Estratégias para seleção de progênies em soja

Abstract

In soybean breeding programs usually the merit of the population is not considered in the selection of progenies or lines. Traditionally the selection of the best genotypes is performed using only the reference generation, i.e., it is not included the evaluations of the previous generations. Thus, the aim of this study was to verify the influence of population merit in the estimation of phenotypic and genetic components and identification of the best soybean lines, and to quantify the efficiency of the sequential analysis in the selective process of soybean lines. Grain yield (bags/ha) and absolute maturity (days) data from soybean breeding program of the Dupont company - Pioneer Division were used. The experiments were carried out in the crop seasons 2012/2013, 2013/2014 e 2015/2016 in 17 cities in the Mato Grosso do Sul, Paraná, Santa Catarina e Rio Grande do Sul states. We estimated the parameters genetic gain with selection, correlation, coincidence index, realized gain and correlated response, considering and ignoring the population merit, under seven selection intensities (1%, 5%, 10%,15%, 20%, 25% e 30%). To evaluate the efficiency of sequential analysis we considered the following strategies: A) analysis considering only the reference generation (VCU trial); B) sequential analysis considering the combination of Wtest and VCU experiments; C) sequential analysis considering the X-test, W-test and VCU experiments; D) sequential analysis considering the Y-test, X-test, Wtest and VCU experiments; E) sequential analysis considering the Z-test, Y-test, X-test, W-test and VCU experiments; F) sequential analysis considering the Xtest and VCU experiments; G) Sequential analysis considering the Z-test, X-test and VCU experiments. In the strategies B, C, D and E were used all data of the experiments, while for strategies F and G only data of thirty common genotypes from the Z-test, X-test and X-VCU experiments were used. Furthermore we consider the following situations to assess the efficiency of the sequential analysis: 1) selection of the top-ten lines in strategy A and check the ranking of the same lines considering the strategies B, C, D and E; 2) Coincidence selection of the top-ten lines in each sequential analysis strategy (B, C, D and E); 3) Selection of the top-ten common lines form experiments Z-test, X-test and VCU. The parameters variance components, heritability and experimental variation coefficient were better estimated when the population merit was considered, providing consequently greater gain with selection for both traits (grain yield and absolute maturity). The coincidence and ranking among the selected progenies considering and ignoring the population merit were of greater in more advanced generations of inbreeding and higher intensity selection. There was change in the ranking of lines when considering the sequential analysis involving the previous generations relative to reference generation, showing its influence on the recommendation of new cultivars. This change is most evident in higher unbalance conditions.