Multivariate analysis of 110 green super rice (GSR) lines for morpho-physiological diversity attributes

Abstract

Genetic diversity present in rice genotypes plays a crucial role in improving breeding strategies and developing high-quality cultivars. This study evaluated genetic diversity, agronomic performance, and disease resistance across 110 unique rice cultivars over two growing seasons (2021–2022) using an alpha-lattice design. Ten morpho-physiological and yield characteristics were assessed along with resistance to brown leaf spot (BLS) in natural field conditions. The combined analysis of variance indicated highly significant genotypic effects (P < 0.01) across all traits, with broad-sense heritability values varying from 8.83% for days to maturity to 45.56% for plant height. K-means clustering effectively categorized genotypes into five distinct groups. Cluster 4 demonstrated the highest yield at 10.64 t ha⁻¹, characterized by early flowering, whereas Cluster 5 attained significant productivity of 10.2 t ha⁻¹ due to its large grain size. The distribution of BLS resistance demonstrated considerable variation across clusters (χ² = 50.1, P < 0.001). Cluster 1 presented 60% highly resistant genotypes, while Cluster 2 displayed 60% resistant genotypes, with no entries classified as susceptible. Notable performers comprised NGSR.18 with a multi-trait score of 70.77%, GSR1 achieving a yield of 12.76 t ha⁻¹, and GSR34, GSR49, GSR50 exhibiting high resistance to BLS. The correlation analysis demonstrated genetic independence between yield and disease resistance, thereby supporting the feasibility of simultaneous improvement. The findings offer significant genetic resources for the development of high-yielding, disease-resistant rice cultivars.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-025-08049-6.

Introduction

Rice (Oryza sativa L.) is currently the most significant crop worldwide, and it is a staple food in many countries worldwide [1]. Rice is cultivated in more than 100 countries and about 168 million hectares is under rice cultivation worldwide with the production of 800 million tons FAOSTAT (FAO, 2023). Global consumption of rice is around 502 million tons during 2020-21, whereas it was 437.18 million tons during last 10 years [2]. Major consumption of rice is in Asia where per capita consumption of rice is 45 kg per year [3].

In Pakistan rice is the second main staple food major source foreign exchange earnings [4]. The share of rice in agricultural value addition is 2.5% whereas its contribution in the total GDP of Pakistan is 0.6%. During 2023–2024, area under rice cultivation was 3.6 million hectares, and the total rice production was 9.9 million tons [5]. Pakistan earns US$ 3.36 billion by exporting half of its rice and the rest half is consumed locally. World population is expected to be 10 billion during 2050 and more diverse and climate resilient crops are required to meet the food security challenges. Future predictions suggest that rice demand will increase by up to 40% more than the present world yield by 2025 [6].

Rice production faces multiple challenges under the changing climatic scenario worldwide. In Pakistan rice production faces multiple challenges i.e. drought stress, heat stress, flood & submergence condition in moon soon season in rice growing areas, biotic factors like bacterial leaf blight and brown leaf spot (BLS) at later stages of rice crop combined with a heat wave [7, 8]. BLS caused by fungus Bipolaris oryzae, is a severe threat to rice crop and leading major reduction in yield and quality, some report suggests that BLS caused huge yield losses of up to 90% when the disease severely impacted Bengal in mid-20th century [6]. Genetic variation in local germplasm is less due to a narrow genetic base for both biotic and abiotic factors. Availability of quality is a major issue. More than 70% of the area under Pakistan rice cultivation consists of coarse rice varieties. Among the coarse rice cultivars, hybrid rice is gaining more popularity but getting fresh hybrid seeds every year is cost intensive and beyond the reach of poor farming community. Open pollinated cultivars are best suited for developing countries and production of basic and certified seeds is within the reach of farming community. Among coarse groups of rice, green super rice (GSR) offers its superior climatic resilience and yield attributing traits. It is developed by collective efforts of scientists from China and IRRI, Philippines.

Before introducing into a breeding program for selection or hybridization as parent, germplasm evaluation for variability and morpho-physiological assessment is a pre-requisite. Genetic variation in the rice gene pool has prime importance for the initiation of a breeding program. Variation in the gene pool is of two types, phenotypic variation (PV) and genetic variation (GV), The genotypic variation represents the heritability that is present within population and between populations. Genetic divergence is needed to choose parental lines for hybridization, define heterotic groups, estimate genetic distance and diversity, and forecast gene action in following generations, all of which depend on germplasm genetic variability [9]. There are several ways to determine variability in large populations with multiple traits datasets but multivariate analysis by using Principal Component Analysis (PCA) [10–12] and cluster analysis is more common analysis in rice morpho-physiological based variability estimation [13, 14].

In the present study, a set of 110 GSR advanced lines including two check cultivars (KSK-434 and IR-06) were acquired from Plant Genetic Resource Institute of Bio Resource Conservation Institute (PGRI-BCI), National Agricultural Research Center (NARC) Islamabad Pakistan. The objective of this study was to evaluate genetic divergence based on morpho-physiological traits and BLS disease response. Best performing genotypes with stable high yield and disease resilience will be identified for future breeding programs as selection and parental lines for hybridization.

Materials and methods

Experimental site and genetic material

The field experiment was carried out during the 2021 and 2022 growing seasons at the Rice Program, Crop Sciences Institute, National Agricultural Research Center (NARC), located in Islamabad, Pakistan (33°42’N, 73°10’E). The research location is situated on the Potohar Plateau, positioned between the Margalla Hills, with an elevation varying from 460 to 610 m above sea level. The experimental material comprised 110 diverse coarse rice genotypes sourced from two institutions: the Bioresource Conservation Institute-Plant Genetic Resource Institute (BCI-PGRI), NARC, Islamabad, and the National Institute of Genomics and Advanced Biotechnology (NIGAB), NARC, Islamabad. The comprehensive list of rice germplasm utilized in this study is provided in Table S1.

The climatic conditions showed seasonal variations with temperatures ranging from 15.27 to 31.57 °C across both years. Rainfall patterns varied significantly, with July receiving the highest precipitation (5.18–5.64 mm) and November the lowest (0.07–2.15 mm). Monthly rainfall ranged from 1.44 to 6.79 mm during the growing period. These environmental conditions supported rice cultivation and natural BLS disease development throughout both experimental seasons (Table 1). 

Table 1.

Climatic data of experimental site during the two growing seasons of 2021 and 2022

Parameter	Rainfall		Temp.		Temp_MAX		Temp_MIN
	2021	2022	2021	2022	2021	2022	2021	2022
May	5.59	3.83	26.43	28.47	38.75	39.86	16.1	16.85
Jun	1.44	2.44	30.57	31.17	42.84	43.68	18.84	16.09
Jul	5.18	5.64	31.57	28.96	44.09	38.55	22.46	23.11
Aug	3.32	6.79	29	27.57	37.72	35.2	20.44	19.27
Sep	6.34	2.17	26.89	25.66	35.76	33.1	19.54	17.97
Oct	3.12	0.52	20.63	21.28	32.08	31.92	9.19	11.01
Nov	0.07	2.15	15.27	15.37	26.24	27.52	6.91	6.06

Source : https://power.larc.nasa.gov/data-access-viewer/

Experimental design and agronomic management

The 110 rice genotypes were organized in an alpha-lattice design involving two replications, consisting of 10 blocks per replication and 11 genotypes per block. This design enabled effective management of spatial variability while accommodating the extensive range of genotypes in the study.

The nursery establishment utilized plastic trays featuring drainage pores, with dimensions of 45.7 cm in length, 30.5 cm in width, and 5 cm in height. Before sowing, rice seeds underwent treatment with Hombre^® pesticide (Bayer) in accordance with the manufacturer’s guidelines. Each tray was filled with silt soil and inoculated with 100 g of treated seeds, with the surface amended using farmyard manure. Standard agronomic practices were employed for nursery management, incorporating recommended fertilizer application, while manual weed removal was conducted to ensure optimal seedling development for subsequent transplanting.

Seedling transplantation was conducted using the puddled method for land preparation. Rice seedlings, aged one month, were transplanted with two seedlings allocated per hill. The spacing between plants and rows was maintained at 20 cm. The plot dimensions were established at three m in length and two m in width, accommodating a total of 10 rows of plants, resulting in 150 plants per plot. Fertilizer was applied at rates of 120 kg ha^− 1 of nitrogen (N), 90 kg ha^− 1 of phosphorus (P₂O₅), and 60 kg ha^− 1 of potassium (K₂O), utilizing urea, diammonium phosphate, and muriate of potash. The full doses of phosphorus and potassium were applied during land preparation, while one-third of the nitrogen was applied at that time. The remaining nitrogen was applied in two splits at 35 and 55 days post-transplanting. Continuous flooding irrigation was administered for 30 days post-transplantation, followed by weekly irrigation intervals thereafter. Irrigation was administered 20–30 days prior to harvesting. Weeding was performed using conventional manual methods.

Agronomic trait measurements

Agronomic data was recorded on ten randomly selected plants from each genotype for the following traits: Plant height (PH, cm) was measured from ground level to the tip of the tallest panicle at physiological maturity. Days to 50% flowering (DTF) was recorded when 50% of the plants in each plot had emerged panicles with visible anthers. Tillers per plant (TPP) were counted at maturity stage by recording the total number of productive tillers bearing panicles. Panicle length (PL, cm) was measured from the base of the panicle to its tip on the main tiller. Grains per panicle (GPP) were determined by counting the total number of filled grains from five randomly selected panicles per plant.

Thousand grain weight (TGW) was measured by counting and weighing 1000 randomly selected filled grains and expressed in grams. Days to maturity (DM) was recorded when 85% of the grains in each plot reached physiological maturity. Paddy yield (PY) was measured as the total grain weight per plot and converted to kg ha^− 1. Normalized Difference Vegetative Index (NDVI) was measured using a handheld NDVI meter during the vegetative growth stage to assess plant vigor. Soil Plant Analysis Development (SPAD) values were recorded using a SPAD-502 chlorophyll meter to estimate leaf chlorophyll content during the active tillering stage.

Assessment of BLS

The assessment of BLS disease was performed under natural field conditions, employing the International Rice Research Institute (IRRI) Standard Evaluation System for Rice [15], without the use of artificial inoculation. Disease assessment utilized a 0–9 scoring scale based on the Standard Evaluation System (SES) for rice [15], determined by the percentage of leaf area affected by BLS lesions. The assessment occurred during the reproductive phase, at which point disease symptoms were most evident. The disease scoring scale and associated resistance categories are presented in Table 2.

Table 2.

BLS disease scoring scale and resistance categories for rice genotype evaluation

Scoring Scale	leaf area covered	Response
0	No infection.	No disease
1	Less than 1%	*HR
2	1–3%	R
3	4–5%	R
4	6–10%	MR
5	11–15%	MR
6	16–25%	MR
7	26–50%	S
8	51–75%	HS
9	76–100%	HS

*HR highly resistant, R resistant, MR moderately resistant, S susceptible, HS highly susceptible

For analysis purposes, genotypes were categorized into six resistance classes: HR (highly resistant), R (resistant), MR (moderately resistant), MS (moderately susceptible), S (susceptible), and HS (highly susceptible). Disease scoring was performed on three randomly selected plants per plot, and the average score was used for statistical analysis.

Data analysis

Experimental design and combined analysis of variance

Analysis of variance (ANOVA) was conducted using a combined alpha-lattice design across two years (2021 and 2022), which were treated as distinct environments for G×E analysis. Homogeneity of variances between years was confirmed using Levene’s test (P > 0.05) [16], supporting the validity of the combined analysis. The statistical model employed was:

where, Y₍_iⱼₖₗ₎ is the observed phenotypic value, µ is the overall mean, y₍ₗ₎ is the fixed effect of the lth environment (year), r₍_i₎₍y₍ₗ₎₎ is the random effect of the ith replication nested within the lth environment, b₍_iⱼ₎₍y₍ₗ₎₎ is the random effect of the jth incomplete block nested within the ith replication and lth environment, g₍ₖ₎ is the fixed effect of the kth genotype, (gy)₍ₖₗ₎ is the fixed effect of genotype × environment interaction, representing differential genotypic performance across environments, e₍_iⱼₖₗ₎ is the random residual error.

Variance components estimation

Genetic parameters were estimated from mean squares obtained from the combined ANOVA across environments [17]. From the combined analysis of variance, variance components were estimated from the mean squares as follows:

Where GMS is the genotype mean square, GEMS is the genotype × environment interaction mean square, EMS is the error mean square, r is the number of replications per environment (r = 2), e is the number of environments/years (e = 2), represents genotypic variance, , represents genotype × environment interaction variance, represents environmental (error) variance, represents total phenotypic variance.

The variance decomposition in the mixed model accounts for both fixed effects (environment, genotype, and G×E interaction) and random effects (replication within environment, incomplete block within replication and environment, and residual error).

Broad-sense heritability (H²) on an entry-mean basis across environments was estimated as:

Data Preparation and quality control

Prior to statistical analysis, all raw data were examined for normality using the Shapiro-Wilk. Data that deviated significantly from normality (P < 0.05) were transformed using appropriate transformations (log, square root, or Box-Cox transformation) to meet the assumptions of parametric tests. Outliers were identified using the interquartile range (IQR) method, where values exceeding Q3 + 1.5×IQR or below Q1–1.5×IQR were flagged and verified against field records.

Multivariate analysis

Descriptive statistics were calculated for all 11 agronomic traits across 110 genotypes, with data distribution visualized through density plots and violin plots combined with box plots. Phenotypic correlations among traits were assessed using Pearson correlation coefficients, with statistical significance tested at P < 0.05 and results displayed as color-coded correlation matrices. Prior to correlation analysis, multicollinearity among traits was assessed using variance inflation factor (VIF), with VIF > 10 indicating severe multicollinearity. Principal component analysis (PCA) was performed to identify patterns of variation among genotypes and reduce data dimensionality. Before conducting PCA, the following assumptions were verified: (1) adequacy of sample size (n = 110 genotypes with 11 variables met the minimum ratio of 10:1), (2) suitability of data for factor analysis assessed using Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (KMO > 0.6 considered acceptable) and Bartlett’s test of sphericity (P < 0.05), and (3) linearity among variables examined through scatterplot matrices.

All trait data were standardized (z-score transformation) prior to PCA to ensure variables with different units and scales contributed equally to the analysis. PCA was based on eigenvalue decomposition: (S - λI)v = 0, where S represents the correlation matrix (due to standardization), λ is the eigenvalue indicating variance explained by each component, and v is the corresponding eigenvector. Principal components with eigenvalues > 1 (Kaiser criterion) were retained, collectively explaining a substantial proportion of the total variance.

Multivariate analysis included hierarchical clustering using Euclidean distance on standardized trait values and Ward’s linkage method to minimize within-cluster variance, and K-means clustering (k = 5) to identify distinct genetic groups. The optimal number of clusters was determined through elbow method (plotting within-cluster sum of squares against number of clusters) and silhouette analysis (average silhouette width > 0.5 indicating reasonable cluster structure). BLS resistance was evaluated using a 1–9 scale and categorized into six classes (HR, R, MR, MS, S, HS). The relationship between genetic clusters and disease resistance was examined through cross-tabulation and chi-square tests. All statistical analyses were performed with α = 0.05 significance level.

Software and packages

All statistical analyses were performed using R version 4.3.0 (R Core Team, 2023) in RStudio version 2023.06.1. Combined analysis of variance for the alpha-lattice design was conducted using “lme4” v1.1–34 for linear mixed models with REML estimation. Data manipulation used “dplyr” v1.1.2, while descriptive statistics and normality tests employed “psych” v2.3.6 and base R functions. Multicollinearity was assessed using Variance Inflation Factor from “car” v3.1–2. Principal component analysis was performed with “FactoMineR” v2.8, and Kaiser-Meyer-Olkin and Bartlett’s tests used “psych” v2.3.6. Hierarchical clustering (Euclidean distance, Ward’s linkage) and K-means clustering utilized “cluster” v2.1.4, with optimal cluster determination via “factoextra” v1.0.7. Visualizations employed “ggplot2” v3.4.2, “corrplot” v0.92, and “pheatmap” v1.0.12. Statistical tests used “agricolae” v1.3–6 and base R, with α = 0.05 significance level.

Results

Significant genetic variation and genotype × environment interactions detected across rice germplasm

The combined analysis of variance across two growing seasons revealed significant variation for most morpho-physiological and yield-related traits among the 110 rice genotypes evaluated (Table 3). Year effects were highly significant (P < 0.01) for 8 out of 10 traits examined, including PH, DTF, GPP, TGW, DM, PY, NDVI, and SPAD chlorophyll readings, with TPP showing moderate significance (P < 0.05). Only PL demonstrated non-significant year effects, suggesting relatively stable expression of this trait across environmental conditions. Genotypic effects were highly significant (P < 0.01) for all ten traits evaluated. The magnitude of genotypic variance components was considerable across all traits, with mean squares ranging from 10.64 for PL to 2593.90 for GPP, providing the genetic basis for the substantial phenotypic diversity observed in descriptive statistics (Descriptive Statistics and Distribution Analysis) and cluster analysis (Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection). The genotype × year interactions were significant for 7 out of 10 traits, with DTF, GPP, TGW, DM, and PY showing highly significant interactions (P < 0.01), while PH and SPAD readings demonstrated moderate significance (P < 0.05). Non-significant genotype × year interactions were observed for TPP, PL, and NDVI.

Table 3.

Combined alpha-lattice ANOVA for 110 rice genotypes across two years (2021–2022) for 10 studied traits

SOV	DF	PH	DTF	TPP	PL	GPP	TGW	DM	PY	NDVI	SPAD
Years	1	342.50**	89.45**	25.60*	8.20ⁿˢ	4521.30**	28.90**	67.80**	5.45**	0.0035**	31.20*
Reps(Years)	2	75.10*	104.50**	37.40*	4.60ⁿˢ	5982.20**	32.80**	289.00**	2.85**	0.0012ⁿˢ	39.70*
Blocks(Reps×Years)	36	52.30ⁿˢ	42.80**	21.20**	5.90ⁿˢ	2234.60*	8.40ⁿˢ	78.50*	1.65**	0.0045ⁿˢ	18.90ⁿˢ
Genotypes	109	247.28**	50.46**	16.58**	10.64**	2593.90**	17.89**	31.10**	3.61**	0.0022**	22.97**
Genotypes × Years	109	87.50*	23.20**	11.20ⁿˢ	7.80ⁿˢ	1456.70**	14.30**	19.80*	2.15**	0.0018ⁿˢ	16.40*
Residual	182	58.9	1.48	9.12	5.15	441.2	8.75	11.05	0.37	0.00082	12.95
Total	439	487.35	98.23	156.84	73.92	18,456.72	142.67	398.28	29.85	0.0142	287.53

SOV Source of Variation, DF Degrees of Freedom, PH denotes Plant Height (cm), DTF Days to Flowering, TPP Tillers Per Plant, PL Panicle Length (cm), GPP Grains Per Panicle, TGW Thousand Grain Weight (g), DM Days to Maturity, PY Paddy Yield (t ha^− 1), NDVI Normalized Difference Vegetation Index, SPAD Chlorophyll Content (unitless). Ns not significant (P > 0.05), *: significant at P < 0.05, **: significant at P < 0.01, and ***: highly significant at P < 0.001

Genetic components of variance and heritability estimates

The estimation of genetic parameters revealed substantial variation in the magnitude of variance components and heritability estimates across the ten morpho-physiological and yield-related traits evaluated (Table 4). Genotypic variance components ranged from 0.0014 for NDVI to 2127.00 for GPP, reflecting the diverse genetic architecture underlying different trait categories. Environmental variance components were consistently larger than their corresponding genotypic components for all traits, with ratios of environmental to genotypic variance ranging from 1.19 for PH to 22.84 for DM. Phenotypic variance, representing the total observable variation, exhibited the widest range among variance components, with values spanning from 0.0052 for NDVI to 7383.92 for GPP. The partitioning of phenotypic variance revealed that environmental factors contributed 54.44% to 91.17% of the total phenotypic variation across traits, while genetic factors accounted for 8.83% to 45.56% of the observed variation.

Table 4.

Estimation of genetic components of variance of morpho-physiological and yield contributing traits - combined alpha-lattice analysis (2021–2022)

SOV	PH	DTF	TPP	PL	GPP	TGW	DM	PY	NDVI	SPAD
(σ²g)	184.85	48.92	7.1	5.22	2127	8.74	19.52	3.22	0.0014	9.5
(σ²e)	220.77	117.26	35.89	7.94	5256.92	33.54	201.5	5.39	0.0038	38.84
(σ²p)	405.62	166.18	42.99	13.16	7383.92	42.28	221.02	8.61	0.0052	48.34
H2 (%)	45.56	29.44	16.52	39.67	28.81	20.67	8.83	37.4	26.92	19.65

σ²g genotypic variance, σ²e environmental variance, σ²p phenotypic variance, H² broad sense heritability (%), SOV Source of Variation, PH Plant Height (cm), DTF Days to Flowering, TPP Tillers Per Plant, PL Panicle Length (cm), GPP Grains Per Panicle, TGW Thousand Grain Weight (g), DM Days to Maturity, PY Paddy Yield (t ha^− 1), NDVI Normalized Difference Vegetation Index, SPAD Chlorophyll Content (unitless)

For H² estimates, the results demonstrated considerable variation among traits, ranging from 8.83% for DM to 45.56% for PH. Based on the magnitude of heritability estimates, traits could be categorized into three distinct groups. High heritability traits (H² > 35%) included PH (45.56%), PL (39.67%), and PY (37.40%), indicating strong genetic control and favorable prospects for direct selection. Moderate heritability traits (H² = 20–35%) comprised DTF (29.44%), GPP (28.81%), and NDVI (26.92%), suggesting moderate genetic influence with reasonable selection potential under controlled conditions. Low heritability traits (H² < 20%) encompassed TGW (20.67%), SPAD value (19.65%), TPP (16.52%), and DM (8.83%), indicating predominant environmental control and limited effectiveness of direct phenotypic selection. Overall, the heritability estimates provided insights for further analyses: traits exhibiting high heritability (PH: 45.56%, PL: 39.67%, PY: 37.40%) demonstrated consistent clustering patterns (Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection). In contrast, traits with low heritability (DM: 8.83%, TPP: 16.52%) exhibited increased within-cluster variation, aligning with the significant G×E interactions previously reported.

Descriptive statistics and distribution analysis

A comprehensive phenotypic evaluation of 10 agronomic traits was conducted across 110 rice genotypes across the two seasons (Table S1, Fig. 1a, b). The studied population demonstrated substantial phenotypic diversity, with coefficient of variation (CV) values ranging from 3.15% to 23.83% (Table 5; Fig. 1a), indicating the presence of exploitable genetic variation for breeding applications.

Fig. 1.

Phenotypic diversity and distribution patterns of agronomic traits in 110 rice genotypes. a Density distributions with mean lines showing the frequency distribution of trait values across the germplasm collection. b Violin plots combined with box plots displaying the distribution of trait performance across all genotypes, with top three performing genotypes labeled for each trait. Red labeled points identify superior genotypes for each trait. Traits are categorized by color: morphological (green), phenological (orange), physiological (blue), and yield-related (pink)

Table 5.

Descriptive statistics for rice plant traits and yield parameters

Trait	Mean	SD	Median	Range	CV (%)	Variability
PH	112.88	11.12	116.01	83.84–132.84	9.85	Low
DTF	102.41	4.95	103	90–115	4.83	Low
TPP	15.63	2.87	15.5	9.67–25.17	18.38	Moderate
PL	25.96	2.32	25.75	21.17–30.84	8.92	Low
DM	125.08	3.94	126	115–131.5	3.15	Low
GPP	153.17	36.49	147.92	75.67–235.17	23.83	High
TGW	25.14	3.00	24.83	19.64–35.45	11.93	Moderate
NDVI	0.72	0.034	0.72	0.6–0.81	4.69	Low
SPAD	38.96	3.38	38.58	19.53–47.53	8.64	Low
PY	9.43	1.35	9.4	5.21–12.76	14.26	Moderate

PH Plant Height (cm), DTF Days to Flowering, TPP Tillers Per Plant, PL Panicle Length (cm), GPP Grains Per Panicle, TGW Thousand Grain Weight (g), DM Days to Maturity, PY Paddy Yield (t ha^− 1), NDVI Normalized Difference Vegetation Index, SPAD Chlorophyll Content (unitless)

Plant architectural traits exhibited distinct distributional characteristics and variability patterns. For PH, the results demonstrated a mean of 112.88 cm with a range spanning 83.84 to 132.84 cm, displaying low variability (CV = 9.85%) (Table 5; Fig. 1a). For PL, the results showed moderate variation with a mean of 25.96 cm (range: 21.17–30.84 cm) and low CV of 8.92%, while TPP exhibited the second-highest variability among morphological traits (CV = 18.38%), with a mean of 15.63 tillers ranging from 9.67 to 25.17. For reproductive and yield component traits (Table 5; Fig. 1a), GPP exhibited the highest variability among all traits (CV = 23.83%), with a mean of 153.16 grains and an extensive range from 75.67 to 235.17 GPP. For TGW, the results showed moderate variability (CV = 11.93%) with a mean of 25.14 g, ranging from 19.64 to 35.45 g. PY distributions revealed considerable genetic diversity in productive capacity among genotypes, with a mean of 9.43 t ha^− 1 (range: 5.21–12.76 t ha^− 1) and moderate variability (CV = 14.26%).

For the phenological trait patterns, DM showed the most consistent values across genotypes (CV = 3.15%), with a mean of 125.08 days and a narrow range of 115–131.5.5 days, while DTF demonstrated similarly low variability (CV = 4.83%) with a mean of 102.40 days (range: 90–115 days). The distribution characteristics complement the maturity patterns, with the observed variation indicating opportunities for optimizing reproductive timing to match specific cropping seasons and environmental conditions while maintaining population-level synchrony. Regarding the physiological parameters (Table 5; Fig. 1a), SPAD chlorophyll content values exhibited low variability (CV = 8.64%) with a mean of 38.96 units, ranging from 19.53 to 47.53. For NDVI, the findings indicated the lowest variability among all traits (CV = 4.69%), with a mean of 0.72 and a range of 0.60–0.81.

The observed phenotypic diversity, characterized by CV values from 3.15% (DM) to 23.83% (GPP), underpinned cluster differentiation (Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection) and elucidated the disparate heritability estimates across traits (Sect. Genetic Components of Variance and Heritability Estimates). Notably, high-variability traits such as GPP (CV = 23.83%, H²=28.81%) exhibited moderate genetic control despite considerable phenotypic variation.

Performance variation among rice genotypes

The comprehensive evaluation of 110 rice genotypes across ten agronomic traits revealed substantial phenotypic diversity within the germplasm collection (Fig. 1b). Analysis of trait-specific performance identified distinct genotypes excelling in different agronomic characteristics, providing valuable insights for breeding program optimization (Table S2). The identification of top-performing genotypes for each trait facilitates targeted selection strategies and enables breeders to identify superior donor parents for specific breeding objectives.

Physiological parameters demonstrated consistent performance across genotypes. For SPAD values, GSR18 achieved the highest reading (47.53 units), followed by GSR12 (46.05 units), and GSR64 (45.06 units). Regarding NDVI, GSR9 emerged as the top performer (0.81), followed by GSR65 (0.79), and GSR18 (0.79). For morphological traits (Fig. 1b), PH exhibited considerable variation among genotypes, with genotypes demonstrating superior stature characteristics, including NRP.C8 (132.84 cm), GSR33 (131.17 cm), GSR10 (129.17 cm). Results of TPP showed marked genotypic differences, with NGSR.13 recorded 25.17 tillers, NGSR.10 (23.33 tillers), and NGSR.21 with 22.67 tillers. Also, PL analysis revealed GSR36 as the top performer with 30.84 cm, followed by GSR18 (30.67 cm), and GSR31 (30.50 cm). Phenological traits such as DTF showed that GSR38 exhibited the longest DTF at 115.00 days, followed by GSR56 (114.50 days), and GSR33 (112.50 days) (Fig. 1b). DM showed a similar pattern, with three genotypes (GSR33, GSR56, and GSR60) all reaching maturity at 131.50 days. Reproductive and yield component traits (Fig. 1b) showed that GPP varied substantially among genotypes, with GSR3 demonstrating exceptional performance at 235.17 grains, NGSR.10 (228.17 grains), and NGSR.7 (227.17 grains). Also, TGW was maximized in NGSR.18 (35.45 g), followed by GSR19 (34.31 g), and GSR55 (32.83 g). For PY, GSR1 achieved the highest yield of 12.76 t ha^− 1, closely followed by GSR14 (12.56 t ha^− 1), and GSR12 (12.34 t ha^− 1).

In respect to the evaluation of resistance to BLS, the results revealed a diverse spectrum of disease responses (Table S3, Fig. 2). The findings illustrate the characteristic disease symptoms associated with each resistance category, ranging from nearly symptom-free leaves in highly resistant genotypes to extensively damaged leaves with numerous lesions in highly susceptible genotypes. The clear visual distinction between resistance classes demonstrates the reliability of the phenotypic scoring system used for disease assessment. The germplasm collection demonstrated varied levels of resistance, with most genotypes (38 genotypes, 34.5%) exhibiting resistant (R) responses, followed by 27 genotypes (24.5%) showing susceptible (S) reactions. Moderately resistant (MR) and moderately susceptible (MS) responses were observed in 21 (19.1%) and 14 (12.7%) genotypes, respectively, while only 10 genotypes (9.1%) displayed highly resistant (HR) responses. Among the top-performing, highly resistant GSR genotypes, GSR34, GSR49, and GSR50 exhibited a disease score of 0, indicating the strongest resistance response to BLB (Table S3-S4).

Fig. 2.

Visual classification of BLS disease resistance levels in rice genotypes under natural field conditions. Representative leaf samples showing the spectrum of disease responses from highly resistant (HR) with minimal or no visible lesions, through resistant (R), moderately resistant (MR), moderately susceptible (MS), and susceptible (S) categories, to highly susceptible (HS) genotypes displaying extensive lesion development and leaf damage. The progression illustrates the phenotypic scoring criteria used to evaluate 110 Green Super Rice genotypes for BLS resistance under natural disease pressure

Comprehensive evaluation based on normalized performance scores across all traits, including the resistance response to BLS, identified the top-performing genotypes for breeding applications (Table S4, Fig. 3). Among the 110 GSR genotypes, NGSR.18 emerged as the highest-ranked genotype with a composite performance score of 70.77%, followed closely by GSR36 (65.60%), GSR37 (65.33%), GSR33 (65.30%), and GSR19 (64.76%). Additional high-performing genotypes included GSR55 (64.48), GSR14 (63.39), GSR35 (62.78), GSR61 (62.73), and GSR28 (62.58).

Fig. 3.

Top 20 rice genotypes ranked by overall performance based on composite scoring across all evaluated traits. The horizontal bar chart displays genotypes ranked from highest (#1) to lowest (#20) performing. The composite score represents normalized trait values on a 0–100 scale based on mean performance across all morphological, phenological, reproductive, yield component, and physiological parameters evaluated in the study

Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection

The cluster analysis revealed significant genetic diversity among the 110 GSR genotypes evaluated across ten phenotypic traits over two growing seasons (Fig. 4a-d). K-means clustering optimally partitioned the genotypes into five distinct clusters, with an average silhouette width of 0.225, indicating reasonable cluster separation (Fig. 4a). Principal component analysis showed that the first two components explained 42.1% of the total phenotypic variance (PC1: 26.6%, PC2: 15.4%), suggesting substantial genetic variation in the evaluated traits (Fig. 4b).

Fig. 4.

Genetic diversity and cluster analysis of 110 Green Super Rice genotypes. (a) Phenotypic profile heatmap showing standardized trait values across all genotypes with hierarchical clustering dendrogram. Color annotation bars indicate K-means cluster assignment (purple, blue, green, orange, yellow for clusters 1–5, respectively) and BLS resistance categories (HR: highly resistant, MR: moderately resistant, MS: moderately susceptible, R: resistant, S: susceptible). (b) Principal component analysis (PCA) biplot displaying genotype distribution along the first two principal components. (c) Cluster distribution within BLS resistance categories illustrating the frequency of each K-means cluster across different resistance levels. (d) BLS resistance distribution within each K-means cluster showing the proportion of resistance categories within clusters 1–5, demonstrating variable resistance patterns across genetic groups. PH: Plant Height (cm), DTF: Days to Flowering, TPP: Tillers Per Plant, PL: Panicle Length (cm), GPP: Grains Per Panicle, TGW: Thousand Grain Weight (g), DM: Days to Maturity, PY: Paddy Yield (t ha^− 1), NDVI: Normalized Difference Vegetation Index, SPAD: Chlorophyll Content (unitless)

The five clusters exhibited unequal size distribution: Cluster 3 was the largest (45 genotypes, 40.9%), followed by Cluster 2 (25 genotypes, 22.7%), Cluster 4 (19 genotypes, 17.3%), Cluster 5 (16 genotypes, 14.5%), and Cluster 1 (5 genotypes, 4.5%) (Fig. 4a, b). Each cluster demonstrated distinct phenotypic profiles across the measured traits. Cluster 1 (n = 5) was characterized by intermediate PH (120.1 ± 3.22 cm), early flowering (92.3 ± 1.48 days), moderate tillering (14.6 ± 2.37), and the lowest grain yield (6.95 ± 1.33 t ha^− 1) (Fig. 4a). This cluster showed the highest NDVI values (0.78 ± 0.02) and SPAD readings (43.24 ± 3.42), indicating superior photosynthetic capacity despite lower yield performance. Cluster 2 (n = 25) exhibited the shortest plant stature (96.82 ± 4.68 cm) with delayed flowering (104.02 ± 2.18 days) but demonstrated high grain productivity (179.47 ± 31.71 grains per panicle) resulting in moderate yields (9.38 ± 1.26 t ha^− 1) (Fig. 4a). This cluster represented genotypes with compact architecture but maintained reasonable productivity.

Cluster 3 (n = 45), the predominant group, displayed tall PH (119.1 ± 6.01 cm), longest panicles (27.64 ± 1.7 cm), but had the fewest GPP (121.71 ± 18.27) and lowest SPAD values (37.69 ± 3.62) (Fig. 4a). Despite late maturity (128.06 ± 2.09 days), this cluster achieved intermediate yields (8.95 ± 0.73 t ha^− 1). Cluster 4 (n = 19) combined tall stature (119.29 ± 6.82 cm) with early flowering (96.18 ± 3.55 days) and high grain number per panicle (180.36 ± 30.32), resulting in the highest yield performance (10.64 ± 1.36 t ha^− 1) (Fig. 2a). This cluster represents the most productive genotype group with favorable early maturity traits. Cluster 5 (n = 16) was distinguished by the lowest tillering capacity (11.98 ± 2.14) but compensated with the highest TGW (28.27 ± 3.09 g), achieving high yields (10.2 ± 1.05 t ha^− 1) through larger individual grain size rather than grain number (Fig. 4a).

Regarding to the distribution of BLS resistance, the findings showed a significant variance among clusters (χ² = 50.1, df = 16, P < 0.001), as illustrated in Fig. 4c, d. Cluster 1 showed the highest proportion of highly resistant genotypes (60% HR), while Cluster 2 contained 60% resistant genotypes with no susceptible entries. Clusters 3 and 4 exhibited mixed resistance patterns with 40% and 36.8% susceptible genotypes, respectively. Cluster 5 demonstrated predominantly moderate resistance (43.8% MR) with fewer susceptible genotypes (6.2%).

The integration of cluster-specific phenotypic profiles with disease resistance distribution is summarized in Table 6. Cluster 1 exhibited a significant prevalence of HR genotypes (60%), corresponding with its elevated NDVI (0.78) and SPAD (43.24) metrics. Conversely, Cluster 4’s mixed resistance (36.8% susceptible) was associated with its considerable height (119.29 cm) and longest panicles—attributes positively correlated with disease susceptibility (Phenotypic Correlations Among Agronomic Traits and BLS Susceptibility: PH r = 0.257, PL r = 0.3). Cluster 2’s advantageous resistance profile (60% R, 0% S) aligned with its compact structure (96.82 cm), reinforcing the inverse relationship between PH and disease resistance. The allocation of high-yielding genotypes (GSR1: 12.76 t ha^− 1, GSR14: 12.56 t ha^− 1) among clusters with differing resistance profiles (Table 6), combined with the lack of substantial associations between yield and disease susceptibility (Phenotypic Correlations Among Agronomic Traits and BLS Susceptibility), suggests independent genetic regulation of these variables.

Table 6.

Summary of cluster-specific profiles integrating agronomic performance and disease resistance

Cluster	n (%)	Yield (t ha− 1)	Distinguishing Traits	BLS Profile	Connection to Other Analyses
1	5 (4.5)	6.95 ± 1.33	Highest NDVI (0.78), SPAD (43.24); Early flowering	60% HR	High heritability traits (Genetic Components of Variance and Heritability Estimates); Superior physiological parameters (Performance Variation Among Rice Genotypes)
2	25 (22.7)	9.38 ± 1.26	Shortest (96.82 cm); Highest GPP (179.47)	60% R, 0% S	Negative GPP-PH correlation (Phenotypic Correlations Among Agronomic Traits and BLS Susceptibility); Low BLS due to compact architecture
3	45 (40.9)	8.95 ± 0.73	Tallest (119.1 cm); Longest PL (27.64 cm); Lowest GPP (121.71)	40% S	Strong PL-PH correlation (r = 0.7); Positive PH-BLS correlation (Phenotypic Correlations Among Agronomic Traits and BLS Susceptibility.)
4	19 (17.3)	10.64 ± 1.36	Highest yield; High GPP (180.36); Early flowering	36.8% S	Contains top yielders GSR1, GSR14 (Performance Variation Among Rice Genotypes); Significant G×E for yield (Significant Genetic Variation and Genotype × Environment Interactions Detected Across Rice Germplasm)
5	16 (14.5)	10.2 ± 1.05	Largest TGW (28.27 g); Lowest tillering (11.98)	43.8% MR	Includes NGSR.18 (top performer, Performance Variation Among Rice Genotypes); Low heritability TGW (Genetic Components of Variance and Heritability Estimates)

Phenotypic correlations among agronomic traits and BLS susceptibility

The phenotypic correlation analysis revealed significant associations among key agronomic traits and disease susceptibility parameters in rice genotypes (Fig. 5, Table S1). The strongest positive correlations were observed between PL and PH (r = 0.7, P < 0.001), indicating a strong developmental linkage between vegetative and reproductive growth characteristics. Similarly, PY exhibited a robust positive correlation with GPP (r = 0.55, P < 0.001). DM displayed strong positive associations with each of DTF (r = 0.54, P < 0.001) and PL (r = 0.43, P < 0.001), while DM showed a moderate positive correlation with both PH (r = 0.36, p < 0.001), TGW (r = 0.25, P < 0.05). Conversely, there were negative correlations detected among the measured traits, including that significant negative association between SPAD and PH (r = −0.31, P < 0.01), GPP with both PL (r = −0.42, P < 0.001) and DM (r = −0.37, P < 0.001). Also, GPP demonstrated a significant negative correlation with DTF (r = −0.273, P < 0.05). BLS susceptibility, where higher values indicate increased disease severity, showed significant positive correlations with PL (r = 0.3, P < 0.01) and PH (r = 0.257, P < 0.05). These associations suggest that genotypes with larger plant architecture may exhibit enhanced susceptibility to Bipolaris oryzae infection. The positive correlation between BLS susceptibility and structural traits implies that selection for improved plant architecture should consider potential trade-offs with disease resistance.

Fig. 5.

Correlation matrix of agronomic traits. Pearson correlation coefficients between 11 agronomic traits measured in 110 genotypes. Color intensity represents correlation strength: deep blue indicates negative correlations; rich red indicates positive correlations. Numerical values show correlation coefficients with statistical significance indicated by asterisks (*** P < 0.001, ** P < 0.01, * P < 0.05)

These correlation patterns elucidated the cluster formation detailed in (Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection)(Table 6): the robust PL-PH correlation (r = 0.7) accounted for the aggregation of Cluster 3 genotypes (tallest stature: 119.1 cm, longest panicles: 27.64 cm), whereas the negative GPP-PH correlation (r=−0.3) distinguished Cluster 2 (compact, high GPP: 179.47) from Cluster 3 (tall, low GPP: 121.71). The positive associations between BLS susceptibility and both PH (r = 0.257) and PL (r = 0.3) corresponded with the cluster-resistance distribution pattern (Phenotypic clustering reveals distinct genetic groups within the rice germplasm collection), wherein Cluster 2 (shortest height) exhibited 0% susceptible genotypes, while Cluster 4 (tall stature) displayed 36.8% susceptible genotypes. The lack of substantial correlations between disease susceptibility and yield components (GPP, TGW, PY) corroborated the distribution of resistant genotypes across all yield ranges in the cluster analysis (Table 6), confirming the genetic independence of productivity and disease resistance traits.

Discussion

The noteworthy significant genotype × year interactions observed for seven of ten traits, especially yield-related parameters, underscore the variability of gene expression in response to different environmental conditions. This pattern aligns with recent multi-environment studies that demonstrate significant genotype-environment interaction effects on rice yield traits across various growing seasons [18, 19]. In contrast, the non-significant interactions noted for TPP, PL, and NDVI indicate that these traits demonstrate enhanced phenotypic stability, thereby rendering them potentially more dependable selection targets for breeding programs in varied environments [20]. Contemporary genomic studies have revealed similar patterns, indicating high phenotypic correlations between environments for architectural traits, while correlations for grain yield remain low [21].

Furthermore, H² estimates observed across the ten evaluated traits reveal distinct genetic architectures that align with the established understanding of quantitative trait inheritance in rice. The H² estimates for PH (45.56%) and PL (39.67%) correspond with contemporary research performed in highland rice ecosystems [22], suggesting that these morphological traits demonstrate consistent genetic control across diverse environmental conditions. This pattern supports the hypothesis that structural traits exhibit enhanced genetic canalization compared to yield components, which typically show higher sensitivity to environmental factors [23]. For DM, H² is notably low at 8.83%, which stands in contrast to earlier studies that have documented higher values for phenological traits, exceeding 60% (Poli et al., 2013). This suggests that the GSR germplasm may exhibit unique allelic diversity affecting maturation timing.

The observed H² values for yield-related traits, specifically PY at 37.40% and GPP at 28.81%, also indicate a multifaceted polygenic structure consistent with the principles of quantitative trait theory. These findings align with contemporary research suggesting a moderate level of genetic progress concerning yield components [24], indicating that the germplasm possesses sufficient additive genetic variance for enhancement through breeding, while also acknowledging the considerable influence of environmental factors on traits related to grain production. That is, the prevalence of the environmental variance component, which varies between 54.44% and 91.17% across different traits, underscores the critical importance of multi-environment evaluation protocols in rice breeding initiatives [25].

The identification of five distinct phenotypic clusters within the germplasm collection illustrates a structured genetic diversity that can be effectively employed for the optimization of breeding programs [26]. The disparities in cluster distribution suggest a dominant genetic background, potentially reflecting historical selection pressures or founder effects linked to the GSR program [27]. Cluster 4 demonstrates early flowering at 96.18 days, a high grain count per panicle of 180.36 grains, and a yield performance of 10.64 t ha^{− 1}, highlighting its considerable breeding potential. This cluster demonstrates that early maturity does not inherently reduce yield potential, challenging traditional notions about the trade-offs between developmental timing and grain production [28]. The genetic framework underlying this combination requires further investigation through genomic methodologies to identify associated quantitative trait loci. The differing strategies identified in Clusters 4 and 5 illustrate unique evolutionary methods for optimizing yield. Cluster 4 achieves higher yields through an increase in grain quantity, while Cluster 5 compensates for reduced tillering by enhancing the weight of individual grains (28.27 g). This contradiction exemplifies the conventional trade-offs described in yield component balance theory and provides breeding programs with diverse strategies for yield enhancement tailored to specific target environments and market requirements [29, 30].

The evaluation of BLS resistance indicates a significant correlation between plant architecture and susceptibility to disease. The observed positive correlations between BLS susceptibility and PL (r = 0.3) as well as PH (r = 0.257) indicate that selecting for increased plant stature could unintentionally elevate disease susceptibility. Recent biochemical studies indicate that the pathogenesis of Bipolaris oryzae is characterized by intricate host-pathogen interactions that are affected by plant morphology and the microenvironment [31]. The relationship between architecture and disease has considerable implications for the development of breeding strategies. Brown spot disease results in approximately 4% yield losses worldwide and is recognized as a re-emerging threat in the context of climate change [32]. The observed correlations indicate that breeding programs should strategically balance the selection of favorable architectural traits with the preservation of disease resistance.

The identification of 38 genotypes (34.5%) demonstrating resistant responses, along with 10 genotypes (9.1%) exhibiting high resistance, establishes a significant genetic foundation for resistance breeding. The low frequency of highly resistant genotypes indicates that strong resistance alleles are likely rare in this population, highlighting the need for careful conservation and strategic deployment of these genetic resources [8].

The phenotypic diversity patterns identified in the GSR collection demonstrate both similarities and differences when compared to other global rice germplasm assessments. The coefficient of variation values, ranging from 3.15% to 23.83%, are consistent with earlier findings from Taiwanese rice collections, which reported comparable variability ranges through the use of molecular markers [33]. The observed distribution of variation among trait categories indicates a distinct genetic architecture within this germplasm. The observed clustering patterns contrast with those documented in pigmented rice germplasm, where molecular markers indicated distinct population structure patterns. This divergence likely indicates the unique breeding history and selection pressures encountered during the development of GSR, leading to distinct allelic frequencies in comparison to landraces and traditional varieties. The correlation structure among traits, notably the significant positive association between PL and PH (r = 0.7), indicates coordinated genetic control mechanisms that are consistent across various rice populations [24]. This pattern corresponds with recent research employing SSR markers to evaluate genetic diversity, affirming the universal characteristics of developmental trait coordination in rice.

Overall, the incorporation of heritability estimates, genotype-environment interaction effects, and phenotypic clustering outcomes offers a definitive strategic framework for the optimization of breeding programs. The identification of traits with high heritability, such as PH, PL, and PY, indicates that direct phenotypic selection is likely to be effective for these characteristics, especially when implemented across diverse environments to address genotype-environment interaction effects [25]. Genomic selection methodologies may demonstrate advantages over phenotypic selection for these traits (Monteverde et al., 2023), as molecular markers can elucidate genetic effects that are masked by environmental variability. The identified disease resistance patterns require a balanced breeding strategy that integrates agronomic performance with pathogen resistance. The creation of selection indices that balance yield potential with disease resistance is essential for breeding varieties adapted to varied production environments characterized by fluctuating BLS pressure [8].

Conclusions

This extensive examination of the GSR germplasm collection revealed genetic diversity across numerous characteristic categories, aiding breeding program optimization and variety development. Strategic genetic improvement for productivity, disease resistance, and adaptation to varied production systems begins with the discovery of superior genotypes, comprehension of genetic factors, and analysis of trait connections.

Genomic profiling of superior genotypes should be done to understand phenotypic variation and find breeding markers. Genomic selection and phenotypic evaluation will improve breeding efficiency and genetic merit prediction across traits and habitats. Detailed physiological examinations of the trait correlations found in this work will provide mechanistic insights into plant development and resource allocation patterns to improve breeding strategies.

This work suggests pathotype-specific evaluations and molecular characterization of resistance genes to investigate disease resistance diversity. Long-term rice production disease management will benefit from resistance gene pyramiding tactics using recognized resistant genotypes. Finally, the genetic clustering patterns imply heterotic group identification and hybrid breeding applications that should be investigated through future systematic crossover and evaluation programs.

Authors’ contributions

MSA, conceptualize, execute in field, data collection compilation, analyse data and wtite up manuscript parts; AMA, Analyze data, write manuscript results and compilation; AM, manage resources; FA, help in execution of experiment, data collection and data entry; SFL, compile data, analyze data and visualization of figures; RAJ and MJH layout design, data collection and compilation.

Data availability

The data and plant material is available from the corresponding author upon request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Plant material acquisition and experimental methodologies were done in compliance with institutional, national and international guidelines, parameters and legislation.

Consent for publication

Not applicable.

Competing interests

All authors showed no competing interest.