Trait profiling and genotype selection in oilseed rape using genotype by trait and genotype by yield*trait approaches

Abstract Selection and breeding for high‐yielding in oilseed rape have always been one of the leading objectives for oilseed rape breeders. This process becomes more complicated when all quantitative traits are considered in selection in addition to grain yield. In the present study, 18 oilseed rape genotypes along with 2 check cultivars (RGS003 and Dalgan) were evaluated across 16 environments (a combination of 2 years and eight locations) in the tropical climate regions of Iran during 2018–2019 and 2019–2020 cropping seasons. The experiments were conducted in a format of randomized complete block design (RCBD) with three replications. The obtained multienvironmental trial data were utilized to conduct multivariate analysis, genotype by trait (GT) biplot, and genotype by yield*trait (GYT) biplot (Breeding, Genetics and Genomics, 1:2019). The GT and GYT biplot accounted for 55.5% and 93.6% of the total variation in the first two main components. Based on multivariate analysis and GT biplot, pod numbers in plant (PNP) and plant height (PH) were chosen as two key traits in spring oilseed rape genotypes for indirect selection due to high variation, strong positive correlation with grain yield (GY), and their high representatively and discriminability in genotype selection. The mean × stability GT biplot represented G10 (SRL‐96‐17) as the superior genotype. Based on the mean × stability GYT biplot, eight above‐average genotypes were identified that took high scores in stability, high‐yielding, and all evaluated quantitative traits at the same time. Based on the superiority index of GYT data, G10 (SRL‐96‐17) and G5 (SRL‐96‐11) indicated the best yield–trait combinations profile and ranked above check cultivars and then selected as superior genotypes. Similarly, cluster analysis using the WARD method also separated eight superior genotypes. Based on the result of the present study, GT ad GYT methodologies are recommended for trait profiling and genotype selection in oilseed rape breeding projects, respectively.


| INTRODUC TI ON
Oilseed rape (Brassica napus L.) is the second worldwide major supplier of edible oil, after soybeans (Liersch et al., 2020). The oilseed rape grain production reached almost 71 million tons during the 2018-2019 cropping season (Liersch et al., 2020). In Iran, however, oilseed rape is the first major oilseed cultivated crop (Amiri Oghan et al., 2016). Based on reported data in FAOSTAT, 289996 t of oilseed rape was harvested in Iran from an area of approximately 140,000 ha during the 2019-2020 cropping season, with an average yield of about 2074 kg/ha (FAO, 2018).
Grain yield is a complex trait and the main purpose of plant breeders is to identify the foundation of relationships between grain yield and other traits to increase grain production (Zulfiqar et al., 2021).
Achieving high yields is one of the most important goals in plant breeding programs. However, there are three key challenges in the process of genotype evaluation that limit the success of plant breeders: first genotype by environment interaction (GE), second the reverse or unfavorable relationship among key traits, and third the high complexity of the key traits like yield (Kendal et al., 2019;Sofi et al., 2022;Yan & Frégeau-Reid, 2018).
Different models are developed to overcome this challenge. The first is independent culling which discards genotypes fail to meet the minimum required value for a trait, regardless of how well the genotype is for other traits. The second is index selection which ranks genotypes based on index values of a linear combination of the target traits (Sofi et al., 2022;Yan & Frégeau-Reid, 2018). The major difficulty associated with these strategies is their high subjectivity. The weight and truncation points can potentially vary from researcher to researcher and from time to time for the same researcher, even for the same dataset (Yan & Frégeau-Reid, 2018).
Regarding GGE-biplot analysis, new methods are developed like a genotype-by-trait (GT) biplot which visualizes the genotype relationship with traits in a biplot to represent the strengths and weaknesses of the genotypes (Yan & Rajcan, 2002). This methodology has been widely used by many researchers to find the traits relationship in different plants (Dehghani et al., 2008;Gouveia et al., 2022;Karahan & Akgun, 2020;Santana et al., 2021;Sofi et al., 2022;Tsenov et al., 2021).
Despite the advantages of GT biplot in identifying the interrelationships among the traits of genotypes and trait profiles, this methodology could not prepare enough results for the breeders about the selection or elimination of genotypes. To wipe out these deficiencies and increase the efficiency of genotype selection, the GYT biplot methodology was developed (Yan & Frégeau-Reid, 2018). This methodology afforded a novel comprehensive and effective approach to evaluate the genotypes based on multiple traits through a graphical ranking of genotypes. The feature of this strategy is combining yield with various target traits which made it possible to represent the strengths and weaknesses of the genotypes at the same time.
Based on the GYT biplot, yield is considered the constant trait in the determination of the efficiency of a genotype by itself, while other traits are valuable to producers only when they are combined with sufficiently good yield levels (Yan & Frégeau-Reid, 2018).
In the present study, 18 spring oilseed rape genotypes along with the two check cultivars were studied based on GT and GYT approaches. This study aimed to define the interrelationship among the traits, the association among genotypes and traits, and the ranking of genotypes based on multiple traits.

| Plant materials and multienvironment trials
A total of 20 open-pollinated spring oilseed rape genotypes (Table 1) were evaluated for their quantitative traits including grain yield (GY), days to start flowering (DSF), days to end flowering (DEF), days to maturity (DM), flowering period (FP), plant height (PH), pod numbers in plant (PNP), grain numbers in pod (GNP), one thousand grain weight rape genotypes for indirect selection due to high variation, strong positive correlation with grain yield (GY), and their high representatively and discriminability in genotype selection. The mean × stability GT biplot represented G10 (SRL-96-17) as the superior genotype. Based on the mean × stability GYT biplot, eight above-average genotypes were identified that took high scores in stability, high-yielding, and all evaluated quantitative traits at the same time. Based on the superiority index of GYT data, G10  and G5 (SRL-96-11) indicated the best yield-trait combinations profile and ranked above check cultivars and then selected as superior genotypes. Similarly, cluster analysis using the WARD method also separated eight superior genotypes. Based on the result of the present study, GT ad GYT methodologies are recommended for trait profiling and genotype selection in oilseed rape breeding projects, respectively.

K E Y W O R D S
GT biplot, GYT biplot, high-yielding, multi environmental trial (TGW). The experiment has been conducted across 16 environments (combination of years and locations) in the tropical climate of Iran including north and south tropical regions (Gorgan, Sari, Moghan, Behbahan, Borazjan, Dezfoul, Zabol, and Hajiabad) during 2018-2019 and 2019-2020 cropping seasons in a format of randomized complete block design (RCBD) with three replications. More description of these environments is presented in Table 2. Each plot consisted of four rows five-meter long with a spacing of 30 cm between rows and 5 cm within rows. The spacing of 60 cm was considered between the plots. Seeds of all genotypes were sown with an experimental sowing machine.
The amount of seed consumption was 6 kg/ha which was sown according to the instructions on the suitable dates in each region.

| Statistical analysis
In this study, 18 new promising oilseed rape spring genotypes along with 2 check spring cultivars were evaluated. Multivariate analysis including Pearson correlation, Ward cluster, and path analysis was conducted in this study to find the relationship between the assessed genotypes traits.

| Genotype by trait (GT) biplot methodology
The data from 16 environments (a combination of 2 years and eight locations) were utilized to prepare the genotype by trait (GT) biplot (Yan & Tinker, 2006) and genotype by yield*trait (GYT) biplot (Yan & Frégeau-Reid, 2018). The Genotype-by-trait (GT) table was created by the mean value of 16 environments' data for each genotype trait.
Genotype-by-trait (GT) biplot methodology visualizes traits and genotype relationships in a biplot to represent the strengths and weaknesses of traits and genotypes (Yan & Rajcan, 2002). Based on the GT biplot, acute, obtuse, and right angles of traits vectors represent positive, negative and zero (no) correlation, respectively. Furthermore, a relatively short vector of the GT biplot indicates the low variation of traits across genotypes and vice versa (Yan & Rajcan, 2002).

| Genotype by yield*trait (GYT) biplot methodology
Using the GT table, the genotype by yield*trait (GYT) biplot (Yan & Frégeau-Reid, 2018) was constructed in this study. To provide the GYT  The GYT data were standardized for each yield-trait combination as represented in the following formula and then integrated to calculate the superiority index (SI) for each genotype (Yan & Frégeau-Reid, 2018). The high value of SI indicated its superiority (Yan & Frégeau-Reid, 2018).
Standardized yield*trait value = Genotype yield _ trait value − Yield _ trait mean of all genotypes Yield _ trait Standard deviation of all genotypes

| Traits profiling based on Pearson correlation and path analysis
The mean of traits (genotype by trait) data across 16 environments (combination of 2 years and eight locations) for 20 oilseed rape genotypes are shown in Table 3. The highest variation belonged to pod numbers in plant (PNP), grain numbers in pod (GNP), and grain yield (GY), respectively. The Pearson correlations among all evaluated traits TA B L E 1 Code, name, and origin of the tested oilseed rape genotypes.

No
Code Name Origin

| Genotype by trait (GT) biplot
The trait-standardized genotype by trait (GT) data was utilized in the current study to represent the trait profiles by GT biplot. The GT biplot accounted for 55.5% of the total variation of data. The first and second principal components accounted for 32.9% and 22.6% of the variation in data, respectively ( Figure 3).

| Uncovering the relationship of traits based on GT biplot
To grouping the studied traits and genotypes, the polygon view or which-won-where biplot of GT data is demonstrated in Figure 2a.

| Determining ideal traits to identify superior genotypes
The studied traits were ranked based on both discriminating ability and representativeness. The length of the trait vectors shows TA B L E 2 Agro-climate description of the tested environments. how well the trait is represented in the biplot. The greater trait vector indicates more discrimination of the trait, while the relatively short vector demonstrates an insufficient variation of the trait across genotypes (Santana et al., 2021;Yan & Frégeau-Reid, 2018). Additionally, to identify an ideal genotype, the GT data were also visualized based on the ATC view of the biplot (Figure 3d). Based on this, the center of concentric circles defines as an ideal trait and the nearest genotypes to this point would be the superior genotypes.
Eventually, genotype G10 that is the nearest to the center of concentric circles considered as the superior one.

| Genotype ranking based on the mean × stability GT biplot
The mean × stability biplot was also used in this study to compare genotypes based on GT data (Figure 3c) > G16 > G8. This finding shows that genotype G10 would be the TA B L E 3 The mean of traits (genotype by trait) data across 16 environments (combination of 2 years and eight locations) for 20 oilseed rape genotypes.

| GYT biplot
The original GT data ( Table 3) were used to create the GYT table through a combination of grain yield (GY) with quantitative traits (Table 4). To this end, traits whose larger values were desirable were multiplied by grain yield (GY), while the traits whose lower values were favorable (including earliness traits) were divided by grain yield (GY).
Therefore, a larger value in the GYT table is always desirable. The data of the GYT table were applied to a graphical display named GYT biplot. Principal component analysis indicated that the first and second principal components accounted for 93.6% of the total variation in the yield-trait combinations data. The first principal component represents 89.2% of the total variation of the data and the second principal accounted for 4.4% of the total variation of the data (Figure 4).

| Determining ideal yield-trait combinations to identify superior genotypes
The which-won-where GYT biplot is divided into seven sectors ( Figure 4a). It was observed that all yield-trait combinations were placed in the same sector with the genotype G10 at the vertex of the sector. This finding shows that genotype G10 was the best in combining grain yield (GY) with other quantitative traits.
As shown in Figure

| Genotype ranking based on the mean × stability GYT biplot and the superiority index
Based on the mean × stability GYT biplot, the best ranked above average (right side of the double-arrowed line) genotypes included: G1 0 > G5 > G20 > G14 > G11 > G2 > G19 > G16 (Figure 4c). Among these superior genotypes, G10 and G5 located in the center of concentric circles show the best yield-trait combinations profile (Figure 4d).  Table 4. The ranking of SI values was similar to the results of mean × stability biplot ranking.

| Cluster analysis of GYT data for grouping the genotypes
Cluster analysis was conducted in this study for 20 studied oilseed rape genotypes based on standardized GYT data ( Figure 5). Cluster analysis separated genotypes into two main groups. The first cluster is composed of eight genotypes with high values of yield-trait combinations; including G10, G5, G20, G14, G11, G2, G19, and G16.
These eight genotypes were ranked as above-average genotypes based on the mean × stability GYT biplot. On the contrary, the second group was composed of lower-than-average scored genotypes based on mean × stability GYT biplot. Therefore, cluster analysis based on standardized GYT data could be an efficient method to identify above-average genotypes similar to mean × stability GYT biplot.

| DISCUSS ION
The breeding of oilseed rape, which aims at achieving higher yield, becomes a complex activity due to the negative correlation among quantitative traits. Therefore, selecting the superior genotypes considering multiple traits is one of the major challenges for oilseed rape breeders. This study aimed to identify the relationship between oilseed rape traits and ranking genotypes based on mega-traits. To this end, a multienvironmental trial experiment was conducted in 16 F I G U R E 2 Path analysis for evaluated traits of 20 oilseed rape genotypes. DEF, days to end flowering; DM, days to maturity; DSF, days to start flowering; FP, flowering period; GNP, grain numbers in pod; GY, grain yield; PH, plant height; PNP, pod numbers in plant; TGW, one thousand grain weight.
environments (a combination of eight locations and 2 years) using genotype by trait (GT) and genotype by yield*trait (GYT) biplot approaches.
In this study, nine quantitative traits were evaluated. There is a wide variation among the studied genotypes, as represented by the high standard deviation values ( Table 2) Accordingly, pod numbers in plant (PNP) and plant height (PH) were detected as key traits in the genotype selection of spring oilseed rape.
GT biplot methodology has been introduced for a long time as a practical strategy for trait profiling and genotype selection in plants based on the interaction between genotypes by traits (Yan & Rajcan, 2002). However, rare studies utilized oilseed rape genotypes in this context. Dehghani et al. utilized the GT biplot for the first time using two environments (sowing date) for five winter oilseed rape genotypes (Dehghani et al., 2008). No other studies applied these approaches in oilseed rape genotypes.  GT biplot represented the relationship between traits and trait profiles of the genotypes (Yan & Rajcan, 2002). Based on the GT biplot, pod numbers in plant (PNP) and plant height (PH) were the most discriminative traits with a positive correlation to grain yield (GY). On the other hand, correlation analysis also confirmed the positive correlation of pod numbers in plant (PNP) and plant height (PH) with grain yield (GY). Instead of these traits, 1000 grain weight was previously found as the closest trait to grain yield in winter oilseed rape (Dehghani et al., 2008). However, the negative correlation of grain yield with phenological traits including days to start flowering and days to end flowering was previously confirmed in winter oilseed rape (Dehghani et al., 2008). PH was also found as the most representative trait in addition to its discriminability. Therefore, PNP and PH were also found as ideal traits for genotype selection based on the GT biplot.
Based on the GT biplot, genotype G10 was found as a superior one. This methodology ranked G2, G12, G5, G14, G6, G16, and G8 as the next superior genotypes that were better than average in their traits profile. The negative or no relationship of grain yield with some quantitative traits like days to start flowering and days to end flowering complicated genotype selection based on traits profiling. On the other hand, the high direct impact of the phenological traits on grain yield (GY) based on Path analysis emphasized that these traits should not be neglected. Additionally, although grain yield (GY) is the major target trait, an equivalent value to other traits is allocated to it. The purpose of plant breeding is to select a genotype with desirable traits as long as they have a good yield (Yan et al., 2019). Based on this paradigm, a new approach that combined yield with the other target traits was developed named genotype by yield*trait (GYT) biplot that completes the deficiencies encountered in the GT biplot (Yan & Frégeau-Reid, 2018). This methodology led us to rank oilseed rape genotypes based on their general advantages over yield by trait combinations. Recent studies of GT and GY*T approaches comparison reported GY*T biplot is more effective in case of multitrait and multienvironment selection as compared to a normal GT biplot (Karahan & Akgun, 2020;Sofi et al., 2022).
The GYT methodology allowed us to constitute a composition between grain yield and each trait, regarding high-yielding genotypes as desirable ones. Thus, the desired genotypes not only were superior in their grain yield but also in terms of the other traits.
A comparison of the total ratio of PC1 and PC2 in total variation indicated a higher value in the GYT biplot (93.6) than GT biplot (55.5), which was consistent with a previous report on barley genotypes (Kendal, 2020). This made the GYT biplot ranking and other related outputs more reliable than the GT biplot.
Based on the results of the GYT biplot, all yield-traits combination vectors are close to each other that demonstrate their positive correlation. This is another advantage of the GYT biplot compared with the GT biplot that aligns all traits to make a strong association through combination with the grain yield (GY) (Yan & Frégeau-Reid, 2018). Thus, genotypes could be ranked based on yield-trait combinations and superior ones could be selected with much higher accuracy considering all the traits.
The other benefit of using the GYT method is the selection of a limited number of important traits for evaluation to reduce the evaluation costs in the field (Mohammadi et al., 2020). Our results indicated a strong correlation among all yield-trait combinations.
Therefore, similar results would be achievable by measuring the fewer number of these traits.
Cluster analysis was also conducted based on genotype by yield*trait data that separated the superior genotypes similar to the GYT biplot and superiority index. Accordingly, the second cluster of genotypes (vertical clustering) included series that all categorized as above average based on the GYT biplot. Additionally, the second cluster of yield*trait combination (horizontal clustering) is composed of traits with direct effects on grain yield (GY) based on path analysis. Therefore, this methodology of cluster analysis led us to separate traits based on their direct and indirect effects in addition to genotype selection.

| CON CLUS ION
This study aimed to determine how quantitative traits of oilseed rape genotypes are associated to increase grain yield. Another purpose of this study was to identify superior genotypes through composition grain yield with quantitative traits for the tropical regions of Iran.
In this study, 20 spring oilseed rape genotypes were utilized in the multienvironmental trials. Based on multivariate analysis and GT biplot, Pod number in plant (PNP) and plant height (PH) were detected as the most essential selection criteria for grain yield enhancement in oilseed rape. The GT biplot facilitated visual comparisons of the traits, but was flawed in genotype ranking due to the negative correlation of some traits with grain yield. Genotype assessment based on F I G U R E 5 Cluster analysis of genotypes and yield-trait combinations based on standardized GYT data using the WARD method. DEF, days to end flowering; DM, days to maturity; DSF, days to start flowering; FP, flowering period; GNP, grain numbers in pod; GY, grain yield; PH, plant height; PNP, pod numbers in plant; TGW, one thousand grain weight. grain yield composition with quantitative traits through GYT methodology eliminated this defect. Accordingly, eight above-average genotypes were detected. Of them, G10, G5, G14, G11, and G2 were ranked above the RGS003 check cultivar, while G10 and G5 ranked above both check cultivars (Dalgan AND RGS003), as the first and second determined genotypes. Therefore, genotypes G10 and G5 were detected as the best genotypes with the highest grain yield and outstanding quantitative traits.

ACK N O WLE D G E M ENTS
We would like to thank the Seed and Plant Improvement Institute of Iran (SPII) for supporting us in conducting the research with project number 0-03-03-293-971358.

FU N D I N G I N FO R M ATI O N
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

I N FO R M ED CO N S ENT
Written informed consent was obtained from all study participants.