Correlations and canonical variables applied to the distinction of soybean cultivars in a tropical environment

Copyright: © 2022 Agronomy Science and Biotechnology. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, since the original author and source are credited. The objective of this study was to evaluate the performance of soybean cultivars through their correlations and canonical variables in a tropical environment. This study was conducted in the municipality of Mineiros, GO, Brazil. The experimental design used was randomized blocks with four replications using 10 soybean cultivars (Bônus, Desafio, Flecha, Foco, ICS7019, M5917, M7110, Power, ST721 and ST797). During the conduct of the experiment, pest control was carried according to good practices and integrated management. At the end of the cycle of each cultivar, 10 plants were collected at random and then the agronomic attributes were taken. The data obtained were submitted to the assumptions of the statistical model, verifying the normality and homogeneity of the residual variances, as well as the additivity of the model. Univariate and multivariate models were used. The analysis were performed on the Rbio software using the R plattform in addition to the Software Genes. According to the summary of analysis of variance, it was observed that all cultivars differed for all traits. The soybean cultivars Flecha and M5917 showed the highest yields in a tropical environment. The cultivars showing a strong correlation between the number of grains per plant and yield. And the canonical correlations tools were efficient and complementary in the data analysis.


INTRODUCTION
The search for genetic material resistant or tolerant to pests, diseases and herbicides always associated with a high productive potential is what large companies seek, which annually launch two to three new materials on the market. The study of the adaptation of cultivars to different microclimates becomes important for the knowledge of the best productive material and its behavior in the face of environmental variations (Gaviraghi, Pellegrin, Werner, Bellé, & Basso, 2018). Furthermore, the use of strategies such as estimating variance components and genetic parameters of aspects of agronomic importance contribute to genotype selection based on the agronomic ideotype and thus promote better crop performance (Barbosa et al., 2021).
In this way, the importance of soybean for Brazil is great, and in order to obtain high levels of yield, the challenge is to plant a cultivar that has good quality, and that best adapts to the region, so given the above, this study aimed evaluate the performance of soybean cultivars through their correlations and canonical variables in a tropical environment.

MATERIAL AND METHODS
The study was conducted at the Luís Eduardo de Oliveira Salles Experimental Farm, belonging to the Mineiros University Center -UNIFIMES, rural area of the municipality of Mineiros, GO, Brazil. Geographically it is at 17º 58 'S latitude and 45º 22' W longitude and approximately 800 m altitude. Average temperature of 22.7 °C and average annual rainfall of 1695 mm, occurring mainly in spring and summer. The experimental area is classified as Aw type (hot to dry) (Köppen & Geiger, 1936).
The experimental design used was randomized blocks with four replications using 10 soybean cultivars (Bônus, Desafio, Flecha, Foco, ICS7019, M5917, M7110, Power, ST721 and ST797). The experimental plots were composed of 4 lines of 5 meters long, with a spacing of 0.45 m between the lines, the useful area of the plot will be 9 m 2 , with a density of 15 seeds per meter of furrow. The main morphoagronomic traits of soybean cultivars were described in (Table 1).
Before planting, pre-planting desiccation was performed (Cobucci, Stefano, & Kluthcouski, J). It was used 450 kg ha -1 of fertilizer 05-25-15 applied in the furrow and in a single dose next to the seeding. Sowing was carried out on November 8, 2018 (Ferreira, Amaral, Silva, Curvelo, & Pereira, 2019). During the conduct of the experiment, the control of pests, diseases and weeds were carried out as they became necessary, respecting good practices and integrated management (Quintela, 2001).
At the end of the cycle of each cultivar, 10 plants were collected at random from the useful area of the experimental plot, thus, the following agronomic attributes of the cultivars were evaluated: POG: pods with one grain in units; PTWG: pods with two grains in units; PTHG: pods with three grains in units; PFG: pods with four grains in units; NGP: number of grains per pod in units; NPP: number of pods per plant in units; GP: number of grains per plant in units; and YI: yield in sc ha -1 . The data obtained were submitted to the assumptions of the statistical model, verifying the normality and homogeneity of the residual variances, as well as the additivity of the model. Afterwards, the analysis of variance and application of the Scott-Knott averaging test was performed, at 5% probability. Subsequently, the variables were subjected to Pearson's linear correlation in order to understand the association trend, with its significance based on a 5% probability by the t test. Canonical correlations were estimated between group 1 (GP and YI) and group 2 (POG, PTWG, PTHG, PFG, NGP and NPP), with significance between the groups of traits assessed based on the chi-square statistic. After the genetic dissimilarity was carried out by the Mahalanobis algorithm, where the residual matrix was weighted, the distance dendrogram was constructed using the UPGMA cluster, then the biplot canonical variables method was used, where it was possible to visualize the general variability of the experiment and the multivariate trends. The analysis were performed on the Rbio software using the R plattform (Bhering, 2017), in addition to the Software Genes (Cruz, 2016).

RESULTS AND DISCUSSION
According to the summary of analysis of variance, it was observed that all cultivars differed for all traits analyzed (p <0.01). Reliability was observed in the values of CV's with low values (CV<10) for NGP: number of grains per pod; medium (CV 10-20) in POG: pods with one grain, PTWG: pods with two grains; PTHG: pods with three grains, NPP: number of pods per plant, GP: number of grains per plant and YI: yield; in addition to high (CV>20) for PFG: pods with four grains (Table 2). It was corroborated with Bohn et al. (2016), Torres, Silva and Teodoro (2015), Castro, Kouri, Alves and Silva-Neto (2014), and Ribeiro et al. (2016).
The cultivars Bônus, Foco and ST797 showed higher averages for the variables of POG and PTWG with values of 14.90 and 24.50 units, respectively. The PTHG average was very expressive for all cultivars with the exception of Bônus, which delivered only 6.45 units plant-1. Bônus, Desafio, Flecha and Foco were the genetic materials that had the highest PFG (0.93 units plant -1 ) ( Table 3). Castro et al. (2014) observed lower means of POG, PTWG and PTHG. Similar results were found by Bohn et al. (2016), which observed that soybean cultivars with higher and lower PTHG, showed statistically similar yield. Thus, these results demonstrate that the plants have the potential to compensate for variations in the number of pods, increasing the weight of grains, thus enabling satisfactory yield.   Regarding the NPP and GP variables, the cultivars Flecha, Foco, M5917 and ST797 presented the highest averages with 57.50 and 125.46 units plant -1 , in this sequence. Superior results were obtained by Torres et al. (2015) and Scheffler, Perleberg, Rodrigues and Kuhn (2016), who found NPP values above 60 pods per plant.
When the YI variable was analyzed, two cultivars Flecha and M5917 stood out, reaching a mean of 99.93 and 105.95 sc ha -1 . Similar results were found by Scheffler et al. (2016) and lower means were found by Borges et al. (2018), Bohn et al. (2016), Doná et al. (2019) attributed the averages below the climatic factor. The wide variation in yield observed between the cultivars evaluated, shows the existence of different levels of adaptability of these materials to local environmental conditions and reinforces the importance of continuing the work of evaluating cultivars . The agronomic performance of soybean genotypes occurs due to genetic constitution, environmental conditions and genotype x environment interaction, therefore, studies are needed to promote the proper positioning of each genotype in specific environments .
One of the genetic breeding strategies, as well as genotype positioning, is the use of Pearson's linear correlation, in order to estimate the direction and degree of linear association between two random traits . Pearson's correlation coefficients arranged in the correlation network revealed 19 correlations among soybean variables. Positive correlations were diagnosed in the pairs (POGxPTWG), (POGxNPP), (POGxGP), (POGxYI), (PTWGxNPP), (PTWGxGP), (PTWGxYI), (PTHGxNGP), (PTHGxNPP), (PTHGxGP), (PTHGxYI), (NPPxGP), (NPPxYI) and (NGPxYI) and negative in pairs (POGxNGP), (PTWGxNGP), (PTHGxPFG), (NGPxNPP) (Figure 1). The association between agronomic traits is important because it allows verifying the degree of interference of a trait on another of economic interest, as well as practicing indirect selection (Zuffo et al., 2016). The use of correlation networks can increase the effectiveness of selection in soybean breeding. It allows to quickly identify the pairs of traits that present correlations of greater magnitude, to determine which groups of variables influence in a more expressive way the most important characters for the breeding program and to identify the groups of correlated variables.
The two canonical correlations and their respective canonical pairs were significant (p≤0.01) by the chi-square test (Figure 2), as well, verified in Zanatto et al. (2016). The high magnitude of the canonical correlation coefficients (r = 1.00 in the first and r = 0.91 in the second) showed a high dependency between the two groups of characters (Figure 2).
The analysis of canonical correlations showed that the increase in POG, PTWG, PTHG and NPP, in addition to the reduction in PFG and NGP, potentiate the increase in the GP variable. However, to raise the YI, it became necessary for the soybean plants to have the highest number of PTWG, PFG and NPP (Figure 2). The correlation 0.77 ** YI PTWG GP NPP PTHG POG NGP PFG studies allow to identify and quantify the associations of morphological and productive characters with the performance of the crops (Carvalho et al., 2015). From the dissimilarity matrix, it was possible to generate a dendrogram using the UPGMA clustering methodology, using the generalized Mahalanobis distance. In this methodology, the distance matrix between individuals in the population is calculated and then the most distant individuals are grouped.
Among the soybean cultivars analyzed in the dissimilarity dendrogram, two groups were generated, with a highlight for the group formed by the cultivars Desafio, M7110 and Power and the other cultivars presenting similar traits being in the second group (Figure 3). Santos et al (2015) and Rigon (2015) also found formation of different groups among soybean cultivars. The formation of groups of soybean allows, to the producers, options in the decision making for the choice of the cultivar.
The canonical axes add up to a total of explanation equivalent to 97% of the total variation of the data. Similar results were evidenced by Szareski et al. (2016), where canonical variables explained 85.05% of the existing genetic variation. The variables GP and NPP showed similarities to each other, where the cultivar Foco presented the highest GP. In the NGP variable, the closest cultivar was M7110, whereas YI was the only variable in the negative axis, where the Bônus cultivar stood out (Figure 4). Silva et al. (2015) concluded that multivariate analysis methodologies are efficient to verify similarities or differences in yield variability, based on the chemical and physical attributes of the soil in the studied area. Also being added the influence of soybean genetic variability and seed treatment on the performance of the initial grubbing of its seedlings.
Univariate analyzes revealed significance between soybean cultivars for all variables analyzed, with Pearson's correlations ranging from 0.36 to 0.93 between positive and negative. In the multivariate, cultivar trends were diagnosed in the canonical variables tool, formation of significant groups in the canonical correlations, as well as distinctions of the dendrogram.