propose the use of artificial neural networks as a viable alternative to conventional prediction methods. We conduct a thorough evaluation of the efficiency of these networks with respect to the prediction of breeding values. Therefore, we considered eight simulated scenarios, and for the purpose of genetic value prediction, seven statistical parameters in addition to the phenotypic mean in a network designed as a multilayer perceptron. After an evaluation of different network configurations, the results demonstrated the superiority of neural networks compared to estimation procedures based on linear models, and indicated high predictive accuracy and network efficiency.