Genetic and yield analysis of F1 and F2 populations of improved soybean varietiesAbstract
This study aimed to evaluate the General Combining Ability (GCA) and Specific Combining Ability (SCA) of soybean germplasm adapted to tropical regions. Ten soybean crosses from six tropical-adapted varieties were performed which revealed substantial genetic variability in seven agronomic traits. Genetic variance, heritability, and correlations for growth and yield-related traits were evaluated. In the F1 generation, high genetic variance was observed for plant height (PH) and number of pods per plant (NPPP), whereas broad-sense heritability was highest for PH (0.82) and 50-seed weight (50SW; 0.72). Number of seeds per plant (NSPP) was strongly influenced by environmental effects. Genetic correlations identified key yield predictors, especially between NPPP and NSPP (r = 0.93), and positive associations of PH with yield components. In F2, Semsa-107 and Huasteca-100 showed superior performance, while Cedrela × Semsa-107 and Cedrela × Huasteca-100 were the best hybrids. Combining ability analysis indicated additive effects in Cedrela, NK-12, and Semsa-107.
Keywords:
Genetic variance components; quantitative genetics; yield-related traits; general and specific combining ability; Glycine max (L.) Merril
INTRODUCTION
Soybean (Glycine max (L.) Merril) is a crop of major global relevance due to its high protein and oil content, making it a key seed for food, feed, and industrial uses worldwide (Pagano and Miransari 2016). Beyond the nutritional and economic value, soybean contributes significantly to agroecosystem functioning through its symbiotic interactions with rhizobia and mycorrhizal fungi, which stimulate soil fertility and nutrient cycling (Pagano and Miransari 2016). Global production is dominated by Brazil, the United States, Argentina, and China, together accounting for more than 80 % of the total output (USDA 2025). However, despite its global importance, soybean productivity and adaptation in emerging production regions, particularly in tropical environments, remain challenging (Dong et al. 2021) and may be curtailed by genotype × environment interactions (Li et al. 2024).
In Central America, limited investment in breeding and technology has constrained the development of local varieties to reach the full productive potential of the crop. Moreover, most major seed companies have prioritized the development and commercialization of transgenic varieties, which may not always align with local production systems or regulatory frameworks, while many farmers face financial limitations that restrict their access to improved seed. Thirdly, global soybean research is largely driven by countries with extensive soybean production systems in other regions of the world (Stupar et al. 2024). These are common constraints in all of Central America and reinforce the need to incorporate broader genetic variability into regional breeding programs to develop new varieties with improved performance, resilience, and quality.
The development of improved varieties depends on the availability of germplasm with high genetic variability and favorable agronomic traits (Hanafiah et al. 2023). Hybridization remains a fundamental strategy in this process, as crosses between genetically divergent parents can generate novel combinations and increase the likelihood of identifying superior segregating progenies (Jiménez and Korpelainen 2013). Such genetic recombination is essential for the expansion of the adaptive potential of soybean into non-traditional production environments, e.g., the tropical regions of Central America. During early selection stages, the estimation of genetic variance components is essential for determining trait heritability and to design efficient breeding strategies (Vogel et al. 2021). Diallel crosses enable plant breeders to evaluate the performance of parental lines, referred to as general combining ability (GCA), as well as the performance of specific hybrid combinations, known as specific combining ability (SCA) (Teodoro et al. 2019). In legumes, both GCA and SCA have been shown to influence yield, indicating complex inheritance patterns involving additive as well as non-additive effects (Ceyhan et al. 2008, Jou-Nteufa and Ceyhan 2024). By identifying parental lines with high combining ability, breeders can accelerate the process of developing varieties with higher yield and adaptability (Rahmadana et al. 2023). In soybean, numerous studies have successfully used combining ability analyses (Gavioli et al. 2006, Souza et al. 2020, Abebe et al. 2023).
Rigorous agronomic evaluation of progenies, supported by experimental designs and statistical tools, is essential to ensure accurate selection. In this context, this study addressed the GCA and SCA of soybean germplasm adapted to tropical regions, and determined genetic variance, heritability, and correlations for growth and yield-related traits. The results will support the development of new genetic combinations and the identification of promising lines with high yield potential and improved adaptation to the environmental conditions of Central America.
MATERIAL AND METHODS
Study location
Hybridization and generation advancement were carried out in a greenhouse of the Agricultural Experimental Station Fabio Baudrit Moreno, affiliated with the Faculty of Agri-Food Sciences of the University of Costa Rica, in Alajuela, in 2022 and 2023. Subsequently, the segregating generations were established and evaluated in the field in Sarapiquí, Heredia province, in 2024.
Plant material
Six commercial soybean varieties (CIGRAS-06, CEA-CH-86, Huasteca-100, Cedrela, Semsa-107, and NK-12) were used as parental genotypes in this study. These six were selected based on their documented adaptation and performance under tropical conditions.
Crossbreeding
In December 2022, seeds of each soybean variety were planted in 6-L pots under controlled conditions at the Fabio Baudrit Moreno Agricultural Experiment Station (lat 10° 00' 26.78" N, long 84° 15' 57.50" W). Thirty-two pots were planted for each variety. In February 2023, during the reproductive stages R1 and R2, 318 crosses were performed manually, according to a partial diallel mating scheme. Eight days after pollination, fruit set was verified to confirm the effectiveness of the hybridizations. Of the 318 hybridizations, 70 resulted in successful pod set.
Evaluation of the F1 generation
In August 2023, seeds from successful crosses were sown individually in pots, resulting in an F1 population comprising 166 plants. At physiological maturity, each plant was harvested separately, and the following agronomic traits were measured: number of nodes per plant (NNPP), number of branches per plant (NBPP), number of pods per plant (NPPP), and plant height (PH, cm).
After harvest, the plant material was transferred to the Grain and Seed Research Center (CIGRAS) of the University of Costa Rica (UCR), where reproductive traits were evaluated, including yield per plant (YPP, g), 50-seed weight (50SW, g) at 12% moisture content, and number of seeds per plant (NSPP). The resulting seed lots were then sieved through a No. 20 round-hole sieve (7.9-mm aperture), packed in 6 × 5 cm Kraft paper bags, and stored at 6 °C and 70% relative humidity in a cold room until use.
Evaluation of the F2 generation
In February 2024, a second experiment was established, under field conditions in Sarapiquí (lat 10° 29' 23" N, long 83° 56' 15" W). F2 seeds were sown in a single-row plot arrangement, of individual 2.5-m long rows, in a randomized block design. Each row contained 20 plants, with a spacing of 0.6 m between rows. Rows of the 10 parental genotypes were also included for comparison. At physiological maturity, the five most vigorous and productive plants per row were selected by visual phenotypic selection. Of these representative plants, the same quantitative variables evaluated in the F1 generation were measured to identify superior genotypes.
Modeling and statistical analyses
For the agronomic variables assessed in the F1 generation (PH, NNPP, NBPP, NPPP, NSPP, 50SW, and YPP), genetic, environmental, and total variance components were estimated. Broad-sense heritability and expected genetic gain were also estimated for each trait. Variance components were estimated using linear mixed models implemented in the “lme4” package (Bates et al. 2015). In addition, a genetic correlation matrix was generated for the entire set of variables using R v4.4.1.
Variance components were estimated using linear mixed-effect models implemented in the lme4 package in R:
Yijk = µ + P i + M j + C ij + ε ijk
Where P i , M j , and C ij represent the random effects of male parent, female parent, and specific cross, respectively. Total genetic variance was estimated as:
σ2G = σ 2 Male + σ 2 Female + σ 2 Cross
and phenotypic variance as:
σ2P = σ 2 G + σ 2 E
Broad-sense heritability was estimated as:
H2 = σ 2 G / σ 2 P
This estimate represents the proportion of total phenotypic variance attributable to genetic effects in the F1 population. Because the evaluated material consisted of F1 hybrids and the model included both parental and cross effects, additive and non-additive sources of variation were not completely separable. Therefore, heritability estimates should be interpreted as approximations based on variance components rather than an exact partitioning of additive genetic variance.
Parental effects were interpreted as proxies for GCA, whereas the cross effect was interpreted as a proxy for SCA, following a mixed-model approach conceptually based on diallel analysis. For the F2 generation, evaluated in a single-row plot design, and for the rows representing the six parental genotypes, analysis of variance (ANOVA) was conducted, followed by mean comparisons using Tukey's HSD test (P = 0.05). These analyses were performed using the “agricolae” package (Mendiburu 2023). General combining ability (GCA) and specific combining ability (SCA) effects were estimated using functions from the “dplyr” package (Wickham et al. 2023), also within the R v4.4.1 program. GCA and SCA effects in the F2 population were estimated following Griffing’s diallel methodology using cross means. GCA effects were interpreted as indicators of additive genetic effects, whereas SCA effects were associated with non-additive genetic effects. Heterosis was calculated relative to the mid-parent value.
RESULTS AND DISCUSSION
Evaluation of the F1 generation
The 10 crosses derived from six improved soybean cultivars generated substantial phenotypic and genetic variability, providing valuable material for soybean breeding programs targeting tropical environments. Analysis of variance components (Table 1) showed marked differences among the seven evaluated traits. Highest genetic variances were observed for PH (2211.18) and NSPP (2968.61). In contrast, environmental variances exceeded genetic variances for most traits, particularly NSPP (5739.58), NPPP (981.32), and YPP (185.70), suggesting a strong environmental influence on trait expression and emphasizing the importance of multi-environment trials for accurately identifying stable, high-performing genotypes. The predominance of non-additive gene effects observed in this study is consistent with a previous report in cowpea, where SCA variance was found to be higher than GCA variance for most yield-related traits (Jou-Nteufa and Ceyhan 2024).
Genetic proportions or estimates of broad-sense heritability (H²), indicated varying degrees of genetic control among traits. Highest values were recorded for PH (0.82) and 50SW (0.72), followed by a moderate value for NNPP (0.52). In contrast, H² estimates were low for NPPP (0.19), YPP (0.20), NSPP (0.34), and especially NBPP (0.12). These patterns suggest a relatively strong genetic control of PH and 50SW under the evaluated conditions. This is particularly relevant because seed number and seed weight are major determinants of grain yield in soybean (Vogel et al. 2021).
Correlation analysis revealed significant and agronomically meaningful associations among yield components. PH showed moderate positive correlations with NPPP (r = 0.4022), NSPP (r = 0.4231), and YPP (r = 0.4455). Likewise, NNPP and NBPP had a strong positive correlation (0.5755). Particularly high correlations were observed between NPPP and NSPP (r = 0.9312), as well as between these traits and YPP (r = 0.8353 and r = 0.8577, respectively). These strong, significant genetic correlations indicate that indirect selection for improved performance via NPPP and NSPP is feasible. It is worth noting that previous studies found no significant correlation between yield components and grain yield (Kuswantoro et al. 2018), suggesting the presence of genotype × environment interactions. These relationships are particularly relevant for the development of multivariate selection indices capable of maximizing genetic gains, especially because seed number is a major predictor of soybean yield (Leite et al. 2018).
On the other hand, correlations were low and negative between 50SW and NNPP (r = −0.1967), NPPP (r = −0.2092), and NSPP (r = −0.2692), and virtually null with YPP (r = −0.0206). The lack of association between 50SW and yield may reflect a compensatory relationship between seed size and seed number. Finally, the moderate correlations of PH (r = 0.4455), NNPP (r = 0.5392), and NBPP (r = 0.5706) with yield are relevant for designing an ideotype that integrates architectural and reproductive efficiency traits to enhance productivity. Previous studies have identified genomic regions associated with plant architecture and seed weight, suggesting partial genetic linkage between these traits (Zhou et al. 2015).
Evaluation of the F2 generation
Differences in the six evaluated variables were observed between the hybrids and their parental lines (Table 2). However, the high standard deviation values reflect substantial intra-population variability, which is expected in segregating generations and can be attributed to both genetic recombination and environmental influences. The Tukey test (α = 0.05) showed overall superior agronomic performance of the parents over the crosses across all evaluated variables. This information, together with genotype data, can be highly valuable for soybean improvement using genomic prediction, enabling equal genetic gains with fewer resources or greater gains with the same resources (Jean et al. 2021).
In terms of PH, parent NK-12 (39.20 cm) and the cross Semsa-107 × Cedrela (47.30 cm) exhibited the lowest and CEA-CH-86 the highest values (110.60 cm). The crosses CIGRAS-06 × CEA-CH-86, NK-12 × CIGRAS-06 and Semsa-107 × CIGRAS-06 had the highest plants (88.7 cm, 84.07 cm, and 88.3 cm, respectively). For NBPP, the parents CEA-CH-86 (10.40) and Cedrela (8.40) reached the highest values, suggesting favorable combining ability for this trait, likely associated with additive genetic effects (Teodoro et al. 2019). The cross Cedrela × Semsa-107 presented the highest NBPP (6.43). For NK-12, NBPP was also relatively high (6.40). The values of the other genotypes were intermediate to low, and lowest for the crosses Semsa-107 × CIGRAS-06 (2.46), CIGRAS-06 × CEA-CH-86 (2.87), and NK-12 × Huasteca-100 (3.00).
For the key performance components NPPP and NSPP, the values of the parents Huasteca-100 and Semsa-107 were consistently the highest, indicating favorable genetic potential for seed production. This variation in yield-related traits confirms previous studies that highlighted the importance of genotype-dependent performance in legumes (Ceyhan et al. 2008, Jou-Nteufa and Ceyhan 2024). Specifically, for NSPP, Huasteca-100 (178.40) and Semsa-107 (158.20) were the best parents, and the crosses Cedrela × Semsa-107 (120.24) and Cedrela × Huasteca-100 (118.23) the best hybrids. For total seed production (per plant), Huasteca-100 (423.80) and Semsa-107 (373.20) were again the top genotypes, followed by Cedrela × Huasteca-100 (274.72) and Cedrela × Semsa-107 (272.74).
Regarding 50SW, parent NK-12 showed the highest value (12.72 g), indicating favorable expression of seed size in this genotype, consistent with reports identifying specific loci associated with this trait in soybean (Diers et al. 2018). The maximum YPP was recorded for parent Semsa-107 (51.45 g), followed by CIGRAS-06 (39.36 g) and Huasteca-100 (38.92 g). Among hybrids, Cedrela × Semsa-107 (33.65 g) and Cedrela × Huasteca-100 (32.28 g) exhibited the highest YPP, although their performance fell short of that of the elite parents. The reduced performance observed in some F₂ populations may reflect the breakdown of favorable gene combinations through segregation and recombination, a phenomenon commonly observed in advanced generations following hybridization (Ramlal et al. 2022).
The evaluation of GCA and SCA revealed the type and magnitude of underlying genetic interactions governing the traits (Table 3). Significant GCA effects were detected for all evaluated traits, (PH, NBPP, NPPP, NSPP, 50SW, and YPP), indicating an important contribution of additive genetic effects to trait expression. In legumes, GCA has been widely associated with additive gene effects controlling yield and quality traits (Ceyhan et al. 2008, Tamüksek and Ceyhan 2024, Jou-Nteufa and Ceyhan 2024). Notably, the GCA values of Cedrela, NK-12, and Semsa-107 were consistently high across multiple traits, identifying them as strong general combiners and valuable parental sources of favorable alleles for parental selection in breeding pipelines.
The SCA analysis, which captures non-additive genetic effects, revealed substantial variability across hybrid combinations, suggesting the involvement of dominance and possibly epistatic interactions (Table 3). In PH, estimates ranged from negative (e.g., Semsa-107 × Cedrela: −20.89) to positive values (e.g., Semsa-107 × CIGRAS-06: 11.49). Similar patterns were observed for yield components: NPPP ranged from −29.77 to 23.40 (Cedrela × Huasteca-100), while NSPP varied from −72.04 to 56.11 (Cedrela × Huasteca-100). In contrast, 50SW exhibited reduced variability (−0.51 to 0.47). Finally, YPP showed SCA values ranging from −8.32 for the cross Semsa-107 × CIGRAS-06 to 5.96 for Cedrela × Semsa-107.
Crosses with consistently positive SCA effects across multiple traits, e.g., Cedrela × Huasteca-100 and CIGRAS-06 × Semsa-107, indicate the presence of favorable non-additive genetic interactions. These interactions can be strategically exploited to identify superior hybrid combinations in early generations, particularly under selection schemes that exploit dominance and epistatic effects (Lewers et al. 1998).
Distinguishing between additive and non-additive genetic effects is important for refining breeding strategies. Notably, some crosses derived from parents with low GCA exhibited positive SCA values. This pattern suggests genetic complementarity, in which specific parental combinations generate advantageous interactions despite their individually modest additive contributions.
Altogether, the combined evidence from genetic variances, heritability estimates, correlation structures, and the partitioning of GCA and SCA effects reinforces the strategic value of this germplasm set for early-generation soybean breeding. The consistent performance of high-GCA parents such as Cedrela, NK-12, and Semsa-107, together with the identification of hybrid combinations exhibiting favorable SCA, particularly Cedrela × Huasteca-100 and CIGRAS-06 × Semsa-107, suggests that both additive and non-additive genetic effects may be exploited to accelerate genetic gain. These results underscore the potential of integrating multivariate selection, parental complementarity, and targeted recombination to assemble ideotypes with enhanced yield performance and adaptability under tropical conditions. The patterns of genetic relationships among the evaluated crosses identified in this study provide a firm foundation for informed decision-making in subsequent selection cycles and for the development of superior cultivars tailored to diverse agroecological environments.
ACKNOWLEDGEMENTS
This study was conducted as part of the doctoral studies of José Israel López-Rodríguez in the Graduate Program in Agricultural Sciences and Natural Resources (PCARN by its acronym in Spanish), University of Costa Rica (UCR). The authors gratefully acknowledge the support of the German Academic Exchange Service (DAAD) and the PCARN Graduate Program. This research was conducted within project 734-C2-609, “Development and Promotion of New Varieties and High-Quality Seed of Costa Rican Soybean,” funded by the Vice Presidency for Research of the University of Costa Rica.
Data Availability Statement
The datasets generated and/or analyzed during the current research are available from the corresponding author upon reasonable request.
REFERENCES
- Abebe AT, Adewale S, Chigeza G, Derera J2023 Diallel analysis of soybean (Glycine max L.) for biomass yield and root characteristics under low phosphorus soil conditions in Western EthiopiaPLoS ONE 18:e0281075
- Bates D, Maechler M, Bolker B, Walker S2015 Fitting linear mixed-effects models using lme4Journal of Statistical Software 67:1-48
- Ceyhan E, Avcı MA, Karadaş S2008 Line × tester analysis in pea (Pisum sativum L.): Identification of superior parents for seed yield and its componentsAfrican Journal of Biotechnology 7:2810-2817
- Diers BW, Specht J, Rainey KM, Cregan P, Song Q, Ramasubramanian V, Graef G, Nelson R, Schapaugh W, Wang D2018 Genetic architecture of soybean yield and agronomic traitsG3: Genes, Genomes, Genetics 8:3367-3375
- Dong L, Fang C, Cheng Q, Su T, Kou K, Kong L, Zhang C, Li H, Hou Z, Zhang Y, Chen L, Yue L, Wang L, Wang K, Li Y, Gan Z, Yuan X, Weller JL, Lu S, Kong F, Liu B2021 Genetic basis and adaptation trajectory of soybean from its temperate origin to tropicsNature Communications 12:5445
- Gavioli EA, Perecin D, Di Mauro AO2006 Analysis of combining ability in soybean cultivarsCrop Breeding and Applied Biotechnology 6:121-128
- Hanafiah DS, Lubis K, Haryati SH, Damanik GM, Limbong MS, Silaen FR, Joshua LA2023 Assembly of soybean genotypes developed through three-way crossSABRAO Journal of Breeding and Genetics 55:940-950
- Jean M, Cober E, O’Donoughue L, Rajcan I, Belzile F2021 Improvement of key agronomical traits in soybean through genomic prediction of superior crossesCrop Science 61:3908-3918
- Jiménez OR, Korpelainen H2013 Preliminary evaluation of F1 generation derived from two common bean landraces (Phaseolus vulgaris) from NicaraguaPlant Breeding 132:205-210
- Jou-Nteufa C, Ceyhan E2024 Genetic analysis of seed yield and some traits in cowpea using diallel analysisTurkish Journal of Agriculture and Forestry 48:278-293
- Kuswantoro H, Artari R, Rahajeng W, Ginting E, Supeno A2018 Genetic variability, heritability, and correlation of some agronomical characters of soybean varietiesBiosaintifika 10:9-15
- Leite WDS, Unêda-Trevisoli SH, Silva FMD, Silva AJD, Mauro AOD2018 Identification of superior genotypes and soybean traits by multivariate analysis and selection indexRevista Ciência Agronômica 49:491-500
- Lewers KS, St Martin SK, Hedges BR, Palmer RG1998 Testcross evaluation of soybean germplasmCrop Science 38:1143-1149
- Li J, Li Y, Agyenim-Boateng KG, Shaibu AS, Liu Y, Feng Y, Qi J, Li B, Zhang S, Sun J2024 Natural variation of domestication-related genes contributed to latitudinal expansion and adaptation in soybeanBMC Plant Biology 24:651
- Mendiburu F2023 agricolae: Statistical procedures for agricultural research. R package version 1.3-7.
- Pagano MC, Miransari M2016 The importance of soybean production worldwide. In Miransari M (ed) Abiotic and biotic stresses in soybean production. Academic Press, Amsterdam, p. 1-26
- Rahmadana F, Hanafiah DS, Siregar LAM, Lubis K, Haryati Haryati2023 The phenotype of the three-way cross soybean (Glycine max [L.] Merril) resultsIOP Conference Series: Earth and Environmental Science 1183:012007
- Ramlal A, Nautiyal A, Baweja P, Kumar Mahto R, Mehta S, Pujari Mallikarunja B, Vijayan R, Saluja S, Kumar V, Kumar Dhiman S, Lal SK, Raju D, Rajendran A2022 Harnessing heterosis and male sterility in soybean [Glycine max (L.) Merril]: A critical revisitFrontiers in Plant Science 13:981768
- Souza RS, Barbosa PAM, Yassue RM, Bornhofen E, Espolador FG, Nazato FM, Vello NA2020 Combining ability for the improvement of vegetable soybeanAgronomy Journal 112:3535-3548
- Stupar RM, Locke AM, Allen DK, Stacey MG, Ma J, Weiss J, O'Rourke JA2024 Soybean genomics research community strategic plan: A vision for 2024-2028The Plant Genome 17:e20516
- Tamüksek Ş and Ceyhan E2024 Genetic analysis of grain yield and its components in green bean for soils with high lime contentBrazilian Archives of Biology and Technology 67:e24231090
- Teodoro LPR, Bhering LL, Gomes BEL, Campos CNS, Baio FHR, Gava R, Silva Júnior CA, Teodoro PE2019 Understanding the combining ability for physiological traits in soybeanPLoS ONE 14:e0226523
-
USDA2025 Soybean explorer. Available at <Available at https://ipad.fas.usda.gov/cropexplorer/cropview/commodityView.aspx?cropid=2222000 >. Accessed on February 21, 2025.
» https://ipad.fas.usda.gov/cropexplorer/cropview/commodityView.aspx?cropid=2222000 - Vogel JT, Liu W, Olhoft P, Crafts-Brandner SJ, Pennycooke JC, Christiansen N2021 Soybean yield formation physiology - a foundation for precision breeding based improvementFrontiers in Plant Science 12:719706
- Wickham H, François R, Henry L, Müller K, Vaughan D2023 dplyr: A grammar of data manipulation. R package version 1.1.4.
- Zhou Z, Jiang Y, Wang Z, Gou Z, Lyu J, Li W, Yu Y, Shu L, Zhao Y, Ma Y, Fang C, Shen Y, Liu T, Li C, Li Q, Wu M, Wang M, Wu Y, Dong Y, Tian Z2015 Resequencing 302 wild and cultivated accessions identifies genes related to domestication and improvement in soybeanNature Biotechnology 33:408-414
Edited by
-
SCIENTIFIC EDITOR:
Luiz Antônio dos Santos Dias
