Genetic dissection and stability analysis of grain protein content and thousand-grain weight in emmer wheat (Triticum turgidum subsp. dicoccum) germplasm

Abstract

Emmer wheat (Triticum turgidum subsp. dicoccum), an allotetraploid species of wheat, is a rich source of nutritional as well as biotic and abiotic stress tolerance traits. It is an important genetic resource for improving grain protein content in durum and bread wheat, yet its genetic basis for protein accumulation and yield-related traits are less studied. In this study, a diverse panel of 150 emmer wheat germplasm accessions conserved in national genebank of India was evaluated for grain protein content (GPC) and thousand grain weight (TGW) across four environments (New Delhi, Pune, Karnal, and Ludhiana) during winter season 2021–2022 using an alpha lattice design. Genotypes were highly significant for GPC and TGW and substantial phenotypic variation (12.16–17.76% for GPC and 25.98 to 38.64 g for TGW) was observed based on pooled data. GPC and TGW were negatively correlated, with principal component analysis revealing opposing contributions of these traits to total phenotypic variation. We identified superior donors for GPC and TGW based on multi-environmental evaluation. Accessions EC578064, EC299074, EC577402, IC534016, and IC32513, exhibited consistently high GPC with moderate TGW across environments, while IC138455 and IC32513 were stable genotypes for GPC. Genome-wide association analysis using 6800 polymorphic SNP markers based on the 35 K Axiom Wheat Breeders’ array identified 21 significant marker–trait associations (MTAs) across environments, out of these eight markers were located on chromosome 5B. Nine stable and novel MTAs were identified, which consistently expressed across locations, highlighting genomic regions with less genotype × environment interaction. These stable MTAs were linked to genes involved in nitrate transport, stress tolerance, grain development, and yield regulation, providing biologically meaningful targets for protein improvement in emmer wheat. Overall, this study resulted in the identification of promising donor genotypes for GPC and TGW, and linked markers and candidate genes for possible deployment into breeding program for the development of high yielding and protein rich wheat cultivars.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-49519-7.

Introduction

Wheat is a staple crop globally, nourishing a major part of the population and serving as a critical component of agriculture. Among different cultivated species, emmer wheat (Triticum turgidum subsp. dicoccum) stands out due to its resilience to harsh environmental conditions and nutrient richness¹. Triticum turgidum subsp. dicoccum (emmer or khapli wheat) is the domesticated form of Triticum dicoccoides, wild emmer, an allotetraploid wheat species. Currently, emmer wheat is only grown in 1% of the world’s total wheat area. Domesticated approximately 10,000–11,000 years ago in the Fertile Crescent, emmer wheat spread globally, serving as a vital source of carbohydrates, proteins, and essential nutrients in human diets^2,3. Its nutritional value, tolerance to adverse climatic conditions, and disease resistance have highlighted its significance as a genetic resource for enhancing food security and nutrition worldwide⁴. Emmer wheat holds promise for modern breeding efforts due to its high grain protein content (GPC), reaching approximately 17.1%^5,6, and its unique gluten composition potentially benefiting individuals with gluten sensitivities.

Biofortification, which enhances crops’ nutritional content through agronomic and genetic approaches, offers a sustainable solution for addressing hidden hunger and achieving the United Nations Sustainable Development Goals⁷. Biofortification is particularly essential for India, a major wheat producer grappling with nutritional deficiencies in staple grains. By leveraging crops like emmer wheat, biofortification strategies aim to improve nutritional security through the incorporation of traits such as high GPC, stress tolerance, and disease resistance into breeding programs. Emmer wheat exhibits substantial variability in grain quality traits, which has been explored in various studies. GPC in emmer wheat has been reported to be significantly higher than that of durum and bread wheat, with a mean value of 16.9%^5,8. Although the yield of emmer wheat is generally lower than that of modern wheat varieties⁸, its nutritional advantages, including higher protein levels and stronger gluten, are important factors for consideration in food baskets and breeding program. For instance, Italian emmer wheat lines have demonstrated exceptional grain protein content and thousand-grain weight⁹. Also, crosses between durum and emmer wheat cultivars have shown improvements in GPC¹⁰.

An efficient assessment of genotype × environment (G×E) interaction is crucial for evaluating the yield potential and stability of genotypes in multi-location trials¹¹. Traditionally, regression-based techniques have been employed to analyse G×E interactions by describing the linear responses of genotypes across varying environments¹². However, these methods often fall short in capturing the complex, non-linear behaviour of genotypic responses. In this context, the Additive Main Effects and Multiplicative Interaction (AMMI) model has emerged as a powerful analytical tool. It integrates additive (main effects) and multiplicative (interaction effects) components, offering a more comprehensive understanding of G×E interactions^13,14. 

Moreover, genotypic performance prediction improves when the random nature of G×E interaction is considered, as emphasised by Piepho et al.¹⁵. This underscores the importance of models that incorporate both fixed and random components to enhance the accuracy of genotype selection.

Breeding wheat varieties with high yields and superior grain quality depends upon mining germplasm for trait discovery using integration of conventional and genomic tools¹⁶. Identifying genes and quantitative trait loci (QTLs) associated with GPC and thousand kernel weight (TKW) in common wheat (Triticum aestivum L.) is crucial for enhancing both yield and nutritional quality. Recent studies have pinpointed several significant QTLs on the A and B sub-genomes that influence these traits. A genome-wide association study (GWAS) involving 280 bread wheat genotypes identified 16 marker-trait associations (MTAs) for GPC and 4 for TKW across the A, B, and D sub-genomes, with the B sub-genome harbouring the highest number of MTAs (10), followed by the A and D sub-genomes (8 each). Key MTAs for GPC were located on chromosomes 1 A, 1B, 4B, 5 A, and 6 A, while significant MTAs for TKW were found on chromosomes 2B and 3 A, with candidate genes such as the NAC domain superfamily, zinc finger RING-H2-type, and serine/threonine-protein kinase LRK10 playing roles in stress tolerance, nutrient remobilisation, and grain development¹⁷. Another study utilising high-density SNP markers identified 12 QTLs for TKW on chromosomes 2 A, 4B, 6 A, and others, with major QTLs like QTKW.iari-4B.1 and QTKW.iari-3B explaining phenotypic variances of 17.53% and 10.18%, respectively, while the QTKW.iari-3B region contains the candidate gene TraesCS3B02G295700, encoding an alpha/beta hydrolase fold protein. Additionally, a stable QTL, QTKW.iari-4D.1, identified on chromosome 4D, explained 12.69% of phenotypic variance and contained the candidate gene TraesCS4D02G039500, encoding an oxysterol-binding protein¹⁸. Further research on a durum × wild emmer wheat mapping population across five environments identified QTLs for GPC on chromosomes 2B and 7 A, and for TKW on chromosomes 2 A and 6 A, emphasizing the genetic complexity and environmental sensitivity of these traits¹⁹. These studies underscore the importance of the A and B sub-genomes in governing GPC and TKW in common wheat, with the identified QTLs and candidate genes serving as valuable targets for marker-assisted selection to develop wheat varieties with improved grain quality and yield.

Advancements in genotyping and genomic technologies have enabled the mapping of genes and QTLs associated with important traits in emmer wheat. For example, genes influencing GPC (QGpc-ndsu·5B1, QGpc-ndsu·5B2, and QGpc-ndsu·5B3) have been mapped in Triticum turgidum var. dicoccoides²⁰. Avni et al.²¹ used a high-density SNP-based map to identify QTLs for grain morphology and yield traits. Maccaferri²² performed a genome-wide association study and high-resolution mapping using a panel of wild and domesticated tetraploid wheats, revealing loci linked to adaptation, root architecture, and nutrient use efficiency. Furthermore, QTLs for GPC and grain weight have been identified on chromosomes 2AS, 6AS, and 7BL in durum wheat²³. Thus, Genome-Wide Association Studies (GWAS) have emerged as powerful tools for unravelling the genetic architecture underlying complex traits through identifying marker-trait associations for several traits, including grain protein content and thousand-kernel weight in wheat^24–26. However, while bread and durum wheat have been extensively studied, emmer wheat remains relatively underexplored, presenting a significant opportunity for further research. This study aims to identify genetic variation and marker-trait associations for grain protein content and thousand-grain weight in diverse emmer wheat genotypes across multiple environments using GWAS, ultimately pinpointing candidate genes associated with key traits along with stability analysis for promising genotypes.

Materials and methods

Planting material and experimentation

A panel of 150 genetically diverse emmer wheat accessions, comprising both indigenous and exotic materials sourced from the national genebank (Supplementary Table 1), was used in this study. All accessions belonged exclusively to the domesticated emmer wheat subspecies (Triticum turgidum subsp. dicoccum), without inclusion of wild emmer wheat (T. turgidum subsp. dicoccoides). Subspecies identity was verified using standard agro-morphological descriptors, including grain features, rachis brittleness, spike disarticulation pattern, and glume morphology, in accordance with established taxonomic criteria. Accordingly, all phenotypic evaluations and genomic analyses presented here are specific to T. turgidum subsp. dicoccum. The panel encompassed released varieties, landraces, germplasm accessions, and elite breeding lines representing indigenous as well as global collections. The diverse panel was grown during the winter season of 2021–2022 at four different locations: ICAR-NBPGR, Pusa Farm, New Delhi (28.63°N, 77.22°E); Agharkar Research Institute, Pune, Experimental Farm, Hol, Maharashtra (18.52°N, 73.85°E); ICAR-Indian Institute of Wheat and Barley Research (IIWBR), Karnal, Haryana (29.68°N, 76.99°E); and Punjab Agricultural University, Ludhiana (30.90°N, 75.80°E). The experiment was designed using an Alpha Lattice Design (ALD) with five main plots; each main plot was further split into 30 subplots of size 1m² with two replications. Recommended agronomic practices, irrigation and fertilisers (120 kg N, 60 kg P₂O₅ and 40 kg K₂O ha^− 1) were followed throughout the experiments uniformly at each location. Complete doses of K₂O and P₂O₅ were applied at the time of sowing. Nitrogen was applied in three split doses, 60 kg N ha^− 1 at sowing, 30 kg N ha^− 1 at crown root initiation and 30 kg N ha^− 1 at booting stages. The experiment was sufficiently irrigated with five irrigations during the crop growth. Weeding was done manually. Spray of fungicide, tilt was done to avoid the infection of rust diseases.

Top ten selected genotypes based on overall BLUP values for Grain protein content. (EC578064, IC32513, IC32502, IC138450, IC138455, IC534016, EC540812, EC299074, EC577402, IC535086) was grown during subsequent year 2022-23 at Delhi, Karnal and Ludhiana following a randomised block design and harvested grains from individual plots were evaluated for grain protein content (GPC) and thousand grain weight (TGW). Standard local agronomic practices for durum wheat (irrigation, fertilizer and plant protection practices) were applied uniformly at all sites.

Phenotyping for 1000 grain weight and grain protein

Twenty representative spikes of each genotype were harvested and threshed manually at maturity. Dehulling was carried out by hand and carefully cleaned by discarding broken grains and foreign material. The grain weight of the seeds was measured by counting and weighing the thousand number of seeds using a standardized high-precision seed counter (Mitsubishi Electric, GOT 2000) attached with a weighing balance (Ishida Co. Ltd.).

Grain samples from each genotype and replication were collected separately from each experimental location and stored at an optimum moisture content of 12–13% prior to further analysis. Samples were homogenised using a Foss Cyclone mill equipped with a 1 mm sieve and then were used for the estimation of total nitrogen content. Flour samples (0.1 g) were weighed using a scientific balance (Sartorius ATX224R), and nitrogen content was determined using the rapid MAX N (Elementar) equipment, which employed the Dumas method (992.3). The percentage of nitrogen in the samples was determined using the formula:

By multiplying the acquired nitrogen percentage by a conversion factor (5.7), the protein content was calculated²⁷.

Phenotypic data analysis

The best linear unbiased predictors (BLUPs) of each genotype for each environment were calculated for the alpha lattice design using META-R Software (v 6.0 https://hdl.handle.net/11529/10201). The calculated BLUPs were then used in the subsequent univariate, bivariate and multivariate analysis. IBM SPSS Statistic (V 20.0; https://www.ibm.com/products/spss-statistics) and R software (v 4.2.3; https://www.r-project.org) were used to calculate analysis of variance, summary statistics, coefficient of variation, heritability, frequency distribution, correlation, and principal component analysis. Two-way cluster diagram showing dendrogram and heatmap were analysed by SAS JMP (version 14; https://www.jmp.com). For stability analysis of 10 selected genotypes for grain protein content and thousand grain weight, AMMI analysis was performed for seven environments at Delhi, Ludhiana, and Karnal for two consecutive years and in Pune for one year using respective R packages such as ggplot, inti and metan. Seven testing environments were designated as K-1, K-2, D-1, D-2, L-1, L-2 and P-1.

Genotyping

Genomic DNA was isolated from 15-day-old seedlings of the emmer wheat diversity panel (150 germplasm accessions) using the CTAB method originally described by Murray and Thompson in 1980²⁸. The wheat association panel was genotyped using the Axiom Wheat Breeders’ Array according to the Affymetrix method characterized by Allen et al.²⁹. A total of 35,153 single nucleotide polymorphism (SNP) markers were processed to obtain high-quality 24,089 informative markers. The filtering and imputation were done using TASSEL (version 5.0; www.maizegenetics.net/tassel). Filtering was done for markers with minor allele frequency (MAF) less than 0.05, missing data greater than 10%, and heterozygosity greater than 30%. The remaining set of 6,800 high-quality SNP markers was used in GWAS analysis.

Linkage disequilibrium (LD) and population structure analysis

The squared allele frequency correlation (r²) between SNP markers was used to estimate LD across sub-genome A and B using TASSEL (version 5.0; www.maizegenetics.net/ tassel) and values were plotted against genetic distance in R (version 4.2.3; https://www.r-project.org) as per Remington et al.³⁰. Population stratification of 150 genotypes of the emmer wheat association mapping panel was done using STRUCTURE software (version 2.3; https://web.stanford.edu/group/pritchardlab/structure.html)³¹. To estimate the number of subpopulations (k) in a hypothetical range of k = 1 to 10, a burn-in of 10,000 iterations was followed by 20,000 Monte Carlo Markov Chain (MCMC). The subpopulation number was calculated using an ad hoc statistic called delta k, which was based on the rate of change in log likelihood of data between successive values³². Molecular-marker based PCA was conducted with the help of GAPIT (https://zzlab.net/GAPIT) developed by Lipka et al.³³.

Genome-wide association study (GWAS)

Association mapping was conducted using 6,800 polymorphic SNP markers and BLUP values of grain protein content and thousand grain weight across environments using three multi-locus models (MLMM, Farm CPU, and BLINK) of the GAPIT (https://zzlab.net/GAPIT) R (version 4.2.3; https://www.r-project.org). Marker–trait associations with –log10(P) ≥ 3.5 (corresponding to P ≤ 3 × 10⁻⁴) were considered putatively significant and used for cross-environment integration. Using this threshold, a total of 70 marker–trait associations were identified across all environments and models (Supplementary Table 4). In addition, a stringent Bonferroni correction was applied to identify highly reliable associations, which were regarded as statistically robust and effective in controlling type I error, following Xu et al.³⁴. The distribution of p-values for significant SNP markers linked to the investigated traits was illustrated through Manhattan plots, which categorise gene coordinates by chromosome. Each plot dot corresponds to a single SNP.

In silico analysis for candidate gene identification

The significantly associated MTAs were checked for their putative candidate genes with default search parameters in the Ensemble Plants database of the bread wheat genome (IWGSC (RefSeq v1.0). To identify putative candidate genes, a distance of 100 kb intervals was selected at both the overlapped regions and also the interval regions flanked on either side of the markers. The genes are further explored for their respective functions from the database (http://plants.ensembl.org/index.html). Further, MG2C (http://mg2c.iask.in/mg2c_v2.0/), Chao,³⁵ was used to provide a graphical depiction of the genomic location of MTA-QTLs.

Results

Phenotypic variation in Dicoccum germplasm for grain protein content and thousand grain weight

Phenotypic evaluation data were collected from four locations: New Delhi (D1), Karnal (K1), Pune (P1), and Ludhiana (L1), for the traits Grain Protein Content (GPC) and Thousand Grain Weight (TGW). The locations were treated as random effects while genotypes as fixed effects. Significant differences among genotypes were observed for both traits across all locations. The combined analysis of variance revealed highly significant differences among locations and a significant genotype × location interaction, demonstrating a strong environmental influence on trait expression and differential genotype responses across environments; notably, the location effect was evaluated against the block (location) mean square, whereas genotype and genotype × location effects were tested against the pooled residual mean square. (Supplementary Table 2).

The frequency distribution curves showed a normal to near-normal distribution for both traits across most of the environments. Normality was further assessed using the Shapiro-Wilk test, where a p-value ≥ 0.05 indicated normal distribution, p-values between 0.001 and 0.05 suggested near-normal distribution, and p-values ≤ 0.001 indicated non-normal distribution (Fig. 1). Descriptive statistics for GPC and TGW are provided in Table 1.

Fig. 1.

Frequency distribution of grain protein content (GPC) and thousand grain weight (TGW) in a panel of 150 Triticum turgidum subsp. dicoccum genotypes evaluated across four environments. Histograms represent trait distributions in different colour for individual locations (Delhi, Karnal, Ludhiana, and Pune), and “Overall” distribution represents pooled phenotypic values across environments. The x-axis indicates trait values (GPC in %, TGW in g), and the y-axis represents relative frequency. Density curves illustrate the distribution pattern for each environment, highlighting environmental effects on trait expression. GPC grain protein content, TGW thousand grain weight.

Table 1.

Descriptive statistics of grain protein content (%) and thousand grain weight (g) in 150 Triticum turgidum subsp. dicoccum accessions based on best linear unbiased predictor (BLUP) values estimated across environments.

Trait	Mean ± SE	Range	SD	PCV%	Heritability	LSD
Delhi GPC	15.4 ± 0.13	11.64 − 18.54	1.55	10.04	0.96	0.65
Karnal GPC	16.05 ± 0.14	12.15–21.46	1.68	10.43	0.71	0.65
Ludhiana GPC	13.87 ± 0.18	11.22–21.78	2.17	15.62	0.98	0.70
Pune GPC	15.13 ± 0.14	11.52–18.52	1.71	11.29	0.97	0.68
Overall GPC	15.11 ± 0.07	12.16 – 17.76	0.90	5.95	0.38	0.66
Delhi TGW	30.98 ± 0.45	17.85–41.54	5.53	17.84	0.83	1.91
Karnal TGW	27.88 ± 0.39	16.34–36.51	4.78	17.16	0.97	1.80
Ludhiana TGW	38.29 ± 0.35	21.05–48.42	4.23	11.05	0.97	1.47
Pune TGW	37.52 ± 0.36	21.65–46.93	4.35	11.6	0.87	1.80
Overall TGW	33.64 ± 0.27	25.98–38.64	3.27	9.74	0.42	1.38

SD standard deviation, PCV phenotypic coefficient of variation, LSD least significant difference.

Grain protein content (%)

The grain protein content (GPC) exhibited significant variation among the genotypes across locations. At Delhi, GPC ranged from 11.06% to 18.54%, with a mean of 15.40%, a PCV of 10.04%, and high heritability of 96%. The genotype EC06902 had the highest BLUP value (18.54%), while IC252503 had the lowest (11.60%). Promising genotypes included EC06902, IC535136, EC11074, IC535144, and EC577404. At Karnal, GPC ranged from 12.15% to 21.46%, with a mean of 16.05%, PCV of 10.43%, and heritability of 71%. The genotype EC06910 had the highest BLUP value (21.46%) and IC551396 the lowest (12.15%), with EC06910, IC535090, IC138455, IC35097, and IC535086 identified as promising genotypes. At Ludhiana, GPC ranged from 11.21% to 21.78%, with a mean of 13.87%, a PCV of 15.62%, and heritability of 98%. EC578064 had the highest BLUP value (20.93%) and IC535088 the lowest (11.22%), with EC578064, EC299074, EC577402, IC535106, and IC535108 identified as promising genotypes. At Pune, GPC ranged from 11.52% to 18.52%, with a mean of 15.13%, a PCV of 11.29%, and a heritability of 97%. IC534016 had the highest BLUP value (18.52%) and IC128392 the lowest (11.52%), with IC534016, IC138450, IC78699, IC535086, and EC11232 identified as promising genotypes. Across locations, GPC ranged from 12.16 to 17.76%, with a mean of 15.11%, PCV of 11.29%, and heritability of 38%. EC578064 (15.34%) had the highest BLUP value, while EC06912 had the lowest (15.00%). Accessions EC578064, EC299074, IC32502, IC138455, and EC540812 were identified as valuable sources for grain protein content (Table 2).

Table 2.

Top and bottom 10 accessions for percent grain protein content (GPC).

Delhi		Karnal		Ludhiana		Pune		Overall
Top Ten Accessions
Accessions	GPC	Accessions	GPC	Accessions	GPC	Accessions	GPC	Accessions	GPC
EC06902	18.54	EC06910	21.46	EC577402	21.78	IC534016	18.52	EC578064	17.76
IC535136	18.19	IC535090	21.40	EC299074	21.67	IC138450	18.05	IC138455	17.14
EC11074	18.07	IC138455	21.19	EC578064	20.93	IC78699	18.01	IC32502	17.12
IC535144	18.03	IC35097	19.87	IC32502	20.08	IC535086	18.01	EC299074	17.08
EC577404	17.85	IC535086	19.14	IC535108	17.94	EC11232	17.99	EC540812	16.99
EC577406	17.75	IC32513	18.60	EC299171	17.50	EC6909	17.93	EC577402	16.96
EC540812	17.74	IC535078	18.33	EC577411	17.06	IC47034	17.69	IC535086	16.80
IC531559	17.65	IC534016	18.32	IC535106	16.94	IC138331	17.63	IC534016	16.56
IC551399	17.63	EC299171	18.31	EC540813	16.91	IC535083	17.47	IC138450	16.49
IC535129	17.62	EC577398	18.20	IC551400	16.89	IC138900	17.45	IC32513	16.45
Bottom Ten Accessions
IC252503	11.60	IC551396	12.15	IC535088	11.22	IC128392	11.52	IC551396	12.16
IC539302	11.61	EC11232	13.26	IC535302	11.26	IC551398	11.54	IC535123	13.37
IC551396	12.18	IC47022B	13.32	IC535113	11.28	IC551396	11.87	IC535302	13.41
IC138472	12.19	EC11074	13.34	EC577410	11.30	IC47040	12.05	IC416358	13.47
IC78699	12.23	IC416358	13.39	IC47037	11.32	IC402045	12.17	IC535113	13.58
IC416358	12.40	EC12941	13.57	IC277713	11.34	IC35170	12.29	IC535120	13.61
IC118763	12.73	EC577411	13.62	EC577409	11.42	IC535105	12.50	IC402020	13.62
IC47037	12.73	IC138471	13.59	IC47048	11.36	IC535302	12.49	EC577410	13.62
IC531969	12.84	IC277713	13.91	EC06902	11.42	IC402020	12.54	IC535070	13.71
IC534621	13.05	EC11071	13.91	EC12954	11.42	IC551399	12.58	EC12941	13.82

Thousand grain weight (g)

The thousand-grain weight (TGW) showed significant variation across locations. At Delhi, TGW ranged from 17.84 to 41.54 g, with a mean value of 30.98 g, a PCV of 17.84%, and a heritability of 83%. The genotype, IC534621 had the highest BLUP value (41.54 g) whereas IC78699 had the lowest (17.85 g), with IC534621, EC08572, IC534586, IC252486, and IC138900 identified as promising genotypes. At Karnal, TGW ranged from 16.34 to 36.51 g, with a mean value of 27.84 g, CV of 17.16%, and heritability of 97%. EC299074 had the highest BLUP value (36.51 g) while IC416358 had the lowest (16.34 g), with EC11071, IC277713, EC577411, IC138471, and EC12941 identified as promising genotypes. At Ludhiana, TGW ranged from 21.05 to 48.42 g, with a mean value of 38.29 g, a PCV of 11.05%, and heritability of 97%. EC577410 was having the highest BLUP value (48.42), while IC252503 had the lowest (21.05), with EC577410, IC78706, IC138472, EC06902, and IC535090 identified as promising genotypes. At Pune, TGW ranged from 21.65 to 46.93 g, with a mean value of 37.52 g, PCV of 11.6%, and heritability of 87%. IC78706 was having the highest BLUP value (46.93 g) while IC252503 had the lowest value (21.65 g), with IC534016, IC138450, IC78699, IC535086, and EC11232 identified as promising genotypes. Across locations, based on pooled data, TGW ranged from 25.98 to 38.64 g, with a mean value of 33.64 g, PCV of 9.74%, and heritability of 42%. IC138472 had the highest BLUP value (38.64 g) whereas IC78699 had the lowest (25.98 g). The genotypes, IC138472, IC78706, EC299074, EC577411, and IC252486 were identified as promising genotypes for TGW (Table 3).

Table 3.

Top and bottom 10 accessions for thousand grain weight (TGW, g).

Delhi		Karnal		Ludhiana		Pune		Overall
Top Ten Accessions
Accessions	TGW	Accessions	TGW	Accessions	TGW	Accessions	TGW	Accessions	TGW
IC534621	41.54	EC299074	36.51	EC577410	48.42	IC78706	46.93	IC138472	38.64
EC08572	41.42	IC535304	33.93	IC78706	48.03	IC138472	46.32	IC78706	38.54
IC534586	40.53	IC113725	33.44	IC138472	46.42	IC535090	45.46	EC299074	37.89
IC252486	40.48	EC06845	33.33	EC06902	46.17	EC299171	45.01	EC577411	37.83
IC138900	40.08	IC78706	33.05	IC535090	45.56	EC06902	44.80	IC252486	37.35
EC577402	39.39	EC577407	32.73	EC299171	45.48	IC534960	44.48	EC577409	37.24
IC138472	38.89	IC593664	32.57	EC299074	45.00	EC12941	43.35	EC08572	37.16
EC577411	38.52	IC252486	32.48	IC534960	44.56	EC577410	43.34	IC138474	37.00
IC551396	37.79	IC138471	32.48	IC138331	44.15	IC138475	43.22	IC534621	36.95
IC539302	37.62	IC47800	32.48	EC12941	43.97	EC577409	43.14	IC138475	36.95
Bottom Ten Accessions
IC78699	17.85	IC416358	16.34	IC252503	21.05	IC252503	21.65	IC78699	25.98
IC535144	23.23	IC35097	19.27	IC138455	24.86	IC138455	24.40	IC35097	28.31
IC535078	23.36	EC06900	20.86	IC78699	27.50	IC138450	25.00	IC138450	28.77
IC535086	23.38	IC32513	21.00	IC138450	27.69	IC78699	25.06	IC252503	29.02
IC35097	23.86	IC35093	21.58	IC32502	29.84	IC32502	29.24	IC32502	29.18
IC535090	24.32	IC78699	22.15	IC535108	30.58	EC577406	29.81	IC138455	29.20
IC535092	24.46	IC535085	22.27	IC212168	31.90	IC212168	30.69	IC535086	30.06
EC540812	24.94	IC47021	22.36	IC535112	32.29	IC535113	30.91	EC577406	30.24
IC47021	25.13	EC577406	22.61	EC577402	32.29	EC540812	31.07	IC535092	30.39
EC299171	25.15	IC47548	22.94	IC535086	32.43	IC535108	31.15	IC35093	30.47

Correlation studies

The Pearson Coefficient of Correlation for GPC and TGW pertaining to each environment was computed and a negative correlation between them was found for each location and pooled data (Table 4). GPC and TGW had significant negative correlations in Delhi (r = -0.238, p = 0.01) and Ludhiana (r = -0.179, p = 0.05). Further, a significant negative correlation (r = -0.193, p = 0.05) for pooled data revealed a general tendency of increasing protein content linked with reduced TGW.

Table 4.

Correlation coefficients between grain protein content and thousand grain weight in 150 Triticum turgidum subsp. dicoccum accessions estimated using best linear unbiased predictor (BLUP) values across environments.

Delhi	Protein	TGW
Protein	1	− 0.238**
TGW	− 0.238**	1
Karnal
Protein	1	− 0.121
TGW	− 0.121	1
Pune
Protein	1	− 0.102
TGW	− 0.102	1
Ludhiana
Protein	1	− 0.179*
TGW	− 0.179*	1
Overall
Protein	1	− 0.193*
TGW	− 0.193*	1

** Correlation is significant at the 0.01 level, * significant at the 0.05 level.

Principal component analysis (PCA)

Principal Component Analysis (PCA) was performed separately for each location and the pooled data, and the cumulative variation was investigated (Supplementary Table 3). At Delhi location, the first and second PCs explained 61.90% and 38.1% of the variance respectively, while at Karnal, the first and second PCs explained 56.05% and 43.95% of the variance, respectively. In Pune, the first PC accounted for 58.97% of the variance, while the second PC explained 41.03% of the variance. In Ludhiana, the first PC explained 55.1% of the variance, and the second PC accounted for 44.9%. When considering all environments together, the first PC explained 59.67% of the total variance, and the second PC contributed 40.33% to the cumulative variation across environments. Analysis of the key contributors to PC1 revealed that GPC and TGW had contributions of 0.70 and − 0.70, respectively, indicating a significant and opposing interaction between these two traits (Fig. 2).

Fig. 2.

Principal component analysis (a) Delhi, (b) Karnal, (c) Pune, (d) Ludhiana and (e) across the environments. Principal component analysis (PCA) illustrating the genetic variation among 150 Triticum turgidum subsp. dicoccum genotypes based on grain protein content and thousand grain weight. The first two principal components (PC1 and PC2) explain the maximum proportion of total phenotypic variance in each panel. Each point represents a genotype, and the spatial separation of points reflects the degree of phenotypic divergence among genotypes under different environmental conditions. PCA plots are shown for individual environments: (a) Delhi, (b) Karnal, (c) Pune, (d) Ludhiana, and for (e) pooled data across environments.

Grouping of genotypes based on GPC and TGW

The hierarchical cluster analysis was performed based on pooled data over the locations for GPC and TGW to generate dendrogram as a two-way clustering heatmap (Fig. 3) that illustrated genetic relationships among 150 emmer wheat accessions. Hierarchical clustering was performed using Ward’s method with Euclidean distance, resulting in six distinct clusters, determined by the optimal cut-point through canonical discriminant analysis in SAS background, which indicated substantial increases in within-cluster variance during agglomeration. The dendrogram explained about the hierarchical arrangement of the genotypes with groupings of similar genotypes as clusters are joined by shorter branches, which highlighted limited variation, whereas comparatively more variation was observed as the distance between genotypes increased. Cluster 4 had the maximum number of genotypes 35, followed by cluster 3 with 34 genotypes. Cluster 2 was the smallest cluster group consisting of only 11 genotypes (Table 5). Cluster 1, 5 and 6 had an intermediate membership of 22, 25 and 23 germplasm accessions, respectively. By visualising colour gradient pattern of heat map for both traits, it was evident that Cluster 1 genotypes are invariably low in protein (blue colour) and high in thousand grain weight (red colour), whereas Cluster 6 genotypes have high grain protein content (red colour) and relatively low grain weight (blue colour).

Fig. 3.

Two-way clusters representing dendrogram and heatmap for 150 emmer wheat accessions based on grain protein content and thousand grain weight employing Euclidean distance and Ward method of clustering. The two-way clusters representing dendrogram and heatmap generated by SAS-JMP software, illustrates the genetic relationships among accessions, with branch lengths representing dissimilarity between clusters and trait value represented for each genotype on X axis. Colours within the heat map range from light blue (least prevalent) to dark red (most prevalent), illustrating the prevalence of a particular trait within a particular genotype A total of six clusters are depicted in different six colours. The dashed horizontal line indicates the cut-off used to define the optimal number of clusters.

Table 5.

Grouping of 150 Triticum turgidum subsp. dicoccum germplasm accessions into distinct clusters based on hierarchical clustering using Ward’s method.

Cluster	Count	Name of Accessions	Protein						TGW
			Delhi	Karnal	Ludhiana	Pune	Overall	Delhi		Karnal	Ludhiana	Pune	Overall
1	22	IC533783, EC06845, IC535093, IC138471, IC593664, EC11071, IC277713, EC12941, EC577410, EC12954, EC11074, IC138418, IC448026, IC112083, EC06902, EC06912, IC535070, IC535082, IC535120, IC535123, IC535133, IC535141	15.20	14.43	12.25	14.72	15.02	29.63		29.28	41.53	40.54	34.71
2	11	IC534586, IC138472, IC138474, IC78706, EC299074, IC138475, EC577407, EC299171, EC577409, EC577411, IC138331	14.96	15.46	15.13	15.64	15.12	35.05		30.88	44.22	43.52	37.19
3	34	IC252486, IC534621, IC416358, IC539302, EC540809, IC530555, IC118763, IC547564, EC11072, EC577401, EC06840, IC535105, IC535130, IC551398, IC402045, IC551396, IC402020, IC535302, IC531969, IC47022B, IC47800, EC11073, IC35170, EC577932, IC47040, IC47037, IC47548, IC535304, EC08572, IC535113, IC535079, IC535138, IC535125, IC535134	14.35	15.64	13.65	13.91	13.19	33.08		28.04	37.00	36.35	33.75
4	35	IC138900, IC534016, IC534587, IC252504, EC299208, EC577400, IC535076, IC535088, IC535090, IC535097, EC577404, IC535124, IC535126, IC535143, IC138897, IC551399, IC551400, IC534960, IC551397, IC113725, IC402012, IC535301, IC47021, IC47026, EC08479, EC11389, EC06908, EC06910, IC535117, IC535140, IC535136, IC535137, IC535142, IC128392, IC138371	16.25	16.90	14.12	14.69	14.69	30.85		27.72	40.28	39.59	34.27
5	25	IC534012, IC138896-A, IC566241, EC577398, IC535078, IC535083, IC535112, IC535129, IC535144, IC402018, IC118774, IC531559, IC535118, IC47048, EC6909, EC11386, EC11232, IC35171, IC35174, IC47034, IC47035, IC535116, IC535092, IC535131, EC12565	16.05	16.60	12.27	16.46	16.46	28.11		27.04	36.72	36.02	32.33
6	23	IC252503, EC578064, EC540812, IC78699, EC577406, EC06900, IC535071, EC577402, IC535106, IC535108, IC535085, IC535086, IC535153, IC212165, IC212168, IC138898, EC540813, IC32513, IC35093, IC35097, IC32502, IC138450, IC138455	15.32	16.58	16.45	16.25	16.25	30.04		25.74	32.88	31.95	31.18

Stability analysis

AMMI analysis for ten genotypes superior for GPC for seven environments was conducted, and AMMI biplots are presented in Fig. 4.

Fig. 4.

Additive Main effects and Multiplicative Interaction (AMMI) biplots illustrating genotype and genotype × environment interaction effects for grain protein content (GPC) and thousand grain weight (TGW). Panels show (a) AMMI1 biplot for GPC, (b) AMMI2 biplot for GPC, (c) AMMI1 biplot for TGW, and (d) AMMI2 biplot for TGW. In AMMI1 biplots, the x-axis represents the main effect (trait mean), while the y-axis represents the first interaction principal component axis (IPCA1). AMMI2 biplots display the first two interaction principal component axes (IPCA1 and IPCA2), enabling visualization of the magnitude and pattern of genotype × environment interactions. Genotypes and environments positioned closer to the origin exhibit greater stability across environments, whereas those farther from the origin indicate stronger interaction effects.

AMMI analysis for grain protein content (GPC)

The AMMI1 biplot for GPC (Fig. 4a) illustrated the relationship between genotype mean performance and the first principal component (PC1), which explained 47.8% of the total G×E interaction variance. Environment K-1 displayed the highest PC1 score, highly discriminative. Genotypes IC32513 and IC577402 displayed high GPC and were specifically suitable to P1. Conversely, genotypes such as IC138455, EC577402, and IC534016, located near the origin, exhibited low interaction with high stability across environments, although with slightly lower GPC value.

The AMMI2 biplot (Fig. 4b), incorporating PC1 (47.8%) and PC2 (21.5%) and cumulatively explaining 69.3% of the interaction variance, offered deeper insights into crossover interactions. Genotypes EC540812, IC138450, and EC299074 were scattered away from the origin, indicating specific adaptability, whereas IC138455, IC577402, and IC534016 were close to the origin, reinforcing their stability. Environments K-1 and L-1 having the largest length exerted strong interaction effects with genotypes. Genotypes EC299074, EC540812 and IC138450 were specifically suited to environment K1, as evident from their directional association.

AMMI analysis for thousand grain weight (TGW)

The AMMI1 biplot for TGW (Fig. 4c) revealed that PC1 captured 48.2% of the G×E interaction variance. Genotypes IC138450, IC32502, and IC32513 showed high PC1 value and were positioned away from the origin, indicating genotype-specific adaptation to D2 and L1. IC138450 was having a high TGW and a high PC1 score. Genotypes such as EC299074, IC534016 and IC138455, although having moderate TGW, showed greater stability due to their proximity to the origin.

The AMMI2 biplot (Fig. 4d) included PC1 (48.2%) and PC2 (26.9%), cumulatively explaining 75.1% of the interaction variance. Genotypes IC32513 and IC32502 are specifically suited to D2. Meanwhile, IC534016, EC299074, and IC535086 clustered near the origin, showing both stable performance and low interaction with environments. Among environments, D-2 and L-2 were highly discriminative, while D-1 and P-1 were more representative environments, better suited for identifying generally adapted genotypes.

Genome-wide spread of SNPs, linkage disequilibrium (LD) and population structure analysis

After initial filtering, a total of 6,800 polymorphic SNP markers were retained. These SNP markers were distributed almost proportionately across the 14 chromosomes, with the highest number of markers on chromosome 1B (723) and lowest on chromosome 4B (270) (Supplementary Fig. 1). The distribution of SNP markers across each chromosome is graphically represented using an SNP marker density plot (Fig. 5a). The squared correlation coefficient (r²) for all 6,800 markers was used to assess linkage disequilibrium (LD) using calculations for expected values of r² under drift equilibrium³⁶. Genome wide LD decayed with genetic distance; the LD decayed to its half at 1.94 Mb for the whole genome, indicating the markers within this LD block exhibit strong linkage disequilibrium (Fig. 5b).

Fig. 5.

Genome-wide marker characteristics of the association mapping panel. (a) SNP density plot showing the distribution of single nucleotide polymorphism (SNP) markers across wheat chromosomes, with marker density represented along physical positions (Mb). This plot illustrates genome coverage and variation in marker distribution across chromosomes. (b) Genome-wide linkage disequilibrium (LD) decay plot depicting the relationship between pairwise LD (measured as ) and physical distance (Mb) between SNP markers. The LD decay curve indicates the rate at which LD declines with increasing physical distance, providing an estimate of the genomic resolution of the association mapping analysis.

The analysis of the population structure revealed that the optimal value for ΔK (delta K) was 2. This indicates that the genotypes in the GWAS panel can be separated into two distinct subpopulations. Population structure inferred by Principal Component Analysis (PCA) revealed two subpopulations in the GWAS panel. The first subpopulation consisted of 118 genotypes (G1), while the second subpopulation included 32 genotypes (G2) with a threshold of 0.8 (Fig. 6).

Fig. 6.

Population structure and genetic stratification of the wheat association mapping panel. (a) Population structure inferred using a model-based clustering approach, shown as a bar plot illustrating the assignment of 150 Triticum turgidum subsp. dicoccum genotypes into two subpopulations, G1 (red) and G2 (green). Each vertical bar represents an individual genotype, and the proportion of color indicates membership probability in each subpopulation. (b) Delta K plot showing the rate of change in the log probability of data across K values ranging from 2 to 9, used to determine the optimal number of subpopulations. (c) Three-dimensional principal component analysis (PCA) based on genome-wide SNP data, illustrating genetic differentiation among genotypes and supporting the inferred population structure.

Genome wide association study

Genome-wide association analysis was conducted using 6,800 polymorphic SNP markers and BLUP estimates of grain protein content (GPC) and thousand grain weight (TGW). Across all environments and three models (MLMM, FarmCPU, and BLINK), a total of 70 marker–trait associations were identified (Supplementary Table 4). A suggestive threshold of −log10(p) ≥ 3.5 was adopted to detect putative associations, acknowledging that strict Bonferroni correction can be overly conservative in studies with moderate marker density and complex quantitative traits, as reported in previous wheat GWAS studies³⁷. Results of GWAS analysis pertaining to all three models employed and for all the locations are presented in the Circular Manhattan plot. When considering the distribution of significant MTAs for GPC and TGW across different environments, both Pune and Ludhiana demonstrated the highest number of MTAs (15) from all the three models for TGW. Whereas for the Karnal location, lowest number of MTA (1) for both GPC and TGW was recorded for the MLMM model as shown in Supplementary Fig. 2. This indicates that the expression of grain traits is influenced by genotype-by-environment interactions, as the expression of the same genotypes varied more or less at different locations.

Circular Manhattan plots and QQ plots were drawn, with the x-axis displaying the SNPs according to their base pair position across the 14 chromosomes and the y-axis displaying the result at each SNP as -log10(P value) (Supplementary Figs. 3 and 4). A total of 21 unique and significant MTAs were identified for both the traits in all the environments (Table 6). Out of these, 5 MTAs for TGW exceeded the Bonferroni-corrected significance threshold (p < 0.05 ; −log10(p) > 5.12). for TGW were identified. Among different chromosomes, the 5B chromosome exhibited the highest number of significant MTAs (6), indicating its prominent role in influencing traits GPC and TGW.

Table 6.

Unique marker–trait associations (MTAs) identified for grain protein content and thousand grain weight across individual environments using different genome-wide association models at a significant threshold, -log10(P) ≥ 3.5.

Trait	Location	Model	MTA	Chr	Position (Mb)	−log10(p)
GPC	Delhi	MLMM	Affx-92963286	2B	476.42	3.78
	Delhi	MLMM	Affx-92326732	3 A	724.05	3.59
	Karnal	MLMM	Affx-92290480	5B	680.81	3.77
	Ludhiana, Pune, Overall	MLMM, BLINK, FARMCPU	Affx-92754611	5B	662.04	3.72
TGW	Delhi	FARMCPU	Affx-92115317	5 A	565.39	4.26
	Delhi	FARMCPU	Affx-92174750	1B	568.94	3.57
	Delhi	FARMCPU	Affx-92237374	1 A	541.73	4.23
	Karnal	MLMM	Affx-92544100	2B	449.84	4.41
	Delhi	FARMCPU	Affx-92696773	5B	571.18	3.57
	Delhi	FARMCPU	Affx-92811991	5B	572.59	4.14
	Delhi	MLMM, BLINK, FARMCPU	Affx-92195490	5B	545.40	6.54
	Overall	MLMM, BLINK, FARMCPU	Affx-92695862	1 A	514.29	11.88
	Overall	FARMCPU	Affx-92688874	3B	760.89	3.64
	Ludhiana, Overall, Pune	BLINK, FARMCPU	Affx-92280193	6 A	31.04	8.68
	Ludhiana, Overall, Pune	BLINK	Affx-92286446	5B	572.39	4.99
	Ludhiana, Overall, Pune	FARMCPU	Affx-92301332	7B	186.79	4.87
	Ludhiana, Overall, Pune	BLINK	Affx-92452373	3B	483.98	9.83
	Ludhiana, Overall, Pune	BLINK	Affx-92546089	2B	741.86	4.62
	Ludhiana, Overall, Pune	FARMCPU, BLINK	Affx-92640446	2 A	539.97	15.25
	Ludhiana, Overall, Pune	FARMCPU	Affx-92167066	6 A	2.95	4.20
	Ludhiana, Overall, Pune	MLMM	Affx-92931355	2B	477.40	4.31

MTAs for grain protein content (GPC)

A total of 4 significant MTAs distributed on the chromosomes 2B, 3 A and 5B were identified for GPC across the environments (Fig. 7). Out of four, two MTAs were identified for the Delhi location on chromosome 2B (Affx-92963286) at 476.41 Mb and 3 A (Affx-92326732) at 724.04 Mb region with the −log10(p) value of 3.78 and 3.59 respectively. Another MTA identified for the Karnal location was located on the chromosome 5B (Affx-92290480) at the 680.81 Mb region with a −log10(p) value of 3.76. A common MTA was identified for Ludhiana, Pune, and Overall environments on chromosome 5B (Affx-92754611) at a region of 662.02 Mb region with the −log10(p) of 3.72. Grain protein content exhibited a highly polygenic architecture with strong environmental and genotype × environment influences. As a result, most detected loci showed small effect sizes, and no MTAs remained significant under stringent Bonferroni correction. Accordingly, associations exceeding a relaxed significance threshold were treated as suggestive and prioritised based on their consistency across environments.

Fig. 7.

Physical map showing the chromosomal distribution of 21 marker–trait associations (MTAs) identified for grain protein content and thousand grain weight in Triticum turgidum subsp. dicoccum. MTAs associated with grain protein content are indicated in red, while those associated with thousand grain weight are shown in blue. SNP positions are plotted based on their physical locations (Mb) on the corresponding wheat chromosomes, illustrating the genomic regions associated with each trait and highlighting both trait-specific and shared chromosomal regions.

MTAs for thousand grain weight (TGW)

A total of 17 MTAs distributed on the chromosomes 1 A, 1B, 2 A, 2B, 3B, 5B, 6 A and 7B were identified for TGW across the environments (Fig. 7). Out of 17 MTAs, 6 MTAs were identified for the Delhi location which includes makers Affx-92115317 on 5 A chromosomes with −log10(p) of 4.26, Affx-92174750 on 1B chromosome with −log10(p) of 3.57, Affx-92237374 on 1 A chromosome with −log10(p) of 4.23 and Affx-92696773, Affx-92811991 and Affx-92195490 on 5B chromosome with −log10(p) of 3.56, 4.14 and 6.53 respectively. One MTA Affx-92544100 was identified for the Karnal location on chromosome 2B with −log10(p) of 4.41. Among 17 MTAs, eight common MTAs were identified for environments, Ludhiana, Pune and Overall, which includes markers Affx-92280193 and Affx-92167066 on chromosome 6 A with −log10(p) of 8.67 and 4.20 respectively. Markers Affx-92546089 and Affx-92931355 were located on chromosome 2B with −log10(p) of 4.62 and 4.31 respectively. Marker Affx-92286446 was located on the chromosome 5B with a −log10(p) of 4.99, whereas marker Affx-92301332 was located on the chromosome 7B with a −log10(p) of 4.86. Marker Affx-92452373 was located on the chromosome 3B with a −log10(p) of 9.82 whereas marker Affx-92640446 was found on chromosome 2 A with −log10(p) 15.25. Further, two markers Affx-92695862 and Affx-92688874 were identified for the overall environment were located on chromosomes 1 A and 3B with their −log10(p) of 11.87 and 3.64, respectively.

Allelic effects of identified genomic regions on their respective phenotypes

Nine stable MTAs were examined to determine the range of phenotypic variations for GPC and TGW. Boxplots showed the phenotypic values linked to three reliable QTNs (1 for GPC and 2 for TGW) with significant effects on their respective traits. Genotypes in the association panel were categorised into two classes based on different allele types for GPC and TGW. A t-test test was performed to check significance of mean difference between both classes. For GPC, SNP Affx-92754611 had two allele C and T with a phenotypic mean of 14.64% and 15.12% respectively, which was statistically significant with p value (0.025). For TGW, marker Affx-92640446 had the favourable allele ‘C’ with 133 genotypes, had T and C with a mean TGW of 32.98 g and 36.85 g. Similarly, another SNP Affx-92931355 had two allele G and A/G, with mean TGW of 36.85 g and 33.59 g (Fig. 8), where G is a favourable allele with 133 genotypes (Fig. 8 and Supplementary Table 4).

Fig. 8.

Allelic effects of three stable marker–trait associations (MTAs) identified for grain protein content (GPC) and thousand grain weight (TGW) across environments. Box plots depict the phenotypic distribution of genotypes carrying alternative alleles at each SNP locus, illustrating the magnitude and direction of allelic effects on the respective traits. Differences between allelic classes indicate the contribution of each allele to trait variation, highlighting favorable alleles associated with increased GPC or TGW. MTA, marker–trait association; SNP, single nucleotide.

In silico analysis for candidate gene identification

In silico analysis identified candidate genes associated with all 21 MTAs. For GPC in the Delhi environment, the candidate gene for the SNP Affx-92963286 included TraesCS2B02G332900, which coded for the signal peptidase complex subunit Spc2/SPCS2 with peptidase activity. Another candidate gene of this SNP, TraesCS2B02G332800, belonged to the Major Facilitator Superfamily and functions in transmembrane transporter activity. For SNP Affx-92326732, the candidate gene TraesCS3A02G498200 belonged to the Protein Kinase domain, with roles in protein kinase activity and ATP binding. The gene of SNP, TraesCS3A02G498300 is associated with the Signal Recognition Particle (SRP54) subunit and the GTPase domain, which is involved in ATP hydrolysis activity. For the Karnal environment, the SNP Affx-92290480 identified the candidate gene TraesCS5B02G517300, which belonged to the 2Fe-2 S ferredoxin-type iron-sulphur binding domain and was involved in electron transfer activity, iron-sulphur cluster binding, and 2 iron, 2 sulphur cluster binding. Another gene for this SNP, TraesCS5B02G517400 corresponded to the mRNA-decapping enzyme subunit 1, which plays a role in enzyme activator activity. Another gene, TraesCS5B02G517200, is associated with the Peptidase M20 dimerization domain, functioning in hydrolase activity. The stable SNP Affx-92754611 is associated with the gene TraesCS5B02G494400, which encodes an RlpA-like protein with a double-psi beta-barrel domain, as well as an expansin and a cellulose-binding-like domain. Another gene, TraesCS5B02G494300, encoded a protein with an NB-ARC domain that was involved in nucleotide binding and protein kinase activity.

In the Delhi environment for the trait TGW, the candidate gene for SNP Affx-92115317, TraesCS5A02G365100, encodes a MOSC protein with an N-terminal beta-barrel and molybdenum cofactor sulphurase, which is involved in catalytic activity, molybdenum ion binding, and pyridoxal phosphate binding. SNP Affx-92174750 was associated with the gene TraesCS1B02G341000, which encoded the Primosome PriB protein. This protein was involved in single-strand DNA binding, playing a crucial role in stranded DNA binding. Additionally, SNP Affx-92237374 is linked to the gene TraesCS1A02G360700, which encodes a protein from the Glycoside hydrolase family 9. This protein exhibited hydrolase activity, specifically hydrolysing O-glycosyl compounds, and possesses cellulase activity. SNP Affx-92696773, associated with the gene TraesCS5B02G392200, encoded a protein kinase domain involved in protein kinase activity. Another gene, TraesCS5B02G392300, encoded an NADPH-dependent FMN reductase-like protein, which was involved in NAD(P)H dehydrogenase (quinone) activity and FMN binding activity. SNP Affx-92811991 linked to the gene TraesCS5B02G395200, encoded a SANT/Myb domain involved in transcription coactivator activity. Further, the gene TraesCS5B02G395100 encoded a strictosidine synthase activity and SNP Affx-92195490, associated with the gene TraesCS5B02G366500, was involved in DNA binding activity, and another gene, TraesCS5B02G366700, encoded a molybdenum cofactor sulphurase C-terminal domain involved in catalytic activity, molybdenum ion binding, and pyridoxal phosphate binding activity. For the Karnal environment, SNP Affx-92544100 linked with the gene TraesCS2B02G314700 encoded WD40 repeat involved in protein binding activity.

The common SNP Affx-92280193 identified in the Ludhiana, Pune and overall environments associated with the gene TraesCS6A02G058700 encoded a LURP-one-related protein. SNP Affx-92286446 is linked to the gene TraesCS5B02G394800 encoding a Cns1/TTC4 protein with a wheel domain involved in Hsp90 protein binding activity. Additionally, the gene TraesCS5B02G394500 associated with this SNP encodes methyltransferase EEF1AKMT1/ZCCHC4 proteins, which are involved in methyltransferase activity. The SNP Affx-92301332 linked to the gene TraesCS7B02G144400 encoded a short-chain dehydrogenase/reductase (SDR) involved in oxidoreductase activity. Another gene associated with this SNP, TraesCS7B02G144500, encoded a potassium transporter protein involved in potassium ion transmembrane transporter activity. The SNP Affx-92546089 associated with the candidate gene TraesCS2B02G544700 encoded for the Peptidase C19, a ubiquitin carboxyl-terminal hydrolase protein involved in cysteine-type deubiquitinase activity. The SNP Affx-92640446 linked to the gene TraesCS2A02G314900, encoded a pentatricopeptide repeat protein. Additionally, the gene TraesCS2A02G314800, located near this SNP, encodes an FKBP-type peptidyl-prolyl cis-trans isomerase domain involved in peptidyl-prolyl cis-trans isomerase activity. The SNP Affx-92167066 linked to the gene TraesCS6A02G007300, encoded a SANT/Myb domain protein. Another gene in close proximity to this SNP, TraesCS6A02G007400 encoded the guanine nucleotide-binding protein (G-protein) alpha subunit, which is involved in G-protein beta/gamma-subunit complex binding, GTPase activity, and guanyl nucleotide binding activity (Supplementary Table 5).

Discussion

Integrating yield and quality traits in modern crop breeding is essential for developing high-yielding nutri-rich cultivars to meet global food security and nutritional needs. Mining of genetic resources plays a crucial role in the identification of novel traits and donor germplasm for genetics and genomics studies. The wheat core set has been developed on the basis of characterising the entire wheat collection conserved in the national genebank collection of India³⁸ for further evaluation and utilisation in future breeding programmes. Enhancing wheat quality is a costly and laborious process, resulting in slow progress in quality improvement programmes³⁹. Bread wheat core sets developed in India were evaluated for grain protein content and agronomic traits at multiple environments⁴⁰ for the identification of high protein bread wheat donor genotypes. Subsequently, emmer wheat was studied for agro-morphological and molecular characterisation^41,42, disease resistance⁴³ and terminal heat stress⁴⁴. Emmer wheat is gaining importance in recent days due to its specific properties having high protein, iron, zinc and fibre content; hence, emmer germplasm was studied for grain quality evaluation along with yield traits. Traditionally, wheat yield and quality improvements have relied on phenotype-based selection through conventional breeding. Now, with the advancement of genomic technology in genome sequencing, marker discovery and marker-based breeding potential, genotype-based approaches complement conventional methods in breeding high cultivars with high yield and nutrition. Identification of novel sources of genes and genomic regions associated with grain quality traits in emmer wheat is expected to accelerate development of improved varieties for quality traits in emmer, durum or other wheat species. It can also be utilised in marker assisted backcross breeding to improve quality traits of the elite cultivars that may be deficient in those traits.

Phenotypic variation and multivariate analysis

Although wheat is an important source for GPC, most of the modern varieties are relatively low in grain protein (10–14%) (Giancaspro et al. 2019). The overall mean of the GPC ranged between 12.16 and 17.76% in the emmer wheat germplasm, which was higher than the bread wheat GPC (10.38–13.00) reported by Rathan et al.⁴⁵. Further, wild emmer wheat (Triticum turgidum ssp. dicoccoides) exhibits extensive genotypic variation in agronomic traits, including grain composition⁴⁶ and tolerance to both abiotic and biotic stresses⁴⁷. This rich genetic diversity offers enormous potential for improving key agronomic and economically important traits in durum and bread wheat. Therefore, cultivated emmer wheat can also be utilised as a valuable resource for GPC improvement in common wheat breeding. The accessions EC578064, EC299074 (from Syria), IC32502 (from MP, India), IC138455 (from Jaipur, Rajasthan, India), and EC540812 (from Ukraine) have been identified as promising genotypes with high protein content across various locations. These genotypes can be utilised in breeding programmes to improve wheat protein content. The Percent Coefficient of Variation (CV) is critical in determining the level of variation in germplasm for various traits. Grain Protein Content (GPC) recorded %PCV in the range of 5.95 to 15.62 and Thousand Grain Weight (TGW) had 9.74 to 17.84% PCV, indicating good amount of variation in the emmer germplasm. These findings are consistent with recent studies on grain protein content and yield published by Peleg et al.⁴⁷, Rawat et al.⁴⁸, and Bartosova and Curna⁴⁹.

Breeders frequently face the issue of determining whether observed differences in a certain feature are the result of environmental or genetic factors. Notably, in our analysis, characteristics like GPC and TGW have high heritability (> 60%) for individual locations, indicating the presence of additive gene action which can be improved through selective breeding. However, overall heritability was moderate to high, which highlighted that these traits are majorly governed by genetic factors in spite of GEI interaction. These findings are consistent with those of Cakmak et al.⁵⁰, Gomez-Becerra et al.⁵¹, and Badakhshan et al.⁵², indicating that these qualities can be improved through targeted selection procedures. Kartseva^53,54 also reported high heritability of more than 80% for GPC. Correlation analysis is a valuable tool for determining the relationship between two variables. In this study, a negative correlation was observed between thousand-grain weight (TGW) and grain protein content (GPC). This association may reflect competition for nitrogen during grain filling, resulting in a trade-off that complicates the simultaneous improvement of both traits. Similar findings were reported by Kumar et al.^16,55, Krishnappa et al.¹⁷, and Trivisiol et al. (2024) highlighting the importance of selective breeding. However, Kartseva et al.⁵⁴ found no correlation between these traits. Principal Component Analysis (PCA) found that the total variation of PC1 exceeded 50% in all five environments and thus explaining major portion of variability.

The AMMI analysis provided critical insights into the genotype × environment interaction (G×E) patterns for grain protein content (GPC) and thousand grain weight (TGW), two essential quality traits in wheat. The ability of the AMMI model to partition and interpret both main effects and interaction components makes it a robust approach for identifying genotypes with stable and superior performance across diverse environments^13,14. For grain protein content (GPC), the AMMI1 biplot revealed that a substantial portion of the GEI (47.8%) was explained by the first principal component (PC1), indicating considerable interaction between genotypes and environments. Location K-1 exhibited the highest PC1 score but a relatively low mean GPC, indicating it is an unstable yet highly discriminating environment. It amplifies genotypic differences, making it useful for identifying responsive genotypes, but it is less suitable for achieving high average protein content. Among high grain protein genotypes, IC138455, EC577402, and IC534016, positioned near the origin of the AMMI1 and AMMI2 biplots, demonstrated low interaction effects and stable performance. These genotypes are ideal candidates for broad adaptation, particularly in breeding programmes targeting consistent nutritional quality across diverse environmental conditions¹⁵. Further, EC540812 and IC138450 were specifically adapted due to their significant interaction with high-interaction environments such as K-1 and L-1 emphasising their discriminative potential for identifying specifically adapted lines.t.

For thousand grain weight (TGW), the AMMI1 and AMMI2 biplots explained 48.2% and 26.9% of the GEI variance, respectively, indicating that G×E interaction played an even more substantial role in this trait. Genotypes IC138450, IC32502, and IC32513 recorded high PC1 but were located away from the origin indicating considerable interaction with environments D-2 and L-1, suggesting their potential utility in environments that match their specific adaptive profiles. This findings aligns with the notion that certain yield components are more environment-dependent and thus require targeted deployment strategies⁵⁶. Conversely, genotypes such as EC540812, EC577402, and IC138455, which exhibited moderate TGW and were positioned close to the origin, were identified as stable performers for TGW across locations. Their stability is particularly valuable for breeding programmes aimed at improving yield consistency under variable environmental conditions. Furthermore, the AMMI2 biplot for TGW revealed that IC32513 and IC32502 were specifically suited to D-2. In contrast, genotype IC535086 clustered near the origin, reaffirming its general adaptability and low sensitivity to environmental variation.

Among the environments, D-2 and L-2 were identified as highly discriminative for TGW and GPC, indicating their potential for selecting genotypes with specific responses. These environments can be leveraged in breeding trials aimed at exploring and exploiting G×E interaction. Meanwhile, environments D-1 and P-1 showed stability and representativeness, making them suitable for evaluating broadly adapted genotypes. Collectively, the AMMI analysis supports the identification of stable, high-performing genotypes such as IC138455, EC577402, and IC32513 for GPC, which are crucial traits in wheat quality improvement. The identification of representative and discriminative environments further enhances the efficiency of multi-environment trials, guiding selection under both general and specific adaptation frameworks. These findings contribute significantly to wheat quality breeding efforts by enabling more informed genotype selection decisions, ultimately facilitating the development of nutritionally enriched and environmentally resilient wheat cultivars.

Population structure and linkage disequilibrium

GWAS panel analysis unveiled two distinct subpopulations. The incorporation of PCA in GWAS is instrumental in identifying and mitigating differences arising from population structure by taking them as a covariate, thereby aiding in the elimination of spurious Marker-Trait Associations (MTAs) attributed to such structural variations. A linkage disequilibrium decay block of 1.94 Mb was observed for the whole genome. The relatively slow LD decay implied that even a low marker density suffices to capture markers in proximity to causal loci⁵⁷. Maccaferri,²² reported LD decay at 195 kb in wild emmer, 1.4 Mb in domesticated emmer, 1.6 Mb in durum landraces, and 4.5 Mb in durum cultivars (r² = 0.2 threshold). In contrast, Bassi et al.⁵⁸ observed a slower LD decay distance of 51.3 Mb in durum wheat. Slower LD decay distances of 22 Mb and 23 Mb, respectively, were found by Kokhmetova et al.⁵⁹ and Vikram et al.⁶⁰ in hexaploid wheat collections from Kazakhstan and Mexican bread wheat landraces. The comparatively higher LD decay may be due to differences in marker density and population structure. The A genome decayed fastest, followed by the B genome, consistent with studies in bread wheat by Bhatta et al.⁶¹.

Marker trait associations

Association mapping in emmer wheat for GPC is not reported earlier, although lines derived from emmer wheat have been studied⁶². We have evaluated the GWAS panel in multi-environments, and the BLUP value was estimated for individual environments as well as across the environments to give a better predicted value. Further GEI estimation is an important factor to identify environment-specific and consistent QTLs. We identified a total of 21 unique MTAs for GPC (4) and TGW (17) under all four environments and pooled data threshold of −log10(P) ≥ 3.5. Although more stringent multiple-testing corrections such as Bonferroni adjustment can reduce false positives, they are often overly conservative in studies involving moderate marker density and complex quantitative traits with small effect sizes. Similar approaches have been adopted in previous wheat GWAS investigations³⁷. Marcotulli et al.⁶³, in tetraploid wheat for protein content have also reported a smaller number of markers and also SNP markers significantly associated (LOD ≥ 3) with protein content. To improve reliability, greater emphasis was placed on loci consistently detected across environments and multiple models rather than on nominal significance alone. Maximum MTAs were mapped on the B (14) sub-genome, whereas 7 were mapped on the A sub-genome. The pattern of sub-genome-wise marker distribution is also similar, as maximum markers were located on sub-genome B, whereas minimum number of markers were located on sub-genome A. In the present study, a suggestive threshold of −log10(P) ≥ 3.5 was applied to detect marker–trait associations.

In earlier studies, a comparable pattern of MTAs and marker distribution across sub-genomes for GPC and TGW traits was observed in common wheat^64,17,45. Specifically, for the GPC trait, three MTAs were identified on the same chromosome 2B as reported by Marcotuli et al.⁶⁵, Cui et al.⁶⁶, Cabral et al.⁶⁷, Nigro et al.²⁵, Shi et al.⁶⁸ and on the 5B chromosome, as reported by Nigro et al.²⁵; however one MTA identified on the 3 A chromosome was a novel MTA.

Thousand grain weight is one of the major contributing traits for grain yield. For TGW, 17 MTAs were identified on the chromosomes 1 A, 1B, 2 A, 2B, 3B, 5 A, 5B, 6 A and 7B. Seven MTAs were environment specific (6 for Delhi and 1 for Karnal). Similar results were observed for MTAs located on the chromosome 1B by Cabral et al.⁶⁷, on 2B by Krishnappa et al.⁶⁹, Zhang et al.⁷⁰ in common wheat. Significant MTAs for GPC and TGW were identified across different environments. Both Pune and Ludhiana environments demonstrated the highest number of MTAs (15) based on all the three models for TGW. Whereas for the Karnal location, lowest number of MTA (1) was identified for both GPC and TGW employing the MLMM model. This indicates that the expression of grain traits is influenced by genotype-by-environment interactions, as the expression pattern of the same genotype varied across different geographical regions. Hence, a multi-environmental study is imperative for finding environment specific as well as consistent MTAs. Also, favourable alleles were identified for stable markers for TGW and GPC that can be used for marker development and marker-assisted selection in breeding programmes.

Previous studies have demonstrated that certain loci controlling grain protein content also influence plant senescence and micronutrient accumulation. For instance, the NAM-B1 locus on chromosome 6B has been reported to enhance grain protein content while simultaneously increasing grain iron and zinc concentrations through accelerated nutrient remobilisation⁴⁶. Such evidence indicates that grain protein content may share partially overlapping genetic pathways with senescence and micronutrient traits. Although the present study focused on GPC and TGW, future investigations incorporating multi-trait GWAS or joint analysis of protein, micronutrients, and senescence-related traits would provide a more comprehensive understanding of their shared genetic architecture.

Candidate gene identification

In the present study, candidate gene analysis revealed that several MTAs for grain protein content and thousand grain weight were associated with genes involved in nutrient transport, energy metabolism, stress response, and developmental regulation. Many of these functional categories are known to exert pleiotropic effects on grain yield components and quality traits, suggesting that the detected loci may influence grain traits through shared physiological and regulatory pathways rather than trait-specific mechanisms alone. The significant MTAs were used to predict putative candidate genes using the Ensemble Plants database. SNP Affx-92963286, linked to the putative candidate gene TraesCS2B02G332800, encodes a Major Facilitator Superfamily (MFS) protein which is involved in the transmembrane transport of various substrates and has been implicated in enhancing plant resistance to pathogens. For example, the Arabidopsis gene AT1G08090 (homologous to PtrMFS24) encodes a key component of the high-affinity nitrate transport system in roots, highlighting the functional diversity of MFS proteins⁷¹. MFS transporters involved in nitrate uptake and redistribution can influence nitrogen availability during grain filling, thereby exerting pleiotropic effects on grain protein content and yield-related traits. SNP Affx-92290480, associated with TraesCS5B02G517300, encodes a 2Fe-2 S ferredoxin-type iron–sulphur binding domain. These domains typically feature iron–sulphur clusters coordinated by conserved cysteine residues, facilitating the transport and transfer of iron and sulphur atoms, essential for various metabolic processes. SNP Affx-92754611, identified as stable across environments, is linked to TraesCS5B02G494300, which encodes an NB-ARC domain-containing protein. These proteins are central to plant immune responses and contribute to enhanced disease resistance, which is crucial for maintaining crop health and yield. The NB-ARC domain is associated with ATPase activity and plays a key role in signal transduction pathways involved in plant growth, development, and responses to biotic and abiotic stresses⁷². Although primarily associated with immune signalling, NB-ARC domain-containing proteins can influence carbon and nitrogen allocation under stress, indirectly affecting grain development and yield stability.

SNP Affx-92115317 for TGW linked to the putative candidate gene TraesCS5A02G365100 codes for the MOSC domain in molybdenum cofactor sulphurase (MCSU), is essential for sulphur transfer in metal-sulphur cluster biosynthesis, supporting crucial enzymatic activities in plant metabolism. Their specific roles were not much studied in crops⁷³. SNP Affx-92237374 is linked to putative candidate gene TraesCS1A02G360700, encodes a member of the glycoside hydrolase family 9 (GH9). In wheat, TaGH9s may target miRNAs, including some known miRNAs (miR164 and miR398), which regulate the level of cellulose by light and phytohormone and play important roles in pollen fertility and anther dehiscence^74,75. Further, SNP Affx-92544100 associated with gene TraesCS2B02G314700 encodes a WD40 repeat. TaWD40D encodes a member of the WD40 family proteins and functions as a positive regulator of plant responses to salt stress and osmotic stress in plants. That may be indirectly involved in increase in grain weight as reported by Hu et al.⁷⁶ and Kong et al.⁷⁷. SNP Affx-92811991 linked to the gene TraesCS5B02G395200 codes for a SANT/Myb domain. MYB proteins, which contain the SANT domain, play significant roles in regulating abiotic stress tolerance in plants. TaMYB31-B may play a role in the development of flowering spikes in wheat^78,79. OsSTRL2 is an atypical strictosidine synthase that plays crucial roles in regulating anther development and pollen wall formation in rice⁸⁰. Stable SNP Affx-92286446 linked with gene TraesCS5B02G394800 codes for Cns1/TTC4, a wheel domain. In Arabidopsis, the Cns1/TTC4 proteins have been implicated in long-term acquired thermotolerance, suggesting that they help the plant survive prolonged exposure to high temperatures by stabilising and refolding damaged proteins⁸¹. The association of a thermotolerance-related gene suggests a potential role of heat stress adaptation mechanisms in stabilizing grain weight across environments. SNP Affx-92301332 associated with the gene TraesCS7B02G144400 encodes a short-chain dehydrogenase/reductase (SDR). For instance, certain SDRs have been linked to reduced grain size and weight in knockout lines in rice. This indicates that these enzymes play a role in the processes that determine grain size and weight, which directly impacts TGW⁸². SNP Affx-92640446, is linked with the gene TraesCS2A02G314900, which codes for pentatricopeptide repeat. PPR proteins are primarily involved in RNA processing and stability. PPR18 is required for the cis-splicing of nad4 intron 1 and found to be essential to seed development in maize as reported by Liu et al.⁸³. Another gene TraesCS2A02G314800, codes for an FKBP-type peptidyl-prolyl cis-trans isomerase domain. Wheat protein wFKBP77 transcript abundance was the highest in mature embryos. Properly folded proteins ensure optimal cellular function, contributing to plant growth and yield⁸⁴ (Table 7).

Table 7.

SNPs associated with grain protein content and thousand grain weight, their corresponding candidate genes, predicted protein domains, and known or proposed biological functions.

SNP ID	Chromosome Location	Gene ID	Encoded Domain/Protein	Function / Significance	Reference(s)
Affx-92963286	2B	TraesCS2B02G332800	Major Facilitator Superfamily (MFS)	Involved in transmembrane transport; implicated in plant pathogen resistance. Arabidopsis homolog involved in nitrate transport.	Diao71
Affx-92754611	5B	TraesCS5B02G494300	NB-ARC domain-containing protein	Central to plant immune responses; ATPase activity; involved in biotic/abiotic stress signalling.	Pan72
Affx-92115317	5 A	TraesCS5A02G365100	MOSC domain (Molybdenum cofactor sulphurase)	Supports sulphur transfer in metal–sulphur cluster biosynthesis; essential for enzyme function.	Filiz73
Affx-92237374	1 A	TraesCS1A02G360700	Glycoside Hydrolase Family 9 (GH9)	Involved in cellulose regulation, pollen fertility, anther dehiscence; may interact with miRNAs.	Bai74, Luo75
Affx-92544100	2B	TraesCS2B02G314700	WD40 repeat	Positive regulator of salt and osmotic stress responses; may contribute to grain weight.	Hu76, Kong et al.77
Affx-92811991	5B	TraesCS5B02G395200	SANT/Myb domain	MYB proteins regulate abiotic stress tolerance; involved in spike development in wheat.	Roy78, Zhao79
Affx-92286446	5B	TraesCS5B02G394800	Cns1/TTC4 (wheel domain)	Implicated in long-term acquired thermotolerance by protein refolding under heat stress.	Fernández-Bautista,81
Affx-92301332	7B	TraesCS7B02G144400	Short-chain dehydrogenase/reductase (SDR)	Linked to grain size and weight; affects TGW via metabolic regulation.	Zheng et al.82
Affx-92640446	2 A	TraesCS2A02G314900	Pentatricopeptide Repeat (PPR)	Involved in RNA processing; essential for seed development (e.g., splicing of nad4 intron in maize).	Liu83

Overall, the candidate genes identified in this study predominantly belong to functional categories associated with nutrient transport, redox and energy metabolism, stress response, and reproductive development. These processes are inherently interconnected and are known to influence multiple agronomic and quality traits simultaneously. The results therefore support a multi-trait and pleiotropic genetic architecture underlying grain protein content and thousand grain weight, emphasising the need for integrated selection strategies in wheat breeding.

Conclusion

The present study revealed substantial phenotypic variation for grain protein content (GPC) and thousand-grain weight (TGW) among 150 Triticum dicoccum genotypes evaluated across multiple environments. Both traits exhibited quantitative inheritance with moderate to high heritability, along with a consistent negative correlation, indicating a breeding trade-off. Genotypes EC578064, EC299074, IC32502, IC138455, and EC540812 were identified as promising donors for GPC, whereas IC138472, IC78706, EC299074, EC577411, and IC252486 were identified for TGW. Genome-wide association analysis detected 21 significant marker–trait associations, including four for GPC and seventeen for TGW, of which nine were stable across environments. These stable loci were associated with candidate genes involved in nutrient transport, metabolic processes, stress response, and grain development. Overall, the results highlight the complex genetic architecture of grain quality and yield-related traits in emmer wheat and underscore the importance of integrating multi-environment phenotyping with molecular approaches. The identified donor genotypes and stable MTAs provide a basis for validation and their potential use in marker-assisted breeding. This innovative approach may lead to significant improvements in breeding programmes, facilitating the development of emmer wheat varieties that not only thrive in diverse environmental conditions but also meet the growing demands for high yielding nutri-rich varieties. Future research should focus on refining these markers and exploring their applicability across other wheat species to enhance global food security^85,86.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

GPC

Grain Protein Content

TGW

Thousand Grain Weight

GWAS

Genome-Wide Association Studies

MTA

Marker Trait Associations

SNP

Single Nucleotide Polymorphism

QTL

Quantitative Trait Loci

BLUPs

Best Linear Unbiased Predictors

CTAB

Cetyltrimethylammonium Bromide

MAF

Minor Allele Frequency

LD

Linkage Disequilibrium

MCMC

Monte Carlo Markov Chain

GAPIT

Genome Association and Prediction Integrated Tool

MLMM

Multiple Loci Mixed Model

BLINK

Bayesian-Information and Linkage-Disequilibrium Iteratively Nested Keyway

FARMCPU

Fixed and Random Model Circulating Probability Unification

IWGSC

International Wheat Genome Sequencing Consortium

CV

Coefficient of Variations

LSD

Least Significant Difference

PCA

Principal Component Analysis

Author contributions

Conceptualization, resources, Investigation, Data curation, Methodology, Formal analysis, Validation, Writing—original draft (JK, UBA); Investigation, Methodology, data analysis, Validation, Writing & editing(RB, UBA, PKN, SM); Conceptualization, Supervision, Writing—review & editing (JK, GPS); Coordinated field trials, editing the manuscript –JK, AG, AS, SN; Conceptualization, genomic data analysis and interpretation—(AKS, SKJ).

Funding

This research was funded jointly by Indian Council of Agricultural Research (ICAR)-Consortium Research Platform on Agrobiodiversity for phenotyping and Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi.

Data availability

The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The germplasm used in this study was sourced from ICAR-NBPGR and handled in compliance with institutional and national biodiversity regulations.