Analysis of combining ability and stability of yield characteristics of upland cotton in two years

Study area overview

The experimental site is located in the southern Tianshan Mountains, Xinjiang Uygur Autonomous Region, China. It is situated at the northern edge of the Taklamakan Desert and the upper reaches of the Tarim River in the 12th Regiment, Alar City, First Division. The site represents a typical artificially developed arid oasis (longitude 80°30′–81°58′E, latitude 40°22′–40°57′N) with a total area of 3,924.32 km² and altitude between 997 and 1,340 m. The climate of the area is classified as warm-temperate continental desert. The soil at the experimental site is sandy loam. The experimental plot length was 10 m, row spacing configuration was (10 cm + 66 cm + 10 cm + 66 cm + 10 cm), and plant spacing was 10.5 cm. Planting was performed using manual broadcasting.

Compared to other regions in Xinjiang Uygur Autonomous Region, Alar City has less snowfall in winter, abundant sunlight and heat resources, significant temperature variations between day and night, and frequent dusty conditions during spring. The average annual temperature is 10.7 °C, with a frost-free period of up to 220 days, making it suitable for cultivating long-staple and fine-staple cotton. Temperature trends for 2021 and 2022 are presented in Figs. 1 and 2, respectively (data sourced from Weather Forecast: https://lishi.tianqi.com/).

Experimental materials

Seven parental lines—cm3, a115-11, Xinluzhong53, Xinluzhong38, Xinluzhong59, Lu917, and XinLuzao62 (numbered sequentially from 1 to 7)—and their 21 hybrid combinations were used in this study. Hybrid combinations were derived through complete diallel crossing [p(p−1)/2], as detailed in Table 1, provided by the College of Plant Sciences, Tarim University.

The seven parental lines and their 21 hybrid combinations were cultivated at the experimental field of the Twelfth Regiment in Aral City in 2021 and 2022. A randomized block design with two replications was employed, using wide-film planting (10 cm + 66 cm + 10 cm), plant spacing of 10.5 cm, and row length of 3 m. Sowing was conducted using mechanical film perforation combined with manual spot sowing, while standard field management practices were uniformly applied.

During the shedding period (mid-October) in both 2021 and 2022, 10 uniformly grown cotton plants from the middle of each plot were selected. Harvested cotton bolls were evaluated in the laboratory for the following traits: boll number per plant (average number of bolls per 10 plants), boll weight (total weight of harvested seed cotton divided by total boll count from 10 plants), seed cotton per plant (total seed cotton harvested divided by 10), leather cotton per plant (average lint cotton after ginning the harvested seed cotton from 10 plants), and seed index (weight of randomly selected 100 cotton seeds).

Trait determination

1. Plant selection: Ten cotton plants with uniform growth from the centre of each plot were selected to minimize edge effects. Plants with consistent plant height, leaf number, and branch number were ensured.

2. Plant marking: The selected plants were tagged or marked for subsequent identification.

3. Information recording: Each plot’s number and the exact locations of selected plants were documented to guarantee data traceability.

4. Measurement of yield-related traits:

Boll number per plant: Total number of bolls per plant was counted, and the average of ten plants was calculated.

Boll weight: The weight of each cotton boll was measured, and the average boll weight per plant was calculated.

Seed cotton per plant: The total seed cotton weight per plant was calculated (boll number per plant multiplied by boll weight), and the average seed cotton weight per plant for ten plants was determined.

Leather cotton per plant: Lint cotton was obtained by ginning seed cotton using a serrated gin, weighed using an electronic balance, and the average lint cotton per plant for ten plants was calculated.

Seed index: One hundred cotton seeds were randomly selected and weighed to obtain the seed index.

Cotton lint percentage: Calculated as the ratio of leather cotton per plant to seed cotton per plant (cotton lint percentage = leather cotton per plant/seed cotton per plant).

Data analysis

Combining ability analysis

Combining ability (GCA and SCA) served as a crucial indicator to evaluate parental and hybrid performance in breeding. The statistical model for observational means obtained from method 2 of diallel hybridization within each environment or year was: ${\displaystyle Y_{ij}=\mu +g_i+g_j+s_{ij}+{\varepsilon }_{ij}.}$

The variety effect is v_ij = g_i + g_j + s_ij, and there is s_ij = s_ji, which means the positive and negative crossing effects are the same. ∑g_i = 0, for each j, ∑S_ij = 0. In the formula, i, j = 1, 2, 3⋯p; when i = j (parental self crossing): Y_ij = μ + 2g_i + ɛ_ij; when i < j (orthogonal combination): Y_ij = μ + g_i + g_j + s_ij + ɛ_ij.

μ is the overall average; ɛ_ij for the error term; it follows the distribution of $N\left(0,{\delta }_e^2\right)$).

The combining ability model was defined as follows: Y_ij = μ + GCA_i + GCA_j + SCA_ij + ɛ_ij.

where µ denotes the overall experimental mean; GCA_i and GCA_j are the GCA effects for parents i and j, respectively; SCA_ij is the SCA effect between parents i and j; and ɛ_ij is the error term.

The expression of combining ability was influenced by genotype (parent) as well as environmental factors, such as year, location, and climatic conditions. In this study, DPS19 software (http://www.dpsw.cn/) was used to determine combining abilities. A fixed-effects model was employed to estimate combining ability effects for each trait across the two-year data. Joint variance analysis of GCA and SCA was conducted across years based on the principle of additivity of sums of squares to evaluate the significance and stability of combining abilities across environments.

Genetic effects analysis

Using the PROC VARCOMP procedure in SAS software, generalized heritability was calculated via a mixed linear model (with year as a fixed factor and genotype as a random factor): ${\displaystyle h^2=\frac{{\delta }_G^2}{{\delta }_G^2+\frac{{\delta }_{GE}^2}{n}+\frac{{\delta }_{\varepsilon }^2}{nr}}\times 100\text{\%}}$

where ${\delta }_G^2$ is the genotype variance; ${\delta }_{GE}^2$ is the genotype-year interaction variance; ${\delta }_{\varepsilon }^2$ is the error variance; n is the number of years; and r is the number of repetitions.

Correlation and stability analysis of combining ability

Correlations between the combining abilities of various traits across years were calculated to identify traits and combinations less affected by environmental variability. Stability comparison coefficients were calculated separately for each trait. By comparing stability coefficients C_g and C_s, parents and hybrids with superior stability were selected to improve breeding efficiency: ${\displaystyle C_g=\frac{{\delta }_{GCA}^2}{{\delta }_{GCA}^2+{\delta }_{GCA\times E}^2}\times 100\text{\%}}$ ${\displaystyle C_g=\frac{{\delta }_{SCA}^2}{{\delta }_{SCA}^2+{\delta }_{SCA\times E}^2}\times 100\text{\%}}$

where C_g and C_s represent the stability coefficients for GCA and SCA, respectively; ${\delta }_{GCA}^2$ and ${\delta }_{GCA\times E}^2$ are the variances of general and specific combining abilities; ${\delta }_{SCA}^2$ and ${\delta }_{SCA\times E}^2$ are the variances of interactions between combining ability and environment.

Stability analysis of hybrid performance

The AMMI genotype-environment interaction model was applied to evaluate hybrid combinations. GGE biplots displayed performance and stability of combining abilities for each trait across years (Zhao et al., 2024). The AMMI model was defined as follows: ${\displaystyle {\mathrm{Y}}_{ger}=\mu +{\alpha }_g+{\beta }_e+\sum _1^n{\lambda }_n{\gamma }_{gn}{\delta }_{en}+{\rho }_g+{\varepsilon }_{ger}}$

where Y_ger is the observed value for the g-th genotype in the e-th environment with r replicates; μ is the overall mean; α_g and β_e denote genotype and environment main effects, respectively; λ_n is the eigenvalue of the n-th principal component; γ_gn and δ_en are genotype and environment scores on the n-th principal component; ρ_g is the residual; ɛ_ger is the experimental error.

Genstat21 software was used for GGE biplot analysis to evaluate combining ability stability and adaptability of parents and hybrid combinations across years.

Normality test of residuals for yield traits across years

The residuals of yield traits underwent normality testing (Table 2). Kolmogorov–Smirnov tests indicated Sig. ≥ 0.05, confirming that the experimental data for both years followed a normal distribution.

Joint analysis of variance of yield traits across years

Joint analysis of variance (ANOVA) results (Table 3) revealed no significant differences in seed index and boll weight among blocks within years, while all other traits showed significant or highly significant differences. Yield traits differed significantly or highly significantly across years, except for seed index, suggesting that yields are considerably influenced by environmental factors. Additionally, all yield traits displayed significant or highly significant differences across genotypes (hybrid combinations) and genotype-year interactions. The pervasive significant interactions between genotype and year indicate substantial environmental influences on yield traits.

Joint ANOVA of combining ability for yield traits across years

ANOVA results for combining ability (Table 4) indicated no significant differences between years for all yield traits. GCA differences among parents were significant or highly significant for all yield traits except cotton lint percentage and boll weight. SCA differences among combinations were highly significant for leather cotton per plant, seed cotton per plant, and boll number per plant. The interaction between year and parental GCA was significant or highly significant for all yield traits, and the interaction between year and SCA of combinations was significant for leather cotton per plant, seed cotton per plant, and boll number per plant.

Analysis of genetic effects on yield traits in upland cotton

Combining ability analysis for yield traits across two years (Table 5) showed that all six yield traits were controlled by both additive and dominant genes. Given that the variance of GCA exceeded that of SCA for all six traits, it can be concluded that additive genetic effects predominantly controlled these traits, complemented by dominance effects. The genotype-environment interaction variance was substantial for all yield traits, indicating strong environmental influences. The highest broad-sense heritability was 70.263% for seed cotton per plant, and the lowest was 56.702% for boll weight.

Correlation and stability analysis of GCA and SCA across years

Pearson correlation coefficients and stability analyses for parental GCA and SCA over two years (Table 6) showed no significant correlations between years for most yield traits except cotton lint percentage, suggesting inconsistent combining ability across years. Stability comparison coefficients for GCA of leather cotton per plant and seed cotton per plant were 51.372% and 55.187%, respectively. Stability coefficients for other traits ranged between 40% and 50%. Stability coefficients for SCA were relatively lower for cotton lint percentage (44.986%) and boll weight (48.645%), whereas coefficients for other traits ranged between 50% and 60%. Except for cotton lint percentage, whose SCA stability coefficient was lower than that of GCA, stability coefficients of SCA were generally higher than those of GCA, indicating that SCA exhibited greater stability than GCA for most traits across different years.

GCA analysis of yield traits for each parent across different years

The GCA of seven upland cotton parents over two years was analysed. Results (Fig. 3) indicated that parents 1, 2, and 4 showed negative GCA effects for boll number per plant in both years, whereas parent 7 exhibited positive effects consistently. The other parents demonstrated inconsistent effects across the two years. For seed cotton per plant, parents 1 and 2 had consistently negative effects, parent 7 consistently showed a positive effect, and the remaining parents exhibited inconsistent effects. Similarly, for leather cotton per plant, parents 1 and 2 consistently displayed negative effects, parent 7 consistently showed a positive effect, and other parents had varying effects. Regarding boll weight, parents 1 and 2 demonstrated consistent positive effects, while parents 3, 5, and 7 showed consistent negative effects. For cotton lint percentage, parents 2 and 3 had consistently positive effects, parents 4, 5, and 6 had consistently negative effects, and others were inconsistent. Finally, for seed index, parents 1 and 6 consistently showed positive effects, parents 3 and 7 consistently had negative effects, and the remaining parents displayed varying effects.

SCA analysis of yield traits for each hybrid combination across different years

SCA of 21 combinations over two years (Fig. 4) revealed consistent positive effects for boll number per plant in combinations 12, 25, and 35, whereas combinations 14, 17, 27, 36, 46, 57, and 67 consistently showed negative effects. For seed cotton per plant, combinations 15, 25, 34, 35, and 37 consistently showed positive effects, whereas combinations 14, 16, 27, 46, 57, and 67 consistently exhibited negative effects. For leather cotton per plant, combinations 15, 25, and 35 consistently demonstrated positive effects, while combinations 14, 17, 27, 46, 57, and 67 showed consistent negative effects. For boll weight, combinations 14, 15, 17, 23, 34, 35, 57, and 67 consistently had positive effects, whereas combinations 27, 45, 46, 47, and 56 consistently exhibited negative effects. In terms of cotton lint percentage, combinations 13, 14, 15, 23, 26, 27, 45, and 56 consistently displayed positive effects, whereas combinations 12, 17, 24, 34, 35, 36, and 46 consistently exhibited negative effects. For seed index, combinations 12, 23, 35, and 67 consistently showed positive effects, while combinations 14, 17, 37, 45, and 56 consistently had negative effects. The remaining combinations demonstrated inconsistent effects across the two years.

Stability analysis of combining ability for yield traits in parents and combinations across years

Stability analysis of GCA

The stability of GCA for seven parents over two years (2017 and 2018) was analysed using GGE biplots (Fig. 5). Parents positioned right of the vertical axis performed above the overall mean, whereas those to the left performed below average. Parents positioned closer to the horizontal axis showed greater stability, while those further away exhibited instability.

Parents 3, 5, and 7 had superior GCA for boll number per plant, with parent 7 being most stable and parents 3 and 5 less stable. For seed cotton per plant, parents 4, 5, and 7 exhibited higher GCA; parent 7 was again most stable, whereas parents 4 and 5 were comparatively unstable. Regarding leather cotton per plant, parents 3, 5, and 7 showed better GCA, with parent 7 displaying the highest stability. For boll weight, parents 1, 2, 4, and 6 had higher GCA; parents 2 and 6 showed greater stability, while parent 4 was the least stable. Parents 2 and 3 had better and stable GCA for cotton lint percentage. Parents 1, 4, and 6 exhibited better GCA for seed index, with parent 1 being most stable and parent 4 least stable.

Stability analysis of SCA

The stability of SCA for 21 combinations in both years was evaluated using GGE biplots (Fig. 6). Combinations on the right side of the vertical axis performed above the overall mean, while those on the left performed below average. Combinations nearer the horizontal axis were more stable; those further away showed less stability.

For boll number per plant, combinations 12, 13, 15, 25, 26, 34, 35, 37, and 56 exhibited higher SCA; combination 12 had the best stability, while combinations 25, 26, 34, and 56 were relatively stable. For seed cotton per plant, combinations 12, 15, 23, 25, 26, 34, 35, 37, and 56 displayed better SCA; combination 15 was most stable, and combinations 12, 23, 25, and 35 showed relatively high stability. Regarding leather cotton per plant, combinations 12, 13, 15, 23, 25, 26, 34, 35, 37, and 56 had superior SCA, with combination 15 being most stable, followed by combinations 12, 23, 25, and 35. For boll weight, combinations 12, 14, 15, 17, 23, 34, 35, 36, 37, 57, and 67 demonstrated higher SCA; combinations 23 and 34 were most stable, while combinations 15 and 67 showed moderate stability. Combinations 13, 14, 15, 16, 23, 26, 27, 45, and 56 had better SCA for cotton lint percentage; combinations 14, 15, 27, and 45 were the most stable, followed by combinations 23 and 26. For seed index, combinations 12, 15, 23, 24, 25, 35, 57, and 67 showed higher SCA; combinations 12 and 35 had the highest stability, whereas combinations 23, 24, and 67 exhibited moderate stability.

Genotype and environmental analysis of upland cotton yield traits

Cotton yield is influenced by multiple factors. Studies by Elsamman et al. (2024) and Galdi et al. (2022) indicated that the environment is a major determinant of cotton yield. In the present study, significant or highly significant yearly differences were found in all traits except seed index, supporting previous findings. Additionally, significant or highly significant differences appeared among genotypes and in genotype-year interactions for all traits. This indicates substantial genotype effects and genotype-by-environment interactions influencing yield traits, aligning with findings by Yehia (2022) and Askander (2020).

Analysis of combining ability, correlation, and stability for yield traits across years

Environmental factors impacted traits across all parental lines, consistent with conclusions by Bhandari et al. (2021). For cotton combining ability, leather cotton per plant, seed cotton per plant, and boll number per plant exhibited significant or highly significant effects in parental GCA, hybrid SCA, and interactions involving year × parental GCA and year × hybrid SCA. In this study, correlations of SCA for six yield traits across years were generally weak. SCA showed greater stability across years compared with GCA. Traits such as leather cotton per plant, seed cotton per plant, and boll number per plant can thus serve as selection criteria for identifying superior hybrids.

Cotton yield traits were influenced by additive and dominant genetic effects. This finding is consistent with conclusions by Feng et al. (2022), Subalakhshmi et al. (2022), and Abdel-Aty et al. (2023), who reported yield traits predominantly controlled by additive effects, complemented by dominant effects. The current analysis also demonstrated significant environmental influence on all traits. Generalized heritability was highest for seed cotton per plant (70.263%) and lowest for boll weight (56.702%). These results can inform breeding programs aiming at improved cotton varieties.

Selection of superior parents and hybrid combinations

Based on GGE biplot analysis, parent 7 (Xinluzao62) demonstrated the best overall GCA performance and stability among seven parents, making it suitable for future breeding programs. Among 21 hybrid combinations, combinations 12 (cm3 × a115-11), 15 (cm3 × Xinluzhong59), 23 (a115-11 ×Xinluzhong53), 25 (a115-11 ×Xinluzhong59), 26 (a115-11 ×Lu917), 34 (Xinluzhong53 ×Xinluzhong38), 35 (Xinluzhong53 ×Xinluzhong59), and 67 (Lu917 ×Xinluzao62) exhibited superior SCA performance and stability. These combinations can undergo further regional field evaluations. If future trials confirm the findings of this study, these hybrids can be recommended for broader regional adoption.