MGIDI-based ideotype selection and genetic variation for wheat agronomic traits and bread-making quality (BMQ) under drought stress condition

Abstract

Enhancing flour quality is key to bread wheat marketability. It requires evaluating agronomic traits, bread-making quality (BMQ), and baking parameters to develop high-yielding varieties. This study assessed genetic variation and the effects of drought stress on these traits in a core collection of wheat cultivars, landraces, and lines. Field experiments used a split-plot design with three replications over two seasons under well-irrigated (WI) and drought-stress (DS) regimes (irrigation ceased at flowering). Genotype-by-irrigation interactions revealed significant genetic heterogeneity under water-limited conditions. Trait ranges varied with regime: the gluten index (GI) and grain hardness (GH) were broader under WI, while protein content, Zeleny sedimentation value (ZSV), water absorption, and dough development time were broader under DS. Drought increased tensile strength, protein content, and dough stability, indicating inconsistent trait responses. Genotype and genotype-by-environment (GGE) biplot analysis revealed that drought-stressed environments (E2, E4) were most discriminating for grain yield (GY) and GI. Several genotypes (Shiraz for GY; Satin for GI) showed close association with E2 and E4 environments, indicating adaptation to drought. A desirability index combining mean performance and stability identified Shiraz (GY) and Satin (GI) as superior for GY and GI, respectively. However, no single genotype consistently excelled across all traits, confirming trait-specific adaptation. Genetic coefficient of variation was higher for BMQ than agronomic traits in both regimes. Heritability was high for ZSV (> 0.90), GI (> 0.80), and spikelet number per spike (> 0.70) across regimes but inconsistent for other traits. Canonical correlation analysis identified two significant pairs, linking ZSV and GH indirectly with yield components. The multi-Trait Genotype–Ideotype Distance Index (MGIDI)-based selection identified top performers (Danesh, Line 181, Shiraz). It also quantified irrigation-dependent outcomes: drought drove gains in GI (+ 60.0%) and yield traits but increased GH (+ 16.1%). In contrast, WI favored yield with reduced GH (-8.7%). These findings highlight MGIDI’s utility in resolving trait trade-offs for environment-specific breeding. The GGE biplot models effectively visualized mega-environment differentiation and genotype stability, providing a complementary framework to MGIDI for selecting wheat varieties with both drought resilience and BMQ. Results reveal inconsistent interrelationships between agronomic and BMQ traits, notably a yield–quality trade-off (GH, GI) under contrasting irrigation regimes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-09175-5.

Introduction

Over the past few decades, climatic variability has led to severe abiotic stresses, with drought stress emerging as a major issue resulting in significant yield losses [1–5]. With the world’s population increasing and facing the adverse effects of climate change, enhancing both grain yield and quality are crucial goals in wheat breeding programs [6–9]. Enhancing flour quality is of paramount importance for the marketability of bread wheat, as most wheat grains are processed into wheat flour [10–13]. During drought stress, the shortage of water and limitations in nutrient uptake and transport can strongly affect the growth, yield, and quality of wheat [14, 15]. The flowering and grain filling stages of wheat are particularly sensitive to drought stress, leading to a reduction in grain yield [16, 17]. Flour quality is primarily determined by grain starch, protein content grain hardness and milling quality [18–30]. The changes in protein and starch content in wheat grains due to drought stress can also influence flour and baking quality [31]. Rekowski et al. [32] found that drought stress during anthesis in different wheat genotypes increased gliadin and glutenin levels in wheat grains, improving bread quality. Another study by Zarea et al. [33] highlighted the interaction between grain yield and protein content with the effects of drought stress in wheat, emphasizing the need to consider the relation between yield and protein related traits under water restricted conditions. Other studies have also shown an increase in protein content coupled with a reduction in grain yield under drought stress in wheat [34, 35].

Conventional plant breeding is widely recognized as a cost-effective and environmentally friendly method for improving bread-making quality (BMQ) parameters [36]. Screening diverse germplasm with ample genetic diversity for bread quality establishes a strong foundation. This helps create wheat varieties with higher protein content and better bread quality [37, 38]. Currently, the grain quality of commercial wheat is subpar, likely due to breeding programs that have prioritized maximum grain yield over the past few decades [39]. However, the genetic variation for grain quality found in wild relatives and progenitors of cultivated wheat can be utilized in breeding programs to enhance grain protein levels and bread-making quality [40]. Furthermore, genetic diversity within bread wheat is essential for both breeding efforts under biotic/abiotic stress and genetic resource preservation [41, 42]. Understanding genetic diversity within plant species helps researchers identify traits. These traits can enhance food production and develop high-yield and high-quality crop varieties [43, 44]. In a study of 35 different genotypes evaluated over two years, significant differences in yield, its components, and gluten content were identified among the genotypes tested [43]. Notably, high heritability, genetic advance, and phenotypic (PCV) and genotypic coefficient of variation (GCV) were observed for yield, its components, and gluten content [43]. Furthermore, it has been shown that genetic variation exists for grain quality traits and that there is an inconsistent relationship between grain yield and grain quality in wheat [45–47]. Developing high-yielding wheat varieties with standard BMQ requires analyzing overlooked relationships. These include links between yield traits, bread-making parameters, and flour baking properties, evaluations often missing in breeding programs. Such comprehensive assessments are crucial for breeders to make informed selections. To address challenges like trait multicollinearity and complex decision-making, we propose the integrated use of multivariate analysis with the Multi-Trait Genotype–Ideotype Distance Index (MGIDI) [48, 49]. MGIDI synthesizes complex phenotypic data including grain yield, kernel weight, protein content, gluten strength, starch levels, and drought tolerance into a single, interpretable index via Principal Component Analysis (PCA) [48, 50, 51]. This approach ranks genotypes by their proximity to an ideal target (ideotype), while factor analysis quantifies individual trait contributions [52]. Ultimately, MGIDI enables efficient identification of genotypes optimally balancing yield, quality, and resilience, accelerating cultivar development for diverse environments [53, 54]. Therefore, the aims of this study were (1) to evaluate the effects of terminal drought on agronomic traits and BMQ, (2) to estimate heritability and genetic variation parameters, (3) to analyze interrelationships among grain yield, flour mixing properties, and BMQ traits, and (4) implement MGIDI-based ideotype selection to resolve trade-offs between wheat grain yield and quality and identify superior genotypes.

Materials and methods

Plant material and experimental design

The study utilized different wheat varieties as plant materials for the assessment of agronomic and bread quality related traits (Supplementary Table S1). The wheat varieties tested in the current study were commercial cultivars traditionally used for cultivation in Iran and landraces collected from different Iranian geographical areas. This core collection was selected for broad genetic variation in yield components and bread‑making quality. Commercial cultivars offer high yield potential and local adaptation whilst landraces carry valuable traits and genes for improvement of stress tolerance and grain quality in commercial cultivars [40, 41]. A split-plot design with three replications was implemented with irrigation treatments (well-irrigated and drought stress) as the main factors and wheat (varieties and landraces) as the sub-plot. The experiment was conducted over two consecutive growing seasons at the research field of the School of Agriculture, Shiraz University, Iran (29°43′N, 52°35′E, 1480 m a.s.l.). Accordingly, the wheat genotypes were evaluated in four environmental conditions consisting of 2 irrigation regimes × 2 growing seasons. Sowing was conducted at a depth of 2 cm in soil in 1 m² plots consisting of four rows with grain spacing of 10 cm. All recommended agronomic practices were followed throughout the growth season. Urea was applied at a rate of 100 kg N ha⁻¹ at sowing and stem elongation stages [33]. Irrigation was estimated based on the soil moisture levels. Soil moisture was monitored gravimetrically at weekly intervals. Three samples were taken in representative locations from the 0–30 cm depth, oven-dried at 105 °C for 48 h, and gravimetric water content was calculated as (fresh weight – dry weight) / dry weight × 100 [14, 62].In the drought stress (DS), the plots received winter precipitation and regular irrigation water until the early flowering stage of the plants in spring when irrigation ceased until the end of the season. No rainfall occurred after the cessation ofirrigation at the early flowering stage. In the well-irrigated (WI)treatment, irrigation was continued every 10 days through the whole growing cycle when the grain reached physiological maturity, regardless of soil moisture. Weeding was performed manually, and herbicides such as TOTAL, 2, 4-D, and paraquat were applied during the growing cycles [14]. Weather information is presented in Supplementary Table S1. Weather data were collected from the Bajgah weather station located at the School of Agriculture, Shiraz University, Iran.

Agronomic traits

Several agronomic traits including plant height (PH; cm), spike length (SL), and spikelet number per spike (NSS) were measured on 10 plants per plot for all genotypes. Days to maturity was recorded by counting the number of days from sowing untill 50% of plants in each replication reached physiological maturity. The grain number per spike (NGS), grain weight per spike (GWS), thousand grain weight (TGW; g) and grain yield (GY) was measured after harvesting the plants. GY (g per plant) was measured by weighing the total grain from 10 plants per plot and the average was used as the grain yield of each genotype [62].

Measurement of quality parameters

To better understand the trade‑off between yield and BMQ using the MGIDI method, we selected 19 out of the original 25 agronomic genotypes (Supplementary Table S2) using an extreme‑phenotype sampling strategy [48, 49]. First, from the 25 genotypes grown under both WW and DS conditions, those with the smallest percentage changes in grain yield (i.e., the most drought‑stable genotypes) were selected. Second, from these, genotypes with very different values for grain hardness (GH) and gluten index (GI) were selected. These 19 genotypes were used for the MGIDI analysis, which combined grain yield, grain hardness, and gluten index. For two additional quality traits including protein content and Zeleny sedimentation value (ZSV), we only measured them in the 10 genotypes that had the highest grain hardness and gluten index values. The remaining 9 genotypes were not tested for protein or ZSV. This approach challenges the MGIDI ideal genotype with diverse trait combinations and helps identify well‑balanced genotypes that maintain both drought resilience and BMQ [50, 51].

Grain hardness (GH)

A number of 19 genotypes with the smallest change in GY between WI and DS conditions were selected to undergo a GH test in the Department of Biosystem, School of Agriculture, Shiraz University, Iran, following a previously described protocol [55]. This test was conducted using a pressure instrument (INSTRON, Iran Expo) instrument. Each single grain was placed in the center of a round plate and pressure was applied until the grain broke. The amount of pressure required to break each wheat grain was recorded for every genotype [55],

where, σ is the banding stress (MPa), F is the rupture force (N), t is the thickness of the grain and, w is the grain width (mm).

Protein content

The protein content of the flour was analyzed using the Kjeldahl method [56]. Specifically, 1 g of Kjeldahl tablet powder was added to 0.2 g of flour, followed by the addition of 5 ml of sulfuric acid to the samples. The samples were then placed in a digestive instrument for 4 h to obtain a clear solution. Subsequently, the clear solution was mixed with 20 ml of distilled water and the nitrogen (%) was measured using an automated Kjeldahl analyzer (K110, Hanon Instruments, Jinan, China). The protein content (%) in the flour was calculated using the following equation [57]:

Gluten index (GI)

The AACC method 38-10.01 [58] was followed to determine the gluten index. Quantification tests were carried out using a roller mill in a flour factory in Shiraz, Fars, Iran. Ten grams (10 g) of flour from each of the 19 genotypes selected for grain hardness were mixed with 4.5 ml of distilled water, then placed on a lace and processed using a GLUTEN INDEX instrument (Glutomatic 2200, Perten Instruments, Hägersten, Sweden). The resulting dough was washed with cold water, and the resulting number and index were considered as the wet gluten. The dough was then placed on filter paper and centrifuged. The amount of dough remaining on the filter paper was considered as high-quality gluten, while the dough that passed through was considered as substandard gluten. The sum of high-quality and substandard gluten was considered as the total gluten. The gluten index was calculated as follows [58],

Zeleny sedimentation value (ZSV)

Of the 19 genotypes, 10 were selected for ZSV analysis based on their higher gluten index and grain hardness and the others excluded from ZSV analysis. The AACC [59] procedure was followed to analyze ZSV. To determine ZSV, 50 ml of bromophenol blue was added to 3.2 g of flour and shaken by hand 12 times before being placed on a shaker for 5 min. Following this, 25 ml of lactic-isopropyl acid was added to the samples and shaken for an additional 5 min. The samples were then placed on a smooth surface and the ZSV was determined after 5 min.

Statistical analysis and genetic variation parameters

In the combined analysis of variance (c-ANOVA) of the data collected from the field experiments, the effects of replication and year were considered as random, and genotype was defined as fixed. Based on the results of c-ANOVA and the significant treatment by year interactions, mean comparisons of traits for genotypes were performed for each growing season. The data analysis was carried out using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA) . However, for estimating genetic related parameters (genotypic and phenotypic variances) in each irrigation condition and year, the variance components were estimated using the expected mean squares (EMS) of sources for sources of variations in the ANOVA for a randomized complete block design (RCBD). The variance components were used for the estimation of heritability in broad sense (h²_bs), phenotypic coefficient of variation (PCV) and genetic coefficient of variation (GCV). Additionally, the data for grain quality related parameters and bread making traits were analyzed based on a completely randomized design (CRD). The formulas used for genetic parameters are as follow [73],

where, , VG, VE and VP represent trait mean, and genetic, environmental and phenotypic variances, respectively.

Simple correlation coefficients were computed for the analysis of the simple relationships among traits in SAS 9.4. Simple correlations show the relation between two variables without considering other variables. In contrast, canonical correlation analysis (CCA) considers the linear combination of all variables tested to analyze interrelationships between sets of variables. The CCA provides information about multidimensional relationships between two sets of variables which is not possible to assess through simple correlation analysis. The CCA was performed to determine associations between two sets (denote U and V) of traits including agronomic (U) and quality related (V) traits using SAS 9.4. The results of CCA are typically straightforward to interpret, as the canonical variables highlight the most correlated pairs and variables unravel their distinct natures.

The effect of drought stress on wheat was also assessed through the determination of percentage changes of traits between the two irrigation treatments as follow,

Principal Component Analysis (PCA) was performed in R program, version 4.3.1 (R Core Team, Vienna, Austria) using the FactoMineR package, version 2.8 (R Foundation for Statistical Computing, Vienna, Austria) and the factoextra package, version 1.0.7 (Alboukadel Kassambara, Marseille, France) to reduce dimensionality and identify patterns in the dataset. Variable contributions were analyzed through factor maps (fviz_pca_var) with gradient coloring (blue-to-red) reflecting their importance to components, while individual sample representation was evaluated using cosine-squared (cos2) values (fviz_pca_ind) [60]. A clustered heatmap analysis was performed using the R packages pheatmap, version 1.0.12 (Raivo Kolde, Tartu, Estonia) and RColorBrewer, version 1.1‑3 (Erich Neuwirth, Graz, Austria) to visualize patterns of trait variation across genotypes across two irrigation regimes. The heatmap employed Euclidean distance-based hierarchical clustering with complete linkage for both rows (genotypes) and columns (traits).

Genotype and genotype-by-environment (GGE) biplots

The GGE biplot analysis for GY, GH, and GI was performed using the metan package [48] in R. For each trait, the data for multi-environment trials were fitted to the GGE model using the gge() function, with four environments as the second dimension: E1 (first year, well‑irrigated), E2 (first year, drought stress), E3 (second year, well‑irrigated), and E4 (second year, drought stress). A desirability index was calculated for each genotype as the difference between mean performance and stability, where lower stability values contribute to a higher desirability score. Genotypes were then ranked separately for mean, stability, and desirability to facilitate selection of superior genotypes combining high performance with high stability. The resulting biplot was visualized with the plot() method to generate “which-won-where” biplots, which include a convex hull to identify the winning genotype in each environment and sector lines to delineate mega-environments.

The multi-trait genotype-ideotype distance index (MGIDI)

The MGIDI is calculated by measuring the Euclidean distance between treatment scores and the ideal treatment. The formula for MGIDI is as follows [61]:

where, γ_ij represents the score of the ith accession in the jth factor (i = 1, 2,…,t; j = 1,2,…,f ), with t and f being the number of accessions and factors, respectively; and γ_j is the jth score of the ideal accession. A lower MGIDI value signifies that an accession is more closely aligned with the ideal accession, indicating favorable values across all calculated indices. The selection differential for all traits was established using a selection intensity of approximately 10%. The ideotype goals are biologically justified. GY was always maximized (goal = 100) [6, 62], GH and GI were minimized under WI (goal = 0) because softer grains improve milling and flour quality [20, 22], and maximized under DS (goal = 100) because harder grains with high GI preserve bread quality under drought [11, 28, 32, 58, 64]. The MGIDI analysis was performed in R program, version 4.3.1 (R Core Team, Vienna, Austria) using the metan package, version 1.18.0 (Tiago Olivoto, Santa Maria, Brazil) to evaluate genotype performance across multiple traits under stress conditions [61]. A mixed-effects model was fitted via gamem() with genotypes (GEN) as random effects, blocks (BLOCK) as replicates, and all remaining traits as response variables. The MGIDI index was computed through the mgidi() function, which calculates distance-based selection scores relative to an ideal genotype. All selection gains (SG%) were computed separately for each irrigation regime using only the phenotypic data from that regime (WI or DS). Therefore, they represent the observed phenotypic response to direct selection within that specific water environment, not predicted genetic gains across environments.

Factor analysis (FA)

FA was used to rank genotypes based on desired trait values using the MGIDI data. Initially, FA was performed using 𝑟X𝑖𝑗 to analyze the correlation structure and reduce the dimensionality of the data. This process provided a more comprehensive understanding of the relationships between RILs and their traits [61].

where, X represents a 𝑝×1 vector of rescaled observations, 𝜇 is a 𝑝×1 vector of standardized means, L is a 𝑝×𝑓 matrix of factorial loadings, 𝑓 is a 𝑝×1 vector of common factors, and ε is a 𝑝×1 vector of residuals. Here, p and f represent to the number of traits and common factors retained, respectively. The eigenvalues and eigenvectors are obtained from the correlation matrix of 𝑟X𝑖𝑗.

Strengths and weaknesses

The proportion of the MGIDI attributed to the ith treatment and explained by the jth factor (𝜔𝑖𝑗) was utilized to assess the efficacy of the treatments. This measure helps to identify the strengths and weaknesses of each treatment as outlined below [61]:

where, 𝐷𝑖𝑗 represents the distance between the ith treatment and the ideal treatment for the jth factor. A low contribution of a factor suggests that the characteristics linked to that factor closely match the desired treatment.

Results

Effects of drought stress on agronomic traits

The results of the c-ANOVA revealed a significant effect of year, demonstrating that mean comparison should be performed for each individual growing season. The results of ANOVA for traits in individual year are shown in Table 1. The means for agronomic traits are shown in Supplementary Tables S2-S5,. The density plots (Fig. 1), which compare the distribution of seven wheat agronomic traits between WI and DS regimes, indicated that drought stress almost reduced traits. For PH, the cultivars Sabalan, Azar2, and Homa showed the highest PH under WI conditions, while Sirvan was the shortest. Dehdasht, Shiraz, and Azar2 were the tallest plants under DS, with Aftab being the shortest. For SL, Baharan and Danesh had the longest spikes under WI conditions, while Shiraz had the longest spikes under DS. Shiraz and Danesh also had the highest NSS under WI conditions, while Homa had the lowest. Under DS, Danesh and Satin had the highest NSS, with Homa having the lowest.

Table 1.

Analysis of variance for morphological traits and grain yield components of 25 wheat genotypes under well-irrigated and drought stress conditions in the crop year 2015-2016

Source	D.F	NSS	PH	SL	NGS	GWS	TGW	GY
	Mean squares (2015–2016)
Block	2	9.713	17.627	4.272	237.079	0.451	42.528	25.493
Irrigation regime (I)	1	8.314**	297.073**	3.654**	36.52**	0.247**	160.0**	43.6**
Error	2	3.441	57.350	1.309	12.716	0.003	33.824	25.880
Genotype (G)	24	25.247**	175.224**	7.050**	195.3**	0.256**	65.142**	9.914**
I × G	24	6.003**	47.508**	1.882**	35.77**	0.10**	25.54**	16.5**
Error	93	4.059	44.663	1.287	39.861	0.104	22.958	5.546
CV (%)		13.7	10.8	14.5	16.9	24.2	13.4	53.0
	Mean squares (2016–2017)
Block	2	4.680	363.802	3.250	38.763	0.221	9.94	0.758
Irrigation regime (I)	1	3.405**	552.92**	0.96**	174.1**	2.0**	661.5**	18.096**
Error	2	5.309	154.283	2.703	22.106	0.038	6.582	0.442
Genotype (G)	24	17.335**	396.6**	5.881**	68.2**	0.15**	114.7**	1.657**
I × G	24	2.769**	149.5**	1.77**	26.4**	0.05*	11.30**	0.634**
Error	93	1.933	186.463	1.705	35.262	0.051	12.609	0.534
CV (%)		9.6	19.8	17.9	22.2	20.4	9.2	19.8

^** significant at the 1% level, respectively

D.F Degree of freedom, NSS Number of spikelets per spike, PH Plant height, SL Spike length, NGS Number of grains per spike, GWS Grain weight per spike, TGW Thousand-grain weight, GY Grain yield per plant

Fig. 1.

Density plots comparing the distribution of seven wheat yield and agronomic traits between well-irrigated (WI) and drought-stressed (DS) regimes. Each panel corresponds to one trait: (A) NSS (number of spikelet per spike), (B) PH (plant height, cm), (C) SL (spike length, cm), (D) NGS (number of grains per spike), (E) GWS (grain weight per spike, g), (F) TGW (thousand-grain weight, g), and (G) GY (grain yield, g). Density curves were estimated using a Gaussian kernel with bandwidth adjustment (adjust = 1.2). Semi-transparent polygons below each curve indicate the distribution density; solid lines show the kernel density estimate. Rug plots (jittered tick marks) below each panel indicate individual data points. The y-axis represents probability density, and the x-axis shows trait values

Shiraz and Danesh had the highest NGS under WI and DS, respectively, while Homa had the lowest under both conditions. Yavaros had the highest GWS, and ΔH-96-8 had the lowest under WI conditions. Under DS, Danesh and Line 181 had the highest GWS, while Aftab had the lowest. The highest TGW was found in Homa and the lowest in Shoush under WI conditions. Under DS, Yavaros and Homa had the highest TGW, while ΔH-96-8 had the lowest. Danesh and Homa presented the highest GY under WI conditions, while Line 181 had the lowest. However, under DS, Line 181 had the highest GY, and Baharan had the lowest.

Drought stress affects bread making quality related parameters

Zeleny sedimentation value, grain protein and hardness

The results of ANOVA revealed significant differences in quality parameters among wheat genotypes (Supplementary Table S6). GI (%) varied between 3.22 and 97.25 under the WI regime and between 4.16 and 48.32 under the DS treatment over two seasons (Table 2). Ghabus showed a higher GI than any other genotype in three out of the four tested environments. Other genotypes showed inconsistent values for GI between WI and DS treatments over seasons.

Table 2.

The average quality characteristics of dough of 19 wheat genotypes under well-irrigated and drought stress conditions in two crop years 2015–2016 and 2016–2017

Genotype	WI1		DS1		WI2		DS2
	GI (%)	GH (N)	GI (%)	GH (N)	GI (%)	GH (N)	GI (%)	GH (N)
M-92-20	3.22	87.32	12.17	167.30	11.15	121.34	10.03	136.42
WS-90-18	6.64	63.62	7.61	142.27	14.79	102.86	11.09	119.95
Aflak	7.06	93.95	4.48	84.71	6.69	98.95	7.74	135.22
Azar 2	24.00	96.52	25.81	123.68	11.50	88.28	11.05	89.05
Aftab	5.93	94.09	6.13	117.11	4.46	88.52	7.15	145.84
Baharan	5.69	111.40	7.14	168.54	6.10	111.13	6.80	147.59
Tigar	12.10	149.94	7.88	125.96	12.87	127.07	11.90	138.01
Danesh	9.71	76.13	20.37	98.08	23.83	79.82	12.67	99.70
Rizhav	6.89	94.77	4.16	151.52	10.65	92.13	14.48	95.17
Satin	97.25	124.31	45.82	92.11	19.42	104.28	10.04	89.78
Saji	11.30	188.97	7.82	193.68	4.50	115.97	6.75	168.07
Sirvan	7.13	127.85	14.04	151.93	7.67	137.34	6.42	155.78
Shiraz	9.84	88.43	15.19	71.55	15.28	126.85	14.48	104.98
Ghabus	15.66	95.50	49.32	108.06	23.24	100.67	14.55	128.05
Cras sabalan	8.71	71.21	34.06	87.63	10.67	86.57	10.61	90.92
Karim	23.37	159.48	6.59	110.91	6.79	111.25	3.91	135.80
Line181	83.52	120.16	5.35	126.42	20.98	133.49	11.10	115.44
Natasha	12.45	211.36	15.63	106.44	13.10	120.15	6.35	133.34
Yavaros	4.33	157.47	36.99	136.35	19.73	146.21	12.32	156.68
LSD (5%)	13.42	40.52	8.12	55.12	9.35	41.07	2.93	33.12

WI1 Well-irrigated (first year), DS1 Drought stress (first year), WI2 Well-irrigated (second year), DS2 Drought stress (second year), GI Gluten index, GH Grain hardness

GH is a quality related trait associated with the milling properties of wheat and the bread-making quality of the resulting milling products. The GH ranged between 63.62 and 211.36 under WI condition and between 71.5 and 193.68 under DS treatment. Tiger, Saji, Sirvan and Yavarous showed higher GH in the tested environments than others.

Genotypes showed significant differences for protein content and ZSV over environments and other dough quality related traits (Supplementary Tables S6 and S7). Protein content in the genotypes tested varied between 13.34 and 18.40% in the WI treatment and between 13.13 and 25.19% in the DS over two seasons (Table 3). Tiger grains accumulated high protein in both WI and DS treatments in the first season whilst their rank for protein was not repeated in the second season. Other genotypes showed inconsistent values for protein content in the tested environments. Significant differences were found between the ZSV values of genotypes (Table 3). The ZSV varied from 10.5 to 26.5% in the WI treatment and from 2.5 to 25.5% in the DS treatment. Genotypes with higher gluten content showed higher ZSV values. Azar 2 had the lowest ZSV in all environments whilst Satin had high ZSV in DS in the first season and WI regime of the second season. The highest ZSV was found in Ghabus under the WI regime. Other genotypes showed inconsistent ZSV across irrigation regimes and season.

Table 3.

The average quality characteristics of dough of 10 wheat genotypes in well-irrigated and drought stress conditions in two crop years 2015-2016 and 2016-2017

Genotype	WI1		DS1		WI2		DS2
	ZSV (%)	PrC (%)	ZSV (%)	PrC (%)	ZSV (%)	PrC (%)	ZSV (%)	PrC (%)
WS-90-18	24.00	13.34	24.50	14.53	23.50	15.20	20.50	16.52
Azar 2	11.50	14.95	2.50	15.25	24.50	13.85	3.00	14.21
Tigar	23.50	18.40	23.50	20.26	25.50	14.48	24.50	14.48
Danesh	21.50	14.27	20.50	15.25	20.00	16.93	20.50	17.98
Satin	24.00	14.95	25.00	21.44	26.00	13.73	22.50	14.67
shiraz	22.50	14.21	17.50	16.65	10.50	15.75	18.50	16.90
Ghabus	26.50	16.65	14.00	18.80	21.50	14.00	17.50	25.19
Cross sabalan	25.50	14.50	20.00	20.95	13.50	17.09	22.50	13.73
Line181	23.00	15.91	22.50	23.44	24.50	13.97	25.50	15.87
Yavaros	17.00	15.51	20.50	17.75	26.00	13.64	24.00	13.13
LSD (5%)	1.22	1.73	2.17	8.01	1.93	3.29	1.73	3.85

WI1 Well-irrigated (first year), DS1 Drought stress (first year), WI2 Well-irrigated (second year), DS2 Drought stress (second year), ZSV Zeleny sedimentation value, PrC Protein content

Traits changes under well-irrigated and drought-stress conditions

Supplementary Tables S8–S10 show the changes (%) in agronomic and BMQ traits between WI and DS for each genotype. Chamran exhibited the highest increase (35.2%) in NSS, while the Shiraz cultivar showed a significant decrease (-22.2%) in DS compared with the WI treatment. In terms of PH, the line no.181 and Cross Sabalan experienced the most significant increases and decreases, respectively, under WI compared to DS conditions. Additionally, the Line 181 had the greatest increase in SL, while the Shiraz cultivar showed the most significant decrease. The NGS underwent substantial changes between WI and DS conditions, with a 30.3% increase for the Chamran cultivar and a -22.4% decrease for the Shiraz cultivar. Changes in GWS varied among cultivars, with the Line 181, Ghabus, and Yavarus showing different responses to drought stress. Similarly, the TGW increased in Aflak and Shoush, while it decreased in Azar2 and Homa under DS. The GY of Line 181 increased significantly by 466.7%, while the Homa cultivar showed a decrease of -74.1%.

Among the bread quality related parameters, the GI showed significant changes in M-92-20 (278%), Yavarus (754%), and Cross Sabalan (291%), compared with Karim (-71.8%) and Satin (-52.9%) between the two irrigation conditions (Supplementary Tables S7 and S8). WS-90-18 (123.6%) and M-92-20 (91.6%) showed substantial increases in GH, while Natasha (-49.6%) and Karim (-30.5%) showed decreases under both conditions. Azar2 (32.7%) and WS-90-18 (-39.4%) experienced significant increases and decreases, respectively, in the ZSV. The Line 181 showed the highest increase in protein index at 44.7%, while the Yavarus had the lowest change (-3.7%).

Heritability of agronomic and bread making related traits

The PCVs exceeded the GCV for all traits examined (Supplementary Tables S11- S14). The GCV ranged from 14.75% to 54.90% for agronomic traits and from 8.75% to 137.50% for grain quality parameters. The results indicated that under WI conditions, the ZSV (99.25%), GI (96.96%), GH (88.59%), protein content (85.53%), TGW (75.37%), and NSS (75.10%) exhibited high heritability. Similarly, under DS, the ZSV (98.94%), GI (92.17%), and NSS (71.99%) showed the highest heritability.

Interrelationship of agronomic and bread making quality traits

The correlation matrices for seven agronomic traits (NSS, PH, SL, NGS, GWS, TGW, and GY) under WI and DS regimes are presented in Fig. 2. Under WI, most yield‑related traits showed moderate to strong positive correlations with NSS. For example, NGS was positively correlated with NSS (r = 0.82, p < 0.01), and GY showed a positive correlation with NSS (r = 0.56, p < 0.01). Similarly, GWS exhibited a strong positive correlation with NSS (r = 0.76, p < 0.01). In contrast, TGW showed a negative correlation with NSS (r = − 0.43, p < 0.05). Under DS, NGS remained positively correlated with NSS (r = 0.76, p < 0.01), as did GWS (r = 0.81, p < 0.01) and SL (r = 0.69, p < 0.01). GY showed a weaker positive correlation with NSS (r = 0.49, p < 0.05). TGW continued to show a negative correlation with NSS (r = − 0.36). Results of the CCA for agronomic traits and bread quality related parameters are presented in Table 4, Supplementary Tables S15-S17. Under WI conditions, the first agronomic canonical variable (U1) represented NGS, GWS, and PH. U1 had a strong direct association with GH and an indirect association with ZSV. However, U2 which was mainly affected by PH and yield components showed high and indirect correlation with GH and a direct correlation with protein content. The first grain quality canonical variable (V1), which showed direct and indirect correlation with GH and ZSV, had a stronger association with NGS and PH under WI condition. V2, the second quality-related canonical variable represented by protein content and GH had direct association with PH and yield components. The results of CCA under DS conditions were inconsistent with those under WI conditions. The U1 agronomic variable, which was defined by GY, SL and NGS had a higher correlation with grain protein content and negative association with GH. U2, representing a negative correlation with yield components, had a stronger direct association with ZSV and GH. V1, the first canonical variable of quality-related traits, showed direct and indirect relations with protein content and GH, respectively. V1 had a direct association with GY, NGS and SL. V2, which was a canonical variable driven by ZSV and GI showed indirect associations with yield components.

Fig. 2.

Correlation matrices of seven agronomic traits, NSS (number of spikelet per spike), PH (plant height, cm), SL (spike length, cm), NGS (number of grains per spike), GWS (grain weight per spike, g), TGW (thousand-grain weight, g), and GY (grain yield, g) under well-irrigated (A) and (B) drought stress conditions. The color scale indicates Pearson correlation coefficients ranging from − 1 (negative correlation, red) to + 1 (positive correlation, blue). Correlation coefficients are shown as numbers inside the squares; symbols (*, **) placed below each coefficient denote statistical significance at p < 0.05, and p < 0.01, respectively

Table 4.

Morphological and qualitative standard canonical variables and their correlation with morphological traits and qualitative traits in well-irrigated conditions

Trait	Coefficient (U1)	Coefficient (U2)	Correlation with U1	Correlation with U2	Correlation with V1	Correlation with V2
NSS	-0.68	2.81	0.19	0.42	0.19	0.33
PH	1.01	0.68	0.90	0.02	0.90	0.01
SL	0.26	-1.88	0.36	0.14	0.36	0.11
NGS	0.75	-1.53	0.37	0.15	0.37	0.12
GWS	-0.08	-0.32	0.46	0.19	0.46	0.15
TGW	-0.38	-0.22	0.27	-0.39	0.27	-0.31
GY	-0.03	0.72	0.41	0.38	0.41	0.30
Canonical Correlation	U1V1: 0.99	U2V2: 0.79
Trait	Coefficient (V1)	Coefficient (V2)	Correlation with V1	Correlation with V2	Correlation with U1	Correlation with V2
ZSV	-0.96	0.30	-0.93	0.34	-0.93	0.27
Prot	0.35	0.94	0.30	0.95	0.30	0.75

NSS Number of spikelets per spike, PH Plant height, SL Spike length, NGS Number of grains per spike, GWS Grain weight per spike, TGW Thousand-grain weight, GY Grain yield per plant, ZSV Zeleny sedimentation, Prot Protein index

Clustering wheat varieties

The agglomerative cluster analysis classified the wheat genotypes into four distinct groups based on both agronomic and two quality-related traits (Table 5 and Supplementary Table S18). The classification of these genotypes into four groups was based on their similarity in traits. Under WI conditions (Supplementary Table S18), Group 1 consisted of M-92-20, WS-90-18, Aflak, Azar2, Aftab, Baharan, Rizhav, Ghabus, Cross sabalan. Group 2 (Karim, Sirvan, Yavaros, Saji, Natasha, Tiger) showed the highest mean for GH. Group 3 (Line 181, Satin) had the highest mean for GI. Group 4 (Shiraz, Danesh) displayed the highest mean values for NSS, PH, SL, NGS, GWS, and GY. No significant differences among groups were observed for TGW under WI conditions.

Table 5.

Analysis of variance and comparison of mean of genotypic groups resulting from cluster analysis for yield traits and quality traits in drought stress conditions in wheat

Trait	Mean squares	P-value	Mean
			Group 1	Group 2	Group 3	Group 4
NSS	7.42	0.0002	14.51c	13.33c	15.85b	17.31a
PH (cm)	58.29	0.0248	64.67a	54.55b	63.75a	64.27a
SL (cm)	2.42	0.0216	7.28bc	6.72c	8.11ab	8.90a
NGS	26.20	0.0046	32.02bc	29.18c	33.72b	37.40a
GWS (g)	0.10	0.0002	1.17b	0.95c	1.18b	1.48a
TGW (g)	20.46	0.0049	36.81a	31.62b	33.88b	34.42ab
GY (g)	3.35	0.0071	3.65b	3.07b	3.59b	5.87a
GI (%)	105.46	0.3367	15.01a	5.89a	20.34a	12.91a
GH (N)	635.50	0.4075	132.56a	127.41a	112.92a	107.59a

Means followed by similar letter(s) in each row are not significantly different at 5% probability level

Group composition under drought stress: Group 1: Danesh, Line 181; Group 2: Satin, Ghabus, Shiraz, Natasha, Tiger; Group 3: Rizhav, Aftab and Group 4: Karim, Sirvan, Yavaros, Saji, M‑92‑20, WS‑90‑18, Aflak, Azar 2, Baharan, Cross Sabalan

NSS Number of spikelets per spike, PH Plant height, SL Spike length, NGS Number of grains per spike, GWS Grain weight per spike, TGW Thousand-grain weight, GY Grain yield per plant, GI Gluten index, GH Grain hardness

Under DS conditions (Table 5), Group 1 (Karim, Sirvan, Yavaros, Saji, M-92-20, WS-90-18, Aflak, Azar 2, Baharan, Cross sabalan) exhibited the highest TGW. Groups 1, 3 and 4 showed similarly high PH (with no significant difference among them), while Group 2 had significantly lower PH. Group 2 comprised Rizhav and Aftab, and Group 3 comprised Satin, Ghabus, Shiraz, Natasha, and Tiger. Group 4 (Danesh, Line 181) had the highest mean values for NSS, SL, NGS, GWS, and GY. No significant differences among groups were found for GI or GH under DS conditions.

Heatmap clustering of agronomic and bread quality traits

In the analysis of wheat genotypes cultivated under WI conditions, hierarchical clustering identified three distinct genotypic clusters: (1) high-yield, low-hardness, and low-gluten index genotypes (e.g., Danesh, Line 181); (2) low GH and low-yield genotypes (e.g., Aftab, Rizhav); and (3) genotypes with exceptional GI (e.g., Ghabus, Satin) Further examination of trait correlations revealed a significant trade-off, indicating that GH and GI were inversely related to GY. This relationship is exemplified by the contrasting performance of Saji (GH: 164.34, GY: 3.88) and Cross Sabalan (GH: 76.28, GY: 4.65). Notably, yield-associated traits, including GY, TGW, and NGS, exhibited positive clustering, with Cross Sabalan achieving the highest overall yield performance (GY: 4.65, TGW: 40.73). In contrast, structural traits such as PH and SL demonstrated minimal correlations with yield outcomes (Fig. 3A).

Fig. 3.

Hierarchically clustered heatmap of wheat genotypes under (A) well-irrigated and (B) drought stressed condition across morphological traits, yield, and parameters related to bread grain quality. Genotypes (rows) and traits (columns) are arranged by hierarchical clustering using Euclidean distance with complete linkage. The color scale represents standardized trait values (z-scores), where blue indicates values below the trait mean, white represents values near the mean, and red indicates values above the trait mean. Traits include: PH (plant height), TGW (thousand grain weight), SL (spike length), GY (grain yield), NSS (number of spikelet per spike), and GWS (grain weight per spike), gluten index, grain hardness. Dendrograms show similarity relationships between genotypes and trait associations

Under DS, hierarchical clustering delineated three distinct groups. Group 1 comprised high-yielding genotypes such as Danesh and Shiraz. Group 2 represented balanced performers exemplified by Azar 2, which showed moderate yield combined with a TGW of 41.20 g. Group 3 included stress-resilient genotypes characterized by high gluten stability, notably Satin with a GI of 58.33 and Line 181 at 52.25. Trait correlations revealed a significant trade-off between yield and gluten strength; high-yielding genotypes like Danesh (GI: 16.77) showed weaker gluten compared to stress-resilient types like Satin. Synergistic relationships were observed between PH and TGW, as demonstrated by Azar 2 with a height of 77.05 cm and a TGW of 41.20 g (Fig. 3B).

Association analysis between traits

The PCA of 19 wheat genotypes under WI conditions revealed that the first two dimensions accounted for 60.3% of the total variability, with Dimension 1 (Dim1) explaining 40.7% and Dimension 2 (Dim2) accounting for 19.6%. Dim1 was primarily influenced by spike architecture traits, including the GWS and the NSS, as well as SL and the NGS. Collectively, these traits contributed more than 40% of the observed variation. In contrast, Dim2 was predominantly shaped by PH and grain quality parameters, with GH contributing 18.3% and the GI contributing 18.2%. The biplot analysis illustrated a distinct genetic trade-off: high-yielding genotypes, such as Cross Sabalan and Line 181, clustered positively along Dim1, while genotypes with a focus on quality, including Ghabus and Satin, were more prominent along Dim2. Notably, PH and SL exhibited minimal contributions to the overall variation, each accounting for less than 8%, thereby indicating their limited role in the structural variation of the genotypes (Fig. 4A).

Fig. 4.

Principal Component Analysis (PCA) biplot illustrating the relationships among genotypes under two distinct conditions: (A) well-irrigated and (B) drought-stressed. The dimensions, Dim1 and Dim2, represent the genotypes, which are identified by numbered points (1–19) corresponding to the following varieties: M-92-20, WS-90-18, Aflak, Azar 2, Aftab, Baharan, Tiger, Danesh, Rizhav, Satin, Saji, Sirvan, Shiraz, Ghabus, Cross Sabalan, Karim, Line 181, Natasha, and Yavaros. The biplot also includes vectors representing key agronomic variables: plant height (PH), thousand grain weight (TGW), spike length (SL), grain yield (GY), number of spikelet per spike (NSS), grain weight per spike (GWS), gluten index, and grain hardness

The PCA of 19 wheat genotypes under DS conditions revealed that the first two dimensions collectively explained 66.1% of the total variability, with Dim1 accounting for 42.9% and Dim2 contributing 23.2%. Dim1 was predominantly associated with NSS, NGS, GY, GWS, and SL. Dim2 was strongly influenced by grain quality traits, with TGW contributing most significantly, followed by PH, GH and SL. The biplot illustrated a clear genetic divergence: high-yielding genotypes (e.g., Cross Sabalan, Line 181) aligned positively along Dim1, while quality-oriented genotypes like Satin and Shiraz were prominent along Dim2 (Fig. 4B).

Integrated GGE biplot and ranking analysis of GY, GH, and GI across four test environments (E1–E4)

Integrated analysis of the ranking data (Supplementary Table S19), GGE biplot panels (Fig. 5), and environment coordinates for GY, GH, and GI revealed the performance and stability of 19 genotypes across four test environments (E1–E4). For GY, Danesh exhibited the highest mean yield (6.10 g) but poor stability (rank 17) and consequently low desirability (3.128), whereas WS‑90‑18 was the most stable (rank 1) with moderate yield and a desirability of 3.558 (Supplementary Table S19). The best desirability for GY was achieved by Shiraz (4.122), combining high mean (4.83 g) and good stability (rank 4). In the GY biplot (PC1 58%, PC2 36.2%), environments E2 and E4 were closely associated with high‑yielding genotypes such as Shiraz, while E1 and E3 were less discriminative (Fig. 5A). For GH, Saji had the highest mean hardness (166.67%) but only moderate stability (rank 11) and a desirability of 142.24, while Ghabus was the most stable (rank 1) with lower hardness (108.04%) and a desirability of 107.77. The top desirability for GH was observed in Yavaros (144.66), which ranked second in mean (148.15%) and fourth in stability. The GH biplot (PC1 59.3%, PC2 26.4%) showed that E2 and E3 were highly representative, with Saji and Yavaros positioned near these environment vectors (Fig. 5B), indicating specific adaptation; E1 and E4 were more distant. For GI, Satin exhibited the highest mean (43.13%) and also the highest desirability (33.306), despite only moderate stability (rank 12). Conversely, Natasha was the most stable (rank 1) with a desirability of 11.547. In the GI biplot (PC1 76%, PC2 21.1%), E4 and E2 were the most discriminating environments, favoring Satin and Shiraz; E1 and E3 were closer to the origin, suggesting weaker interaction (Fig. 5C).

Fig. 5.

Genotype and genotype-by-environment (GGE) biplots for grain yield (GY), grain hardness (GH), and gluten index (GI) of 19 wheat genotypes tested across four environments (E1–E4); E1 (first year, well‑irrigated), E2 (first year, drought stress), E3 (second year, well‑irrigated), and E4 (second year, drought stress). (A) GY biplot, (B) GH biplot, and (C) GI biplot Genotypes are shown as labeled points; environments are shown as labeled vectors (E1–E4). The dashed circle (unit circle) indicates the ideal environment or genotype position. Genotypes in the top 20% (based on desirability) are highlighted in black; other genotypes are shown in grey

Ranking of genotypes based on MGIDI

Under WI conditions, MGIDI analysis ranked genotypes by multi-trait performance (yield, morphology, grain quality). It identified Danesh (G08, MGIDI: 1.0), Line 181 (G17: 1.2), and Shiraz (G13: 1.8) as top selections (lowest MGIDI scores) using a cut-off of 2.5. These genotypes aligned well with the ideotype, mainly driven by FA1 (yield traits: NSS, SL, GWS, GY). They achieved significant gains in GY and GWS, while FA2 (grain quality) showed a trade‑off with reduced GH. The strengths assessment confirmed optimal adaptation to yield-related traits, positioning G08 as the most balanced genotype (Fig. 6A, B).

Fig. 6.

(A) The ranking of wheat genotypes based on the MGIDI index, which is derived from morphological traits, yield, and parameters related to bread grain quality, is presented in ascending order across two years under well-irrigated conditions. The selected lines, identified based on this index, are highlighted in red. The central red circle indicates the cut-off point established according to the selection pressure applied. Furthermore, a comprehensive assessment of the strengths and weaknesses of the selected wheat genotypes over the two years is illustrated. This is depicted in well-irrigated conditions (B), showcasing the proportion of each factor (FA) contributing to the computed MGIDI index. A smaller proportion attributed to a factor (closer to the outer edge) signifies that the traits within that factor are more aligned with the ideotype. Panel C shows the MGIDI-based ranking under drought stress, while panel D illustrates the corresponding strengths and weaknesses analysis for the selected genotypes under drought stress over the two years. The dashed line represents the theoretical value, assuming equal contribution from all factors. The genotypes are coded as follows: G01: M-92-20, G02: WS-90-18, G03: Aflak, G04: Azar 2, G05: Aftab, G06: Baharan, G07: Tiger, G08: Danesh, G09: Rizhav, G10: Satin, G11: Saji, G12: Sirvan, G13: Shiraz, G14: Ghabus, G15: Cross Sabalan, G16: Karim, G17: Line 181, G18: Natasha, G19: Yavaros

Under DS, MGIDI analysis identified Natasha (G18), Line 181 (G17), and Shiraz (G13) as top genotypes (lowest MGIDI scores, cut‑off ~ 2.5). Three factors explained 81.3% of the variability: FA1 (yield‑spike traits: SL, GY), FA2 (grain quality: GH), and FA3 (gluten quality: GI). Selected genotypes achieved significant gains in yield traits (NSS and SL) and GI but showed increased GH, highlighting a stress-induced trade-off. The strengths assessment revealed that G18 and G13 excelled in FA1 (yield adaptation), while G17 prioritized FA3 (gluten quality) (Fig. 6C, D).

The MGIDI selection outcomes are summarized in Table 6. The SG (%) which is environment-specific derived from the phenotypic data of each irrigation regime independently and reflect the direct selection differential for that water condition. Under WI conditions, selection increased all yield‑related traits (NSS, SL, NGS, GWS, GY) with SGs ranging from 1.28% (NGS) to 12.4% (GWS) and 10.3% (GY). GH and GI were reduced (‑8.71% and ‑1.01%, respectively), aligning with the ideotype goal of minimizing GH and GI under WI. Under DS conditions, selection increased NSS (13.8%), SL (14.4%), and GY (6.13%). Notably, GH increased by 16.1% and GI by 60.0%, consistent with the ideotype goals of maximizing both traits under DS. These gains were primarily associated with FA1 (yield traits) under both regimes, while quality traits were linked to FA2 under WI and to FA2 (GH) and FA3 (GI) under DS (Table 6).

Table 6.

Linkage of factors to traits, mean of the original population (Xo), mean of the selected genotype (Xs), selection differential (SD), selection differential percentage (SD%), selection gains (SG), selection gain percentage (SG%), objectives, and goal based on the multi-trait genotype–ideotype distance index (MGIDI) index for morphological traits, yield, and bread grain quality related parameters in wheat genotypes across two years in well-irrigated and in drought stress conditions

Mean performance
Trait	Factor	Xo	Xs	SD	SD%	SG	SG%	Objective	Goal
Well-irrigated
NSS	FA1	14.60	16.30	1.67	11.40	1.11	7.61	Increase	100
SL	FA1	7.26	8.36	1.10	15.20	0.75	10.40	Increase	100
NGS	FA1	32	33.40	1.37	4.27	0.40	1.28	Increase	100
GWS	FA1	1.12	1.34	0.22	19.90	0.13	12.40	Increase	100
GY	FA1	3.63	4.20	0.58	16	0.37	10.30	Increase	100
PH	FA2	63.70	64.40	0.73	1.15	0.30	0.47	Increase	100
Grain hardness	FA2	121	107	-14.20	-11.70	-10.60	-8.71	Increase	0
Gluten index	FA2	13.90	13.80	-0.14	-1.03	-0.14	-1.01	Increase	0
Drought-stress
NSS	FA1	14.50	16.80	2.28	15.70	2	13.80	Increase	100
SL	FA1	7.43	8.61	1.18	15.90	1.07	14.40	Increase	100
GY	FA1	3.98	4.44	0.45	11.50	0.24	6.13	Increase	100
PH	FA2	65.70	67.10	1.38	2.10	0.61	0.93	Increase	100
NGS	FA2	30.30	32.20	1.84	6.06	0.67	2.21	Increase	100
GWS	FA2	1.32	1.34	0.02	1.81	0.004	0.36	Increase	100
Grain hardness	FA2	110	130	19.80	17.90	17.70	16.10	Increase	100
Gluten index	FA3	15.60	25.3	9.71	62.20	9.37	60	Increase	100

NSS Number of spikelets per spike, SL Spike length, NGS Number of grains per spike, GWS Grain weight per spike, GY Grain yield per plant, PH Plant height

Discussion

The results of this study indicated that drought stress at the reproductive stage of wheat growth imposed inconsistent effects on agronomic and BMQ traits. The GGE biplot and desirability analysis confirmed that no single genotype consistently combined high mean performance with high stability across GY, GH, and GI, underscoring the trait-specific nature of drought adaptation. This inconsistency arises from the differential sensitivity of physiological and biochemical processes to water limitation. GY is highly sensitive to water deficit during grain filling due to reduced photosynthesis, impaired assimilate partitioning, and accelerated leaf senescence [62]. In contrast, protein content and gluten strength often increase through nitrogen remobilization from vegetative tissues to the grain, combined with a concentration effect caused by reduced starch accumulation under drought [63]. The differential accumulation of proteins under drought stress is governed by specific biochemical pathways. Water deficit during grain filling downregulates starch synthesis enzymes such as ADP‑glucose pyrophosphorylase and soluble starch synthase, while simultaneously upregulating genes involved in storage protein synthesis, including glutenins and gliadins [11, 16, 26, 66]. This shift in carbon and nitrogen partitioning is mediated by abscisic acid (ABA) signaling, which promotes the remobilization of vegetative nitrogen reserves to the grain in the form of amino acids, particularly glutamine and proline [1, 14, 77]. Consequently, the observed increase in protein content and gluten strength under drought stress results from both a concentration effect (due to reduced starch accumulation) and active nitrogen remobilization, whereas the inconsistent responses among genotypes reflect allelic variation in the efficiency of these remobilization pathways and in the heat‑shock protein networks that protect glutenin macropolymer assembly [32, 65]. Furthermore, drought stress alters the balance between carbon and nitrogen metabolism, favoring the synthesis of storage proteins (glutenins and gliadins) over carbohydrates in some genotypes, while others exhibit disrupted protein assembly due to heat shock protein inhibition [32, 65]. Moreover, genotype‑specific adaptive responses [11, 32] and seasonal variations in temperature and rainfall modulate trait expression, leading to strong genotype‑by‑environment interactions (GEI) [68, 74]. Together, these opposing responses at the physiological, metabolic, and genetic levels explain the observed inconsistencies across genotypes and environments.

Our results demonstrated that there is potential to breed for high baking quality under DS conditions and there are trade-offs between breeding for agronomic performance and breeding for end use baking quality in wheat. The GY biplot revealed that environments E2 and E4 (DS conditions) were most discriminating for GY, indicating that water limitation amplifies genetic differences in yield potential and provides a more effective selection environment than WI conditions. Unravelling the challenges related to the interrelationship of agronomic and grain quality related traits is important in breeding programs aimed at genetic improvement of crop varieties [44]. Efforts for the simultaneous improvement of GY and bread quality in wheat depend on access to ample genetic diversity under variable environmental conditions. Drought reduced GY but increased several BMQ traits. It has been shown that when compared with grain quality related traits, GY is more sensitive to drought at the post-anthesis stage than in the early-stage of growth [62]. In agreement with the results of our study, drought stress during the reproductive stages of growth increased protein accumulation and reduced starch content in wheat grains in the study of Ullah et al. [63] study. Furthermore, protein accumulation affects GH and starch damage, which in turn affects milling quality and resistance to milling in GH [64–66]. The GH biplot showed that environments E2 and E3 were highly representative, indicating that selection for hardness can be reliably performed in such environments. The close association of specific genotypes with these environmental vectors points to local adaptation rather than broad stability, a biologically meaningful pattern for breeders targeting specific production zones. Our findings on the inconsistent responses of genotypes across irrigation regimes and seasons highlight the complex genetic control of both GI and GY. As shown by Line 181, which showed the highest increase in GY under drought while also maintaining a high GI, the negative correlation between yield and bread quality traits is not absolute. This suggests that such trade-offs can be overcome by identifying and pyramiding specific adaptive alleles. However, for most other genotypes, GI values remained inconsistent across environments, likely due to G×E interactions. GI is a complex trait shaped by both genetic factors and environmental conditions, particularly water availability and temperature during grain filling. Under DS, some genotypes may enhance gluten strength adaptively (e.g., through increased glutenin synthesis), whereas others may exhibit reduced gluten quality because of reduced protein accumulation. Moreover, seasonal variations such as differences in temperature and rainfall between the two growing seasons can further modulate the expression of gluten‑related genes. Consequently, genotypes that remain stable across environments are rare, and the observed inconsistency highlights the differential sensitivity of gluten index to environmental fluctuations [11, 32]. The GI biplot demonstrated that the most discriminating environments were again the drought‑stressed ones (E4 and E2). This indicates that water limitation helps to identify genetic differences in gluten strength, making these environments ideal for selecting superior bread‑making quality under water‑limited scenarios. The desirability index effectively identified genotypes that balance high performance and reliability.

As expected, the higher variation of GH, protein, ZSV, dough water absorption and development time among varieties in DS compared with the WI conditions suggested the importance of interactions between irrigation water treatments and the traits. Several varieties (i.e. Danesh, Ghabus, Saji, Sirvan, Satin, Shiraz, Tiger) of our study had high values for traits under both the WI and DS conditions. These data suggest the selection for high yielding genotypes under drought stress is possible. The range of flour water absorption in our study was larger (53.9–64.3%) than the range (61.76–68.4%) identified in the study of Aydoğan et al. [67]. Water absorption, which is influenced by gluten content and starch, is correlated with dough stability [67]. In the study of Kaszuba et al. [30], the results indicated that protein and wet gluten content, flour water absorption and crumb porosity are reliable parameters for quality assessment of wheat grains and that these parameters are recommended for breeders, bakers and cultivation testing authorities. In a study of a core collection of wheat varieties with different origins in Mediterranean regions, grain related quality traits showed no specific relation pattern with environmental/climate changes [68]. Although our wheat varieties showed large variation for GY and BMQ parameters, no single variety had all desirable quality-related parameters, which complicates selection for the genetic improvement of bread quality. Comparison of wheat varieties for dough/flour quality based on a single parameter is challenging as selection based on one trait may lead to the loss of favorable genes for other quality related traits. These results demonstrated that a more comprehensive metric is required for baking quality assessment in wheat. The absence of a single superior genotype across all quality parameters underscores the necessity for a balanced breeding strategy. The use of multivariate selection indices such as MGIDI, which can integrate multiple weighted traits, is therefore crucial for enhancing both yield and complex quality profiles simultaneously.

Our study identified high genetic variation and high heritability for several bread‑making and agronomic traits. Therefore, selecting varieties with superior quality traits and high GY from crosses between extreme genotypes will result in high genetic gain for each target trait and facilitate the development of new varieties. Heritability, which is defined as the proportion of genetic components in phenotypic variation is often used to quantify the precision of individual selection for the improvement of a trait. The results revealed that ZSV, GI and NSS had high heritability in both irrigation regimes, demonstrating the meaningful contribution of genetic components and the low effect of environmental variance in the phenotype of these traits. The GCV values and heritability estimates were similar to those identified for BMQ parameters by Rao et al. [34] in wheat. Our results showed that grain quality parameters had higher heritability than agronomic traits, which indicates the resilience of quality related traits against the changing environmental conditions between the two irrigation regimes The results of Ul-Haq et al. [69] indicated high broad-sense heritability for gluten index and protein content in wheat but these traits declined with increasing grain number per spike. The high heritability for key quality traits (ZSV, GI) under both water status conditions is particularly positive for breeders. It suggests that genetic gains for these traits can be achieved consistently through selection, even in unpredictable drought-prone environments, provided sufficient genetic diversity is present in the breeding population.

Analysis of simple correlations indicated positive but non-significant correlations between GY and several quality related parameters in our study. Other studies have shown inconsistent correlations between agronomic/growth traits and BMQ parameter in wheat [38, 69]. The CCA in our study demonstrated inconsistent interrelationships of agronomic and BMQ parameters in response to two irrigation regimes. In the WI treatment, GH and protein index showed direct (positive) relationships with GY components and PH but indirect relations with ZSV were observed. However, in the DS treatment of our study, GY components indicated direct relations with protein and ZSV and indirect associations with GH and GI. The shift from a yield‑hardness association under irrigation to a yield‑protein association under drought, as revealed by CCA, aligns with the GGE biplot observation that environment‑specific selection is necessary. This shows that the same genotype may present different response depending on which trait–environment combination is targeted. Groos et al. [70] demonstrated that bread-making scores are poorly correlated among environments, as they are poorly predicted by multiple regression on dough rheology parameters and by protein content. An evaluation of 50 years breeding for bread quality in UK has shown that continuous selection for decreased protein content has been compensated for by increased gluten quality for the improvement of BMQ in wheat [71]. The traits protein content and ZSV, having positive relation with yield components and high heritability can be utilized for indirect selection and genetic improvement of the end-use quality. This inconsistency in the interrelationships of the two sets of agronomic and quality related traits under the two irrigation regimes suggests the importance of including trait by environment interactions and considering a trade-off between traits in decisions for the improvement of both GY and BMQ as end use property in wheat. The GGE biplots further support this conclusion by visually mapping which environments favor which trait combinations, thereby enabling breeders to match genotypes to target production zones. The shift from a yield-hardness association under irrigation to a yield-protein association under drought, as revealed by CCA, is a critical finding. It indicates that the physiological and genetic pathways linking yield and bread quality traits are fundamentally altered by drought stress. This environmental change of trait relationships means that selection strategies must be specific; ideotypes for irrigated versus rainfed systems will likely differ in their optimal trait combinations.

Hierarchical clustering under WI conditions revealed three genotypic groups. A fundamental trade‑off emerged: high‑yield genotypes (Cross Sabalan with GY of 4.65 and GH of 76.28) showed reduced GH and GI. In contrast, quality‑focused genotypes (e.g., Satin) reduced yield. This demonstrates the “dilution effect” driven by carbon partitioning toward starch over storage proteins [68, 72].From a biochemical perspective, minimizing GH under well-irrigated conditions is associated with reduced grain glassiness and lower levels of friabilin proteins, particularly puroindolines, which are surface-active proteins that weaken the starch-protein interface [20, 28]. The observed reduction in GH (− 8.71%) under WI reflects a metabolic priority toward starch biosynthesis via increased activity of granule-bound starch synthase, which dilutes the relative proportion of hard endosperm proteins [22, 25]. This trade-off is biologically grounded, when water is abundant, carbon flux is directed toward carbohydrate storage to maximize yield, whereas nitrogen flux to sulfur-rich glutenin polymers is constrained by competition from starch synthesis [6, 76]. The GGE biplot patterns corroborate these clusters: high‑yielding genotypes under WI were associated with E2 and E4 (drought environments), while quality‑focused genotypes aligned with specific environment vectors, confirming that the trade‑off is environment‑dependent. Under DS, this trade-off intensified: high-yielding genotypes (e.g., Danesh: GY 6.40) showed severely compromised gluten strength (GI: 16.77), whereas stress-resilient genotypes (e.g., Satin: GI 58.33; Line 181: 52.25) maintained quality at the expense of yield, reflecting prioritization of nitrogen remobilization and osmotic adjustment prioritization [73]. The GH biplot further highlighted this pattern, as high‑hardness genotypes under stress were tightly linked to E2 and E3, indicating that these environments are ideal for selecting stress‑tolerant, high‑quality genotypes. Drought also reshaped trait relationships, generating a synergistic link between PH and TGW (e.g., Azar 2: height 77.05 cm, TGW 41.20 g), suggesting taller phenotypes enhance hydraulic conductivity and photoassimilate supply under water scarcity [68]. The positive correlation between PH and TGW under drought reflects a biochemical and hydraulic mechanism. Taller phenotypes maintain greater xylem conductance and hydraulic lift capacity, which sustains photosynthate transport to developing grains under water‑limited conditions [74, 75]. This is supported by higher activities of sucrose synthase and invertase in the grain sink of taller genotypes, enabling continued starch biosynthesis despite reduced overall water availability [64, 66]. Furthermore, taller plants often exhibit enhanced osmotic adjustment through accumulation of compatible solutes such as proline, glycine betaine, and trehalose, which protect enzyme function and maintain turgor pressure during grain filling [1, 14, 77]. The genotype Azar 2 exemplifies this mechanism, combining above‑average plant height with high TGW under drought, suggesting that breeding for moderate height with efficient osmotic adjustment may partially decouple the traditional trade‑off between height and yield stability. Critically, Line 181 emerged as a uniquely adaptable genotype, clustering with high-yield types under irrigation yet aligning with quality specialists under drought, indicating pleiotropic adaptations that merit genomic interrogation for climate-resilient breeding [74, 75]. This unique behavior of Line 181 was also evident in the GGE biplots, as it is located in the central position of the GY biplot (moderate yield and stability) whilst shifted toward the high‑performance sector in the GI biplot under drought‑favoring environments. Such cross‑trait adaptability is rare and valuable for breeding for multiple environments. The identification of distinct stable clusters under each irrigation regime provides a practical basis for parental selection. Breeders targeting drought resilience might focus on crossing Group 4 (high yield) with Group 3 (high quality) to introgress favorable alleles, while the stability of Group 3 for GI makes it a valuable donor for quality traits.

The MGIDI analysis revealed strongly contrasting selection priorities across environments. Under WI, top genotypes (Danesh, MGIDI 1.0; Line 181, 1.2) achieved significant yield gains (GY + 10.3%, GWS + 12.4%) but reduced GH (-8.71%). This demonstrates successful dilution‑effect management, where carbon flux favors starch over storage proteins [76]. The contrasting ideotype for GH and GI between WI and DS conditions is biologically reasonable. Under WI, minimizing GH (‑8.71%) and GI (‑1.01%) reflects the fact that softer grains reduce milling costs and improve flour uniformity [20, 22]. The small reduction in GI was acceptable because the selected WI genotypes already had adequate gluten strength. Under DS, maximizing hardness (+ 16.1%) and GI (+ 60.0%) is advantageous: harder grains correlate with higher protein and stronger gluten [11, 32, 64], becoming critical for bread quality when yield is compromised [62]. This environment‑dependent shift confirms that MGIDI can effectively adapt selection criteria to target environments. The GGE biplot results strongly support this shift: under DS, as the most discriminating environments for GI were E4 and E2. These environments are precisely the environments where MGIDI selected for high GI. Thus, the biplot visually validates the MGIDI‑derived selection strategy. The biochemical basis for increased GI and GH under drought resides in stress‑induced changes in protein polymerization and disulfide bond formation. Drought stress enhances the expression of protein disulfide isomerase and other chaperone proteins that facilitate the formation of high‑molecular‑weight glutenin subunits (HMW‑GS) into stable glutenin macropolymers [11, 18, 31]. Concurrently, water limitation increases the proportion of ω‑gliadins and γ‑gliadins relative to α/β‑gliadins, which alters dough extensibility and resistance [32, 58]. The increase in GH under drought is linked to higher grain protein content and greater endosperm vitreosity, as harder grains result from tighter packing of starch granules within a continuous protein matrix [20, 28]. This stress‑induced hardening is a drought-adaptive property because harder grains resist mechanical damage during milling and produce flours with superior water absorption, which is critical for dough stability under suboptimal processing conditions [12, 29].

Under drought, MGIDI selection inverted quality objectives. Top performers (Natasha, Line 181, Shiraz) showed exceptional GI gains (+ 60.0%) despite increased hardness (+ 16.1%). This shows stress‑induced reallocation of nitrogen reserves to gluten polymers and compensatory mechanisms that maintain end‑use quality at the expense of yield [77, 78]. Line 181’s dual prominence (ranked 2 irrigated, 3 drought) signifies unique genomic adaptability. It excels in yield architecture (FA1) under optimal conditions while prioritizing gluten stability (FA3) under stress. This pleiotropic adaptation may be linked to TaGS2-1 A allele expression, which modulates spike fertility and glutenin synthesis [79]. The consistency between MGIDI rankings and the GGE‑derived desirability ranks validates the reliability of both approaches. The desirability index, being a simple linear function, offers a clear alternative to more complex indices without sacrificing predictive value. The 81.3% variability captured by three drought factors (vs. two in irrigation) underscores environmental modulation of trait hierarchies, where FA1 (yield-spike) and FA3 (gluten quality) become decoupled under stress, necessitating distinct ideotypes [80]. These results compel breeding programs to adopt environment-specific weightings: minimizing hardness in irrigated systems while strategically tolerating its increase in drought-prone regions where gluten strength ensures marketability [81]. The application of the MGIDI index effectively quantified the trade‑offs and revealed genotype-specific strengths, providing a clear plan for selection. The fact that different trait factors explained variation in different environments powerfully suggests against a generic breeding objective. Future work should involve applying differential weighting strategies within the MGIDI framework to represent the economic or processing priority of traits (i.e. higher weight for gluten index in premium bread wheat) in specific target environments.

Conclusions

The results of this study indicate that no single variety exhibited all desirable BMQ traits and GY under both irrigation regimes. Adaptation to drought was trait‑specific, high‑yielding genotypes were typically unstable with poor GH or GI, while stable genotypes rarely achieved the highest means. Therefore, selection must rely on indices that integrate both mean performance and stability. High GCV and heritability values for ZSV and GI under both irrigation conditions suggest a strong genetic influence, indicating that selection for these traits could lead to significant genetic improvement in wheat quality. Drought‑stressed environments (E2 and E4) consistently discriminated genotypes for GY and GI, while well‑irrigated environments (E1 and E3) contributed little additional information. Therefore, breeders should give priority to drought stressed environments when screening early‑generation material for drought tolerance and high quality. Selection weights must be optimized to the environment. In well-irrigated systems, grain hardness should be minimized, whereas in drought-prone regions, higher hardness can be tolerated because it is associated with the gluten strength required for market acceptance. Our results indicated that terminal drought established the inherent trade-off between grain yield and bread-making quality in wheat, driven by competing resource allocation to starch versus storage proteins. MGIDI‑based ideotype selection navigated this conflict effectively. It identified environment‑optimized genotypes: under WI, Danesh and Line 181 maximized yield gains while reducing GH; under DS, Natasha, Line 181, and Shiraz achieved exceptional GI improvement despite increased hardness. Line 181 exhibited unique adaptive plasticity: under well-irrigated conditions it performed in a yield-oriented mode, but under drought it switched to a quality-focused mode. This dual behavior, validated by both MGIDI and GGE biplots, makes Line 181 a priority genetic resource for climate-resilient breeding. Line 181’s dual prominence underscores its unique adaptive plasticity, thriving in yield (FA1) or quality (FA3) modes across environments. For breeding programs, we advocate divergent selection weightings: hardness minimization in high-input systems versus strategic acceptance in drought-affected regions where gluten strength is paramount. The combined use of GGE biplots and the desirability index provides a strong framework for selecting superior genotypes across both irrigated and drought‑stressed conditions. Overall, our study demonstrated three main points: (1) a more comprehensive metric is needed for BMQ assessments; (2) available potential to breed for high baking quality under DS conditions; and (3) a trade‑off between breeding for agronomic performance and end‑use BMQ as these two sets of traits showed inconsistent relationships.

Abbreviations

DS

Drought stress

WI

Well irrigated

PH

Plant height

SL

Spike length

NSS

Spikelet number per spike

NGS

Grain number per spike

GWS

Grain weight per spike

TGW

Thousand grain weight

GY

Grain yield

GH

Grain hardness

GI

Gluten index

ZSV

Zeleny sedimentation value

MGIDI

Multi-trait genotype–ideotype distance index

GGE

Genotype and genotype by environment

PCA

Principal component analysis

CCA

Canonical correlation analysis

c-ANOVA

combined analysis of variance

EMS

Expected mean squares

RCBD

Randomized complete block design

h²_bs

heritability in broad sense

PCV

Phenotypic coefficient of variation

GCV

Genetic coefficient of variation

CRD

Completely randomized design

FA

Factor analysis

AACC

American Association of Cereal Chemists

GEI

Genotype × environment interaction

SG

Selection gains

BMQ

Bread making quality

E1

Environment 1 (first year well irrigated)

E2

Environment 2 (first year drought stress)

E3

Environment 3 (second year well irrigated)

E4

Environment 4 (second year drought stress)

U1

First agronomic canonical variable

U2

Second agronomic canonical variable

V1

First quality-related canonical variable

V2

Second quality-related canonical variable

Dim1

First dimension from PCA

Dim2

Second dimension from PCA

FA1

Factor 1 from factor analysis

FA2

Factor 2 from factor analysis

FA3

Factor 3 from factor analysis

Authors’ contributions

MK: performed the experiments, prepared samples, and collected the data, MF: contributed to bread making quality measurements and editing the final draft of the manuscript, TMM: wrote the initial draft of the manuscript, CR: reviewed the final draft of the manuscript, MS: involved in data analysis and reviewed the initial draft, MZ: wrote the initial draft of the manuscript, HP: supervised the research work, RN: contributed to the field works and managements, AD: advisor of the work and edited the text, AM and TH: helped in research work and edited the text of the manuscript, BH: conceptualized the research, supervised the work, contributed to writing and reviewing the initial and final draft of the manuscript. All authors read and approved the final draft of the manuscript.

Funding

Not applicable.

Data availability

Data will be made available on request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.