Modificação da metodologia de ‘m + a’ e uso do pedigree para seleção precoce de progênies de soja

Abstract

High efforts are now being made at the stages of soybean crop improvement, which is constantly expanding in the country and in the world. Therefore, when generating segregating progenies, it is necessary a costly and time consuming phase of homozygosis increase to identify those with potential for extraction of new strains. The m + a methodology, which estimates the concentration of favorable alleles in two consecutive generations, is an alternative for this, however, demands a lot of resources to perform. A modification of this methodology to circumvent these “barriers” can be very useful for breeding programs. In this context, the objective was to propose a modification to this methodology in order to make it more easily executable. To obtain the phenotypic data, simultaneous experiments were performed, required by the traditional methodology, and sequential, by the modified methodology, with three consecutive levels of homozygosity: F2, F3 and F4. These experiments were carried out in the localities of Luiziana/PR and Campo Mourão/PR, in randomized blocks with three replications and as treatments the 20 progenies in the aforementioned homozygosity levels plus four commercial controls. To obtain the data used in the traditional methodology, progenies at the three homozygosity levels were evaluated simultaneously in the two mentioned locations. To generate data for the proposed modification of the methodology, sequential experiments with these levels were installed at different times, together with the checks in each one of them. Data regarding progenies kinship were used for matrix assembly and creation of predictive scenarios. The traits evaluated for both methodologies were: days to flowering (DF); plant height (PH); days to maturation (DM); 1000 seed weight (TSW) and yield (YLD). The data were analyzed by the R software. Using estimates of the best linear unbiased estimation (BLUEs) of progenies at homozygosity levels, superior correlations were obtained between F2 and F3 generations for most of the characters, except for PMS character in both methodologies. The correlations obtained with the estimates of m + a were lower than those observed with the BLUEs in both methodologies. Contrasts with positive or negative correlations also coincided between the methodologies. Estimates of m + a between F2 and F3 generations to predict progeny averages in F4 did not show strong enough correlations to make accurate inferences about the latter generation based on those. The predictive scenarios elaborated with the incorporation of the kinship matrix for the generation of BLUPs were also not informative to predict the behavior of the generations. There was similarity between the results obtained by the traditional m + a methodology and the proposal suggested in this study based on the magnitudes and directions (signals) of the generated estimates. However, the traditional methodology generated higher predictive capacities for YLD and PH, while the modified one for DM and TSW, which were also the highest correlations obtained overall. It is very likely that neglected epistatic interactions in the design of the method were crucial for the expression of highly complex characters and, as a result, the predictions showed low accuracy for the progenies. Finally, it was found that over the generations it is better to evaluate to select progenies based on adjusted averages/BLUEs than with estimates of m + a, either by the traditional methodology or by the proposed modification. For more consistent and complementary results regarding the methodology, the use of genotyping data for pedigree incorporation in the analysis is suggested.