A high-density consensus genetic map was generated by integrating seven reference genetic maps using the R package LPmerge (Endelman and Plomion, 2014). The detailed LPmerge code is listed in Supplementary Data S1. The integrated maps included: (i) “Mungbean-Berken × ACC41-RIL” (Humphry et al., 2005); (ii) “Mungbean-JP211874 × JP229096-BC₁F₁” (Isemura et al., 2012); (iii) “Mungbean-Berken × ACC41-RIL” (Wu et al., 2014); (iv) “Mungbean-VC2917 × ZL-RIL” (Liu et al., 2017); (v) “Mungbean-Huaye1 × Zijing1-F₂” (Wang et al., 2017); (vi) “Mungbean-Dahuaye × Jilv9-RIL” (Wang et al., 2020); (vii) “Mungbean-Sulu16-10 × Weilu11-F₂” (Ye et al., 2021).

Following map integration, the consensus map and initial QTL data were imported into BioMercator V4.2.3 for analysis (Supplementary Table S1). Initial QTLs were projected onto the consensus map using the QTLProj module (Veyrieras et al., 2007). QTL meta-analysis was performed via the Veyrieras two-step method to identify MQTLs (Sosnowski et al., 2012). Firstly, QTL clustering was performed on each chromosome using BioMercator’s standard parameters. The optimal number of MQTLs was determined by comparing five statistical criteria: Akaike information criterion (AIC), AIC correction, AIC3, Bayesian information criterion (BIC), and Average Weight of Evidence (AWE). The model with the frequent values across all criteria was selected. Second, Second, the 95% confidence interval (CI) and peak position of each MQTL were defined based on the best-fit model from the first stage. Initial QTLs falling within the MQTL CI were integrated, while those failing to meet the minimum AIC threshold were excluded (Aloryi et al., 2024). Information on the files corresponding to the first step (_model.txt) and the second step (_table.txt) is provided in Supplementary Data S2. MQTLs were systematically named according to their chromosomal locations (e.g., MQTLLG1-1, MQTLLG1-2). The phenotypic variance explained (PVE) by each MQTL was calculated as the mean PVE of its constituent QTLs.

Literature searches were conducted across databases including Google Scholar (https://scholar.google.com/), PubMed (https://pubmed.ncbi.nlm.nih.gov/), and CNKI (https://www.cnki.net/) for the systematic search for QTL studies on yield and yield-related traits in mungbean published from 2005 to present. From these studies, we extracted the following parameters: parental composition of the population, population type and size, measured traits, molecular marker types, QTL flanking markers, logarithm of odds (LOD) scores, phenotypic variance explained (PVE) or R ², and confidence interval (CI) (Supplementary Tables S2, S3). QTLs lacking reported LOD scores were assigned a default value of 3, while those with missing PVE values were excluded. For several QTLs with missing CI information, CI (95%) was calculated using population-specific formulas (Darvasi and Soller, 1997; Guo et al., 2006).

(1) CI = 287   /   n   ×  PVE  for DH populations

These initial QTLs were associated with 71 different traits, categorized into nine groups: (i) growth period-related traits (22 traits such as branching stage, days to flowering, days to harvest, and days to maturity); (ii) yield traits (yield per block, yield index, yield per plant, biomass index and biomass); (iii) branch number (branch number per plant and number of branches); (iv) leaf related traits (chlorophyll content, leaf width, maximum leaf area, and other 8 traits); (v) hundred-grain weight; (vi) plant height and plant height related traits (plant height, number of nods, main stem length, plant height index and stem internode length); (vii) pod related traits (pod number per plant, total number of pods, pod length and other 7 traits); (viii) seed related traits (hard seedness, seed diameter, seed length and other 9 traits); (ix) other traits (12 traits such as main stem thickness, growth habit, germinating percentage) (Supplementary Table S4).

To determine the physical location of these MQTLs, we performed BLASTn alignment of MQTL flanking marker sequences against the V. radiata L. reference genome (Kang et al., 2014) available via EnsemblPlants (https://plants.ensembl.org/). However, due to insufficient sequence information for most flanking markers, direct physical localization was unfeasible. We therefore calculated physical locations using the following genetic-to-physical conversion formula (Prakash et al., 2022): Physical location  bp = Chromosome   physical   length   bp chromosome   g enetic   length   cM × Genetic location  cM

Chromosomal physical lengths were obtained from the mungbean reference genome, while genetic lengths were derived from our consensus map constructed. For GWAS validation, we compiled marker-trait associations (MTAs) from six GWAS studies associated with yield and yield-related traits in mungbean (published in 2018–2023), recording population size, traits, marker types, and MTA counts (Table 2). MQTLs were considered validated if their physical intervals overlapped with ≥1 MTA.

To identify the OrMQTLs for yield and yield-related traits in mungbean, pigeonpea, and common bean, the following steps were taken: (i) identify conserved regions in the genomes of mungbean, pigeonpea, and common bean using synteny and colinearity analyses with the ‘BioMart’ tool in the EnsemblPlants database (https://plants.ensembl.org/biomart/); (ii) screen for pigeonpea and common bean orthologs within the region of the MQTL in mungbean genomes; (iii) compare the physical positions of the pigeonpea and common bean orthologs to the MQTL regions of the corresponding yield and yield-related traits (Halladakeri et al., 2023; Izquierdo et al., 2023), and consider the MQTLs of pigeonpea and common bean that contain at least four of the corresponding genes as OrMQTLs for mungbean. The synteny analysis between mungbean, pigeonpea, and common bean genomes was plotted using Tbtools software (Chen et al., 2020).

According to the criteria developed by Löffler et al. (2009), MQTLs with genetic distances <2 cM, containing at least four initial QTLs from different studies, and with PVE >10% were screened as breeder’s MQTLs for candidate gene mining. The information of gene models within the breeder’s MQTL regions was searched using the EnsemblPlants database (https://plants.ensembl.org/). To identify candidate genes, a comparative genomics approach was implemented to mine orthologs of yield-related genes from rice and Arabidopsis thaliana within the breeder’s MQTL regions of the mungbean genome.

A high-density consensus genetic map was constructed using the R package LPmerge to integrate seven previously published genetic maps in mungbean. The consensus genetic map spanned a genetic distance of 1,679.58 cM and contained 3,497 markers, with an average genetic distance between markers of 0.48 cM (Table 1; Supplementary Table S1). The genetic length of individual linkage groups ranged from 87.80 cM (LG11) to 246.60 cM (LG2), with an average length of 152.69 cM across all linkage groups. The number of markers per linkage group varied from 212 (LG11) to 426 (LG1), with an average of 317.91 markers per linkage group. The marker density of individual linkage groups ranged from 1.40 (LG4) to 3.32 (LG5) markers per cM, with an overall average density of 2.22 markers per cM (Figure 1e; Table 1).

A total of 660 QTLs from 18 independent studies (2005–2023) involving QTL mapping for yield and yield-related traits in mungbean were collected during 2005–2023 for meta-analysis (Supplementary Tables S2, S3). These studies included a total of 21 different QTL mapping populations, including 12 RIL populations, 6 F₂ populations, 2 F₃ populations, and 1 backcross population, with population sizes ranging from 100 to 261 (Supplementary Table S2). The yield and yield-related traits analyzed were categorized into nine types (Supplementary Table S4). Among these, the number of QTLs associated with growth period-related traits, pod-related traits, plant height, hundred-grain weight and grain-related traits was relatively high, accounting for 20.3%, 15.61%, 14.7%, 14.24% and 12.58% of the total number of QTLs, respectively, while the remaining traits represented a smaller proportion (Figure 2a). These QTLs were unevenly distributed across the linkage groups, with LG2 containing the highest number of QTL at 14.70% (97/660), LG8 having the lowest at 5.90% (39/660), and the remaining linkage groups ranging from 46 to 82 QTLs each (Figure 1a,c; Figure 2b). The logarithm of odds (LOD) scores for these QTLs ranged from 2 to 97.25, with the majority (63.6%) falling between 2 and 6 (Figure 2c). The phenotypic variance explained (PVE) by individual QTLs ranged from 1.32% to 99.51%, with an average value of 14.2%, and 41.5% were distributed within 5%–10% range (Figure 1b; Figure 2d).

Among the 660 initial QTLs for yield and yield-related traits, 590 were screened for mapping onto the consensus genetic map after excluding QTLs with missing PVE values or flanking markers for meta-analysis. Of these 590 initial QTLs, meta-analysis integrated 553 QTLs into 72 MQTLs, while 37 QTLs remained as individual QTLs without overlapping any MQTL (Supplementary Table S5). The MQTLs were unevenly distributed across the linkage groups, with the number of MQTLs ranging from 5 in LG6, LG8, and LG9 to 9 in LG1 (Figure 1d). Each MQTL contained at least 2 initial QTLs, and 64 MQTLs consisted of no fewer than 3 initial QTLs, with MQTLLG4-6 comprising up to 46 initial QTLs (Figure 3a; Supplementary Table S5). Among the 72 MQTLs, 75% (54/72) were associated with at least three yield and yield-related traits, and MQTLLG2-5 and MQTLLG4-6 affected 18 yield and yield-related traits simultaneously (Figure 3a; Supplementary Table S5). The confidence intervals (CIs) of these MQTLs ranged from 0.09 to 8.28 cM, with an average CI of 1.21 cM, representing a 6.26-fold reduction compared to the average CI of the initial QTL. The reduction in CI varied significantly across all linkage groups, with the largest reduction observed in LG9 (12.59-fold) and the smallest in LG3 (2.16-fold) (Figure 3b).

A total of 5,563 marker-trait associations (MTAs) were identified from six GWAS studies on yield and yield-related traits in mungbean published since 2018 to date (Table 2; Supplementary Table S6). The physical locations of these MTAs were compared with the 72 MQTLs. Among the 72 MQTLs, 20 were co-located with at least one MTA from the GWAS. Most of these MQTLs (7/20) were co-located with only one MTA, while MQTLLG7-2 (625 MTAs), MQTLLG7-5 (47 MTAs), and MQTLLG11-4 (98 MTAs) were co-located with multiple MTAs. Among the 20 GWAS-validated MQTLs, 11 were validated by only one of the six GWAS studies, five MQTLs (MQTLLG6-1, MQTLLG6-3, MQTLLG6-5, MQTLLG7-5, and MQTLLG10-1) were validated in two GWAS studies, three MQTLs (MQTLLG7-2, MQTLLG8-2 and MQTLLG11-4) were validated in three GWAS studies, and one MQTL (MQTLLG8-3) was validated in four GWAS studies (Supplementary Table S6).

Synteny analysis of the conserved genomic regions between mungbean, pigeonpea and common bean revealed that mungbean has 1,404 orthologs with pigeonpea and 1,504 orthologs with common bean (Supplementary Table S7). Among the 552 conserved genomic regions identified between mungbean and common bean, 15 common bean MQTLs were identified as orthologous MQTLs (OrMQTL) to 22 mungbean MQTLs. For example, the mungbean MQTLLG8-3 was orthologous to two common bean MQTLs (MQTL-YC7.5 and MQTL-YC8.3), and the MQTLLG6-4 was isogenic to three common bean MQTLs (MQTL YC1.2, MQTL-YC6.1 and MQTL-YC8.3). Additionally, the number of conserved gene models between OrMQTLs in mungbean and common bean ranged from 4 (MQTLLG7-6) to 108 (MQTLLG6-4), with 11 OrMQTLs containing at least 20 conserved gene models. Among the 33 MQTLs for agronomic traits, fertility restoration, and seed quality traits in pigeonpea, only 18 conserved gene models were identified in mungbean. This limited number was attributed to the small number of gene models within the MQTL regions, and no OrMQTLs were identified between mungbean and pigeonpea (Figure 4; Supplementary Table S7).

Twenty breeder’s MQTLs were screened for candidate gene mining in MQTL. These breeder’s MQTLs were associated with several mungbean yield and yield-related traits, indicating that key candidate genes controlling these traits might be located within these breeder’s MQTL regions (Supplementary Table S5). A total of 339 gene models were identified within these breeder’s MQTL regions, with the number of gene models per breeder’s MQTL ranged from 1 (MQTLLG3-4, MQTLLG9-4, and MQTLLG9-5) to 125 (MQTLLG7-2) (Supplementary Table S8). To further explore candidate genes affecting yield-related traits in mungbean, 22 mungbean orthologs of Arabidopsis and rice genes related to yield traits such as tiller number, starch synthesis, seed germination, and plant height were identified within the breeder’s MQTL regions through homology comparison with rice and Arabidopsis. Among these, 10 orthologs were derived from rice and 13 from A. thaliana, with one gene being identical to a rice gene (Supplementary Table S9).

Integrating 660 QTLs from diverse studies revealed genomic hotspots on LG2 and LG4 (Figure 2b), with over 14% of QTLs clustering on LG2. This uneven distribution aligns with synteny patterns in common bean (Izquierdo et al., 2023), suggesting conserved selection pressure on these regions. Notably, the preponderance of QTLs associated with growth period-related traits, pod-related traits, plant height, and grain weight highlights these as key determinants of yield potential (Figure 2a). This observation aligns with reported genetic control mechanisms in other legume crops (Izquierdo et al., 2023). These QTL were unevenly distributed across linkage groups (Figure 2b), suggests the existence of genomic hotspots with concentrated genetic potential for yield improvement in mungbean.

As the first QTL meta-analysis in mungbean, our study integrated seven published genetic maps to construct a high-density consensus genetic map (Supplementary Table S1). Through integration of seven published genetic maps, we constructed a high-density consensus genetic map (Supplementary Table S1), mapping most of the initial QTLs (590 out of 660) to this consensus map, and 72 MQTLs were identified via meta-analysis (Figures 1D,E). The statistical confidence of MQTLs correlates positively with the number of underlying component QTLs (Quraishi et al., 2017). Notably, 75% of our MQTLs incorporated ≥3 initial QTLs, with 58.3% (42/72) containing ≥6 component QTLs (Figure 3a), suggesting that these MQTLs have high reliability. Moreover, the average CI of MQTLs was reduced 6.26-fold compared to that of initial QTLs (Figure 3b). From these, we identified 20 breeder’s MQTLs characterized by high contributions (PVE >10%), narrow CIs (<2 cM), and more initial QTLs (≥4). Significantly, most of these breeder’s MQTLs exhibited pleiotropic effects on multiple yield-related traits, with three key loci (MQTLLG2-5, MQTLLG4-6, and MQTLLG9-4) influencing ≥10 yield-related traits (Supplementary Table S5), highlighting their potential for simultaneous improvement of multiple yield components.

GWAS based on linkage disequilibrium can detect minor-effect alleles that are missed in biparental populations. MTAs identified in GWAS can be used to validate candidate genes for QTL mapping (Han et al., 2018; Sallam et al., 2022). Using GWAS-MTA to validate the accuracy of MQTL results has been reported in QTL meta-analysis in legume crops such as soybean, pigeonpea, and common bean (Chen et al., 2021; Halladakeri et al., 2023; Izquierdo et al., 2023). In the present study, 27.8% (20/72) of the MQTL were validated in six GWAS studies published in recent years for yield and yield-related traits in mungbean (Supplementary Table S6). Of these validated MQTLs, the majority (7/20) showed overlap with a single MTA, while 55% (11/20) were supported by only one GWAS dataset. Notably, four critical MQTLs (MQTLLG7-2, MQTLLG8-2, MQTLLG8-3, and MQTLLG11-4) exhibited co-located with multiple MTAs across ≥3 GWAS studies (Supplementary Table S6), strongly suggesting that these genomic regions likely harbor key genetic determinants of yield and yield-related traits. These high-confidence loci (e.g., MQTLLG7-2 and MQTLLG8-3) represent prime targets for MAS, particularly due to their pleiotropic effects on ≥10 traits (Supplementary Table S5).

Comparative synteny analysis revealed conserved OrMQTLs associated with agronomically important traits among legume crops including mungbean, pigeonpea, and common bean (Halladakeri et al., 2023; Izquierdo et al., 2023). Specifically, we identified 22 conserved OrMQTLs sharing synteny between mungbean and common bean genomes (Figure 4), suggesting strong evolutionary selection pressures to preserve these regions due to their functional importance in legume biology. These conserved loci harbor numerous uncharacterized genes in mungbean that represent promising targets for future functional studies. Characterizing genes within OrMQTLs could elucidate the genetic networks regulating yield and yield-related traits across legume species and provide insights into the molecular basis of trait evolution in legume crops. Furthermore, molecular markers derived from these syntenic regions hold potential for accelerating marker-assisted breeding strategies aimed at enhancing yield in mungbean (Saini et al., 2022).

As traditional model plants, A. thaliana and rice possess extensively characterized genomic resources, making synteny analysis between mungbean and these species a valuable strategy for identifying candidate genes associated with agronomically important traits (Gaut, 2002). To precisely map MQTLs to physical genomic positions, we applied the methodology of Prakash et al. (2022), which calculates physical coordinates using chromosomal genetic-to-physical length ratios, as direct alignment of MQTL-flanking markers to the mungbean reference genome was hindered by low sequence identity and alignment scores. Within the 20 breeder’s MQTL regions, we identified 339 gene models, including 22 mungbean orthologs of known regulators of tiller number, starch synthesis, seed germination, plant height, and yield in Arabidopsis and rice (Supplementary Table S9). Notably, the MQTLLG7-2 region harbors Vradi07g07390, an ortholog of the rice amino acid transporter gene OsLHT1 that critically influences plant growth and yield (Wang et al., 2019), and Vradi07g07630, which shares homology with the rice ABA receptor gene OsPYL/RCAR10 implicated in seed germination (Kim et al., 2012). Furthermore, the MQTLLG5-3 region contains Vradi05g08600, a functional ortholog of both the rice leaf morphology gene YUCCA6 (Zhang et al., 2021), and and the Arabidopsis flowering time regulator YUC8 (Ståldal et al., 2012), suggesting pleiotropic roles in yield-related traits. The findings not only advances our understanding of the genetic control of yield in mungbean but also provides a foundation for the development of molecular markers and the implementation of MAS in mungbean breeding programs.