Genome-wide association study identifies loci controlling grain yield and protein content in mungbean (Vigna radiata (L.) R. Wilczek var. radiata) using DArTseq SNP markers Authors: Chiamaka C. Obasi, Kwadwo Ofori, Abraham A. Ngwuta, Emmanuel A. Adjei, Ramakrishnan M. Nair, Sunday N. Obasi, John S. Y. Eleblu, Eucharia A. Obidiebube, Eugene O. Ojua, Joseph I. Ulasi, Henry Okolie, Ya-Ping Lin Published: 2026 ABSTRACT BackgroundMungbean (Vigna radiata (L.) R. Wilczek var. radiata) is an important grain legume valued for its high protein content and ability to improve soil fertility through biological nitrogen fixation. Despite its growing importance in tropical agriculture, genetic improvement of yield and protein content remains limited due to insufficient knowledge of the genomic regions controlling these traits. Genome-wide association studies (GWAS) using high-density single nucleotide polymorphism (SNP) markers provide a powerful approach to dissect the genetic architecture of complex traits in crop species. This study aimed to identify genomic loci associated with grain yield and protein content in mungbean using Diversity Arrays Technology sequencing (DArTseq) SNP markers.ResultsA panel of 120 mungbean genotypes was evaluated across four environments (Awka dry season, Awka rainy season, Uyo dry season, and Uyo rainy season), revealing substantial phenotypic variation: grain yield ranged from 24.00 to 7810.90 kg ha-1 (means 229.10–2333.66 kg ha-¹) with strong genotype × environment (G×E) interactions, while protein content varied narrowly (22.10–24.25%; means ~ 23%). Genotyping using DArTseq generated 24,870 SNPs, filtered to 5,037 high-quality markers across 11 chromosomes (avg. PIC 0.27, MAF 0.24). Population structure analysis identified two major genetic clusters (ΔK = 2), PCA explained ~ 23% variation, and LD decayed to r²=0.2 at ~ 368 kb. Multi-locus GWAS (mrMLM, FASTmrMLM, etc.) identified 15 significant marker-trait associations (MTAs) for protein content (Chr1-2908989, Chr3-7994659, Chr2-809453, Chr5-3185830; genes in amino acid metabolism, transcription regulation [LOB/SRS factors], protein degradation [F-box], ribosomal function) and 15 for yield (Chr3-9213031, Chr3-6148955, Chr4-19860552/Chr4-19961226, Chr5-16189831, Chr6-30896900; genes in carbon metabolism [methyltransferase], brassinosteroid signaling [BKI1], cell wall development [fasciclin-like], ATPase/energy metabolism, thylakoid/photosynthesis [THF1]). Environment-specific MTAs highlighted G×E effects, with Q-Q plots confirming model control.ConclusionsThe study reveals the polygenic, environment-dependent genetic architecture of yield (strong G×E) and stable protein content in mungbean, with moderate LD enabling high-resolution mapping. Identified MTAs, distributed across chromosomes, and candidate genes (transcriptional/protein metabolism for protein; hormonal/cellular/energy pathways for yield) provide robust genomic resources for marker-assisted selection (MAS) and breeding high-yielding, nutritionally enhanced varieties adapted to diverse agro-ecologies.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-026-09201-6. Background Mungbean (Vigna radiata (L.) R. Wilczek var. radiata) is an important grain legume widely cultivated in Asia and increasingly gaining importance in Africa due to its short growth cycle, nutritional value, and adaptability to diverse agro-ecological conditions [1]. The crop serves as a major source of plant-based protein and essential micronutrients, particularly in regions where access to animal protein is limited [2]. In addition to its nutritional value, mungbean contributes to sustainable agricultural systems through biological nitrogen fixation, which improves soil fertility and enhances productivity in intercropping systems [3, 4]. Consequently, improving yield and seed protein content in mungbean has become an important objective in breeding programs aimed at enhancing food and nutritional security. Despite its agronomic potential, mungbean productivity remains relatively low in many regions due to limited genetic improvement and insufficient understanding of the genetic basis of key agronomic traits such as yield and protein content. Yield is a complex quantitative trait influenced by multiple genes and environmental factors, while seed protein concentration is a key determinant of nutritional quality. The simultaneous improvement of these traits requires a deeper understanding of their genetic architecture and the identification of genomic regions controlling trait variation. Recent studies have demonstrated that genomic approaches can accelerate the identification of loci associated with agronomic and nutritional traits in Mungbean [5–7]. Genetic diversity within crop germplasm provides the foundation for crop improvement and breeding. Assessing the extent of genetic variation and population structure among mungbean accessions is essential for identifying useful parental lines and maximizing genetic gain in breeding programs. Molecular marker technologies, particularly single nucleotide polymorphism (SNP) markers, have significantly improved the ability to characterize genetic diversity and population structure in crop species [8, 9]. Previous studies have reported substantial genetic variation among mungbean germplasm collections, emphasizing the importance of exploiting this diversity in breeding programs [10, 11]. High-throughput genotyping platforms such as Diversity Arrays Technology sequencing (DArTseq) have further enhanced genome-wide marker discovery and enabled large-scale genetic analysis in crop species [12]. Genome-wide association studies (GWAS) have emerged as a powerful approach for dissecting the genetic basis of complex traits in plants. Unlike traditional linkage mapping, GWAS utilizes historical recombination events within diverse germplasm panels to detect marker–trait associations at higher resolution. The integration of high-density SNP markers with advanced statistical models has improved the detection of quantitative trait loci (QTLs) associated with agronomic traits in many crops. Multi-locus GWAS methods, such as the multi-locus random-SNP-effect mixed linear model (mrMLM), increase statistical power and reduce false-positive associations by simultaneously considering multiple loci while controlling for population structure and kinship [13, 14]. Recent GWAS studies in mungbean have identified genomic regions associated with important agronomic traits including seed size, flowering time, disease resistance, and nutrient-use efficiency [15–17] For example [6], reported several loci associated with seed size traits that contribute to yield improvement, while [5] identified genomic regions controlling key agronomic traits through large-scale resequencing of mungbean landraces. These studies demonstrate the potential of genomic tools to accelerate mungbean improvement through marker-assisted selection (MAS). However, information on genomic regions controlling yield and protein content in diverse mungbean germplasm remains limited, particularly under different environmental conditions. Understanding population structure and linkage disequilibrium patterns is essential for accurate GWAS analysis. Population stratification can lead to spurious marker–trait associations if not properly accounted for, while the extent of linkage disequilibrium determines the resolution of association mapping. Previous studies have shown that mungbean germplasm exhibits varying levels of population structure and linkage disequilibrium depending on geographical origin and breeding history [18]. Therefore, integrating analyses of genetic diversity, population structure, and linkage disequilibrium with GWAS provides a comprehensive framework for identifying genomic regions controlling complex traits. Earlier mungbean GWAS studies largely focused on single traits such as seed size, flowering time, or agronomic traits evaluated under limited environments or using single-locus approaches. Han et al. [5] identified loci for agronomic traits using re-sequenced germplasm, while Liu et al. [6] focused primarily on seed size traits. Similarly, Akhtar et al. [15] and Reddy et al. [16] reported genomic regions associated with agronomic and yield-related traits, but these studies did not comprehensively evaluate protein content together with grain yield across contrasting seasonal and geographical environments. Therefore, in this study, a panel of mungbean genotypes was evaluated across multiple environments and genotyped using DArTseq-derived SNP markers to investigate the genetic basis of yield and protein content. The objectives were to (i) assess genetic diversity and population structure within the mungbean panel, (ii) examine linkage disequilibrium patterns across the genome, and (iii) identify significant marker–trait associations for yield and protein content through genome-wide association analysis. The results provide valuable genomic resources for marker-assisted selection and support the development of high-yielding and nutritionally improved mungbean varieties adapted to diverse agro-ecological environments. Materials and methods Plant materials and experimental design A panel of 120 mungbean (Vigna radiata (L.) R. Wilczek var. radiata) genotypes was obtained from the World Vegetable Center (WorldVeg) under the International Mungbean Improvement Network (IMIN) programme, as part of the WorldVeg mini-core collection (Table S1). This collection represents a genetically diverse subset of mungbean germplasm conserved and distributed for research and breeding purposes. The plant materials were provided with accompanying passport data and characterization information, and their identity was confirmed based on these records. No plant materials were collected from the wild, and the study was conducted in accordance with relevant institutional and international guidelines governing the use of plant genetic resources. The genotypes were evaluated across four environments representing contrasting agro-ecological conditions: Awka (dry season), Awka (rainy season), Uyo (dry season), and Uyo (rainy season). These environments were selected to capture environmental variability influencing agronomic performance and to enable assessment of genotype × environment interactions affecting yield and protein content. Awka (6.21°N, 7.07°E) and Uyo (5.05°N, 7.93°E) represent humid tropical agro-ecologies with bimodal rainfall patterns. Annual rainfall ranges from 1800 to 2500 mm, with mean temperatures of 25–30 °C. Experiments were conducted during the 2022 and 2023 cropping seasons on sandy loam soils. Field experiments were established using an alpha-lattice design with two replications to improve experimental precision and minimize environmental error. Each plot consisted of two rows of 3 m length, with inter-row spacing of 50 cm and intra-row spacing of 20 cm. Standard agronomic practices for mungbean cultivation were applied uniformly across all locations and seasons to ensure optimal crop establishment and growth. Leaf sampling and DNA extraction Leaf samples for genotyping were collected approximately five weeks after planting. Young leaf tissues from three tagged plants per genotype were punched using a leaf puncher and placed into 96-well plates. The samples were oven-dried at 80 °C and shipped to Diversity Arrays Technology Pty Ltd (DArT P/L), Canberra, Australia, for genotyping. Genomic DNA was extracted using the QIAGEN DNeasy Plant Micro Kit (Qiagen, Valencia, CA, USA) following the manufacturer’s instructions. DNA integrity was assessed by 0.8% agarose gel electrophoresis using lambda DNA standards. DNA concentration was quantified using a Qubit fluorometer, while purity was evaluated spectrophotometrically at 230, 260, and 280 nm. DNA samples with concentrations exceeding 100 ng µL⁻¹ and 260/280 absorbance ratios close to 1.8 were retained for genotyping. SNP discovery and DArTseq genotyping Genotyping was performed using DArTseq™ technology, a genotyping-by-sequencing platform that combines genome complexity reduction with next-generation sequencing to simultaneously discover and genotype genome-wide single nucleotide polymorphisms (SNPs) [19, 20]. The method involves restriction enzyme digestion of genomic DNA followed by adaptor ligation and selective amplification of informative genomic fragments. Complexity reduction in DArTseq involves restriction enzyme digestion; however, specific enzyme combinations and sequencing depth parameters are proprietary to DArT P/L. Sequencing was carried out on an Illumina platform, and SNP calling and primary quality filtering were performed by Diversity Arrays Technology Pty Ltd (Canberra, Australia). SNP markers generated using the DArTseq platform were provided with chromosomal positions based on the mungbean reference genome implemented within the DArT pipeline [19]. These positions were used for genome-wide association analysis and candidate gene identification. A total of 24,870 SNP markers were initially generated for the mungbean panel. SNP filtering and marker quality control Quality filtering was conducted to retain high-confidence markers suitable for downstream analyses. SNP markers were filtered according to the following criteria:Call rate ≥ 80%Minor allele frequency (MAF) ≥ 0.05Reproducibility ≥ 95% Call rate ≥ 80% Call rate ≥ 80% Minor allele frequency (MAF) ≥ 0.05 Minor allele frequency (MAF) ≥ 0.05 Reproducibility ≥ 95% Reproducibility ≥ 95% Markers with excessive missing data or low polymorphism were removed from the dataset. The filtered SNP dataset was subsequently used for population genetic analyses, linkage disequilibrium estimation, and genome-wide association studies. Determination of protein content Seed protein content was determined using the Kjeldahl method following established procedures [5, 21]. Approximately 2 g of finely ground seed sample was digested with 5 g Kjeldahl catalyst (anhydrous sodium sulfate) and 25 mL concentrated sulfuric acid (H₂SO₄) until a clear solution was obtained. After digestion and cooling, the solution was diluted to 250 mL with distilled water. During distillation, 5 mL of digest was mixed with 5 mL of 60% sodium hydroxide solution, releasing ammonia that was absorbed in 2% boric acid solution containing methyl red indicator. The distillate was titrated against 0.049 M sulfuric acid until the endpoint was reached. A reagent blank was included in each batch. A nitrogen-to-protein conversion factor of 6.25 was used as a standard approach in legume studies, although it may slightly overestimate protein content in mungbean. Determination of yield and yield components At physiological maturity, plants were harvested from the net plot area to determine grain yield and associated yield components. Grain yield per plot was recorded in grams and converted to kg ha-¹. Hundred-seed weight (100-SW) was determined by counting and weighing 100 seeds in triplicate. Border rows were excluded during harvest to minimize edge effects and ensure accurate yield estimation. Phenotypic data analysis Phenotypic data from the four environments were analyzed using R software (version 4.2.1) [22]. Adjusted genotype values were estimated using linear mixed models implemented in the agricolae package [23]. Phenotypic observations for grain yield and seed protein content were first subjected to environment-specific mixed-model analysis to correct for field design effects. Within each environment, replication and incomplete block effects were fitted as random terms, while genotype effects were estimated after adjustment for local spatial heterogeneity. This procedure improved precision by reducing residual variance attributable to block-to-block differences. Subsequently, a combined analysis across environments was performed using linear mixed model. The mixed linear model used for combined analysis across environments was:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:Yijk=\mu\:+Gi+Ej+\left(GxE\right)ij+Rk\left(Ej\right)+\epsilon\:ijk$$\end{document} Where: Yijk = observed phenotype µ = overall mean Gi = genotype effect Ej = environment effect (G×E)ij = genotype × environment interaction Rk(Ej) = replication nested within environment εijk = residual error In the model, block effects and test genotype effects were specified as random factors, whereas check varieties were fitted as fixed effects. Because the test genotypes were modeled as random effects, the resulting adjusted genotype estimates are Best Linear Unbiased Predictions (BLUPs). In contrast, estimates obtained for the fixed check varieties correspond to Best Linear Unbiased Estimates (BLUEs). Across-environment genotype BLUPs were extracted and subsequently used as phenotypic inputs for the genome-wide association study (GWAS), as BLUPs provide shrinkage-adjusted predictions that improve precision under unbalanced data structures and multi-environment trials. Broad-sense heritability (H²) was estimated across environments using variance components derived from the mixed linear model as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{H}^{2}=\frac{{\upsigma\:}2\mathrm{G}}{{\upsigma\:}2\mathrm{G}+\frac{{\upsigma\:}2\mathrm{G}\mathrm{E}}{\mathrm{e}}+\frac{{\upsigma\:}2\mathrm{e}}{\mathrm{r}\mathrm{e}}}$$\end{document} Where, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\upsigma\:}2\mathrm{G}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\upsigma\:}2\mathrm{G}\mathrm{E}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\upsigma\:}2\mathrm{e}$$\end{document} represent genotypic, genotype × environment, and residual variances, respectively; e is the number of environments and r the number of replications. Population structure analysis Population structure among mungbean genotypes was inferred using a Bayesian clustering approach implemented in STRUCTURE v2.3.4 [24]. The analysis employed an admixture model with correlated allele frequencies, allowing individuals to have mixed ancestry from multiple genetic clusters. The number of potential clusters (K) was evaluated from K = 1 to K = 10, with ten independent runs performed for each K value. Each run consisted of a burn-in period of 100,000 iterations followed by 100,000 Markov Chain Monte Carlo (MCMC) iterations. The most probable number of clusters was determined using the ΔK method of [25] implemented in Structure Harvester. Membership coefficients were averaged across replicate runs to minimize label switching and ensure stable cluster assignment. Principal Component Analysis (PCA) To further explore genetic relationships among genotypes without prior population assignment, principal component analysis (PCA) was conducted using GenAlEx v6.5 [26]. PCA provides a complementary multivariate approach for visualizing genetic variation and population stratification within the germplasm panel. The first two principal components were used to generate PCA biplots, illustrating clustering patterns and genetic relationships among mungbean genotypes. Linkage disequilibrium analysis Genome-wide linkage disequilibrium (LD) patterns were analyzed to determine the extent of non-random association between SNP markers across the mungbean genome. LD was estimated using the squared correlation coefficient (r²) between pairs of SNP markers. An LD heatmap was generated to visualize the strength of linkage among markers across the genome. Genome-Wide Association Study (GWAS) Genome-wide association analysis was conducted using five multi-locus models implemented in the mrMLM R package [27]: mrMLM [13], FASTmrMLM [28], FASTmrEMMA [14], pLARmEB [27], ISIS EM-BLASSO [29]. These models improve the detection of marker–trait associations by incorporating polygenic background control and empirical Bayes estimation, thereby reducing false-positive associations and increasing statistical power. For declaring significant marker–trait associations (MTAs), a threshold of LOD ≥ 3.0 (equivalent to approximately − log₁₀(P) ≥ 3.0, P ≤ 0.001) was applied, following the default recommendation for multi-locus GWAS models in the mrMLM framework and previous association mapping studies. This threshold provides a balance between minimizing false positives and maintaining adequate power to detect loci of moderate effect. Only SNPs exceeding this criterion were considered significant and retained for further interpretation. Quantile–quantile (Q–Q) plots were used to evaluate the effectiveness of the models in controlling population structure and false discovery rates. Manhattan plots were generated to visualize the genomic distribution of significant marker–trait associations. Population structure information derived from STRUCTURE analysis was incorporated into the GWAS models to reduce spurious associations arising from population stratification. Candidate gene identification and functional annotation Following the identification of significant marker–trait associations, zoom mapping was performed to refine genomic regions surrounding associated SNP markers. Putative candidate genes located within these regions were identified using General Feature Format (GFF3) genome annotation files [30]. Functional annotation of candidate genes was performed using the InterPro database of the European Molecular Biology Laboratory–European Bioinformatics Institute (EMBL-EBI) to infer biological processes potentially associated with yield and protein accumulation [31]. Candidate genes were identified within a ± 368 kb window around each significant SNP, based on the observed genome-wide LD decay distance. Results Phenotypic variation and analysis of variance for studied traits Grain yield and protein content exhibited considerable phenotypic variation among mungbean genotypes across the four environments (Table 1). Grain yield ranged from 24.00 to 7810.90 kg ha⁻¹, with the lowest mean yield recorded in Awka Dry (229.10 kg ha⁻¹) and the highest in Uyo Rain (2333.66 kg ha⁻¹). Substantial variability in yield was further reflected by the relatively large standard deviations observed across environments. In contrast, protein content showed relatively narrow variation, ranging from 22.10% to 24.25%, with mean values consistently around 23% across environments, indicating greater stability compared to grain yield. Combined analysis of variance revealed highly significant (p < 0.001) effects of genotype (G), environment (E), and genotype × environment interaction (G×E) for both traits (Table 2). The environmental effect accounted for the largest proportion of variation for grain yield, highlighting the strong influence of environmental conditions on trait expression. Notably, the G×E interaction mean square exceeded the genotypic mean square for grain yield, indicating substantial variation in genotype performance across environments. Table 1Descriptive statistics for yield and protein content across four environmentsEnvironmentTraitMinimumMaximumMean ± SDᵃAwka DryYield (kg ha⁻¹)24.001100.00229.10 ± 192.61Awka RainYield (kg ha⁻¹)775.082879.161403.98 ± 406.74Uyo DryYield (kg ha⁻¹)27.605649.10465.88 ± 439.54Uyo RainYield (kg ha⁻¹)283.507810.902333.66 ± 1393.52Awka DryProtein (%)22.1024.1023.02 ± 0.63Awka RainProtein (%)22.2324.2523.02 ± 0.63Uyo DryProtein (%)22.3524.2523.00 ± 0.63Uyo RainProtein (%)22.3524.2523.04 ± 0.63ᵃSD Standard deviation Table 1Descriptive statistics for yield and protein content across four environmentsEnvironmentTraitMinimumMaximumMean ± SDᵃAwka DryYield (kg ha⁻¹)24.001100.00229.10 ± 192.61Awka RainYield (kg ha⁻¹)775.082879.161403.98 ± 406.74Uyo DryYield (kg ha⁻¹)27.605649.10465.88 ± 439.54Uyo RainYield (kg ha⁻¹)283.507810.902333.66 ± 1393.52Awka DryProtein (%)22.1024.1023.02 ± 0.63Awka RainProtein (%)22.2324.2523.02 ± 0.63Uyo DryProtein (%)22.3524.2523.00 ± 0.63Uyo RainProtein (%)22.3524.2523.04 ± 0.63ᵃSD Standard deviation Descriptive statistics for yield and protein content across four environments ᵃSD Standard deviation Table 2Combined ANOVA for yield and protein contentSource of VariationdfYield (Mean Square)Protein Content (Mean Square)Genotype (G)1191.125 × 108 ***70.16 ***Environment (E)36.749 × 108 ***0.22 *G × E3572.445 × 108 ***19.02 *Residual4802.202 × 1026.67Significance: *** p < 0.001 Table 2Combined ANOVA for yield and protein contentSource of VariationdfYield (Mean Square)Protein Content (Mean Square)Genotype (G)1191.125 × 108 ***70.16 ***Environment (E)36.749 × 108 ***0.22 *G × E3572.445 × 108 ***19.02 *Residual4802.202 × 1026.67Significance: *** p < 0.001 Combined ANOVA for yield and protein content Significance: *** p < 0.001 Although genotype and G×E effects were significant for protein content, the magnitude of variation was comparatively smaller than that observed for grain yield, consistent with the narrower phenotypic range across environments. Broad-sense heritability (H²) estimates were 0.28 for grain yield and 0.90 for protein content, indicating low to moderate genetic control for yield and strong genetic control for protein content. SNP marker statistics and genome coverage A total of 24,870 single nucleotide polymorphism (SNP) markers were generated using the DArTseq genotyping platform. After quality filtering based on call rate, reproducibility, and minor allele frequency thresholds, 5,037 high-quality SNP markers were retained for downstream analysis. These markers were distributed across the 11 chromosomes of the mungbean genome, providing adequate genome coverage for association mapping. Chromosome-wise distribution of SNP markers revealed uneven marker density across the genome. Chromosome 1 contained the highest number of SNP markers (689; 13.68%), followed by chromosome 8 (649; 12.88%) and chromosome 6 (581; 11.53%), whereas chromosome 10 had the lowest number of markers (287; 5.70%). The average polymorphic information content (PIC) across markers was 0.27, indicating moderate marker informativeness. Minor allele frequency (MAF) values ranged from 0.23 to 0.27, with an average of 0.24. Observed heterozygosity (Ho = 0.13) was lower than expected heterozygosity (He = 0.33), reflecting the predominantly self-pollinating reproductive system of mungbean. Overall, the genome-wide distribution of SNP markers suggests that the marker dataset provides adequate coverage for detecting marker–trait associations through genome-wide association analysis. Detailed chromosome-wise statistics of SNP markers and their distribution across the mungbean chromosomes is presented in Table 3. The filtered SNP dataset was subsequently used for population structure analysis, linkage disequilibrium estimation, and genome-wide association mapping. Table 3Summary of SNP markers used for genome-wide analysesChrSNPSGDPICMAFHoHe16890.330.270.240.050.3424100.330.270.240.100.3232960.360.290.270.040.3643470.340.270.250.230.3354450.330.270.240.120.3365810.330.270.250.110.3575340.320.260.230.100.3186490.320.260.230.240.3194030.350.280.270.120.37102870.320.260.230.150.32113960.310.250.230.220.32Total/Average50370.330.270.240.130.33 Table 3Summary of SNP markers used for genome-wide analysesChrSNPSGDPICMAFHoHe16890.330.270.240.050.3424100.330.270.240.100.3232960.360.290.270.040.3643470.340.270.250.230.3354450.330.270.240.120.3365810.330.270.250.110.3575340.320.260.230.100.3186490.320.260.230.240.3194030.350.280.270.120.37102870.320.260.230.150.32113960.310.250.230.220.32Total/Average50370.330.270.240.130.33 Summary of SNP markers used for genome-wide analyses Population structure of the mungbean germplasm panel Figures 1 and 2 indicate clear population stratification among the 120 mungbean genotypes. The ΔK analysis identified K = 2 as the optimal number of subpopulations, confirming the presence of two major genetic clusters. STRUCTURE bar plots show that most genotypes belong predominantly to one cluster, while a few admixed lines suggest shared ancestry. PCA further validated this grouping, with PC1 (15.23%) and PC2 (7.98%) explaining 23.21% of total variation and clearly separating the genotypes into two distinct groups. Fig. 1a ΔK plot showing optimal K value; (b) STRUCTURE bar plots for K = 2 a ΔK plot showing optimal K value; (b) STRUCTURE bar plots for K = 2 Fig. 2PCA Biplot of 120 mungbean genotypes based on SNP markers PCA Biplot of 120 mungbean genotypes based on SNP markers Linkage disequilibrium patterns in the mungbean genome Genome-wide linkage disequilibrium (LD) analysis revealed the extent of non-random association between SNP markers across the mungbean genome. The LD heatmap (Fig. 3) showed varying levels of linkage among markers, with stronger LD observed among closely linked loci and weaker LD between distant markers. The LD decay curve (Fig. 4) indicated a gradual decline in LD with increasing physical distance between markers. The average genome-wide LD decayed to r² = 0.2 at approximately 368 kb, suggesting moderate LD extent in the studied mungbean panel. This level of LD decay indicates that the marker density used in this study is sufficient for detecting marker–trait associations with reasonable mapping resolution. The observed LD pattern is consistent with expectations for predominantly self-pollinating species, where reduced recombination leads to extended LD blocks. Overall, the LD structure observed in this study supports the reliability of subsequent genome-wide association analysis and provides insight into the genomic resolution achievable for identifying loci associated with grain yield and protein content. Fig. 3Linkage Disequilibrium (LD) heatmap of mungbean genome used for the study Linkage Disequilibrium (LD) heatmap of mungbean genome used for the study Fig. 4Genome-wide LD decay curve showing the decline of LD with physical distance (r² = 0.2 at approx. 368 kb) Genome-wide LD decay curve showing the decline of LD with physical distance (r² = 0.2 at approx. 368 kb) Genome-wide association scans for studied traits Genome-wide association analysis was conducted to identify SNP markers associated with grain yield and protein content across four environments (Awka Dry, Awka Rain, Uyo Dry, and Uyo Rain) using multiple multi-locus GWAS models, including mrMLM, FASTmrMLM, FASTmrEMMA, pLARmEB, and ISIS EM-BLASSO. Combined Manhattan and quantile–quantile (Q–Q) plots for protein content and grain yield across the four environments are presented in Figs. 5 and 6, respectively. Panels (a–d) correspond to Awka multiple multi-locus GWAS models, including mrMLM, FASTmrMLM, FASTmrEMMA, pLARmEB, and ISIS EM-BLASSO. Combined Manhattan and quantile–quantile (Q–Q) plots for protein content and grain yield across the four environments are presented in Figs. 5 and 6, respectively. Panels (a–d) correspond to Awka Dry, Awka Rain, Uyo Dry, and Uyo Rain environments. Across all environments, most SNP markers followed the expected null distribution, while significant deviations were observed for loci associated with the studied traits, indicating effective control of false-positive associations. A total of 15 unique markers–trait associations (MTAs) were identified for protein content and 15 MTAs for grain yield across all environments (Table 4). The distribution of significant SNPs varied across environments, reflecting environmental influences on trait expression. Several loci were detected in specific environments, while others showed consistency across different analytical models, suggesting stable genomic regions associated with the studied traits. Fig. 5a Manhattan and Q–Q plots EVT1 Awka dry Protein b Manhattan and Q–Q plots EVT1 Awka dry yield c Manhattan and Q–Q plots EVT2 Awka rain protein d Manhattan and Q–Q plots EVT2 Awka rain yield Fig. 6a Manhattan and Q-Q plots EVT3 Uyo dry Protein. b Manhattan and Q-Q plots EVT3 Uyo dry yield. c Manhattan and Q-Q plots EVT4 Uyo rain Protein. d Manhattan and Q-Q plots EVT4 Uyo rain yield a Manhattan and Q–Q plots EVT1 Awka dry Protein b Manhattan and Q–Q plots EVT1 Awka dry yield c Manhattan and Q–Q plots EVT2 Awka rain protein d Manhattan and Q–Q plots EVT2 Awka rain yield a Manhattan and Q-Q plots EVT3 Uyo dry Protein. b Manhattan and Q-Q plots EVT3 Uyo dry yield. c Manhattan and Q-Q plots EVT4 Uyo rain Protein. d Manhattan and Q-Q plots EVT4 Uyo rain yield Table 4An overview of the major SNPs in a panel of 120 mungbean genotypes that describe various genomic regions linked to the traits under study for four environmentsTraitsEnvironmentMethodsMarkersChrPos (Mbp)Effect-Log10(P)R²MAFVar(error)ProteinAwka DrymrMLMChr1_290898912.9090.08724.757112.710.48310.0447ProteinAwka DrymrMLMChr3_799465937.995-0.08764.690512.490.41530.0447ProteinAwka DryFASTmrMLMChr6_17778739617.7794.00E-043.75120.000.20830.0472ProteinAwka DrypLARmEBChr1_12963638112.964-0.06393.80124.980.2250.0474ProteinAwka RainypLARmEBChr2_80945320.8098.00E-044.6420.000.1500.0925ProteinAwka RainypLARmEBChr5_318583053.1860.05916.20642.840.31670.0925ProteinAwka RainypLARmEBChr5_25318711525.3190.05533.97372.860.45830.0925ProteinAwka RainyISIS EM-BLASSOChr8_30664570830.6650.11685.658710.410.28330.0889ProteinUyo DrypLARmEBChr6_56950360.570-0.0754.54252.340.1250.0814ProteinUyo DrypLARmEBChr6_360508263.6052.90E-054.95970.000.20830.0814ProteinUyo DrypLARmEBChr8_31395733831.3960.09484.43556.680.26670.0814ProteinUyo DryISIS EM-BLASSOChr6_117876761.1791.00E-044.720.000.0500.097ProteinUyo DryISIS EM-BLASSOChr7_54291218754.2911.00E-043.78690.000.05830.097ProteinUyo RainypLARmEBChr2_24615287224.6151.39E-054.02980.000.39170.0868ProteinUyo RainypLARmEBChr6_31253548631.2540.07044.91754.070.31670.0868YieldAwka DrymrMLMChr3_921303139.213-47.9044.2178.130.203416586.58YieldAwka DryFASTmrMLMChr4_19961226419.961-59.4604.66017.560.108315122.58YieldAwka RainymrMLMChr3_614895536.149-261.4024.166712.280.3136422006.4YieldAwka RainyISIS EM-BLASSOChr5_15655284515.655204.9574.83925.920.2167363282.5YieldAwka RainyISIS EM-BLASSOChr6_29354314629.354-258.7543.83134.250.0833363282.5YieldUyo DrymrMLMChr2_21263131221.263-256.5096.028413.080.059362749.04YieldUyo DrymrMLMChr2_295655922.957-152.1263.90477.540.093262749.04YieldUyo DrymrMLMChr4_19860552419.86193.3993.85637.400.372962749.04YieldUyo DrymrMLMChr5_16189831516.190-272.4546.032312.760.050862749.04YieldUyo DrypLARmEBChr1_12467544112.46865.3214.40980.570.216754604.7YieldUyo DrypLARmEBChr8_42327370842.327-147.154.96151.060.066754604.7YieldUyo RainymrMLMChr5_587153455.872502.2994.01297.950.0763666349.4YieldUyo RainymrMLMChr6_30896900630.897-448.2013.95299.570.1356666349.4YieldUyo RainymrMLMChr8_44276921844.277-549.8253.89866.850.0593666349.4YieldUyo RainymrMLMChr9_10783673910.784-432.8343.85786.950.1017666349.4 Table 4An overview of the major SNPs in a panel of 120 mungbean genotypes that describe various genomic regions linked to the traits under study for four environmentsTraitsEnvironmentMethodsMarkersChrPos (Mbp)Effect-Log10(P)R²MAFVar(error)ProteinAwka DrymrMLMChr1_290898912.9090.08724.757112.710.48310.0447ProteinAwka DrymrMLMChr3_799465937.995-0.08764.690512.490.41530.0447ProteinAwka DryFASTmrMLMChr6_17778739617.7794.00E-043.75120.000.20830.0472ProteinAwka DrypLARmEBChr1_12963638112.964-0.06393.80124.980.2250.0474ProteinAwka RainypLARmEBChr2_80945320.8098.00E-044.6420.000.1500.0925ProteinAwka RainypLARmEBChr5_318583053.1860.05916.20642.840.31670.0925ProteinAwka RainypLARmEBChr5_25318711525.3190.05533.97372.860.45830.0925ProteinAwka RainyISIS EM-BLASSOChr8_30664570830.6650.11685.658710.410.28330.0889ProteinUyo DrypLARmEBChr6_56950360.570-0.0754.54252.340.1250.0814ProteinUyo DrypLARmEBChr6_360508263.6052.90E-054.95970.000.20830.0814ProteinUyo DrypLARmEBChr8_31395733831.3960.09484.43556.680.26670.0814ProteinUyo DryISIS EM-BLASSOChr6_117876761.1791.00E-044.720.000.0500.097ProteinUyo DryISIS EM-BLASSOChr7_54291218754.2911.00E-043.78690.000.05830.097ProteinUyo RainypLARmEBChr2_24615287224.6151.39E-054.02980.000.39170.0868ProteinUyo RainypLARmEBChr6_31253548631.2540.07044.91754.070.31670.0868YieldAwka DrymrMLMChr3_921303139.213-47.9044.2178.130.203416586.58YieldAwka DryFASTmrMLMChr4_19961226419.961-59.4604.66017.560.108315122.58YieldAwka RainymrMLMChr3_614895536.149-261.4024.166712.280.3136422006.4YieldAwka RainyISIS EM-BLASSOChr5_15655284515.655204.9574.83925.920.2167363282.5YieldAwka RainyISIS EM-BLASSOChr6_29354314629.354-258.7543.83134.250.0833363282.5YieldUyo DrymrMLMChr2_21263131221.263-256.5096.028413.080.059362749.04YieldUyo DrymrMLMChr2_295655922.957-152.1263.90477.540.093262749.04YieldUyo DrymrMLMChr4_19860552419.86193.3993.85637.400.372962749.04YieldUyo DrymrMLMChr5_16189831516.190-272.4546.032312.760.050862749.04YieldUyo DrypLARmEBChr1_12467544112.46865.3214.40980.570.216754604.7YieldUyo DrypLARmEBChr8_42327370842.327-147.154.96151.060.066754604.7YieldUyo RainymrMLMChr5_587153455.872502.2994.01297.950.0763666349.4YieldUyo RainymrMLMChr6_30896900630.897-448.2013.95299.570.1356666349.4YieldUyo RainymrMLMChr8_44276921844.277-549.8253.89866.850.0593666349.4YieldUyo RainymrMLMChr9_10783673910.784-432.8343.85786.950.1017666349.4 An overview of the major SNPs in a panel of 120 mungbean genotypes that describe various genomic regions linked to the traits under study for four environments Marker–trait associations for protein content Genome-wide association analysis identified 15 significant markers–trait associations (MTAs) for protein content across the four environments (Table 4), demonstrating the polygenic and environment-dependent nature of protein accumulation in mungbean. In the Awka Dry environment, four significant SNPs were detected on chromosomes 1, 3, and 6, including Chr1-2908989 and Chr3-7994659, which showed relatively high phenotypic variance explained (R² ≈ 12%). Additional loci such as Chr6-17778739 and Chr1-12963638 were also identified, indicating the contribution of multiple genomic regions to protein variation. In the Awka Rain environment, four MTAs were identified on chromosomes 2, 5, and 8. Notably, Chr5-3185830 exhibited a relatively strong association signal (− log10P = 6.21), while Chr2-809453 and Chr5-25318711 were also associated with protein content, suggesting the involvement of diverse genomic regions under favorable moisture conditions. In the Uyo Dry environment, five MTAs were detected, primarily on chromosomes 6, 7, and 8. These included loci such as Chr6-569503, Chr6-3605082, and Chr8-31395733, indicating substantial genetic variation contributing to protein content under dry conditions. Additional markers identified through ISIS EM-BLASSO further support the robustness of these associations. In the Uyo Rain environment, two MTAs were identified on chromosomes 2 and 6 (Chr2-24615287 and Chr6-31253548), indicating comparatively fewer but significant loci contributing to protein content under this environment. Overall, the identified MTAs were distributed across multiple chromosomes, with varying levels of statistical significance and phenotypic effects, highlighting the complex genetic architecture of protein content in mungbean and the influence of environmental conditions on trait expression. Marker–trait associations for grain yield Genome-wide association analysis identified 15 significant marker–trait associations (MTAs) for grain yield across the four environments (Table 4), indicating a complex and environment-dependent genetic architecture for yield in mungbean. In the Awka Dry environment, two significant MTAs were detected on chromosomes 3 and 4 (Chr3-9213031 and Chr4-19961226), with both loci showing moderate phenotypic effects. In the Awka Rain environment, three MTAs were identified on chromosomes 3, 5, and 6, including Chr3-6148955, Chr5-15655284, and Chr6-29354314. These loci exhibited relatively strong association signals, reflecting the influence of favorable environmental conditions on yield expression. In the Uyo Dry environment, six MTAs were detected across chromosomes 1, 2, 4, 5, and 8. Notable loci included Chr2-21263131 and Chr5-16189831, which showed relatively high −log10(P) values and phenotypic variance explained (R²), indicating their importance in yield determination under dry conditions. Additional loci such as Chr2-2956559, Chr4-19860552, Chr1-12467544, and Chr8-42327370 further highlight the polygenic nature of yield. In the Uyo Rain environment, four MTAs were identified on chromosomes 5, 6, 8, and 9, including Chr5-5871534, Chr6-30896900, Chr8-44276921, and Chr9-10783673, suggesting multiple genomic regions contributing to yield under high-moisture conditions. Overall, the identified MTAs for grain yield were distributed across multiple chromosomes, with varying levels of statistical significance, effect sizes, and phenotypic variance explained, reflecting the quantitative nature of yield and its sensitivity to environmental variation. Identification of candidate genes associated with yield and protein content Putative candidate genes associated with significant marker–trait associations (MTAs) for grain yield and protein content were identified based on their proximity to significant SNPs (Table 5). The identified genes were distributed across multiple chromosomes and were functionally diverse, reflecting the complex genetic architecture underlying the studied traits. Table 5Gene identification for the significant SNPs for studied traitsEnvironmentTraitMarkerChrPosition (bp)Putative FunctionReferenceAwka DryProteinChr1_290898912,908,989EC 6.2 ligase (amino acid metabolism)[32]Awka DryProteinChr3_799465937,994,659MRN complex protein (DNA repair)[33]Awka DryProteinChr6_17778739617,778,739LOB domain transcription factor[14]Awka DryProteinChr1_12963638112,963,638LRR receptor-like kinase[34]Awka DryYieldChr3_921303139,213,031Methyltransferase (carbon metabolism)[35]Awka DryYieldChr4_19961226419,961,226Uncharacterized protein—Awka RainProteinChr2_8094532809,453SRS transcription factor[36]Awka RainProteinChr5_318583053,185,830F-box protein (protein degradation)[37]Awka RainProteinChr5_25318711525,318,711Transcription elongation factor[38]Awka RainProteinChr8_30664570830,664,570Ribosomal protein L4[39]Awka RainYieldChr3_614895536,148,955Polynucleotide phosphorylase (PNP)[40]Awka RainYieldChr5_15655284515,655,284Embryogenesis abundant protein[41]Awka RainYieldChr6_29354314629,354,314Microtubule-associated protein (TPX2-like)[42]Uyo DryProteinChr6_5695036569,503Zinc finger transcription factor[14]Uyo DryProteinChr6_360508263,605,082Uncharacterized protein—Uyo DryProteinChr8_31395733831,395,733Uncharacterized protein—Uyo DryProteinChr6_117876761,178,767Protease (peptide metabolism)[43]Uyo DryProteinChr7_54291218754,291,218Uncharacterized protein—Uyo DryYieldChr2_21263131221,263,131Uncharacterized protein—Uyo DryYieldChr2_295655922,956,559BKI1 (brassinosteroid signaling)[44]Uyo DryYieldChr4_19860552419,860,552Uncharacterized protein—Uyo DryYieldChr5_16189831516,189,831Fasciclin-like arabinogalactan protein[45]Uyo DryYieldChr1_12467544112,467,544Uncharacterized protein—Uyo DryYieldChr8_42327370842,327,370Uncharacterized protein—Uyo RainProteinChr2_24615287224,615,287Uncharacterized protein—Uyo RainProteinChr6_31253548631,253,548CONSTANS-like zinc finger protein[46]Uyo RainYieldChr5_587153455,871,534ATPase family protein (CDC48-like)[47]Uyo RainYieldChr6_30896900630,896,900Thylakoid formation protein (THF1)[48]Uyo RainYieldChr8_44276921844,276,921Transcription elongation factor[38]Uyo RainYieldChr9_10783673910,783,673Mitochondrial carrier protein[49]*Chr Chromosome Table 5Gene identification for the significant SNPs for studied traitsEnvironmentTraitMarkerChrPosition (bp)Putative FunctionReferenceAwka DryProteinChr1_290898912,908,989EC 6.2 ligase (amino acid metabolism)[32]Awka DryProteinChr3_799465937,994,659MRN complex protein (DNA repair)[33]Awka DryProteinChr6_17778739617,778,739LOB domain transcription factor[14]Awka DryProteinChr1_12963638112,963,638LRR receptor-like kinase[34]Awka DryYieldChr3_921303139,213,031Methyltransferase (carbon metabolism)[35]Awka DryYieldChr4_19961226419,961,226Uncharacterized protein—Awka RainProteinChr2_8094532809,453SRS transcription factor[36]Awka RainProteinChr5_318583053,185,830F-box protein (protein degradation)[37]Awka RainProteinChr5_25318711525,318,711Transcription elongation factor[38]Awka RainProteinChr8_30664570830,664,570Ribosomal protein L4[39]Awka RainYieldChr3_614895536,148,955Polynucleotide phosphorylase (PNP)[40]Awka RainYieldChr5_15655284515,655,284Embryogenesis abundant protein[41]Awka RainYieldChr6_29354314629,354,314Microtubule-associated protein (TPX2-like)[42]Uyo DryProteinChr6_5695036569,503Zinc finger transcription factor[14]Uyo DryProteinChr6_360508263,605,082Uncharacterized protein—Uyo DryProteinChr8_31395733831,395,733Uncharacterized protein—Uyo DryProteinChr6_117876761,178,767Protease (peptide metabolism)[43]Uyo DryProteinChr7_54291218754,291,218Uncharacterized protein—Uyo DryYieldChr2_21263131221,263,131Uncharacterized protein—Uyo DryYieldChr2_295655922,956,559BKI1 (brassinosteroid signaling)[44]Uyo DryYieldChr4_19860552419,860,552Uncharacterized protein—Uyo DryYieldChr5_16189831516,189,831Fasciclin-like arabinogalactan protein[45]Uyo DryYieldChr1_12467544112,467,544Uncharacterized protein—Uyo DryYieldChr8_42327370842,327,370Uncharacterized protein—Uyo RainProteinChr2_24615287224,615,287Uncharacterized protein—Uyo RainProteinChr6_31253548631,253,548CONSTANS-like zinc finger protein[46]Uyo RainYieldChr5_587153455,871,534ATPase family protein (CDC48-like)[47]Uyo RainYieldChr6_30896900630,896,900Thylakoid formation protein (THF1)[48]Uyo RainYieldChr8_44276921844,276,921Transcription elongation factor[38]Uyo RainYieldChr9_10783673910,783,673Mitochondrial carrier protein[49]*Chr Chromosome Gene identification for the significant SNPs for studied traits *Chr Chromosome For protein content, several candidate genes were related to transcriptional regulation, protein metabolism, and cellular processes. Notably, loci linked to transcription factors such as the LOB domain protein (Chr6_17778739) and SRS-type transcription factor (Chr2_809453) suggest a regulatory role in protein accumulation. In addition, genes associated with protein turnover and synthesis, including F-box proteins (Chr5_3185830) and ribosomal proteins (Chr8_30664570), highlight the importance of post-translational regulation and translational machinery in determining protein content. The identification of a CONSTANS-like zinc finger protein (Chr6_31253548) further suggests potential links between developmental regulation and protein biosynthesis. For grain yield, several candidate genes were associated with growth regulation, cellular metabolism, and stress adaptation. For instance, the identification of a brassinosteroid signaling inhibitor (BKI1; Chr2_2956559) and a fasciclin-like arabinogalactan protein (Chr5_16189831) suggests roles in cell expansion, signaling, and structural development, which are critical for yield formation. Similarly, genes involved in energy metabolism and cellular maintenance, such as ATPase family proteins (Chr5_5871534) and mitochondrial carrier proteins (Chr9_10783673), indicate the importance of energy efficiency and metabolic balance in yield expression. Additionally, genes associated with cellular division and structural organization, including microtubule-associated proteins (Chr6_29354314) and thylakoid formation proteins (Chr6_30896900), highlight the role of cellular integrity and photosynthetic capacity in yield determination. The presence of transcription elongation factors and methyltransferases further supports the involvement of gene regulation and metabolic pathways in controlling grain yield. The identified candidate genes suggest that both protein content and grain yield in mungbean are controlled by multiple biological pathways, including transcriptional regulation, protein synthesis and degradation, hormonal signaling, and cellular metabolism. These findings provide a foundation for future functional validation and marker-assisted selection in mungbean breeding programs. Discussion Phenotypic variation and analysis of variance for studied traits Substantial phenotypic variation was observed among the mungbean genotypes for grain yield and protein content across the test environments, indicating the presence of exploitable genetic diversity for improvement. Significant effects of genotype, environment, and genotype × environment (G×E) interaction confirm that both inherited factors and environmental conditions influenced trait expression. Similar patterns have been reported in mungbean, where yield and quality traits are strongly modified by seasonal and location effects [2, 15, 17]. Phenotypic variation is essential for genome-wide association studies, as it enables the detection of genomic regions controlling quantitative traits [50]. The broad variation observed for grain yield reflects its quantitative nature and dependence on multiple component traits. Earlier studies also described grain yield as a highly complex trait controlled by many loci with strong environmental sensitivity [5, 16, 51]. In contrast, protein content showed relatively narrower but meaningful variation, suggesting moderate genetic control with some environmental modulation. This agrees with reports that nutritional traits in legumes are generally more stable than yield, though still influenced by genotype and management conditions [1, 7]. The significant G×E interaction emphasizes the need for multi-environment testing during mungbean breeding [51]. Genotypes selected in a single environment may not maintain superiority under contrasting conditions. Therefore, breeders should combine stability analysis with marker-assisted selection to identify genotypes carrying favorable alleles that perform consistently across locations. The broad-sense heritability estimates revealed distinct differences in the genetic control of the studied traits. Grain yield exhibited low to moderate heritability (H² = 0.28), indicating a substantial influence of environmental factors and genotype × environment interaction on its expression. This is consistent with the complex, polygenic nature of yield traits, which are highly sensitive to environmental variation. In contrast, protein content showed very high heritability (H² = 0.90), suggesting strong genetic control and greater stability across environments. Traits with high heritability are generally more responsive to selection, indicating that protein content can be effectively improved through breeding. These differences in heritability also provide important context for the GWAS results. The higher heritability of protein content likely contributed to the detection of more stable marker–trait associations, whereas the lower heritability of grain yield may have reduced the power to identify consistent genetic signals due to environmental noise. SNP marker statistics and genome coverage High-throughput SNP genotyping using DArTseq generated a robust dataset for genome-wide analysis, with 5,037 high-quality markers distributed across the mungbean genome. Comparable marker densities have been effectively used in previous mungbean diversity and GWAS studies [8, 15]. The moderate polymorphic information content and minor allele frequency values observed indicate that the markers were sufficiently informative for genetic analysis. The lower observed heterozygosity relative to expected heterozygosity reflects the predominantly self-pollinating nature of mungbean and is consistent with previous reports [18]. Population structure of the germplasm panel Population structure analysis revealed two distinct genetic clusters (K = 2), supported by both STRUCTURE and principal component analyses, indicating moderate genetic stratification within the germplasm panel. The moderate level of genetic diversity observed further supports the suitability of the germplasm panel for association mapping [52]. Similar moderate stratification has been reported in mungbean germplasm collections originating from diverse agro-ecological zones [10–12]. The presence of admixed individuals suggests shared ancestry and possible gene flow among accessions. Accounting for population structure is essential in GWAS to reduce false-positive associations arising from confounding genetic background effects [53, 54]. The incorporation of population structure into the multi-locus GWAS models used in this study improved the reliability of the detected marker–trait associations. The present study improved the reliability of detected loci, consistent with the recommendations of Wang et al. [13]. Linkage disequilibrium and mapping resolution The observed linkage disequilibrium (LD) decay of approximately 368 kb suggests that the marker density achieved in this study was adequate to capture associations with causal loci. Adequate marker coverage is critical in GWAS, as it increases the likelihood that markers are in linkage disequilibrium with functional variants controlling traits [55]. Similar LD decay patterns have been reported in mungbean and other legumes [18]. The LD structure observed in this study therefore supports the reliability of the identified marker–trait associations and the genomic resolution achieved. This pattern is typical of self-pollinating species, where recombination events are less frequent, leading to extended LD blocks. Genome-wide association findings for protein content Genome-wide association analysis identified 15 significant marker–trait associations (MTAs) for protein content across multiple environments, confirming that protein accumulation in mungbean is controlled by multiple genomic regions. The detection of loci across different chromosomes highlights the polygenic nature of this trait. Previous studies in legumes have likewise shown that seed nutritional quality is governed by multiple loci with additive effects [6, 7]. Several identified loci were associated with genes involved in transcriptional regulation and protein metabolism, including LOB domain transcription factors, SRS transcription factors, F-box proteins, and ribosomal proteins. These gene categories are known to regulate gene expression, protein synthesis, and protein degradation, suggesting that protein accumulation is influenced by coordinated transcriptional and post-translational mechanisms. The presence of environment-specific loci indicates that environmental conditions influence the expression of protein-related genes, although the relatively narrow phenotypic range observed suggests a degree of genetic stability. Similar findings have been reported in legumes, where seed protein content is controlled by multiple loci and metabolic pathways [56, 57]. Genome-wide association findings for grain yield A total of 15 significant MTAs were identified for grain yield across the evaluated environments, confirming the complex and environment-dependent genetic architecture of yield in mungbean. Grain yield is influenced by multiple physiological processes, including growth, resource allocation, and stress responses. Similar multi-locus control of yield and yield components has been reported by Akhtar et al. [15], Reddy et al. [16], and Han et al. [5]. The polygenic nature of grain yield observed in this study suggests that genome-enabled breeding approaches, such as genomic selection, may be particularly useful for accelerating genetic gain in mungbean, especially for complex traits controlled by numerous loci with small effects [58]. Several yield-associated loci were linked to genes involved in hormonal signaling, structural development, and energy metabolism. Notably, genes such as BKI1 (brassinosteroid signaling), fasciclin-like arabinogalactan proteins (cell wall development), ATPase family proteins (energy metabolism), and THF1 (photosynthetic regulation) were identified. These genes are associated with processes that directly influence plant growth, biomass accumulation, and reproductive success. The identification of multiple environment-specific loci further supports the strong genotype × environment interaction observed for grain yield, indicating that different genetic mechanisms may be expressed under varying environmental conditions. Similar findings have been reported in previous GWAS studies in legumes [57, 59]. Candidate gene identification Candidate gene analysis revealed that loci associated with grain yield and protein content are linked to genes involved in diverse biological processes, including transcriptional regulation, signal transduction, metabolic activity, cellular development, and energy metabolism. The presence of transcription factors such as SRS, LOB domain proteins, and CONSTANS-like proteins suggests that regulatory mechanisms play a central role in controlling both protein accumulation and plant development. This agrees with recent reviews emphasizing that productivity and nutritional traits in mungbean are controlled by interconnected physiological pathways rather than single genes [4, 7, 17]. Genes associated with protein metabolism, including F-box proteins and ribosomal components, indicate that protein synthesis and degradation pathways contribute significantly to seed protein content. Similarly, genes involved in hormonal signaling and structural development, such as BKI1 and fasciclin-like proteins, highlight the importance of signaling pathways and cell wall dynamics in yield determination. Additional genes associated with cellular processes, including microtubule-associated proteins, transcription elongation factors, and mitochondrial carrier proteins, further emphasize the role of cellular organization and energy metabolism in trait expression. The diversity of functional categories represented by the identified candidate genes supports the polygenic nature of the studied traits and highlights the involvement of multiple biological pathways in regulating grain yield and protein content. Conclusion This study provides comprehensive insights into the genetic architecture of grain yield and protein content in mungbean using multi-locus genome-wide association analysis across diverse environments. Significant phenotypic variation was observed for both traits, with grain yield showing strong environmental influence and protein content exhibiting relatively stable expression. A total of 15 significant markers–trait associations (MTAs) were identified for grain yield and 15 for protein content, distributed across multiple chromosomes, confirming the polygenic nature of these traits. The detection of environment-specific loci further highlights the importance of genotype × environment interactions, particularly for grain yield. Candidate gene analysis revealed that loci associated with protein content are linked to genes involved in transcriptional regulation and protein metabolism, while yield-associated loci are linked to genes involved in hormonal signaling, structural development, and energy metabolism. These findings provide valuable insights into the biological pathways underlying these traits. The identified MTAs and candidate genes represent promising genomic resources for marker-assisted selection and genomic-assisted breeding aimed at improving yield and nutritional quality in mungbean. Future studies focusing on functional validation of these candidate genes will further enhance their application in mungbean improvement programs. Supplementary Information Supplementary Material 1. Supplementary Material 1.