Wild rice O. rufipogon Griff. of Raiganj was used as one of the parental lines (donor parent). This wild rice variety grows naturally in the shallow marshy land/ditches of the Raiganj block, Uttar Dinajpur district, West Bengal, India, at latitude 25.62 °N and longitude 88.12 °E, and elevation of 40 m (130 ft). This wild rice variety is highly shattering in nature and fully spreading in the habitat with an annual growth pattern, and it produces awned spikelets on the spreading panicles and disperses the mature seeds. Its grain and hull colors are red and black, respectively. Two well-adapted farmer’s varieties of O. sativa L. subspecies indica cultivar, Badshabhog and Chenga, were used as the parents in this interspecific hybridization. Badshabhog has white aromatic grains with a straw-colored husk, whereas Chenga has a blackish husk with nonaromatic brown grains. Both the farmer’s rice varieties (Badshabhog and Chenga) have been deposited in the National Rice Gene Bank, NBPGR-ICAR, Govt. of India, New Delhi, for conservation purpose with indigenous collection numbers (IC No-0652950 for Chenga and IC no-0652952 for Badshabhog).

The mean pooled data obtained from two kharif seasons (2022 and 2023) were used for biometrical analysis. The genotypic and phenotypic variation, broad-sense heritability, genetic advance, and genotypic and phenotypic correlation coefficients were estimated. Genetic diversity analysis was performed following D² statistics proposed by Mahalanobis (1936). The RIL genotypes were classified into several clusters by Tocher’s method using Mahalanobis D² distance statistics (Rao, 1952). The broad-sense heritability (H%) and other genetic variability parameters were calculated using the standard methods (Johnson et al., 1955; Allard, 1960). Phenotypic coefficients of correlation were calculated based on Burton and de Vane’s formula (1953). The multivariate PCA was utilized to estimate the relative contribution of various traits to the total variability based on the original concept of Pearson (Hotelling, 1933). Statistical analyses were carried out using various software applications, such as SPSSv-22, XLSTAT, PAST4.03, Origin 2024, and R4.4.1.

The following formulas were used to calculate the genetic variability parameters:

Broad-sense heritability of the breeding lines was estimated using the formula by Allard (1960). Broad sense heritability  H = σ 2 g σ 2 p × 100 . where (H) = broad-sense heritability; σ²p = phenotypic variance; σ²g = genotypic variance. Genotypic variance (σ²g) = (MS₂ –MS₃)/b; error variance (σ²e) = MS₃; MS₂ = mean square of populations; MS₃ = mean square of error; b = number of blocks. Phenotypic variance  σ p 2 = σ g 2 + σ e 2 . Genotypic coefficient of variation  GCV = σ g 2 / X − × 100 . Phenotypic coefficient of variation  PCV = σ p 2 / X − × 100 .

X^¯ = the mean of the trait.

Different variance components such as the phenotypic coefficients of variation (PCV) and genotypic coefficient of variation (GCV) were estimated according to the method of Burton and de Vane (1953). Environmental variance was calculated by the formula suggested by Burton and de Vane (1953). The estimation of genetic advance (GA) and the genetic advance as percentage of the mean (GAM) were calculated according to the method described by Johnson et al. (1955). Genetic Advance  GA = k × σ 2 p × σ 2 g σ 2 p . Genetic Advance as Percentage of Mean  GAM = GA X × 100 .

where GA = genetic advance; K = standardized selection differential at 5% selection intensity (k = 2.063); σ²p = phenotypic variance; σ²g = genotypic variance; GAM = genetic advance as percentage of the mean; × = grand mean of a character.

The alkali spreading value (ASV) (on a scale of 1–7) was measured according to the standard method (Little et al., 1958). A low ASV corresponds to a high gelatinization temperature (GT), and conversely, a high ASV indicates a low GT (Little et al., 1958). Sensory-based aroma (on a scale of 0–3) was evaluated using the standard procedure (Sood and Siddiq, 1978). The gel consistency (GC) was measured as per the standard protocol (Little et al., 1958).

The phenol reaction of the seed coat was tested according to the protocol (Kumar et al., 2021). Freshly harvested grains were collected. Fifteen healthy grains of each cultivar and breeding lines were soaked in 1.5% aqueous phenol solution for 24 h. After that, the solutions were drained, and the grains were air-dried. The hull color was then recorded unstained and stained as compared to the control treatment, in which the grains were treated with distilled water.

Anthocyanin pigments were qualitatively identified from the grains of black rice lines (BW23 and CW16) according to the standard methods (Bhuvaneswari et al., 2020). In brief, the dried, pigmented black rice grain samples (1g) were ground with 5 mL of 70% aqueous methanol at room temperature. After centrifugation at 10,000 × g for 10 min, the extracts were filtered (0.22 μm) before HR-LCMS-QTOF analysis at the SAIF, IIT Bombay, India. The instrument used was HRLCMS QTOF (Agilent Technologies, United States), the data acquisition software was Agilent MassHunter, and the data processing software was Agilent MassHunter Qualitative Analysis B.06; the column was an ZORBAX Eclipse Plus-C18 150 × 2.1 MM, 5 microns (Agilent). The following solvents were used: solvent A: 0.1% formic acid in Milli-Q water and solvent B: acetonitrile. The instrument scanned over the mass (m)/charge (z) range of 100–1,100 in both the positive and negative ion modes.

Total amino acid (TAA) identification was performed using standard protocols (Liyanaarachchi et al., 2020; Tyagi et al., 2022). Briefly, 100 mg of black rice flour (BW23 and CW16) were hydrolyzed in 10 mL of 6N HCI at 110 °C for 24 h. Approximately 20 µL of the solution was taken from the hydrolyzed samples and evaporated by speed vac. Then, it was reconstituted by adding 50 µL of 0.1 N HCl. From this extract, 1 µL of the sample was loaded into the LCMS system for amino acid profiling, along with standard amino acids. Amino acid identification and quantitative analysis were performed with an HR-LCMS-QTOF mass spectrometer (Agilent Technologies, United States; SAIF, IIT Bombay, India) with the following parameters: dual ion source AJS ESI, HiP sampler, binary pump, and diode-array detection (DAD) with gradient elution in a Q-TOF column comp (Poroshell HPH-C18, 2.7 µ, 4.6 × 100 mm).

We have developed two RIL populations (100 genotypes in each population) through interspecific hybridization, namely, BWF (O. sativa cv. Badshabhog × O. rufipogon) and CWF (O. sativa Cv. Chenga × O. rufipogon), to enhance the genetic base of the rice cultivars. In our cross, we have used only AA genome-containing genotypes within the primary gene pool of Oryza species (O. sativa, 2n = 24, AA genome and O. rufipogon, 2n = 24, AA genome), which is why the progeny lines (F1) showed recombination stability; however, hybrid sterility/hybrid breakdown was sometimes observed in the F1 generation. Due to this hybrid sterility/hybrid breakdown criterion, many F1 plants could not grow properly and died. In this study, the analysis of variance (ANOVA) revealed significant (p < 0.001) differences among the 100 RILs for all 15 measured yield-related agro-morphological traits (Table 1). These findings indicate the presence of ample genetic variations among the genotypes in the RIL populations. The 15 yield-related agro-morphological traits were evaluated for two consecutive kharif seasons (2022 and 2023), and some of the RIL genotypes exhibited superior performance in terms of PY and with pigmented grain quality (Tables 2, 3). An unexpected range of phenotypic variation was recorded among the RILs of BWF and CWF (Figure 1; Tables 2, 3). The shortest PH of only 60 cm was recorded in the BW98 line with a small FFL of 15.17 cm and a width of only 5.07 mm with small shattered seeds (Table 3). In contrast, the tallest PH (204 cm) was observed in line BW97 with shattered seeds. The grain pericarp color in both the RILs (BWF and CWF) varied from white, brown, red, and greenish to black, with distinctive grain quality parameters viz., ASV, GT, GC, and aroma (Figure 1; Supplementary Tables S1, S2).

The genetic parameters pertaining to the degree of variability among the RIL genotypes (BWF and CWF) were estimated (Table 4) using the GCV, PCV, H%, GA, and GAM. For the majority of the characters, the magnitude of PCV was significantly greater than that of GCV, indicating the role of genetic factors for trait development. Differences between GCV and PCV were less in both the RIL populations (BWF and CWF), indicating a higher correlation between phenotype and genotype, less environmental effect, and a larger role of genetic factors in these traits’ expression (Table 4). H% was high (>80%) in all the characters studied, except FLL (74.42%), which indicates little environmental influence. The H% for traits ranged from 74.42% (FLL) to 98.01% (GrWt) in BWF lines and from 76.58% (GB) to 98.71% (GrPn) in CWF lines. H% was found to be more than 90% high in eleven traits out of fifteen, such as PH, PnWt, GrPn, GL, GB, KL, KB, GrWt, HD, MT, and PY, in BWF population. Whereas in CWF population, H% more than 90% was recorded for the following traits: FLW, PnWt, GrPn, GrWt, GL, KL, and KB (Table 4). High heritability (>80%) in combination with high GA (>20) was observed for the following traits: PH, PnL, GrPn, GrWt, and PY in the BWF population and PH, PnL, GrPn, and KB in the CWF population, suggesting additive gene action for the characteristics. It was observed that PnWt, GrPn, GrWt, and Till had high GAM (>40%) in the BWF RIL and PnWt, GrPn, and PY in the CWF RIL population.

The correlation coefficients among 15 traits for both the RIL genotypes (100 lines) are shown in Figures 2A, B. GrPn, PnWt, GrWt, and GL were positively correlated with YP in BWF, and PnL, PnWt, GrPn, and GB were positively correlated with YP in CWF (Figures 2A, B). In contrast, HD and MT were negatively correlated with YP in BWF, and KL, HD, and MT were negatively correlated with YP in CWF. The highest value of positive correlation was observed (0.88) between the traits KL and GL, followed by GL and GrWt (0.79), GL, and PnWt (0.61). HD was negatively correlated with three traits: GrWt (−0.22), GB (−0.0.23), and KB (−0.03) in BWF. The highest positive correlation (0.76) was found between HD and MT and between GL and KL (0.61) in CWF. PY was positively correlated with GrPn and PnWt in both the RILs (BWF and CWF) while being negatively correlated with HD and MT. The correlation analysis, therefore, indicates that GrPn, PnL, PnWt, GrWt, and GL are the most important traits that need to be considered in the production of high-yield breeding lines.

The UPGMA dendrogram grouped 100 BWF RILs into 5 clusters and 100 CWF RILs into 4 clusters (Figures 2C, D). The cluster means for 15 traits among 100 BWF RILs and 100 CWF RILs are presented in Table 5. The BWF dendrogram cluster I consists of one genotype of only wild rice (blue) and is grouped separately, whereas cluster II contains a single genotype of the shortest PH of only 60 cm (yellow); clusters III, IV, and V were polygenotypic, comprising 72 genotypes, eight genotypes, and 24 genotypes, respectively (gray, purple, and greenish colors). Similarly, 100 genotypes of CWF RIL population were grouped into four clusters: I, II, III, and IV, with 1, 8, 52, and 42 genotypes, respectively (different colors in the dendrogram).

Mahalanobis D² statistic is widely used to analyze the relative contribution of various yield components to total divergence, and it also classifies different genotypes into suitable clusters based on their genetic distances (D² values) following Tocher’s method. Mahalanobis D² statistic estimates the relative contribution of several components at the intra- and intercluster levels, and genotypes derived from widely divergent clusters are likely to form heterotic combinations. The 100 genotypes of BWF RIL populations were grouped into seven clusters following Tocher’s method based on Mahalanobis D² distance values (Tables 6 and 7). Among them, five clusters are polygenotypic (i.e., I, IV, V, VI, and VII), whereas clusters II and III were monogenotypic and predicted uniqueness in the genes (Supplementary Table S3). The average intra- and intercluster D² distance values are represented in Table 6. Intra-cluster D² values ranged from 0.00 (clusters II and III) to 2,233.83 (cluster VI), followed by clusters IV (1,688.55), V (1,567.85), VII (1,299.42), and I (952.77). The highest intra-cluster distance (2,233.83) in cluster VI indicates wide genetic variation among the genotypes belonging to these clusters, and cluster I recorded the lowest intra-cluster distance (952.77), suggesting a closer relationship and low degree of diversity among the genotypes of this cluster. Clusters II and III consisted of only one genotype each; hence, they lacked intra-cluster distance (0.00). The largest intercluster distances were found between clusters IV and VII (25,865.50) in the BWF lines, indicating that genotypes in cluster IV were far diverse from those of VII. The least distance was observed between clusters I and II (4,047.36), which indicated that genotypes included in these cluster were closely related (Table 6). The cluster-wise mean values for the 15 characters in BWF are presented in Table 7. These are helpful to assess the superiority of the clusters during the improvement of characters through a hybridization program. The cluster mean values showed a wide range of variation for the majority of the characters undertaken in the present study. It was observed that cluster II had recorded the highest mean values for most of the traits, followed by clusters IV, VI, and VII (Table 7). The contribution of different traits to total divergence is depicted in Table 7. The trait GrPn (19.00) showed the maximum contribution toward genetic divergence, followed by GrWt (13.90), KB (13.60), GB (10.50), and PH (8.50). Out of the 15 agro-morphological traits studied, only five traits (GrPn, GrWt, KB, GB, and PH) provided the maximum contribution (65.50%) toward total divergence (Table 7). On the other hand, based on the D² matrix, 100 CWF RILs were grouped into 11 Tocher’s clusters, of which seven were multi-genotypic and four were mono-genotypic (Supplementary Table S4). Each of the 15 traits that contributed to the overall genetic divergence in the CWF was categorized and displayed in Table 8. The contribution toward the total variation was the maximum for GrWt (24.32), followed by the other traits (Table 8). Cluster XI had the highest PY (34.95 g), with the maximum contributions from GrWt, GL, KB, PH, GB, PnL, GrPn, and PnWt. Moreover, PY benefited most from clusters VIII (32.01 g) and X (33.65 g) (Table 9). The average intra- and intercluster distances within the 11 clusters indicate the degree of divergence within and between the groups (Tables 8 and 9). The largest intercluster distances (25,817.49) were found between cluster II (CW27, CW71, CW95, CW23, CW16, CW84, CW85, and CW81) and cluster IV (CW44, CW48, CW26, and CW31), containing genotypes that were found to be the most divergent with the maximum intercluster distance. According to the D² cluster matrix, cluster VII had the largest intra-cluster distance (2942.63) with RIL genotypes CW20 and CW34. The maximum heterosis would result from a cross between genotypes from clusters II and IV, which had the greatest genetic distance (25,817.49).

In the present study, two distinct RIL populations comprising 100 genotypes (BWF and CWF) were developed and characterized using 15 yield-related agro-morphological traits that indicated the presence of gigantic variability for these traits. ANOVA was employed to assess the significance of phenotypic variation in 15 yield-related agronomic traits across the BWF and CWF populations. The results indicated significant genotypic differences (p < 0.001) for most traits, confirming the presence of substantial genetic variation introgressed from O. rufipogon (significant at p < 0.001) (Table 1). The results were consistent with the general notion that the larger the divergence between the parental genotypes is, the higher the heterosis in crosses will be (Falconer, 1964) (Tables 2 and 3). The PY was recorded with the mean values 14.95 g, 22.65 g, and 10.66 g in the control black Chakhao, parent 1 Badshabhog, and parent 2 wild rice (O. rufipogon), respectively, whereas BWF RIL genotypes displayed PY that ranged from 5.00 g to 61.89 g with the mean value of 28.62 g (Table 2). In the case of CWF RIL genotypes, PY was recorded with mean values 14.95 g, 24.79 g, and 10.66 g in the control black Chakhao, parent 1 Chenga, and parent 2 wild rice (O. rufipogon), respectively, and the CWF RIL genotypes displayed PY that ranged from 10.66 g to 43.87 g with a mean value of 22.80 g (Table 3). At least 15 breeding lines out of the 100 BWF RIL populations were considered as promising lines due to better performance related to PY during two kharif seasons (2022 and 2023) with early maturity times (125 days–135 days) than the control black variety Chakhao, Manipur (153 days and plant height of 156.55 cm). The following breeding lines showed high PY: BW6 (33.48 g), BW18 (40.37 g), BW23 (61.89 g), BW24 (56.59 g), BW25 (51.93 g), BW26 (34.19 g), BW33 (32.48 g), BW44 (31.70 g), BW50 (42.74 g), BW77 (37.67 g), BW83 (50.22 g), BW88 (35.94 g), BW90 (43.93 g), BW91 (26.30 g), and BW99 (60.89 g) from the BWF population and the genotypes CW1 (37.99 g), CW11 (37.11 g), CW16 (43.87 g), CW20 (25.70 g), CW23 (26.32 g), CW39 (32.59 g), CW69 (27.86 g), CW78 (32.44 g), CW79 (37.76 g), CW80 (29.11 g), CW90 (33.65 g), CW94 (29.62 g), CW95 (22.78 g), CW97 (25.50 g), and CW98 (23.44 g) from the CWF population (Tables 2 and 3). Our finding was consistent with the earlier report that yield was enhanced when local rice Chakhao Poireiton (purple) was crossed with HYV Sahbhagi Dhan (white) and Shasharang (light brown) (Lap et al., 2024).

A wide range of phenotypic diversity was observed in both the RIL populations and showed transgressive segregation with respect to GL, grain number per panicle, GrWt, PnL, and PnWt (Figure 1; Tables 2 and 3; Supplementary Tables S1 and S2). The results signify that both the RIL populations (BWF and CWF) have shown substantial genetic variation among the genotypes (Tables 2 and 3). The nature and magnitude of genetic divergence prevailing in both the RIL populations (BWF and CWF) were estimated by various multivariate statistical tools, such as genetic variability parameters (broad-sense heritability), Mahalanobis D² statistic, PCA, and cluster analysis using standard formulae. The present results were consistent with the earlier findings that genetic divergence prevailed in our pre-breeding RIL populations (Faysal et al., 2022; At-Ul-Karim et al., 2022; Mondal et al., 2024; Mudhale et al., 2024; Bordoloi et al., 2024). Success in crop improvement generally depends on the magnitude of genetic variability and/or diversity and the extent to which the desirable characteristics are heritable (Mondal et al., 2024). Heritability is crucial in determining a trait’s response to selection and predicting the transmission of desirable characteristics from parents to offspring during breeding (Acquaah, 2009). The traits with high heritability such as PH, active tillering, filled grain per plant, GL, FLW, PnL, and GrWt have been widely reported to effect rice yield (At-Ul-Karim et al., 2022). The yield of rice is controlled by three key components: the number of effective panicles, the number of grains per panicle, and grain weight (Zheng et al., 2024). Agronomic traits such as GrWt and PH have been widely used for the improvement of rice yield in breeding programs (Hernandez-Soto et al., 2021). In this study, moderate-to-high heritability (74.42%–98.71%) was observed in both the RILs F_8:9 (BWF and CWF), indicating moderate-to-high level of genetic control of the traits associated with yield parameters (At-Ul-Karim et al., 2022; Hernandez-Soto et al., 2021; Zheng et al., 2024), and it subsequently showed increased yield in the RILs during field trials in two kharif seasons (2022 and 2023) (Tables 2–4). RIL genotypes (BWF and CWF) showed high yield potential due to high heritability of all the yield-enhancing characters evaluated in the present study (PH, active tillering, filled grain per plant, GL, FLW, PnL, and GrWt) (Tables 2 and 3). High heritability (>80%) with high GA (>20) was observed for the traits PH, PnL, GrPn, GrWt, and PY in BWF lines and PH, PnL, GrPn, and KB in CWF lines, suggesting additive gene action for the characteristics and that these traits could contribute largely to the yield improvement of pre-breeding lines (Tables 2–4). High heritability (H %) for yield-related traits in the BWF population (H % = 80 for PH, GrWt, PY, GrPn, and PnL) indicated strong genetic control, which is likely due to novel alleles from O. rufipogon. Similarly, the CWF population exhibited high heritability for the traits PH, PnL, GrPn, and KB (H % = 80), suggesting that introgression enhanced the genetic contribution to phenotypic variation, facilitating selection for improved traits. High heritability (>90%) in combination with the high score of genetic advance as percent of mean (>40%) was also observed for the traits PnWt, GrPn, and PY in the CWF RIL population, indicating strong additive genetic effects for increasing yield (Table 4). High heritability indicates strong genetic control, and genetic advance highlights the potential for trait improvement, which are all driven by O. rufipogon alleles. Many genes/QTLs of yield-enhancing traits (spikelet number, grain number per panicle, PnL, GrWt, grain size, grain yield, and GL) were identified and introgressed into the elite rice germplasm from the progenitor wild rice for enhancing the yield characters (McCouch et al., 2007; Gaikwad et al., 2021; Roy and Shil, 2020a; Xu et al., 2021; Malik et al., 2022; Seck et al., 2023). Significant genetic gain can be achieved in improving varieties by utilizing novel genes of the neglected wild rice to restore the genetic diversity and allelic variation lost during domestication (Siddiq and Vemireddy, 2021; Eizenga et al., 2024). Superior genes/QTLs of agronomic importance from wild rice (O. rufipogon) can be directly incorporated into breeding programs to generate pre-breeding material, which will serve as a valuable germplasm resource for rice breeding (Henry, 2022; Zhang J. et al., 2022; Bedford et al., 2023; Zheng et al., 2024). Similarly, these types of yield-enhancing traits (allelic variants/genes/QTLs) must have been introgressed into our pre-breeding lines from untapped wild rice (O. rufipogon); otherwise, both the RILs (BWF and CWF) would not show such high heritability with high genetic advance for yield characteristics, and they would have also exhibited high heterotic phenotypic features (Tables 2–4). Increased phenotypic variance observed in both the RIL populations compared with the parental lines could be attributed to the increased level of transgression of these yield-related components and gene interactions (Tables 2 and 3). Therefore, our pre-breeding materials (RILs) are valuable resources for rice improvement programs that may provide a powerful tool for broadening the genetic base of breeding materials to improve rice productivity, including climate change-resilient phenotypes along with high yield and grain quality. The improvement of rice grain quality has become an important breeding target in almost all rice breeding programs since the early 1980s (Mackill and Khush, 2018). Some of the grain qualities (ASV, GT, GC, and aroma) were evaluated in the present study and are depicted in Supplementary Tables S1 and S2.

The RIL population of BWF was grouped into 5 clusters according to the mean cluster values based on 15 traits, and the CWF RIL population was grouped into four clusters, indicating the relationship among the 100 genotypes (Table 5; Figures 2H, G). Our present findings are consistent with the earlier studies of Ahmad et al. (2015) and Bordoloi et al. (2024).

The Mahalanobis D² test is a multivariate statistical tool used to measure the genetic divergence or distance between genotypes based on multiple traits. It is particularly valuable for assessing diversity based on Mahalanobis D² values. Genotypes with large D² values are genetically divergent. Mahalanobis D² distance values can be used to group genotypes into several clusters using Tocher’s method. The BWF RIL population was grouped into 7 clusters and the CWF RIL population was grouped into eleven distinct clusters based on Tocher’s method using Mahalanobis D² distance values, reflecting the successful introgression of wild alleles (Tables 6 and 7). The average intercluster distances were observed to be greater than the average intra-cluster distances, suggesting that the genotypes of both the RIL populations (BWF and CWF) possess a greater degree of genetic diversity. The largest intercluster distances were found between clusters IV and VII (25,865.50) in BWF lines, indicating that genotypes in cluster IV were far diverse from those of cluster VII. The largest intercluster distances (25,817.49) were found between clusters II and IV in the CWF RIL population. The maximum intercluster distance indicated wide diversity, whereas the minimum suggested a close relationship between the groups (At-Ul-Karim et al., 2022). Genotypes with the largest genetic distance in yield-attributing parameters would result in the complementation of gene effects in the hybridization program, and these were detected in the present RILs (Tables 6 and 7). Out of the 15 agro-morphological traits studied, only 5 traits (GrPn, GrWt, KB, GB, and PH) provided the maximum contribution (65.50%) toward the total divergence in the BWF RIL population (Table 7). The contribution toward the total variation was the maximum for GrPn (19.00), followed by GrWt (13.90), KB (13.60), GB (10.50), GL (8.80), PH (8.50), PnWt (5.30), KL (5.00), and PY (4.30) in the CWF RIL population (Table 8). Overall, we have observed that the main characteristics that helped express our study’s diversity were the traits PH, GrPn, GrWt, GB, KL, PY, and KB (Tables 8 and 9). These traits should be taken into consideration while selecting parents for hybridization. Mahalanobis distance-based clustering pattern of both the RIL genotypes (BWF and CWF) into several groups confirmed the quantum of diversity present in the developed pre-breeding lines and provides a scope for its exploitation through breeding for yield improvement in rice (Tables 6–9). We know that the transfer of genes governing desirable traits from wild relatives to cultivated rice is an important strategy in rice breeding (Sanchez et al., 2013; Qiao et al., 2016). The narrow genetic base of modern rice varieties has led to yield plateaus, making it essential to introduce genetic diversity to overcome these barriers. Moreover, the genetic bottleneck that occurred during the domestication of cultivated rice from its immediate ancestral progenitor of wild rice O. rufipogon has played a major role in reducing allelic diversity by at least 50%–60% in cultivated rice than in wild rice O. rufipogon, leading to the loss of genetic variability with yield potentiality (Xiao et al., 1996; Tanksley and McCouch, 1997; Brar and Khush, 2006). Wild rice O. rufipogon has been considered as a reservoir of many untapped gene/QTLs for important agronomic traits, such as yield, quality, nutritional characteristics, and resistance to biotic and abiotic stresses, and it can be utilized in the pre-breeding program for broadening the genetic base of the released varieties to break the yield plateaus (Tanksley and McCouch, 1997; McCouch et al., 2007; Tester and Langridge, 2010; Sanchez et al., 2013; Brar and Khush, 2018). The transfer of genes controlling desirable traits (yield and grain quality) from the wild relatives O. rufipogon to cultivated rice is an important strategy in rice breeding (Tian et al., 2006; Huang et al., 2012B; Qiao et al., 2016; Gaikwad et al., 2021; Zhang B. et al., 2022). Pre-breeding facilitates the introgression of desirable traits such as yield potential, nutritional quality, and resistance to biotic and abiotic stresses (Rao et al., 2020; Bin Rahman and Zhang, 2022; Alam et al., 2024). In this study, we corroborate the results of earlier findings that wild rice serves as a valuable genetic resource for enhancing rice varieties via widening the genetic base through introgression of hidden alien gene/QTLs for high yield potential (Tanksley and McCouch, 1997; Henry, 2022; Bedford et al., 2023; Zheng et al., 2024). A complex trait such as grain yield is controlled by many genes along with being influenced by the environment and is related to other traits such as plant types, growth duration, and other yield-component traits (Mudhale et al., 2024; Bordoloi et al., 2024). The present reports are consistent with the earlier analyses that higher intercluster distances existed in the breeding lines, indicating wide trait variability and genetic divergence (Faysal et al., 2022; At-Ul-Karim et al., 2022; Mondal et al., 2024). Both the RIL populations (BWF and CWF) have shown wide genetic divergence in respect to the 15 characters studied, signifying that the genetic base has broadened as a result of interspecific hybridization through the introgression of untapped genetic components from the underutilized wild rice (O. rufipogon) germplasm of India (Tables 2 and 3). The yield-enhancing associated traits such as spikelet number, grain number, grain size, grain weight, and PnL have been introgressed into the populations developed from crosses of O. sativa × O. rufipogon, making O. rufipogon an ideal germplasm for mining yield-enhancing loci (McCouch et al., 2007; Malik et al., 2022; Gaikwad et al., 2021). These traits also contribute toward increasing the creation of high genetic variation and diversity in our pre-breeding RILs (BWF and CWF) (Tables 6–9).

PCA was used to explore the genetic diversity and population structure in the BWF and CWF populations. It was utilized to categorize all the yield-attributing traits into distinct PCs, thereby revealing the individual traits’ contributions to genetic divergence. The first four components in our study were considered the primary PCs as they showed the greatest variability with eigenvalues greater than >1 in the BWF RIL population (Table 10) and six components in the CWF population (Table 11). It showed that the maximum variation was present in the first two PCs (BWF and CWF), and hence, selection of genotypes from these PCs will be useful for obtaining higher genetic variation with higher yields (Figure 2; Tables 10 and 11). Breeders can use selection to influence such significant traits in the divergence analysis of BWF and CWF RIL populations. In this study, only PH and PnWt exhibited positive values in the first four PCs associated with divergence in the BWF RIL population (Table 10). PC1 and PC2 in the PCA-bi-plot diagram showed the dispersion and nature of diversity for both variables and genotypes (Figure 2). The cumulative variance of 73.74% by the first four axes with an eigenvalue > 1.00 indicates that the identified traits significantly influenced the RIL’s phenotype and could effectively be used for selecting among them in the BWF genotypes (Table 10). The bi-plot analysis showed the relationships between the morphological traits among the tested RIL genotypes of BWF and CWF (Figure 2). The traits influencing PC1 were PnWt, GrWt, GL, GB, KL, and PY. These results also support the GCV estimates for PnWt, GrWt, GL, GB, KL, and PY; the first three traits, along with GrPn, KB, and PH, also corroborated Mahalanobis distance-based divergence in the present study. PCA and Mahalanobis D² analyses further validate the introduction of novel genetic diversity, as RILs exhibit distinct genetic profiles and increased divergence from the cultivated parents. These findings demonstrate that O. rufipogon introgression successfully broadens the genetic base, enhancing the diversity and the potential for developing resilient, high-yielding rice varieties (Tables 2 and 3). Based on the comparison of the 100 RILs of BWF based on PCA bi-plot analysis, the RILs BW18, BW23, BW24, BW25, BW44, BW52, BW77, BW83, BW88, BW90, and BW99 were superior for PnWt, GrWt, PY, GL, GB, and KL. Hence, these results of PCA will be of great benefit to the breeder for identifying parents and the selection of characters for future hybridization programs for varietal improvement. The genetic diversity of the breeding lines (BWF and CWF) was clarified, and component traits contributing to variability were broken down through the combination of PCA; this could provide the framework for a well-run hybridization program. The length of a trait’s vector in PCA represents its contribution to the overall divergence; the longer the vector, the larger the contribution (Figure 2). All the traits exhibited the maximum length and contributed maximally to the total diversity. These results were in conformity with the findings of Bhuvaneswari et al. (2020) and Mondal et al. (2024). The dispersion of RILs across PCA axes confirmed the introduction of novel genetic diversity, broadening the genetic base beyond that of O. sativa.