Assessment of Efficiency of Breeding Methods in Accelerating Genetic Gain in Rice

Abstract

The pedigree, bulk, and single-seed descent-based rapid generation advance methods are commonly practiced breeding methods in rice. But the efficiency of these breeding methods in enhancing genetic gain has not been investigated yet. In this study, we compared the pedigree and bulk method-derived breeding lines of five crosses with RGA-derived lines. The RGA method was found to be almost two times more efficient in capturing high-yielding lines with a high breeding value and thus accelerated genetic gain much more than the bulk and pedigree methods. The RGA method is not only more efficient but also significantly cheaper (~24%) compared to pedigree methods. The cost per kilogram of genetic gain in yield for the RGA lines is almost 3 times lower than the bulk method and 4.5 times lower than the pedigree method, and it can be achieved in half the time required for line development with either the bulk or pedigree method.

1. Introduction

The use of a superior germplasm, appropriate breeding methods, and careful evaluation are the prerequisites for crop improvement [1,2,3]. The breeding method used for managing the segregation of populations for crop improvement is determined by the mechanism of reproduction of the individuals, the objectives, and the resources available for the breeding program. An efficient method of selection is required to screen the breeding populations to identify elite lines with desirable trait combinations.

The commonly used breeding methods for the improvement of self-pollinated crops are the pedigree, bulk, single-seed descent (SSD), and double haploid methods. Among these, the pedigree method is widely used by breeders across the world [4,5]. This method emphasizes the selection of a single plant from a segregating population until later generations prior to yield testing, which allows the elimination of inferior genotypes before advancing to the next generations. Thereby, the genetic base in the advanced generation lines becomes narrower [6]. In addition, the accuracy of selection in early generations is low [6,7], and this may incur more costs for land use and manpower for the handling of a large number of segregating lines [8,9]. In contrast, the bulk method, in which single plant selection is practiced at later generations before extensive yield testing [7], is comparatively cheaper but genetic variation is greatly compromised. Nonetheless, it has demonstrated significant potential in the breeding of cereals for marginal environments, where limited seed availability poses a challenge. Another popular breeding method is the double haploid method, which can be used to rapidly produce true breeding lines. This method shortens the breeding cycle via the immediate fixation of homozygosity and thereby helps to enhance genetic gain. But it has an inherent weakness relating to its selection efficiency, as the double haploid population lacks allelic recombination. On the other hand, the SSD-based rapid generation advance (RGA) method allows a single progeny to be grown from each plant and does not emphasize single plant selection until the lines become genetically homozygous (usually after F₅ or F₆). Thus, it provides an opportunity for advancing generations in the greenhouse by reducing the time required for inbred line development [10]. Plants grown in the greenhouse or screenhouse flower quicker compared to those grown in normal field conditions. Dense planting in field nurseries, as described by Cobb et al. [11] and Rahman et al. [12], also helps to accelerate early flowering. This method shortens the breeding cycle and leads to the faster recycling of the superior lines in the crossing program; thus, the genetic gain can be enhanced more quickly when using this method compared to other methods. A method called speed breeding [13], which uses LED-controlled specific adjustments in light intensity and quality, has been found to further shorten the flowering time in rice [14]. Nonetheless, only shortening the cycle time may not be sufficient to enhance the genetic gain unless superior parents with a higher breeding value are selected and used to make high-value crosses.

The breeding value (BV) is the sum of the additive genetic effects of all alleles [15] present in the genome of a parental line. The BV can be estimated using either pedigree information or genome sequence information modulated with the phenotypic performance of the breeding lines. The BV estimated using genomic sequence or genotyping information is called the genomic estimated breeding value (GEBV). The genetic variation in quantitatively inherited traits governed by small-effect QTLs and genes can be captured using the GEBV [16]. Thus, parents with higher GEBVs are considered to be better parents. The pedigree breeding method, which favors selection for the traits controlled by major genes, might not be as efficient as the RGA method in capturing the variations in small-effect QTLs or genes.

The breeders’ equation [17] suggests that genetic gain, which measures the rate of genetic improvement per unit time or per unit investment, can be accelerated by reducing the time required for isolating and recycling superior progenies as parents in the breeding program. Although the pedigree, bulk, and SSD methods are used by breeders for the genetic improvement of rice [18,19,20], their efficiency in producing superior progenies has not been well investigated. Mishra et al. [20] compared 25 pedigree lines, 10 SSD lines, and 10 bulk lines derived from a cross in the F₃ and F₄ generations and found that the SSD and bulk methods were more efficient in generating higher-yielding lines compared to the pedigree methods. In a study by Kanbar et al. [19] involving 20 families of a single cross produced using each of the pedigree, modified bulk, and SSD methods, the pedigree method was found to be superior compared to the SSD method regarding grain yield. Interestingly, this conclusion was based on a comparison of the population means and not based on multiple comparison tests. The same applies for the study conducted by Mishra et al. [20]. The comparison of F₆ lines developed through four breeding methods, namely the pedigree, modified pedigree, bulk, and SSD methods, from two crosses by Fahim et al. [18] was more systematic than of Mishra et al. [20] and Kanbar et al. [19]. They observed that none of the best lines derived from either the pedigree, modified pedigree, or bulk methods were better than the lines produced with the SSD method. This study, however, compared the lines generated using selection methods (i.e., pedigree, modified pedigree, and bulk methods) with 100 randomly chosen SSD lines where selection was not practiced at all at any stage for any trait, which caused bias in the mean comparison. Our study was therefore undertaken to compare the best lines derived from each of the pedigree, bulk, and SSD methods for realizing the efficiency of generating high-breeding-value parents with higher selection accuracy and to estimate the associated cost incurred for the line development and evaluation, which in turn helped to attain a higher genetic gain.

2. Materials and Methods

2.1. Plant Materials

A total of 100 breeding lines derived using pedigree method, 99 breeding lines derived using bulk method, and 500 breeding lines derived using SSD-based RGA from five crosses were used in this study. In the case of the pedigree and bulk methods, 15 to 24 superior breeding lines were taken from each cross based on the visual phenotypic superiority, while 100 RGA-derived lines from each cross were taken for this study (

Table 1. Summary of the breeding materials used for investigating the efficiency of the breeding methods.

Population ID	Cross Name	Parentage	No. of F6 Lines Tested
			Pedigree	Bulk	RGA
Cross 1	IR 107982	IR05N412/IRRI 168	21	18	100
Cross 2	IR 107971	IR09A228/IR06N155	19	20	100
Cross 3	IR 107980	IR10F559/IRRI 119	20	22	100
Cross 4	IR 107989	IR11N202/IRRI 168	20	24	100
Cross 5	IR 108000	PR 37246-2-3-2-1-1-2-1/IRRI 154	20	15	100
	Total		100	99	500

). The details of the development of breeding lines can be seen in Supplementary Figure S1.

2.2. Evaluation of the Breeding Lines

Seven hundred breeding lines along with seven check varieties (IRRI 104, IRRI 123, IRRI 146, IRRI 154, IRRI 156, IRRI 168, and IRRI 181) were evaluated to estimate the phenotypic performance at Los Banos, IRRI, and Nueva Ecija, PhilRice, during the wet season (July to November) in the Philippines. The trial utilized a partially replicated design (p-rep design), with 20% of the entries replicated twice. The experimental setup consisted of 912 plots, each covering an area of 3.2 m² in size, arranged in a grid of 12 rows by 76 columns. Check varieties were randomized within each row. Seedlings aged 21 days were transplanted at a spacing of 20 cm × 20 cm, with two seedlings per hill. At crop maturity, plant height was recorded from five representative hills, measuring the length from the base of the hill to the tip of the panicle of the tallest tillers. The days to flowering (FLW) were recorded when 50% of the hills of a plot were at heading. Plot yield and grain moisture (%) content were recorded from the entire plot harvest after processing and drying of the grains. The grain yield (t ha⁻¹) was estimated using the formula given below:

$$\mathrm{Y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}\left(\mathrm{t}{\mathrm{h}\mathrm{a}}^{-1}\right)=10\times \mathrm{P}\mathrm{l}\mathrm{o}\mathrm{t} \mathrm{Y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}\left(\mathrm{k}\mathrm{g}\right)\times (100-\%\mathrm{M}\mathrm{C})/86/\mathrm{P}\mathrm{l}\mathrm{o}\mathrm{t} \mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a} \left({\mathrm{m}}^2\right)$$

(1)

2.3. Genotyping and Data Filtering

Green leaf tissue samples from a representative plant of each breeding line and six check varieties grown at Los Banos were collected in labelled glassine bags at 4–5 weeks after transplanting and stored immediately on ice. The leaf samples were stored in a −80 °C freezer until processing for the genotyping. Genotyping using a custom amplicon panel comprising genome-wide 995 SNP markers named as 1K-RiCA panel V 1.0 [21] was conducted at an outsourcing genotyping service laboratory, Agriplex Inc. (Cleveland, OH, USA) with the help of the IRRI Genotyping Services Laboratory. The leaf tissues were fridge-dried and two discs of 4.5 mm in diameter of each entry were taken in each well of a deep-well plate to send to the Agriplex for genotyping. The DNA extraction and subsequent quality check was performed before genotyping. The genotyping data were filtered using TASSEL v5.0 [22]. The plants with more than 15% of heterozygous loci were removed. Also, the markers with more than 15% of missing values along with a minor allele frequency below 0.05 were removed. After filtering, a total of 889 markers were retained for further downstream analysis.

2.4. Extraction of Best Linear Unbiased Predictor (BLUP)

A two-stage mixed linear model analysis [23,24] was performed using grain yield data as a response variable to estimate the BLUP value of each line avoiding spatial variations [25] across the locations. In the first stage, each trial was analyzed separately and BLUP values were extracted per location using the following baseline mixed model:

$$Y_{ij}=\mu +g_i+\dots \dots \dots +{\mathsf{\epsilon }}_{ij}$$

(2)

where $Y_{ij}$ represents grain yield for ith observation, μ is the overall mean, and $g_i$ is the random effect of ith genotype with iid g_i∼N(0, Iσ²g), where σ²g is the genetic variance and ${\mathsf{\epsilon }}_{ij}$ is the residual error with iid ${\mathsf{\epsilon }}_{ij}$∼N(0, Rσ²ε).

The rows, columns, and replicates were considered as the blocking factors and included as random in the model to determine which factors led to the lowest Bayesian information criterion (BIC) [26,27]. The model with the lowest BIC was selected and used to extract the BLUP value for each line, and their prediction error variances (PEV) were obtained for each environment. The reliability of the BLUPs was estimated according to r² = 1 − PEV σ²g.

In the second stage, the BLUP values obtained from the first stage of the model were de-regressed by dividing by the reliability as described in Garrick et al. [28] and used as a response variable in the second stage of the model analysis. The de-regressed BLUPs for yield within each environment were modeled according to Bates et al. [29]. The model used is as follows:

$$Y_{ij}=\mu +g_i+e_j+{\mathsf{\epsilon }}_{ijk}$$

(3)

where, $Y_{ij}$ is the de-regressed BLUP of each line in environment j, μ is the overall mean, and $g_i$ is a random effect of the line i with $g_i$∼N(0, Aσ²g), where σ²g is the genetic variance and A is the additive genetic relationship matrix, e_j is the fixed effect of the environment j, and ${\epsilon }_{ij}$ is the residual error with ${\mathsf{\epsilon }}_{ij}$ εij∼N(0, Rσ²ε), where R is a matrix proportional to the residual error covariance matrix and σ²ε is the error variance. To account for the heterogeneous error variance, the diagonal of R was adjusted to 1/r². The R package lme4 [30] was used to implement the models.

2.5. Estimation of Genomic Estimated Breeding Values

The genomic estimated breeding value (GEBV) of each line was extracted by integrating a kinship matrix encompassing all breeding lines and performance BLUP into the rrBLUP model [31]. The kinship matrix was constructed from the genotyping data obtained from 995 genome-wide SNP markers, using the A.mat() function within the rrBLUP package. Subsequently, the mixed.solve() function of rrBLUP was employed to calculate the GEBV for each line with respect to the grain yield.

2.6. Mean Comparison Analysis

During line fixation stages in the pedigree and bulk methods, selection was performed to isolate superior lines, but in the case of RGA, no selection was practiced. Thus, to create a similar selection environment, 100 superior RGA lines were selected out of 500 lines (taking the top 20 lines from each of the five crosses) based on the yield performance across the sites and were designated hereafter as the “selected RGA lines”. The mean of these selected RGA lines was then subjected to a mean comparison with the lines derived from the pedigree and bulk breeding methods. Tukey’s HSD test was performed using the Agricolae Package in R [32] for assessing the significance of the mean difference between the breeding methods.

2.7. Estimation of Timelapse for the Development and Identification of Superior Lines

The duration spanning from hybridization through to yield testing delineated the time required to isolate superior lines via each respective breeding method. The breeding lines used in this study were generated from the crosses made during the wet season (WS) of 2013. Subsequently, hybridity testing and the production of F₂ seeds from the true F₁ progeny were executed during the dry season (DS) of 2014. For both the pedigree and bulk methods, the advancement of segregating generations from F₂ to F₆ took place from 2014WS to 2016WS. Similarly, within the RGA method, the progression of F₂ to F₆ occurred from 2014WS to 2015WS. Across all three methods, a seed amplification season preceded the yield testing.

2.8. Assessment of Genetic Gain

The genetic gain (ΔG, g/year) in grain yield was calculated for the pedigree, bulk, and RGA methods using the following equation derived from Falconer [33].

$$∆G=\frac{i\times r \times \sigma }{L}$$

(4)

where i is the selection intensity as a function of the proportion of the population selected, r is the squared root of yield heritability, and L is the generation length in years. Here, σ is the genetic standard deviation, which was calculated as the standard deviation of GEBVs from Equation (3).

2.9. Estimation of Cost for the Line Development

The cost of line development (USD line⁻¹) for each breeding method adhered to a full cost recovery policy. This encompassed all breeding operations associated with line development, including hybridization, genotyping for parental quality control, cross confirmation, line fixation, and line evaluation. The calculation also factored in various expenses such as the salary shares of scientific and research support personnel, general operational cost encompassing office and field/greenhouse supplies, expenses for the different services involved in line development and evaluation, and depreciation costs of the equipment utilized in these processes. To determine the cost for development of each line, the total incurred cost for the aforementioned items was divided by the number of selected lines (100 lines) tested for this study. In the case of the RGA-derived lines, the cost for developing 500 lines was considered as the cost for the development of 100 selected RGA lines.

2.10. Estimation of Selection Differential

The selection differential in terms of cost (USD) per kilogram of grain yield was calculated by dividing the line development cost by the deviation of the mean of all 500 tested RGA lines from the mean of the pedigree, bulk, and selected RGA lines. For a detailed breakdown of the calculation, please refer to Supplementary Excel Sheet S1.

3. Results

3.1. Effect of Breeding Methods on Population Mean

The analysis of variances conducted on the breeding lines for grain yield, plant height, and days to flowering revealed significant variations (p < 0.001) both within and among the breeding methods and within and among the locations, except for plant height. Additionally, the interaction between breeding methods and breeding populations was significant for all three traits, as well as between breeding populations and locations for days to flowering. Furthermore, significant interactions were observed between the breeding method and location for yield, and among breeding methods, breeding populations, and locations for yield. However, the interaction between the population and location for yield was found to be non-significant (Supplementary Table S1).

The mean values of the breeding lines within each cross displayed significant variability (p < 0.05) among the breeding methods for all three traits across different locations (

Table 2. Comparison of mean performance of the breeding lines derived from different breeding methods for days to flowering, plant height, and yield.

	Days to Flowering	Plant Height (cm)	Yield (t ha−1)
Across five crosses
Pedigree lines	92 a	123.81 a	3.81 b
Bulk lines	91 a	116.23 b	3.79 b
Selected RGA lines	88 b	113.74 c	4.12 a
Individual crosses
Cross 1
Pedigree lines	92 a	126.51 a	3.76 b
Bulk lines	90 ab	126.94 a	3.89 b
Selected RGA lines	89 b	113.99 b	4.34 a
Cross 2
Pedigree lines	92 a	115.00 b	3.69 b
Bulk lines	90 ab	130.02 a	3.86 b
Selected RGA lines	89 b	116.54 b	4.17 a
Cross 3
Pedigree lines	91 a	113.24 a	4.04 a
Bulk lines	90 a	115.93 a	3.89 a
Selected RGA lines	86 b	111.85 a	4.12 a
Cross 4
Pedigree lines	91 a	115.96 b	3.76 b
Bulk lines	90 a	124.26 a	3.89 b
Selected RGA lines	88 b	112.31 b	4.18 a
Cross 5
Pedigree lines	89 a	110.97 c	4.04 a
Bulk lines	89 a	121.74 a	3.89 b
Selected RGA lines	87 b	114.98 b	3.81 b

Note: Significance at the 5% level of probability; entries sharing the same letters indicated a non-significant difference.

). Across the crosses, the days to flowering ranged from 88 to 92 days, with the selected RGA lines exhibiting the lowest values, which differed significantly from the other two methods. Within individual crosses, the days to flowering varied from 89 to 92 days in Cross 1 and Cross 2, 86 to 91 days in Cross 3, 88 to 91 days in Cross 4 and 87 to 89 days in Cross 5. Similarly, the lowest plant height (113.74 cm) was observed in the RGA lines across the crosses, contrasting with the highest value in the pedigree lines (123.81 cm). Notably, in Cross 1 and Cross 3, the RGA lines exhibited the lowest plant height. Across all crosses and locations, the tested breeding lines yielded between 3.81 and 4.12 t ha⁻¹, with the bulk lines recording the lowest mean grain yield and the selected RGA lines achieving the highest value, significantly differing from the mean values of the other two methods. However, the mean value of the bulk lines did not significantly differ from that of the pedigree method. In terms of individual crosses, the mean grain yield values in the selected RGA lines were significantly higher than in any other methods, except for Cross 5. However, the mean yield of all 500 RGA lines stood at 3.68 t ha⁻¹ across the locations. Further details on the mean comparison among the pedigree, bulk, and selected RGA lines with all 500 RGA lines can be found in Supplementary Excel Sheet S2.

3.2. Effect of Breeding Methods on Breeding Value

Significant variation in the mean GEBV for grain yield was observed among different breeding methods. The highest breeding value for yield (0.656 t ha⁻¹) was recorded for RGA-derived lines, whereas the lowest (0.514 t ha⁻¹) was observed for lines derived from the bulk method. Notably, when considering individual crosses (as detailed in

Table 3. Comparison of mean genomic estimated breeding value (GEBV) of the breeding lines derived from different breeding methods for days to flowering, plant height, and yield.

	Days to Flowering	Plant Height (cm)	Yield (t ha−1)
Across five crosses
Pedigree lines	3.86 a	6.55 b	0.579 b
Bulk lines	3.47 a	15.93 a	0.514 c
Selected RGA lines	−1.30 b	2.88 c	0.656 a
Individual crosses
Cross 1
Pedigree lines	9.51 a	18.92 a	0.339 b
Bulk lines	10.32 a	19.85 a	0.111 c
Selected RGA lines	−1.55 b	4.07 b	0.482 a
Cross 2
Pedigree lines	4.04 a	5.13 b	0.389 b
Bulk lines	1.88 b	23.48 a	0.397 ab
Selected RGA lines	0.40 b	5.17 b	0.483 a
Cross 3
Pedigree lines	3.21 a	2.89 b	0.797 a
Bulk lines	1.83 b	6.16 a	0.668 b
Selected RGA lines	−2.50 c	1.62 b	0.779 a
Cross 4
Pedigree lines	2.44 a	6.23 b	0.684 b
Bulk lines	2.26 a	16.34 a	0.582 c
Selected RGA lines	−1.42 b	1.12 c	0.835 a
Cross 5
Pedigree lines	0.38 a	0.21 c	0.642 c
Bulk lines	0.74 a	13.64 a	0.779 a
Selected RGA lines	−1.72 b	4.35 b	0.724 b

Note: Significance at the 5% level of probability; entries sharing the same letters indicated a non-significant difference.

) the mean GEBVs displayed distinct patterns. The RGA lines exhibited the highest mean GEBV for yield across most crosses, with Cross 5 being an exception where the second highest GEBV was noted. The bulk method yielded the lowest mean GEBV for yield in Crosses 1, 3, and 4, and the pedigree method in Crosses 2 and 5.

Similarly, the mean GEBV for the days to flowering varied significantly between breeding methods and across crosses. Across all five crosses, the RGA method consistently yielded the lowest GEBV for the days to flowering, while the pedigree method exhibited the highest values. Notably, the bulk method showed GEBVs similar to those of the pedigree method in Crosses 1, 4, and 5. Regarding plant height, a similar trend emerged, with the mean GEBV being the lowest with the RGA method and highest with the bulk method across all crosses. In specific crosses, such as Crosses 1, 2, 3, and 4, the RGA method yielded the lowest GEBVs for plant height, although Crosses 2 and 3 also shared the lowest values with the pedigree method. Conversely, the bulk method consistently demonstrated the highest GEBVs for plant height across all crosses.

3.3. Effect of Breeding Methods in Generating Superior Lines

The average yield of the selected 100 breeding lines comprising the top 20 lines from each cross derived from different breeding methods varied remarkably among the individual crosses (Figure 1 and Figure S2). In Cross 1, the breeding lines derived from the pedigree method yielded 4.25 t ha⁻¹ to 5.03 t ha⁻¹, while the lines of the bulk method yielded 4.29 t ha⁻¹ to 5.07 t ha⁻¹, and the lines of RGA method yielded 4.84 t ha⁻¹ to 5.05 t ha⁻¹. The breeding lines of Cross 2 derived through pedigree, bulk, and RGA method yielded 4.04–5.05 t ha⁻¹, 4.5–5.04 t ha⁻¹, and 4.81–5.35 t ha⁻¹, respectively. In the case of Cross 3, the pedigree lines yielded 4.66 t ha⁻¹ to 5.19 t ha⁻¹, while the bulk lines yielded 4.29 t ha⁻¹ to 5.12 t ha⁻¹ and the RGA lines yielded 4.58 t ha⁻¹ to 5.17 t ha⁻¹. The pedigree, bulk, and RGA-derived lines of Cross 4 yielded 4.49 t ha⁻¹ to 5.08 t ha⁻¹, 4.36 t ha⁻¹ to 5.17 t ha⁻¹, and 4.96 t ha⁻¹ to 5.16 t ha⁻¹, respectively. The breeding lines developed through the pedigree, bulk, and RGA methods from Cross 5 yielded 4.45–5.09 t ha⁻¹, 4.36–5.17 t ha⁻¹, and 4.87–5.14 t ha⁻¹, respectively. However, the best check variety, which was identified to be IRRI 154, yielded 4.5 t ha⁻¹.

The breeding methods also showed variable efficiency in generating superior lines compared to the highest-yielding check varieties. Among the top twenty breeding lines chosen from each breeding method, 10 lines from the pedigree method and 12 lines from the bulk method achieved yields surpassing the highest-yielding check variety IRRI 154 by 0.5 t ha⁻¹ or more. Remarkably, all 20 RGA lines demonstrated a yield advantage of at least 0.5 t ha⁻¹ over IRRI 154 (

Table 4. Performance of the top twenty breeding lines derived from each of the pedigree, bulk, and RGA breeding methods.

Pedigree					Bulk					RGA
Designation	Yield (t ha−1)	SE (t ha−1)	FLW (day)	HT (cm)	Designation	Yield (t ha−1)	SE (t ha−1)	FLW (day)	HT (cm)	Designation	Yield (t ha−1)	SE (t ha−1)	FLW (day)	HT (cm)
IR 108000-B-40-3-2-1	5.19	0.109	92	108.9	IR 107976-B-B-B-B-8	5.17	0.266	88	116.7	IR 107989-B-B RGA-B RGA-174	5.35	0.277	92	116.8
IR 108000-B-39-3-1-1	5.12	0.357	89	115.3	IR 107971-B-B-B-B-26	5.17	0.096	89	115.2	IR 107989-B-B RGA-B RGA-282	5.19	0.248	86	111.7
IR 108000-B-49-3-3-1	5.10	0.108	92	116.4	IR 108000-B-B-B-B-19	5.13	0.035	90	120.1	IR 108000-B-B RGA-B RGA-60	5.17	0.565	90	117.1
IR 107976-B-9-2-1-1	5.09	0.215	89	119.9	IR 107971-B-B-B-B-15	5.12	0.131	92	128.4	IR 107971-B-B RGA-B RGA-582	5.16	0.053	86	112.4
IR 108000-B-51-1-2-1	5.09	0.078	90	113.0	IR 108000-B-B-B-B-13	5.12	0.063	84	118.1	IR 107971-B-B RGA-B RGA-98	5.15	0.369	87	114.3
IR 107971-B-20-1-2-1	5.09	0.029	90	111.2	IR 107976-B-B-B-B-16	5.09	0.240	90	128.4	IR 107976-B-B RGA-B RGA-415	5.14	0.228	87	117.4
IR 107989-B-39-3-2-1	5.05	0.168	94	113.7	IR 107982-B-B-B-B-5	5.08	0.255	96	122.5	IR 107971-B-B RGA-B RGA-195	5.14	0.424	88	113.7
IR 107982-B-10-3-1-1	5.03	0.452	97	126.5	IR 107971-B-B-B-B-35	5.08	0.288	93	125.7	IR 107971-B-B RGA-B RGA-74	5.12	0.593	87	115.3
IR 107976-B-19-1-1-1	5.03	0.045	90	107.3	IR 108000-B-B-B-B-36	5.07	0.280	89	110.5	IR 107971-B-B RGA-B RGA-265	5.10	0.211	89	116.1
IR 108000-B-54-1-3-1	5.02	0.108	91	114.8	IR 107976-B-B-B-B-19	5.06	0.254	87	120.2	IR 107971-B-B RGA-B RGA-101	5.09	0.117	89	114.3
IR 108000-B-58-1-2-1	4.99	0.211	90	110.8	IR 107989-B-B-B-B-24	5.05	0.138	92	136.8	IR 107976-B-B RGA-B RGA-440	5.09	0.338	88	115.7
IR 107976-B-20-1-1-1	4.97	0.495	88	106.9	IR 107989-B-B-B-B-10	5.01	0.345	90	135.1	IR 107971-B-B RGA-B RGA-190	5.08	0.250	87	112.0
IR 107971-B-9-3-3-1	4.95	0.083	90	111.9	IR 107976-B-B-B-B-11	4.98	0.111	86	117.2	IR 107976-B-B RGA-B RGA-542	5.06	0.319	86	115.1
IR 108000-B-36-2-1-1	4.95	0.174	93	112.9	IR 107982-B-B-B-B-31	4.96	0.741	97	132.1	IR 107976-B-B RGA-B RGA-291	5.05	0.038	87	118.4
IR 107976-B-21-3-3-1	4.94	0.096	90	115.5	IR 107976-B-B-B-B-21	4.96	0.056	89	130.8	IR 107971-B-B RGA-B RGA-377	5.05	0.476	86	110.9
IR 107971-B-28-1-1-1	4.92	0.014	93	110.5	IR 107971-B-B-B-B-36	4.96	0.196	87	116.0	IR 107982-B-B RGA-B RGA-123	5.05	0.365	85	126.2
IR 108000-B-36-1-1-1	4.92	0.438	89	111.6	IR 107989-B-B-B-B-17	4.95	0.343	93	133.1	IR 107971-B-B RGA-B RGA-151	5.05	0.515	89	110.8
IR 107989-B-18-1-1-1	4.92	0.571	90	116.6	IR 107976-B-B-B-B-3	4.94	0.249	91	121.3	IR 107982-B-B RGA-B RGA-266	5.05	0.014	82	103.8
IR 107976-B-19-3-2-1	4.90	0.140	89	102.7	IR 107971-B-B-B-B-21	4.93	0.331	86	114.9	IR 107989-B-B RGA-B RGA-375	5.04	0.004	87	117.5
IR 107976-B-19-2-2-1	4.89	0.079	89	109.0	IR 108000-B-B-B-B-16	4.93	0.189	92	120.6	IR 107989-B-B RGA-B RGA-372	5.04	0.180	87	111.7
IRRI 104	3.37	0.415	75	90.4	IRRI 104	3.37	0.415	75	90.4	IRRI 104	3.37	0.415	75	90.4
IRRI 123	3.88	0.273	89	116.4	IRRI 123	3.88	0.273	89	116.4	IRRI 123	3.88	0.273	89	116.4
IRRI 154	4.50	0.151	90	118.7	IRRI 146	4.50	0.151	90	118.7	IRRI 146	4.50	0.151	90	118.7
IRRI 156	4.06	0.223	89	121.9	IRRI 154	4.06	0.223	89	121.9	IRRI 154	4.06	0.223	89	121.9
IRRI 168	4.06	0.251	91	123.4	IRRI 156	4.06	0.251	91	123.4	IRRI 156	4.06	0.251	91	123.4
IRRI 181	3.71	0.322	92	123.1	IRRI 168	3.71	0.322	92	123.1	IRRI 168	3.71	0.322	92	123.1
Average	4.75	0.105	90.0	113.4		4.77	0.107	89.5	121.4		4.83	0.113	87.3	114.0

SE = standard error, FLW = days to flowering, HT = plant height (in cm).

).

3.4. Estimation of Timelapse for the Development and Isolation of Superior Lines

In this study, the production and confirmation of true F₁ progenies necessitated a span of two seasons (equivalent to 1.0 year), inclusive of pre-processing time across all three breeding methods (

Table 5. Time required for the developing superior lines from different breeding methods.

Breeding Methods	Pedigree	Bulk	RGA
Establishment of crossing blocks, parental purification, and hybridization	0.5 year	0.5 year	0.5 year
Establishment of F1 confirmation nursery, conducting hybridity tests, and production of F2 seeds from true F1 plants	0.5 year	0.5 year	0.5 year
Advancement of segregating F2–F5 generations (line fixation)	2.0 years	2.0 years	1.0 years
Seed amplification	0.5 year	0.5 year	0.5 year
Yield testing	0.5 year	0.5 year	0.5 year
Total	4.0 years	4.0 years	3.0 years

). The pedigree and bulk methods required two years, totaling four seasons, for the line fixation stage (F₂–F₅ stage). Subsequently, a single season (approximately six months) sufficed for seed amplification followed by an additional season dedicated solely to yield testing. In total, 4.0 years were invested in developing and isolating superior lines using these two breeding methods. In contrast, the time requirements for all stages in the RGA method remained consistent, except for the line fixation stage, which only necessitated one year instead of two. Consequently, the RGA method streamlined the process, requiring only 3.0 years for the isolation of superior lines and subsequent recycling within the breeding program.

3.5. Comparative Costs for the Superior Line Development and Relative Selection Differential

Table 6. Estimated cost for line development in different breeding methods.

Item-Wise Expense (USD)	Pedigree	Bulk	RGA
Personnel cost	3778.66	2332.18	2695.23
General operational cost (office supplies, Farm/laboratory supplies, contractual services, others)	3301.28	1652.28	1483.55
Service center charge (land rental, services, equipment, others)	1103.62	681.16	1610.02
Equipment depreciation cost	11.47	7.08	265.36
Operational support service	202.89	125.22	315.44
Total cost *	8397.92	4797.92	6369.6
Total no. of selected lines	100	100	100
Cost per line (including evaluation cost)	83.98	47.98	63.69
Cost per plant per generation (line development)	0.24	0.12	0.19
Selection differential (USD/kg grain)	64.60	43.62	14.48

* The estimated costs include all costs incurred for the genotyping for quality control and hybridity test of F₁, line fixation, and evaluation of 100 lines each of the pedigree and bulk methods and 500 RGA lines.

shows the cost incurred for the line development and isolation of superior lines using different breeding methods. The total expenses, encompassing scientific personnel salaries, operational costs, service charges and operational support were calculated for each breeding method. The pedigree and bulk method incurred USD 8397.92 and USD 4797.92, respectively, to develop and evaluate 100 breeding lines each. Meanwhile, USD 6369.6 was allocated to develop and evaluate 500 RGA lines, resulting in the selection of 100 superior lines (20 lines from each of five crosses) after yield and other attribute-based selection. Consequently, the costs per superior line development and isolation were USD 83.98, USD 47.98, and USD 63.69 for the pedigree, bulk, and RGA methods, respectively. However, the cost per plant per generation stood at USD 0.24 for pedigree, USD 0.12 for bulk, and USD 0.19 for RGA method. In terms of cost per kilogram of grain yield, the selection differentials were USD 64.60 for pedigree, USD 43.12 for bulk, and USD 14.46 for the RGA method, respectively.

3.6. Estimation of Genetic Gain

The genetic gain estimates for the grain yield in breeding lines derived from different breeding methods were determined using common selection schemes, namely the selection of 1%, 5%, and 10% superior lines from the base population.

Table 7. Estimates of genetic gain for the grain yield of breeding lines developed through different breeding methods.

Breeding Method	Selection (%)	Selection Intensity (i)	Selection Accuracy (r)	Genetic Standard Deviation (σg)	Length of Breeding Cycle, L (year)	Genetic Gain, ∆G (kg ha−1 year−1)
Bulk	1	2.665	0.539	0.302	4	108.48
	5	2.063	0.539	0.302	4	83.98
	10	1.755	0.539	0.302	4	71.44
Pedigree	1	2.665	0.539	0.297	4	106.71
	5	2.063	0.539	0.297	4	82.61
	10	1.755	0.539	0.297	4	70.28
RGA	1	2.665	0.539	0.420	3	201.36
	5	2.063	0.539	0.420	3	155.88
	10	1.755	0.539	0.420	3	132.60

illustrates that as the selection percentage increases, the genetic gain decreases across all breeding methods.

In the bulk method, the genetic gain was 108.48 kg ha⁻¹ year⁻¹, 83.98 kg ha⁻¹ year⁻¹, and 71.44 kg ha⁻¹ year⁻¹ for 1%, 5%, and 10% selection, respectively. For the pedigree method, the genetic gain estimates were 106.71 kg ha⁻¹ year⁻¹, 82.61 kg ha⁻¹ year⁻¹, and 70.28 kg ha⁻¹ year⁻¹ for 1%, 5%, and 10% selection, respectively. Conversely, employing a 1%, 5%, and 10% selection intensity, the RGA method exhibited higher genetic gains of 201.36 kg ha⁻¹ year⁻¹, 155.88 kg ha⁻¹ year⁻¹, and 132.60 kg ha⁻¹ year⁻¹, respectively.

4. Discussion

The selection pressure for identifying superior progenies from segregating generations is influenced by the levels of genetic variation present in the breeding population. The extent of genetic variability depends upon the breeding method employed to develop the segregating population. While the selection pressure in the early generations offers an opportunity to select superior lines, known as transgressive segregants, it narrows the genetic variability within the breeding population [6].

Selection after a few generations of selfing offers a greater potential for capturing transgressive segregants that exhibit a substantial response to selection for quantitatively inherited traits such as yield. According to Kearsey and Pooni [8], selection at early generations yields a minimal response, whereas selection after a few generations of inbreeding can enhance the response by up to five times. In this study, we investigated the relative efficiency of the pedigree, bulk, and single seed descent-based RGA methods in responding to selection.

The significant mean sum of squared values for yield and other traits across different breeding methods, along with their interaction with the breeding population (progenies derived from a cross), underscores the considerable influence of breeding methods on the performance of lines across various crosses and locations.

The pedigree and bulk methods inherently exhibit bias due to breeders’ preference for superior progenies during line fixation, stemming from the practice of individual plant selection in the early segregating generations. In contrast, no selection is practiced during line fixation in the case of RGA, resulting in the retention of inferior lines during generation advancement [7], consequently leading to a lower population mean. The low average yield (3.68 t ha⁻¹) in the base population of RGA lines (Supplementary Table S2) supports these observations. To ensure a uniform selection environment across all breeding methods in this study, selection was imposed on the base RGA lines for yield, nullifying breeders’ biasness for mean comparison.

The mean difference clearly demonstrates the significant influence of breeding methods on yield, flowering time, and plant height. The mean grain yield of the selected RGA lines across the crosses significantly surpassed those of the pedigree and bulk lines (Table 2). Additionally, the mean difference in breeding methods in individual crosses for yield and other traits were also significant.

A computer simulation study conducted by Van Oeveren and Stam [34] highlighted the superiority of the SSD method over the pedigree method for low to moderately heritable traits. Fahim et al. [18] observed the superiority of RGA lines, albeit comparing them with the lines derived from the pedigree and bulk methods, which involve selection during line fixation, while the base population of RGA lines experienced no selection. Conversely, Kanbar et al. [19] reported the superiority of pedigree methods over RGA methods for grain yield. However, both reports were based on single plant yield, failing to accurately reflect the true performance of the lines. Therefore, to ensure a realistic and meaningful mean comparison, selection was imposed on the base RGA population prior to comparing with the mean of breeding lines derived from the pedigree and bulk methods.

The selection of progenies from the breeding population at segregating generations serves to eliminate inferior progenies or underperforming lines, consequently reducing the genetic variability among the population. The pedigree and bulk methods enable selection at early segregating generations, yet this approach does not guarantee the development of high-yielding lines from subsequent yield trials. Conversely, RGA lines developed through the SSD method exhibit a broad spectrum of genetic variability, as this method refrains from pre-testing selection, allowing for extensive yield testing with fixed lines. Consequently, the likelihood of obtaining superior lines is higher with the RGA method compared to the pedigree and bulk methods.

In our study, we observed that among the top 20 breeding lines from each of the pedigree and bulk breeding methods, 10–12 lines outperformed the best check, IRRI 154, by at least 0.5 t ha⁻¹. However, in the case of RGA-derived lines, all of the top 20 lines exhibited at least a 0.5 t ha⁻¹ higher yield than IRRI 154. This result clearly indicates that the probability of obtaining superior lines from the RGA method is nearly twice as high compared to the pedigree and bulk methods (Table 4 and Figure 1). Collard et al. [7] also observed a higher frequency of transgressive and superior lines compared to the check variety, IRRI 154, in a study involving two breeding populations derived through SSD-based RGA methods. Additionally, Collard et al. [35] reported the superiority of SSD-based RGA over the pedigree and bulk methods.

The breeding value holds paramount importance in the recycling of parental lines within breeding programs to enhance genetic gain [11]. Given that GEBV serves as an estimate of additive genetic effects from all favorable alleles, a higher GEBV is typically indicative of a superior parent. However, the accuracy of GEBV estimation is crucial. The mean comparison utilizing GEBV estimates revealed significant differences in the mean breeding values of breeding lines derived from the pedigree, bulk, and RGA methods. Notably, the mean breeding values were the highest for the selected RGA lines, although they were the lowest for the initial 500 base RGA lines, as detailed in Supplementary Table S2, in comparison to the means of the pedigree, bulk, and selected RGA lines. These findings suggest that selection for a higher yield in the RGA lines facilitated the accumulation of favorable alleles governing the trait, consequently resulting in elevated breeding values in the selected RGA lines.

In general, the cost of line development in RGA is often considered more economical compared to other methods as reported by Fahim et al. [18], Collard et al. [7], and Cobb et al. [11]. In our investigation, we assessed the cost of line development, encompassing expenses for hybridization, line fixation, and initial yield evaluation accounting for all components necessary for full cost recovery. We found that the bulk method incurred the lowest line development cost, while the pedigree method incurred the highest. Specifically, the cost for developing and identifying superior RGA lines was approximately 24% less than that of the pedigree methods (Table 6). However, Fahim et al. [18] reported that the RGA method was 10 times cheaper compared to the pedigree method.

The higher cost of RGA lines compared to the bulk method stemmed from the inclusion of the line evaluation expense to screen out inferior ones. A significant drawback of the SSD method is the retention of a substantial number of inferior lines alongside superior ones during line fixation, necessitating their removal before comparing breeding methods that favor the selection of superior lines. Although the cost of developing selected RGA lines fell between the costs of the pedigree and bulk methods, the selection differential in terms of cost per kg of grain yield, reflecting the actual response to selection, was almost 4.5 times lower compared to the pedigree method and three times lower compared to the bulk method (Table 6). Collard et al. [7] suggested that a larger response to selection could be achieved in the half the time required for either the pedigree or bulk methods with RGA.

In this study, it took two years to identify superior F₆ lines using either the pedigree or bulk method, while only one year was sufficient for line fixation up to the F₆ generation following an early harvest and immediate seeding schedule in the RGA method (Table 5). Still et al. [36] reported that pre-mature seeds harvested 10 days after flowering germinate after imbibing water. Considering the period for hybridization, seed amplification, and yield testing, the breeding cycle time for the pedigree and bulk methods was 4.0 years and 3.0 years for the RGA method. If field RGA techniques [11,12] are used for line fixation, the breeding cycle time may increase slightly but the cost per kilogram yield will decrease significantly [7]. Despite the risks posed by natural calamities such as typhoons, cyclones, or tropical storms that can damage field nurseries, which can also occur with greenhouse RGA, the breeding cycle time can be further shortened using controlled speed breeding techniques [13,14,37,38,39]. However, this would greatly increase the cost of line fixation.

The breeder’s equation [17] suggests that by shortening the breeding cycle time and reducing the cost for per kilogram yield, the genetic gain can be enhanced through selection. In our study, we found that the lines developed through the RGA method displayed the highest rate of genetic gain for yield, almost twice that of the pedigree and bulk methods (Table 7). The RGA method not only reduced the breeding cycle time but also decreased the cost per kilogram yield in terms of the selection differential, thereby increasing the rate of genetic gain compared to the other breeding methods. However, the rate of genetic gain depends on many other factors involved in the breeding process. Integrating advanced molecular and omics tools in the RGA-based breeding programs might further accelerate the recovery of lost genetic variation [40,41] that may occur during early generation yield testing and selection processes.

5. Conclusions

The various breeding methods significantly influenced the agronomic performance of the developed lines. The selected RGA lines exhibited a superior performance in grain yield compared to lines developed from the bulk and pedigree methods. While the line development cost in RGA is relatively higher than that in the bulk method, it remains substantially more affordable than the pedigree method. Moreover, the cost per kilogram genetic gain attributed to the selection differential in the yield of the selected RGA lines is 3 times cheaper than the bulk method and 4.5 times cheaper than the pedigree method and can be achieved in half the time required for either the bulk or pedigree method. These findings are poised to inspire and catalyze resource-constraint public sector breeding programs to embrace SSD-based RGA breeding methods, thereby expediting genetic enhancements in rice cultivation.