Multivariate analysis of phenotypic diversity among advanced breeding lines for culm and yield related traits in rice (Oryza sativa L.)

Introduction

Rice (Oryza sativa L.) is the primary staple food for more than half of the world's population and plays a pivotal role in global food and nutritional security. Sustaining rice productivity under changing climatic conditions and increasing labour scarcity remains one of the foremost challenges confronting modern agriculture. While substantial progress has been achieved through the development of high-yielding semi-dwarf cultivars, further genetic improvement is constrained by complex interactions among yield components, plant architecture, and environmental factors. (Peng et al., 2000; Hirano et al., 2017) [11,18] . Consequently, breeding strategies have increasingly shifted towards the development of superior genotypes possessing an optimal combination of high grain yield, lodging resistance and agronomically desirable traits rather than improving individual characters independently. Lodging is among the most detrimental constraints limiting rice productivity, particularly under intensive cultivation systems where high fertilizer inputs, improved irrigation, and favourable growing conditions promote vigorous vegetative growth. Lodging causes irreversible stem bending or breakage, leading to substantial reductions in grain yield, grain quality, harvest efficiency and economic returns. Yield losses due to lodging have been reported to range from 10% to over 80%, depending on the severity, growth stage, genotype, and prevailing environmental conditions (Setter et al., 1997; Berry et al., 2004). The increasing frequency of extreme weather events, including heavy rainfall and strong winds associated with climate change, has further intensified the occurrence of lodging in rice-growing regions, underscoring the necessity of developing cultivars with improved structural stability while maintaining high yield potential (Lang et al., 2012; Luo et al., 2022) [5,14,15,20] .

Grain yield and lodging resistance are quantitatively inherited traits governed by numerous component characters and is profoundly influenced by genotype × environment interactions. These traits are highly interrelated and often exhibit antagonistic or compensatory relationships, making simultaneous selection particularly challenging in breeding programmes (Akshay et al.,2024a) [1] . Understanding the collective contribution of lodging associated and yield related traits is essential for designing effective selection strategies aimed at developing climate-resilient, high-yielding rice cultivars with superior structural stability. Conventional univariate analyses evaluate each trait independently and therefore may overlook the underlying relationships among correlated variables, resulting in inefficient selection decisions when multiple complex traits are involved (Varanasi et al., 2023) [23] . Advanced breeding lines developed from multiple elite parental background provide an enriched genetic base by combining favorable alleles from diverse genetic backgrounds. Such breeding populations generally exhibit wide phenotypic variability, increased recombination and novel trait combinations, thereby offering greater opportunities for identifying superior lines with desirable combinations of lodging resistance and grain yield. However, the large number of correlated agronomic traits measured in advanced breeding populations necessitates robust statistical approaches capable of simultaneously evaluating multiple variables and identifying the principal sources of variation. Multivariate statistical techniques have become indispensable tools in modern plant breeding because they enable comprehensive evaluation of complex datasets involving numerous correlated traits. Among these approaches, principal component analysis (PCA) is one of the most widely adopted dimension-reduction techniques for summarizing the total variability into a few orthogonal principal components without substantial loss of information (Jolliffe, 1986) [12] . PCA facilitates the identification of key traits contributing to overall phenotypic variation, reveals relationships among traits, classifies breeding materials based on multivariate similarity, and assists breeders in selecting ideotypes possessing favorable combinations of economically important characteristics (Akshay et al., 2024b, Anderson, 1972) [2,3] . By reducing redundancy among correlated variables, PCA provides valuable biological insights into trait associations and enhances the efficiency of selection in advanced breeding populations. Therefore, the present study was undertaken to evaluate advanced rice breeding lines developed from multiple parental combinations using a randomized block design (RBD) under field conditions. Multivariate analysis through principal component analysis was employed to quantify phenotypic diversity, identify the major traits contributing to overall variability, elucidate relationships among lodging resistance and grain yield attributes, and identify promising breeding lines possessing desirable trait combinations for future varietal development and parental selection in rice improvement programmes.

Materials and methods

A total of 30 advanced breeding lines (ABLs) developed from diverse parental combinations, were evaluated during the Kharif 2025 season at the Rice Research Unit (RRU), Agricultural Research Institute (ARI), PJTAU, Rajendranagr. The experiment was conducted under irrigated field conditions using a Randomized Block Design (RBD) with two replications. Seeds of each breeding line were initially raised in a nursery and transplanted to the main field 27 days after sowing at a spacing of 20 × 15 cm. All recommended agronomic and crop management practices were followed uniformly throughout the crop growth period to ensure optimum plant establishment and minimize environmental variation. Observations were recorded on three randomly selected plants from each experimental plot for lodging related morphological and yield contributing traits. The traits evaluated includes days to 50% flowering (DFF), plant height (PH, cm), culm length (CL, cm), internode length (INL, cm), culm diameter (CD, mm), culm thickness (CT, mm), panicle length (PL, cm), panicle weight (PW, g), number of grains per panicle (GNP), test weight (TW, g), and grain yield per plant (GYP, g). Plant height (cm) was measured from the plant base to the panicle tip. panicle number, grain number, and panicle weight, Internode length, culm length (cm) were recorded as the panicle-bearing tillers, filled grains per panicle, and weight of a fully matured panicle, the length of the fourth basal internode, the distance from the plant base to the base of the panicle respectively. Pushing resistance was measured by applying a prostrate tester perpendicular to the plant at 20-25 cm above the ground and pushing it to a 45° angle. The fourth basal internode was used for the measurement of culm morphological characteristics. Culm diameter (CD) and culm thickness (CT) were determined following the procedures described by Ookawa et [4,16,17] using Vernier calipers. Culm diameter was calculated as: Culm diameter (CD) = (a 1 +b 1 )/2

Culm thickness was calculated as

Where, a₂ and b₂ denote the inner diameters of the minor and major axes, respectively and a₁ and b₁ represent the outer diameters of the minor and major axes, respectively. Analysis of phenotypic diversity among the advanced breeding lines was performed using Principal Component Analysis (PCA) to identify the major traits contributing to variation in lodging resistance and grain yield. Mean values of all measured traits were subjected to PCA after standardization to eliminate scale differences among variables. The analysis was carried out using R software (Ver. 4.3.2). Scree plots were generated to determine the proportion of variance explained by each principal component, while PCA biplots were used to visualize genotype dispersion and trait associations.

Results and discussion

Principal component analysis revealed the major sources of phenotypic variation

Principal component analysis (PCA) was employed to elucidate the multivariate relationships among lodging related and yield associated traits evaluated in thirty advanced breeding lines. The analysis generated twelve principal components, of which the first four possessed eigenvalues greater than unity, satisfying the Kaiser criterion, and collectively explained 68.76% of the total phenotypic variation (Table 1). The first principal component (PC1) exhibited the highest eigenvalue (2.75) and accounted for 22.94% of the total variation, followed by PC2 (2.47; 20.61%), PC3 (1.71; 14.22%) and PC4 (1.32; 11.00%). Together, the first two principal components explained 43.54% of the total variation, indicating that a substantial proportion of the phenotypic diversity among the evaluated breeding lines could be effectively represented in a twodimensional ordination space (Figure 1). The scree plot displayed a distinct elbow after PC4, followed by a gradual decline in eigenvalues, confirming that the first four principal components adequately captured the biologically meaningful variation within the breeding population. Similar patterns have been reported in rice and other cereal crops, ( [8,19] where the first few principal components efficiently summarize complex variation among correlated agronomic traits. The relatively high cumulative variance explained by the retained components reflects the broad phenotypic diversity and provides a strong basis for identifying key selection traits and superior breeding lines.

Trait loadings, variable contributions and quality of representation identified the principal determinants of genetic divergence

The PCA loading matrix, variable contribution analysis and squared cosine (cos²) values collectively identified the major traits responsible for phenotypic divergence among the advanced breeding lines (Table 2; Figures 2). [18] . The second principal component (PC2), accounting for 20.61% of the total variation, was predominantly governed by culm strength (loading = 0.566; contribution = 32.04%; cos² = 0.792), followed by panicle length (0.541; 29.26%; cos² = 0.724), internode length (0.357; 12.76%; cos² = 0.315), grain number per panicle (0.341;11.6%; cos² = 0.287) panicle weight (0.234; 5.49%; cos² =0.136). similar results are reported by (Reddy et al., 2019) [19] . The high loading, contribution and cos² values of culm strength clearly indicated that stem mechanical strength was the most influential trait defining this component and played a pivotal role in discriminating the breeding lines. Plant height (-0.060) and culm length (-0.112) and test weight (-0.174) exhibited negative loadings on PC2, suggesting that improvements in stem strength may be achieved without necessarily increasing plant stature. Similar findings have been reported by (Ookawa et al., 2010) [17] who demonstrated that strengthening culm mechanical properties significantly enhances lodging resistance without compromising grain yield.

The third principal component (PC3), explaining 14.22% of the total variation, was mainly characterized by culm thickness (loading = -0.670; contribution = 44.92%) and culm diameter (-0.544; 29.62%). Increased culm diameter and thickness improve bending resistance by enhancing the second moment of area and strengthening supporting tissues, thereby contributing directly to lodging resistance (Kashiwagi and Ishimaru, 2004; Ookawa et al., 2010). Similar results were recorded by Gupte et al. (2023) [9,13,17] . Although these traits exhibited comparatively lower cos² values on the first two principal components, their substantial contribution to PC3 confirms their biological importance and highlights the need to consider stem anatomical characteristics alongside plant architecture and mechanical strength during selection. Overall, the loading matrix, contribution percentages and cos² analyses consistently identified culm length, plant height, culm strength, culm thickness and culm diameter as the principal contributors to multivariate variation. These traits therefore constitute the most informative selection criteria for simultaneous improvement of lodging resistance and grain yield in advanced rice breeding populations. [5,6,19,21,22,23] .

Distribution of advanced breeding lines and breeding implications

The PCA score plot revealed a broad distribution of the advanced breeding lines across the principal component space, indicating substantial phenotypic diversity generated through recombination of multiple parental lines (Figure 4). Genotypes positioned farther from the origin exhibited greater divergence, whereas those located near the centre displayed relatively balanced expression of the evaluated traits. The PCA biplot further illustrated the relationships between breeding lines and trait vectors, enabling identification of superior genotypes and potential donor parents (Figure 5). Among the evaluated materials, three breeding lines viz., ABL-1, ABL-7 and ABL-8 were positioned in close proximity to the vectors representing culm strength, grain number per panicle, panicle weight and panicle length, indicating favourable combinations of lodging resistance and yield-contributing traits. Their association with these key variables suggested superior structural stability coupled with enhanced sink capacity, making them promising candidates for advanced yield evaluation and potential donor parents in future breeding programmes. Conversely four breeding lines viz., ABL-2, ABL-6 and ABL-11 were more closely associated with plant height and culm length, reflecting that their taller plant architecture may increase lodging susceptibility under intensive cultivation. Several breeding lines, including ABL-14, ABL-15, ABL-16, ABL-18 and ABL-19, were positioned away from the major trait vectors, indicating comparatively lower expression of both lodgingresistant and yield-associated characteristics, whereas lines located near the origin exhibited intermediate performance across the measured traits. Similar observation were recorded by susmi et al. (2025) [21] . The orientation of the trait vectors further revealed important biological relationships among the evaluated variables. The close alignment of grain number per panicle, panicle weight and panicle length indicated strong positive associations among these yield components, while their positive relationship with culm strength suggested that enhanced stem mechanical stability can be achieved without compromising grain productivity. In contrast, the opposite orientation of plant height and culm length relative to the major yield vectors indicates that excessive vegetative growth may adversely affect reproductive efficiency and increase lodging susceptibility. These findings reinforce the importance of selecting genotypes possessing moderate plant stature, stronger culms and superior panicle architecture for the development of high-yielding, lodging-resistant rice cultivars. Overall, the PCA effectively identified the principal traits governing genetic divergence and highlighted three breeding lines viz., ABL-1, ABL-7 and ABL-8 as promising breeding materials for future varietal development and strategic parental selection.

Conclusion

Principal component analysis effectively unraveled the complex interrelationships among lodging-related and yieldassociated traits in the evaluated advanced rice breeding lines, with the first four principal components explaining 68.76% of the total phenotypic variation. The analysis identified culm length, plant height, culm strength, culm thickness, culm diameter, grain number per panicle and panicle weight as the principal contributors to genetic divergence, underscoring their importance as key selection criteria for simultaneous improvement of lodging resistance and grain yield. The positive association of culm strength with major yield-contributing traits demonstrated that enhanced stem mechanical stability can be combined with improved productivity, providing a strong foundation for developing superior rice ideotypes. Among the evaluated materials, ABL-1, ABL-7 and ABL-8 emerged as promising breeding lines owing to their favourable association with both lodging resistance and yield attributes, making them valuable candidates for advanced multilocation evaluation and potential donor parents in future breeding programmes.

Overall, the study demonstrated that PCA is a powerful multivariate tool for trait prioritization, parental selection and ideotype breeding, thereby accelerating the development of high-yielding, lodging-resistant and climate-resilient rice cultivars through efficient multi-trait selection.

Acknowledgement

AcknowledgementThe author sincerely acknowledge the guidance provided by Dr. V. Hemalatha, Associate Dean, Agricultural college, Palem, PJTAU and Dr. B. Laxmi prasanna, scientist, RRU, ARI, PJTAU, Dr. SNCVL Pushpavalli, Assistant Professor, IBT, PJTAU for providing all of the necessary facilities for our investigation.

References

1. Akshay M, Krishna L, Badri J, Barbadikar KM, Rao DS. Genetic variability and trait association studies of International Journal of Advanced Biochemistry Research https://www.biochemjournal.com internode length and strong culm-related traits in inter sub-specific cross derived recombinant inbred lines in rice (Oryza sativa L.). Journal of Advances in Biology & Biotechnology. 2024;27(8):306-314.
2. Akshay M, Krishna L, Badri J, Barbadikar KM, Rao DS. Unraveling promising genetic variability associated with strong culm and yield parameters in inter sub-specific cross derived recombinant inbred lines in rice (Oryza sativa L.). International Journal of Plant & Soil Science. 2024;36(9):700-711.
3. Anderson TW. An Introduction to Multivariate Analysis. New Delhi: Wiley Eastern Pvt. Ltd.; 1972.
4. Badri J, Padmashree R, Anilkumar C, Mamidi A, Isetty SR, Swamy AVSR, et al. Genome-wide association studies for a comprehensive understanding of the genetic architecture of culm strength and yield traits in rice. Frontiers in Plant Science. 2024;14:1298083.
5. Berry PM, Sterling M, Spink JH, Baker CJ, SylvesterBradley R, Mooney SJ, et al. Understanding and reducing lodging in cereals. Advances in Agronomy. 2004;84:217-271.
6. Chakravorty A, Ghosh PD, Sahu PK. Multivariate analysis of phenotypic diversity of landraces of rice of West Bengal. American Journal of Experimental Agriculture. 2012;3(1):110-123.
7. Chigira K, Yamasaki M, Adachi S, Nagano AJ, Ookawa T. Identification of novel quantitative trait loci for culm thickness of rice derived from strong-culm landrace in Japan, Omachi. Rice. 2023;16(1):31.
8. Deepika K, Lavuri K, Rathod S, Yeshala CM, Jukanti AK, Reddy SN, et al. Multivariate analysis of geographically diverse rice germplasm for genetic improvement of yield, dormancy, and shattering-related traits. Plant Genetic Resources. 2021;19(2):144-152.
9. Gupte VP, Manonmani S, Nivedha R, Suresh R, Senthil Kumar G, Raveendran M. Genetic diversity studies and identification of donors for lodging resistance in rice (Oryza sativa L.). Electronic Journal of Plant Breeding. 2023;14(3):1158-1166.
10. Haranal AA, Eswari KB, Badri J, Rao PR. Yield trait association studies through principal component analysis in a bi-parental population derived from indica/tropical japonica inter-subspecific cross in rice (Oryza sativa L.). Journal of Experimental Agriculture International. 2024;46(8):651-657.
11. Hirano K, Ordonio RL, Matsuoka M. Engineering the lodging resistance mechanism of post-Green Revolution rice to meet future demands. Proceedings of the Japan Academy, Series B, Physical and Biological Sciences. 2017;93:220-233.
12. Jolliffe IT. Graphical representation of data using principal components. In: Principal Component Analysis. Springer Series in Statistics. New York: Springer; 1986. p. 63-95.
13. Kashiwagi T, Ishimaru K. Identification and functional analysis of a locus for improvement of lodging resistance in rice. Plant Physiology. 2004;134(2):676-683.
14. Lang Y, Yang X, Wang M, Zhu Q. Effects of lodging at different filling stages on rice yield and grain quality. Rice Science. 2012;19(4):315-319.
15. Luo X, Wu Z, Fu L, Dan Z, Yuan Z, Liang T, et al. Evaluation of lodging resistance in rice based on an optimized parameter from lodging index. Crop Science. 2022;62(3):1318-1332.
16. Mamidi A, Krishna L, Badri J, Barbadikar KM, Durbha SR, Sai Prasad SV, et al. Navigating rice culm resilience: High-throughput quantitative trait locus mapping in indica and tropical japonica derived population. Plant Science. 2025;362:112756.
17. Ookawa T, Hobo T, Yano M, Murata K, Ando T, Miura H, et al. New approach for rice improvement using a pleiotropic QTL gene for lodging resistance and yield. Nature Communications. 2010;1(1):1-11.
18. Peng S, Laza RC, Visperas RM, Sanico AL, Cassman KG, Khush GS. Grain yield of rice cultivars and lines developed in the Philippines since 1966. Crop Science. 2000;40:307-314.
19. Reddy VRP, Vemireddy LR, Srividhya A, Reddy HK, Siddiq EA. Genetic analysis of culm strength and its related traits in rice (Oryza sativa L.). Applied Biological Research. 2019;21(2):166-175.
20. Setter TL, Laureles EV, Mazaredo AM. Lodging reduces yield of rice by self-shading and reductions in canopy photosynthesis. Field Crops Research. 1997;49(2-3):95106.
21. Susmi TG, Devasena N, Ramchander S, Kalaiarasi P, Pramitha L, Petchiammal KI, et al. Morphological characterization of strong culm traits and their association with yield performance in rice (Oryza sativa L.). Electronic Journal of Plant Breeding. 2025;16(2):557-564.
22. Tiwari DN, Tripathi SR, Tripathi MP, Khatri N, Bastola BR. Genetic variability and correlation coefficients of major traits in early maturing rice under rainfed lowland environments of Nepal. Advances in Agriculture. 2019;2019:1-9.
23. Varanasi YVP, Isetty SR, Revadi P, Balakrishnan D, Hajira S, Prasad MS, et al. Molecular and morphological characterization of introgression lines with resistance to bacterial leaf blight and blast in rice. Plants. 2023;12(16):3012.
24. Yadav S, Singh UM, Naik SM, Venkateshwarlu C, Ramayya PJ, Raman KA, et al. Molecular mapping of QTLs associated with lodging resistance in dry directseeded rice (Oryza sativa L.). Frontiers in Plant Science. 2017;8:1431.