Indonesia is the world’s fourth-largest rice consumer, with an average annual per capita consumption of 93.79 kg (Ministry of Agriculture, 2023; USDA, 2023). Indonesia’s population is predicted to increase by 27.04% in 2045. This is likely to increase rice demand by 14.99% (Zainul et al., 2021). The future challenge is meeting rice demand due to climate change. Indonesian rice production is predicted to decline by up to 14% due to rising temperatures, changing rainfall patterns and an increase in extreme weather events (Ansari et al., 2021; Octania, 2021). Therefore, new rice breeding methods are needed to maintain and increase rice production. One way to mitigate the negative impacts of climate change is to develop high-yielding and early-harvesting rice varieties. This will result in lower water requirements and a lower risk of mortality due to drought or heat stress and the ability to cultivate more than once a year. This aligns with the primary goal of plant breeding: producing varieties with high yield potential while ensuring food security and economic resilience (Wang et al., 2021; Yun, 2023).
       
Methodically selecting progenitors and understanding the genetic pathways that govern desired traits are essential. This is useful for selecting offspring that carry information on the inheritance of agronomic traits (Muthoni and Shimelis, 2020). Combining ability analysis is one of the most effective genetic tools because it can assess the additive effects of genes through General Combining Ability (GCA) and non-additive effects or interactions through Specific Combining Ability (SCA) (Falconer and Mackay, 1996; Acquaah, 2012). This provides clearer information about how genes influence each plant trait. This knowledge is crucial for formulating effective breeding strategies, especially in determining whether to focus selection on early generations or exploit heterosis through hybrid variety development (Sztepanacz and Blows, 2015).
               
Studies combining capability analysis with multi-trait evaluation to simultaneously increase yield and early harvesting in rice are still rare in Indonesia. The purpose of the study was to evaluate combining ability and estimate the genetic parameters of F1 seeds from various parents produced through crossbreeding. The benefits of this research are expected to accelerate the development of superior rice varieties with high yields, early harvesting and adaptability to climate change. 

The plant material consisted of nine rice genotypes selected as parental lines (four superior genotypes and five genotypes with early harvesting traits). These genotypes were used as female and male parents, respectively, to produce F1 seeds and served as controls for the experiment. The mating design used was the North Carolina II (NC II) design, which systematically combines female and male parents to produce 20 cross-pairs (Acquaah, 2012). Verification of the F₁ seeds from the crosses was validated using Simple Sequence Repeat (SSR) molecular markers to ensure that the seeds were of the F₁ genotype. In this study, only 18 F₁ genotypes were used, because the remaining two genotypes were suspected of selfing (Aisya et al., 2024) (Table 1).


       
The study was located at the Greenhouse of the Faculty of Agriculture, Universitas Gadjah Mada, Yogyakarta, Indonesia, from September 2022 to March 2024. Artificial hybridization was performed using selected parental genotypes based on predicted flowering periods to ensure synchronized anthesis timing. Five groups of seeds from each parent were planted sequentially with a two-week interval between groups. Emasculation was performed in the afternoon before pollination to prevent self-pollination. A vacuum emasculation procedure was used to quickly and gently remove the anthers from the flowers. This process was carried out at different plant ages depending on the genotype: G2 at 90 days after sowing (DAS), G4 at 91 DAS, G7 at 101 DAS, G10 at 87 DAS, SR Super at 84 DAS, Inbrida M70D at 87 DAS, Trisakti at 89 DAS, MSP 17 Beras Merah at 90 DAS and Genjah Blora at 82 DAS (Aisya et al., 2024). Pollination success was assessed on the fifth day after pollination and F1 seeds were collected 30 days later.
       
Evaluation of F₁ genotypes was conducted using an Augmented Randomized Complete Block Design (ARCBD) with three rows and six columns. Genotype was the fixed effect and block was the random effect (Haines, 2021). Seedlings were first germinated in Petri dishes for seven days. Seedlings were transferred to soil treated with fungicide for 21 days. Finally, seedlings were transplanted into pots containing a 1:1 mixture of soil and organic fertilizer. The fertilizers used were Urea, Ammonium Sulfate (ZA), Single Superphosphate (SP-36) and Potassium Chloride (KCl) at rates of 300, 100, 150 and 150 kg ha^-¹, corresponding to 1.88, 0.63, 0.94 and 0.94 g per pot, respectively. Urea was applied three times at 14, 28 and 42 days after transplanting (DAT). ZA, SP-36 and KCl were applied once at 14 DAT. Watering was carried out daily. Integrated pest and disease management practices were implemented to control pests and diseases. This includes insecticides, fungicides and bactericides containing pymetrozine, mancozeb and copper hydroxide, respectively (Aisya et al., 2024).
       
The observations of agronomic traits were conducted based on the Standard Evaluation System for Rice (SES) including the number of productive tillers (panicles clump^-1), total tillers number, plant height (cm), flowering time (days after sowing/DAS), harvesting time (DAS), number of grains per panicle, grain fertility (%), 100-grain weight (g) and grain weight per plant (g clump^-1) (IRRI, 2013).
       
Data were analyzed using a generalized linear mixed model (GLMM) and Dunnett’s test (p<0.05) was used to compare each genotype with predefined control parents with high-yield traits. All statistical analyses were performed in SAS OnDemand for Academics software (SAS Institute Inc, 2021). Additionally, correlation matrices and hierarchical heatmaps were generated in RStudio software to visualize trait relationships and clustering patterns using the pheatmap, tidyverse and ggplot2 packages (Welham et al., 2015; Kassambara, 2017; R Core Team, 2017; Wijayanti et al., 2023).

General and specific combining ability
 
The combining ability analysis revealed three distinct sources of variation: General Combining Ability (GCA) from female parents, GCA from male parents and Specific Combining Ability (SCA) in F₁ seeds. GCA effects were significant only for 100-seed weight from female parents, whereas SCA effects were significant for all observed traits (Table 2).


       
Maternal-effect tests revealed that female parents had no significant effect on any of the measured traits (Table 3).  The lack of maternal effects corroborates that the combining ability estimates were unbiased, fulfilling a crucial assumption for precise genetic interpretation using the NC II mating design (Muthoni and Shimelis, 2020). The findings of this study shed important light on the genetic regulation of rice traits associated with high-yielding and early-harvesting, which is essential for breeding strategies aimed at adapting to climate change. Focused breeding strategies are guided by analyses of general combining ability (GCA) and specific combining ability (SCA), which clarify the type and degree of gene action underlying these traits (Abdel-Aty et al., 2022). All traits exhibit significant SCA effects, indicating that non-additive gene activity, particularly dominance effects, plays a significant role in their inheritance (Chen et al., 2019).


       
GCA is the mean performance of a parent across all its hybrid combinations (Chen et al., 2019). Female genotypes G2, G4, G7 and G10 had a significant impact on GCA for 100-seed weight (Table 4). Traits with substantial GCA effects were further investigated to determine parental contributions. G4 had the most potent positive GCA effect on 100-seed weight. This is more likely to pass this trait on to its offspring. This makes it a good choice for breeding operations. A higher GCA effect value for 100-seed weight is considered more advantageous. Genotypes with high GCA effects indicate a strong ability to pass the target trait on to their offspring.


       
In contrast to GCA, SCA effects were significant for all evaluated traits, highlighting the predominance of non-additive gene action in the studied population. Negative SCA effects were considered favourable for flowering and harvesting time because shorter duration is preferred, whereas positive SCA effects were advantageous for yield components. G4/GB had the least favourable SCA effect on flowering (-14.56) and harvest time (-15.58). G4/MSP had the most positive SCA effect on plant height with a score of 17.07. G2/GB had the highest positive SCA for productive tillers (5.25) and total tillers (5.56). G4/GB had the longest panicles, the highest number of grains per panicle and the best grain filling with values of 2.85 cm, 23.38 and 11.51%, respectively. G7/GB had the highest SCA for 100-grain weight (0.10) and G4/GB had the highest grain weight per plant (GWP) (13.0) (Table 5).


       
Variance component analysis revealed that dominance variance exceeded additive variance for all traits. The ratios of additive to dominance variance were all less than 1. The additive variance was not greater than dominance variance for any trait. Consequently, narrow-sense heritability estimates were predominantly low. Yadav and Yadav (2023) informed that the high heritability is more than 0.3, moderate heritability is between 0.1 and 0.3 and low heritability is less than 0.1. Twelve traits showed low heritability, seven traits showed moderate heritability and just one trait showed high heritability. The 100-seed weight was the only trait with high narrow-sense heritability (Table 6). This supports a significant GCA effect, indicating that additive genetic selection can lead to substantial improvements in performance. The wide genetic variability and high heritability of various rice yield and quality traits indicate significant opportunities for improvement through direct selection (Saravanan et al., 2024). Furthermore, Chandramohan et al. (2016) also confirmed that genetic diversity among rice genotypes provides a strong basis for sustainable breeding and yield improvement programs.


       
The consistent dominant variance values   were higher than the additive variances for all traits. This trend suggests that the right parental pairing has a greater impact on hybrid performance than the overall effectiveness of each parent. Only the 100-seed weight showed significant GCA for all traits analyzed, while G4 was identified as the effective parent for improving this trait. The importance of additive gene effects in trait expression is evident by the significant GCA effect. This indicates consistent parental contributions across all hybrid combinations (Fang et al., 2019; Kadium and Svyantek, 2023). The high narrow-sense heritability (h² = 0.35) of the 100-seed weight indicates the suitability of the trait as a selection target in early breeding generations. It suggests that it can be reliably transmitted to offspring (Table 6). Traits with high narrow-sense heritability tend to yield higher expected genetic gains through selection (Fang et al., 2019).
 
F₁ genotypes’ performance compared to high-yielding parents
 
The LS-Means differences between each F₁ genotype and the mean of the control female parent (G2, G4, G7 and G10) were presented in Fig 1. The horizontal axis represents the F₁ genotypes resulting from the cross. In contrast, the vertical axis shows the LS-Means values   compared to the control female parent. A rising histogram indicates a higher value than the mean of the control female parent, while a falling histogram indicates a lower value. The light blue area in the middle indicates the 95% confidence interval for the assessment. Histograms outside this area indicate a statistically significant difference.


       
The G4/GB cross showed the shortest flowering duration and highest yield compared to the control group, with an average LS-Means value of 34 days after planting (DAP) for flowering and 93 DAP for harvest. On the other hand, not all F₁ genotypes differed significantly from their female parent in terms of yield components. These results indicate that G4/GB has a significantly earlier harvest and has comparable yield potential to the superior parent. This makes it a suitable candidate for breeding programs aimed at developing high-yielding and early-harvesting varieties.
       
Correlation analysis revealed that the examined qualities were interrelated in various ways. A strong negative correlation was observed between flowering and harvesting time (r = -0.94***). On the other hand, the flowering time was negatively correlated with plant height (r = -0.47*) and grain weight per plant (r = -0.41*). Likewise, harvesting time showed negative associations with the plant height (r = -0.46*), grain weight per plant (r = -0.48*), panicle length (r = -0.39*) and number of grains per panicle (r = -0.43*) (Fig 2a). These associations suggest that earlier flowering and harvesting were typically associated with improvements in yield components. The correlation matrix revealed a strong positive correlation between flowering and harvesting times, whereas all yield components exhibited either significant or insignificant negative correlations with these times.


       
The cluster heatmap showed that the observed variables were grouped into two main clusters (Fig 2b). Cluster 1 consisted of flowering time and harvest time, while Cluster 2 encompassed all other observed traits. The cluster heatmap grouped the F₁ genotypes into two main clusters. The G4/GB genotype was the only one in cluster 1, distinguishing it from the others. G4/GB showed the earliest flowering and harvesting time, as well as relatively high yield component traits, especially grain weight per plant, panicle length and number of grains per panicle. The second cluster was divided into two subclusters. The first subcluster consisted of G7/TRI, G10/SR, G2/MSP17, G2/M70D and G4/TRI, which were generally characterized by relatively low yield components. In addition, the second sub-cluster includes G7/MSP17, G7/SR, G2/SR, G7/GB, G10/TRI, G2/GB, G4/SR, G4/M70D, G10/M70D, G7/M70D, G4/MSP17 and G10/GB, all of which were characterized by relatively high yield components.
       
A study found a negative correlation between specific yield components and harvest time. Variables such as panicle length and grain weight were significantly negatively correlated with flowering/harvest time. However, this can be explained by advances in rice genetics, where the Ef-cd locus has been shown to shorten the harvest period without sacrificing yield (Fang et al., 2019). These results highlight the need for selection processes that consider both inter-trait correlations and combining ability. Furthermore, combining physiological trait modelling with genomic selection can accelerate the identification of new parental combinations with improved climate responsiveness. Our understanding of heterosis and trait inheritance in rice breeding efforts will be enhanced by examining the molecular mechanisms underlying the observed non-additive effects.

G4 was selected as the parent with the most significant GCA effect on 100-seed weight. G4/GB demonstrated the best SCA effect on earlier flowering and harvesting, as well as improved performance in terms of total grains per panicle and grain weight per plant. G4/GB can be a good candidate for hybrid development.

The research for this article was fully funded by the National Research and Innovation Agency and the Indonesia Endowment Funds for Education (LPDP), Ministry of Finance, through the Program Riset dan Inovasi untuk Indonesia Maju Gelombang 3 grant-in-aid scheme (No. 30/IV/KS/05/2023; 2243/UN1/DITLIT/Dit-Lit/PT.01.03/2023).
 
Disclaimers
 
The views and conclusions expressed in this article are solely those of the authors and do not necessarily represent the views of their affiliated institutions. The authors are responsible for the accuracy and completeness of the information provided, but do not accept any liability for any direct or indirect losses resulting from the use of this content.
 
Informed consent
 
No animals are used in this research.

The authors declare that they have no conflict of interest.