Whole genome sequencing-based multi-locus association mapping for kernel iron, zinc and protein content in groundnut Authors: U Nikhil Sagar, Sejal Parmar, Sunil S Gangurde, Vinay Sharma, Arun K Pandey, D Khaja Mohinuddin, Namita Dube, Ramesh S Bhat, K John, Muga D Sreevalli, P Sandhya Rani, Kuldeep Singh, Rajeev K Varshney, Manish K Pandey Published: 2026 ABSTRACT AbstractMalnutrition is a major global challenge, especially in the developing regions, where improving the nutritional content of staple crops is an important step towards alleviating hidden hunger. Groundnut, a nutrient rich legume, contains several essential nutrients, high protein, essential amino-acids and vitamins required for human health. In this study, multi-season phenotyping data for kernel iron (Fe), zinc (Zn) and protein content (PC), along with whole genome re-sequencing (WGRS) data from a mini-core collection, were used to perform genome-wide association study (GWAS) analysis. Phenotypic variability analysis revealed a large variation in Fe (7.6 – 42.8 ppm), Zn (10.9 – 62.4 ppm) and PC (12.7 – 33.6%). GWAS analysis identified a total of 15 marker-trait associations (MTAs) and 28 candidate genes for pooled season data, and 44 MTAs and 62 candidate genes for individual seasons. Key candidate genes like MYB transcription factor (Arahy.QI0PHV, Arahy.1I6ZSS), Zn finger MYM type protein, RING finger MYM type protein (Arahy.7P97F6, Arahy.9R964H, Arahy.I3B88T) and NAC domain protein (Arahy.LV3APC), were found to be associated with the Fe and Zn homeostasis pathway. In addition, genes related to protein homeostasis, such as Protein kinase family protein (Arahy.4D7KBI), and E3 ubiquitin-protein ligase (Arahy.PE3CF6), were identified within significant MTA regions. These findings provide basis for the detection and characterization of potential candidate genes associated with nutritional quality traits. Haplo-pheno analysis revealed that accessions with superior haplotypes for kernel Fe and Zn content were predominantly found in the Spanish Bunch/Valencia Bunch types, whereas the accessions with inferior haplotypes were more common in the Virginia Bunch/Virginia Runner types. Single nucleotide polymorphism (SNP)-based KASP (Kompetitive Allele Specific Polymerase Chain Reaction) markers for 9 MTAs were designed and validated for Fe and Zn. Among these, three markers (snpAH00636 and snpAH00641 for Fe, and snpAH00644 for Zn) showed polymorphism, and could be utilized in genomics-assisted breeding to develop nutrient-rich groundnut varieties.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57575-2.Keywords: Candidate genes, Fe and Zn homeostasis pathway, KASP markers, Marker trait associations, Nutritional qualitySubject terms: Biotechnology, Computational biology and bioinformatics, Genetics, Molecular biology, Plant sciences Introduction Groundnut (Arachis hypogaea L.), often referred to as the "King of oilseeds", is valued for its high oil content, nutritional value, and economic importance. It is also referred to as peanut, earthnut, monkey nut, wonder nut, and goobernut1. Globally, it is the fourth-most important source of vegetable protein. Groundnut is a self-pollinated, segmental allotetraploid crop with minimal molecular-level variation2. Taxonomically, it belongs to the family Leguminosae and sub-family Papilionaceae3. Originating in South America, it is now widely cultivated across tropical, sub-tropical, and semi-arid regions of the world. Currently, groundnut is grown on over 32.7 million hectares across more than 100 countries, producing approximately 53.9 million tonnes, with an average productivity of 1.64 tonnes ha−14. China and India are the leading producers, contributing 18.38 million tonnes (34.3%) and 10.13 million tonnes (17.4%) respectively, followed by Nigeria- 4.28 million tonnes (8%), USA- 2.53 million tonnes (5%) and Sudan- 2.50 million tonnes (5%), Among these, the USA records the highest productivity of 4469.26 kg ha−14. Groundnut, as a nutrient rich legume crop, contains upto 28% protein, 50% oil, and 20% carbohydrates5. Moreover, it is also an important source of monounsaturated and polyunsaturated fatty acids, bioactive compounds such as polyphenols, flavonoids, isoflavones, and several B-complex vitamins, and antioxidants, including resveratrol and p-coumaric acid. Raw groundnut contains 4.58 mg of Fe, 3.27 mg of Zn and 26 g of PC per 100 g. Fe and Zn are important micronutrients required for human growth and development, and also serve as cofactors for proteins like haemoglobin and cytochromes5. Malnutrition, or "hidden hunger," particularly Fe and Zn deficiencies, affects nearly three billion people worldwide6, with a significant proportion of women and children in India suffering from anemia. Due to its nutritional richness, groundnut and its products can be promoted as nutrient-dense foods to address protein, energy, and micronutrient deficiencies in vulnerable populations7. For example, ready-to-use therapeutic foods (RUTFs) such as Plumpy’Nut®, derived from groundnut, have been highly effective in saving the lives of malnourished children in several regions8. Recent advances in genomic resources have opened new avenues for enhancing the nutritional profile of groundnut. These resources include whole-genome resequencing (WGRS), skim sequencing, mid-density platforms, genotyping-by-sequencing (GBS), restriction site-associated sequencing (RADSeq), double- digest restriction site-associated sequencing (ddRADSeq), and specific length amplified fragment sequencing (SLAF-seq); reference genomes for diploid wild progenitors (A. duranensis and A. ipaensis) and cultivated tetraploid genomes (A. hypogaea subsp. hypogaea and A. hypogaea subsp. fastigiata); and diagnostic markers such as simple sequence repeats (SSRs), insertions and deletions (InDels), and SNPs9. Genomic tools successfully contributed in the development of several varieties with traits of interest, such as high oleic acid, resistance to rust and late leaf spot (LLS)10. High and mid-density genotyping arrays/assays have further enabled the construction of genetic maps, identification of genomic regions and candidate genes via GWAS, and the development of diagnostic markers to accelerate breeding programs5. These tools also facilitate various applications, such as assessing genetic purity in seed systems and gene bank germplasm, foreground and background selection in backcross breeding, trait mapping, and genomic selection in groundnut10. Groundnut is consumed as a part of the human diet, making it one of the most nutrient-dense food crops11. Despite its nutritional value, further enrichment of micronutrients like Fe and Zn, and PC are essential for improving human health, particularly in developing and underdeveloped regions12. Integrating genomic tools with robust phenotyping approaches in groundnut breeding programs offers promising opportunities to enhance nutritional quality. Groundnut biofortification efforts can be further strengthened by identifying and characterizing candidate genes associated with nutrient accumulation. Several key candidate genes encoding transcription factors such as NAC, MYB were reported to regulate seed and pod development, plant growth, and photosynthesis, under Fe deficiency conditions13. Proteins encoded by MYB transcription factor, RING finger protein, and NAC transcription factor were reported to have a key role in Fe and Zn homeostasis pathways in seeds5. Earlier studies also identified homologs such as IRT1, FIT1, bZIP23, and OPT3, in groundnut and other legumes, which are involved in Fe transport from leaves to roots14. Previous studies in cereals identified several genes or gene families including YSL, NRAMP, ZIP, Heavy metal ATPase, cation diffusion facilitator family, nicotianamine synthase, and nicotianamine aminotransferase, responsible for Fe and Zn homeostasis15. Nowadays, GWAS has proven to be highly effective for mapping complex traits across a wide range of populations, including natural and multi-parent breeding populations with varying genetic bases. The main challenge of GWAS is to minimize false positives caused by population structure and family relatedness16. Traditional models such as Mixed Linear Model (MLM) attempts to reduce this challenge by including these two factors as covariates17, but cause overfitting, resulting in false positives and missing key loci. Multi-locus models offer an advantage by capturing additional loci18. For example, Multiple Loci Mixed Linear Model (MLMM) incorporates related markers as cofactors in the mixed model by stepwise regression19. More recently, Bayesian-information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK) model has shown greater statistical power by not requiring causal variants to be distributed across the genome, unlike the Settlement of MLM Under Progressively Exclusive Relationship (SUPER) model and the Fixed and Random Model Circulating Probability Unification (FarmCPU) model, and also reduces the number of false positives. The FarmCPU model, in particular, increases statistical power by reducing confounding between testing markers and kinship while controlling false positives20. In this context, GWAS was performed using WGRS data from the mini-core collection, combined with phenotypic data generated for Fe, Zn and PC. Haplotype information is beneficial for GWAS to identify marker-phenotype associations21, to compensate bi-allelic limitation of SNP markers, and greatly enhance QTL detection efficiency22,23. Moreover, the haplotype-based analysis helps to control false positives and reveals the complex mechanism of causal haplotypes more effectively compared to individual SNPs24. Hence, this study was conducted to identify superior haplotypes, in addition to significant MTAs and candidate genes associated with Fe, Zn and PC homeostasis in groundnut kernels. These candidate genes, encoding various important regulatory proteins, can serve as targets for the development and validation of molecular markers. The discovery of superior haplotypes and donor genotypes marks the foundational steps to facilitate haplotype-based breeding (HBB) for enhancing kernel Fe, Zn, and PC in groundnut. Materials and methods Multi-environment phenotyping for Fe, Zn and PC This study used 178 lines from the groundnut mini-core collection developed by ICRISAT Genebank25 (Supplementary Table 1). Experiments were conducted at ICRISAT, Patancheru, India (L1) across five seasons; rainy 2009 (S2), post-rainy 2009–10 (S3), post rainy 2010–11 (S4)22, and later during rainy 2020 (S5) and rainy 2021 (S6) for estimation of Fe and Zn (ppm) (Supplementary Table 2, 5). PC was estimated during four seasons at ICRISAT (L1)- rainy 2008 (S1), rainy 2009 (S2)26, rainy 2020 (S5) and rainy 2021 (S6). Additionally, PC data for two rainy seasons; 2008 (S1) and 2009 (S2) was also generated at Dharwad (L2)27. Frequency distribution graphs (violin plots) and Pearson correlation plot for Fe, Zn and PC were generated using “SRplot”28, while the principal component analysis (PCA)- biplot was produced using the “Factoextra”29 package in R software. Fe and Zn contents were measured using the ICP-OES method30. For each genotype, 5–10 gm of seeds (15–20 seeds) were used. Approximately 0.3 g of oven-dried samples were placed in 50 ml labelled polypropylene tubes. Nitric acid (2 ml) and hydrogen peroxide (0.5 ml) were added, and the tubes allowed to withstand at the room temperature for overnight. Next day, tubes were vortexed and digested in a block for 30 min at 80 °C followed by 120 min for 125 °C. Caps were loosened after the first 30 min to equalize pressure. After cooling to room temperature, 25 ml distilled water was added and samples were mixed on an orbital shaker, filtered, and the supernatant analyzed using ICP-OES5. The phenotypic data for Fe and Zn in S2—S4 were analyzed at Charles Renard analytical laboratory, ICRISAT, while S5 and S6 samples were analyzed at NCML laboratory, Hyderabad. The PC (%) was quantified using Near infrared-reflectance spectroscopy (NIRS). The groundnut samples (approximately 60 g) were scanned with a diode array analyzer at wavelengths from 950 to 1650 nm. Absorbance readings were recorded in 5 nm increments, and each sample was scanned three times to ensure accuracy5. Samples from two seasons S1 and S2, collected from L1 and L2, were analyzed using NIRS. The estimation of PC during the two seasons, S5 and S6 was done using NIRS at NCML laboratory, Hyderabad. Genomic DNA extraction, sequencing and SNP calling The NucleoSpin® Plant II kit (Macherey–Nagel, Germany) was used to extract the genomic DNA from 100 mg of leaf sample tissue. 500 μL of lysis buffer was used to homogenize the leaf sample, and then 10 μL RNase was added to remove the RNA impurities. The samples were incubated at 65ºC for 1 h in water bath, then centrifuged at 5500 rpm for 10 min. A separate tube was used to collect the supernatant, and binding buffer of 450 μL was added. This whole mixture was centrifuged at 6000 rpm for 1 min after being filtered using a NucleoSpin plant MN column. In the final step, pre-heated elution buffer of 50 μL was added to the column’s membrane filter, incubated at 65ºC for 5 min, and centrifuged at 6000 rpm for 1 min to elute the DNA31. 0.8% agarose gel was used to check the DNA quality whereas the quantity of DNA was assessed using NanoDrop 8000 Spectrophotometer. The TruSeq Library Prep Kit (Illumina, USA) was used to generate paired-end sequencing libraries with a 500 bp insert size to represent each genotype. Short reads (150 bp) with paired ends were generated at the Illumina HiSeq 2500 platform. Sequence data is provided under bio-project ID (PRJNA1002116, PRJNA490835 and PRJNA490832). After sequencing, adapter sequences were eliminated, and fastp v0.20.1 was used to filter low-quality reads32. Reads which had more than 5% undetermined nucleotides ("N") or greater than 20% low-quality bases (Phred quality score < 15) were not included. BWA-MEM v0.7.12-r103933 was used to align the remaining high-quality reads with the reference genome of the cultivated Arachis hypogaea cv. Tifrunner (Tifrunner.gnm1.KYV3)34, using the parameters mem -M -R. SAMtools v1.21 was used to sort the alignments, and Picard v2.18.11 (http://picard.sourceforge) was used to detect duplicate reads. SOAPnuke v1.5.6 was used to filter the raw reads of each line. GATK v4.3.0.0 was used to perform SNP and InDel calling utilizing a joint calling approach35. Primarily HaplotypeCaller was used to generate genomic variant call format (GVCF) files for each sample, using the parameters: -T HaplotypeCaller -ERC GVCF –phred-scaled-global-read-mismapping-rate 45. CombineGVCFs were then used to carry out joint variant calling. Based on GATK best practices, VariantFiltration was used to refine the SNP set using the following hard filtering criteria: QD < 2.0, MQ < 40.0, FS > 60.0, ReadPosRankSum < − 8.0, MQRankSum < − 12.5, and SOR > 3.0. Using BamDeal v0.22 (https://github.com/BGI-shenzhen/BamDeal), depth was computed in 100-kb sliding windows in order to remove variants identified in regions with unusual sequencing depth. SnpEff v4.3t was used with default parameters to annotate the synonymous and non-synonymous SNPs. VCFtools v0.1.13 was used to filter out SNPs with a minor allele frequency below 0.05 or a missing rate larger than 0.2, using the parameters -max-missing 0.2 –maf 0.05. GWAS for Fe, Zn and PC GWAS analysis was conducted using multi-season phenotyping data from 178 mini-core lines and 2,55,144 polymorphic SNPs from WGRS data. The Genome Association and Prediction Integrated Tool (GAPIT version 4.3.3)36 in “R” software package was employed to identify significant MTAs. In this study, several popular GWAS models were tested, including BLINK, MLMM, SUPER, the Compressed MLM (CMLM), FarmCPU, the General Linear Model (GLM) and MLM. The significant MTAs were determined using Bonferroni correction, (p ≤ 0.05/N, where N = number of SNPs)37. Thresholds were set at 1.96 × 10⁻⁷ to remove false associations. Manhattan plots were generated, and MTAs above threshold were considered significant. Quantile–quantile (Q–Q) plots were used to identify the best-fit models for each trait and to confirm non-random associations of MTA loci with Fe, Zn, and PC. Identification of candidate genes from MTAs for Fe, Zn and PC Candidate genes associated with the significant MTAs were identified using the GFF3 annotation file of Arachis hypogaea cv. Tifrunner (Tifrunner V1) obtained from peanutbase (https://www.peanutbase.org). Genomic regions associated with traits were examined, and genes with known roles in Fe, Zn, and PC homeostasis were prioritized. The expression of the candidate genes was examined using the Arachis hypogaea gene expression atlas (AhGEA)38. Expression data were analysed for 16 tissues (Pre-soaked seeds, emerging radicle, cotyledons, embryo, root seedling, leaves at vegetative stage, root at vegetative stage, shoot at vegetative stage, shoot seedling, immature bud, seeds at 5 days after maturity (DAM), seeds at 15 DAM, seeds at 25 DAM, immature pod wall, mature pod wall, nodules). This analysis provides insights into functional relevance of candidate genes in homeostasis of Fe, Zn and PC. KASP marker development and validation The SNPs associated with Fe were identified on 6 different chromosomes- (Ah03, Ah05, Ah14, Ah18, Ah19, Ah20), while one SNP associated with Zn was located on chromosome Ah05. These SNPs showed polymorphism between alleles associated with high vs. low Fe and Zn. SNPs were used to design KASP markers based on 300-bp flanking sequences39 (Supplementary Table 3). For each SNP marker, KASP assays were designed using two allele-specific forward primers and one common reverse primer. The developed KASP markers were validated in a diverse ICRISAT germplasm panel comprising 30 high-Fe/Zn lines and 20 low-Fe/Zn lines (Supplementary Table 4). Haplotype analysis, Haplo-pheno analysis and identification of superior haplotypes for Fe, Zn and PC The subset of up to ten representative SNPs that are in linkage disequilibrium (LD) with the lead SNP discovered by the GWAS study, were used to conduct the haplotype analysis. These representative SNPs were selected to cover a range of LD levels with the significant SNP: low LD (r2 > 0.30 and ≤ 0.50), moderate LD (r2 > 0.50 and ≤ 0.70), and high LD (r2 > 0.70). This LD binning method captured the unique haplotype block patterns around the lead SNP by identifying the SNPs with various degrees of correlation. These representative SNPs are used to define the haplotype groups based on the unique genotype combinations observed within the panel ensuring that all distinct patterns were fully represented. The identified haplotype groups were visualised using FlapJack software40. "Haplo-pheno" analysis was then carried out to evaluate the phenotypic effect of each haplotype. The phenotypic means of all the haplotype groups- which included at least three genotypes, were compared across multiple ranges. One-way analysis of variance (ANOVA), followed by the Tukey’s Honest Significant Difference (HSD) post-hoc test is performed to determine the significant differences among the groups. The haplotype or haplotype groups with the highest/preferred phenotypic mean values were considered as superior haplotypes and the haplotype groups with lowest mean values were considered as inferior haplotypes. Results Phenotypic variability, correlation and multi-variate association among Fe, Zn and PC The phenotypic distributions of Fe, Zn and PC across multiple seasons were visualized using violin plots (Fig. 1A), which showed wide variability. The mean values of Fe content varied across seasons as 21.9 ppm in S2, 29.8 ppm in S3, 23 ppm in S4, 26.7 ppm in S5 and 19.6 ppm in S6. Similarly, mean Zn values were 27.7 ppm in S2, 43 ppm in S3, 34.8 ppm in S4, 35.5 ppm in S5 and 27.2 ppm in S6. Likewise, the PC also exhibited considerable variation, with mean values of 25.1% in S1, 22.8% in S2 at L1 and 23.9% in S1, 23.3% in S2 at L2, 23.6% in S5 and 22.7% in S6 at L1. Across all seasons, Zn consistently exceeded Fe levels. Fe ranged from 7.56 ppm in S6 to 42.8 ppm in the S3, whereas Zn ranged from 10.9 ppm in S6 and to 62.35 ppm in S5. The PC values ranged from a minimum value of 12.74% in S1 at L2 to a maximum of 33.59% in S5 at L1. Phenotypic variation for Fe, Zn and PC in the diverse mini-core set. (A) Violin plots showing the distribution of Fe (ppm), Zn (ppm), and PC (%) for various genotypes across different years under both rainy (R) and post-rainy (PR) seasons. The box plot represents the average frequency of distribution of genotypes in a particular season (B) The correlation matrix indicating relationships among Fe, Zn and PC across seasons and locations, with significant positive correlation between Fe and Zn in pooled post-rainy season (r = 0.7) (C) Principal component analysis (PCA) demonstrating the distribution of the groundnut accessions (each dot on the plot represents an accession), based on the first two principal components, with PCA1 accounting for 54.6% of the variance and PCA2 accounting for 31.5%, cumulatively explaining 86.1% of the total variability. Correlation analysis (Fig. 1B) showed a significant positive correlation among Fe, Zn and PC. The strongest correlation was observed between Fe (post rainy) and Zn (post rainy) (r = 0.7, p ≤ 0.05), followed by Fe (post rainy) and Zn (rainy) (r = 0.43, p ≤ 0.05). Partial but significant positive associations were also observed between Fe (rainy) and PC (rainy- ICRISAT) (r = 0.15, p ≤ 0.05), followed by Fe (rainy) and PC (rainy- Dharwad) (r = 0.16, p ≤ 0.05). Similar associations were observed between Zn (rainy) and PC (rainy- L1) (r = 0.12, p ≤ 0.05), and between Zn (rainy) and PC (rainy- L2) (r = 0.09, p ≤ 0.05). PCA was performed using the three traits. The scatter plot (Fig. 1C) depicts the distribution of genotypes into two subgroups, revealed variations within these subgroups. The first principal component (PCA 1) (x-axis) accounted for 54.6% of total variance, while the second principal component (PCA 2) (y-axis) accounted for 31.5%, together explaining 86.1% of the total variation. The Fe and Zn across all seasons contributed positively to PCA 1, whereas PC contributed positively to both PCA 1 and PCA 2. The Fe and Zn contents were nearly parallel to each other, indicating a strong positive correlation, whereas, the PC was oriented differently from both the Fe and Zn, indicating a weaker positive correlation. Identification of MTAs GWAS analysis with average data considering the mean of the accessions across different environments identified 15 significant MTAs for the three traits (4 MTAs for Fe, 4 for Zn, and 7 for PC) (Table 1). The highest number of MTAs were observed on chromosome Ah17 (3 MTAs), followed by 2 MTAs each on chromosomes Ah02, Ah13, and Ah19. Using seasons wise data, a total of 44 significant MTAs were detected- 20 MTAs for Fe, 13 for Zn and 11 for PC (Table 2). The maximum number of MTAs were identified on chromosome Ah03 (8 MTAs), followed by Ah17 (5 MTAs) and Ah05 and Ah19 chromosomes (4 MTAs each). Table 1List of MTAs detected for Fe, Zn and PC using different models for pooled season data.TraitsSignificant MTAsChromosome noPosition (bp)Allele (Ref)Allele (Alt)P-valueSNP effectPVE %ModelsGene IDGene annotationIronArahy02_5010027Ah025010027GC1.23E−092.6126.57MLMM, SUPER, BLINKArahy.LV3APCNAC domain protein (NAC domain)Arahy.BEB4AAP-loop containing nucleoside triphosphate hydrolases superfamily proteinArahy13_17399102Ah1317399102AT3.96E−082.2811.27BLINKArahy.7P97F6Zn finger, RING/FYVE/PHD-type, P-loop containing nucleoside triphosphate hydrolaseArahy.LJ25B0Glycine-rich cell wall structural proteinArahy19_27281401Ah1927281401CT1.80E−092.3818.18MLMM, SUPERArahy.QI0PHVMYB transcription factorArahy.QM0PMGZn finger MYM-type proteinArahy20_80483965Ah2080483965CG1.20E−08− 3.7614.04BLINKArahy.BF7CWQUncharacterized proteinArahy.11RBMIProtein serine/threonine kinase, Protein kinase-like domainZincArahy02_80149889Ah0280149889GA1.38E−07− 2.3514.83FarmCPU, GLM, MLM, CMLMArahy.M50AVMAutophagy-related proteinArahy.9R964HZn finger MYM-type proteinArahy16_114034439Ah16114034439AG1.19E−14− 4.0320.49FarmCPU, GLM, MLM, CMLMArahy.J1F3AYP-loop containing nucleoside triphosphate hydrolaseArahy.1I6ZSSMYB family transcription factorArahy17_102733285Ah17102733285CA3.28E−092.730.71FarmCPU, GLM, MLM, CMLM, BLINKArahy.6K40U8Dolichyl-diphosphooligosaccharide–protein glycosyltransferase subunitArahy.XM7FE2Maestro heat-like repeat-containing proteinArahy18_127902116Ah18127902116CA2.40E−102.4912.48FarmCPU, GLM, MLM, CMLMArahy.V0ITCCUnknown proteinArahy.I3B88TRING finger and CHY Zn finger proteinProteinArahy03_125540699Ah03125540699TG4.70E−123.3116.05FarmCPU, GLM, MLM, CMLM, BLINKArahy.VMLN7LProtein serine/threonine kinasesArahy.B3BEZ6protein serine/threonine kinasesArahy11_48679516Ah1148679516GC1.99E−082.325.42FarmCPU, GLM, MLM, CMLMArahy.Y46FKNUnknown proteinArahy12_56160933Ah1256160933TG4.94E−09− 1.314.03BLINKArahy.V0ZW9LNEP1-interacting proteinArahy.B5VTGGNucleotide-binding, nucleic acid bindingArahy13_58948078Ah1358948078TG2.67E−08− 1.621.84FarmCPU, GLM, MLM, CMLMArahy.5L4UK7Nudix hydrolaseArahy.0IU8HIProtein kinase family proteinArahy17_10878383Ah1710878383AG5.44E−100.947.03BLINKArahy.KX8RC42-oxoglutarate (2OG) and Fe (II)-dependent oxygenase superfamily proteinArahy.4D7KBIProtein kinase superfamily proteinArahy17_110193095Ah17110193095GA3.83E−08− 2.586.74CMLM, FarmCPU, GLM, MLMArahy.PE3CF6E3 ubiquitin-protein ligaseArahy19_28004976Ah1928004976TC5.72E−103.104.17CMLM, FarmCPU, GLM, MLMArahy.QVG73IPhosphomevalonate kinaseArahy.X0A8D9Aspartic proteinase; Aspartic peptidase, Saposin-like, Aspartic peptidase domain List of MTAs detected for Fe, Zn and PC using different models for pooled season data. Table 2List of MTAs detected for Fe, Zn and PC using different models for season-wise data.SeasonTraitSignificant MTAsType of variantChromosome NoPosition (bp)Allele (Ref)Allele (Alt)P ValueSNP effectPVE %ModelsGene IDGene annotationsS3IronAh19_121223186Intergenic regionAh19121223186TC9.82E−081.912.51SUPER, MLMMArahy.H66LJ0Polyprotein Citrus endogenous pararetrovirusArahy.H3HGUTMYB transcription factorS3IronAh20_80483965Intergenic regionAh2080483965CG1.53E−10− 2.6953.43BLINK, FarmCPUArahy.BF7CWQUncharacterized proteinArahy.11RBMIATP-binding/protein serine/threonine kinaseS2IronAh03_28583444Downstream gene variantAh0328583444CT1.47E−131.4415.37BLINK, SUPER, MLMMArahy.9BBP6DL-lactate/malate dehydrogenaseS2IronAh03_105594763Intergenic regionAh03105594763CT4.84E−121.741.24FarmCPU, BLINK, MLMMArahy.H3WGQZUnknown proteinArahy.MYB15FProtein LURP-one-related 17-likeS2IronAh04_18847124Intergenic regionAh0418847124CT3.77E−122.069.68BLINKArahy.CT4NDIHeavy metal transport/detoxification superfamily proteinArahy.I3MG1XHistone-lysine N-methyltransferase ATXR2-like isoformS2IronAh05_10145205Intergenic regionAh0510145205GA8.47E−08− 1.531.12SUPERArahy.XN4TZ5Plasma membrane mannitol transporterArahy.GVE3NJUnknown proteinS2IronAh10_32538050Intergenic regionAh1032538050TA2.78E−081.402.33FarmCPUArahy.ZME7HSUncharacterized proteinArahy.DGD0VBOxysterol-binding protein-related proteinS2IronAh11_72760600Intergenic regionAh1172760600GA5.53E−101.699.36BLINK, MLMMArahy.W1WWEBHomolog of nucleolar protein (Nop domain)Arahy.XPYE9KATP phosphoribosyl transferaseS2IronAh17_7608976Intergenic regionAh177608976GA2.91E−081.162.55FarmCPUArahy.4B5N5UReceptor-like protein kinase; protein serine/threonine kinaseArahy.6Q0Z4RZinc finger MYM-type proteinS2IronAh17_131484790Upstream gene variantAh17131484790CT1.35E−081.501.99FarmCPUArahy.Q0I0E2NADPH-cytochrome family reductaseS2IronAh18_122103781Intergenic regionAh18122103781CT1.10E−072.1114.63SUPERArahy.Q4GIHGZinc finger MYM-type proteinArahy.USIX8K5-methylthioadenosine/S-adenosylhomocysteine deaminaseS2IronAh20_70265096Intergenic regionAh2070265096GC9.49E−09− 0.4622.38BLINKArahy.ZBIL8CCyclophilin-like peptidyl-prolyl cis–trans isomerase family proteinArahy.UCCR5RActin-binding FH2 (Formin Homology) proteinS4IronAh09_107364323Upstream gene variantAh09107364323CT5.29E−081.095.44SUPERArahy.CUT3DDTranscriptional activator DEMETER-like, DNA glycosylaseS4IronAh13_58751943Intergenic regionAh1358751943AT8.47E−081.387.93SUPERArahy.B2ND5QUnknown proteinArahy.5L4UK7Nudix hydrolase homolog 14S4IronAh17_12433761Intergenic regionAh1712433761CT3.44E−081.4412.23SUPERArahy.13YYB8Unknown proteinArahy.3DA9LKTransport inhibitor response like proteinS4IronAh18_55925435Intergenic regionAh1855925435TC2.54E−092.3618.75MLMM, SUPERArahy.6ZMA90Novel plant snare 12Arahy.MQHC27Actin-binding FH2 (Formin Homology) proteinS4IronAh19_135254456Intergenic regionAh19135254456CT9.11E−081.041.03SUPERArahy.2U0JEXUDP-Glycosyltransferase superfamily proteinArahy.999YA8Calcium-dependent lipid-binding family proteinS4IronAh19_145627456Intergenic regionAh19145627456TC1.74E−08− 1.005.35SUPERArahy.72B7PKL-type lectin-domain containing receptor kinaseArahy.I6PPQFProbable protein phosphataseS5IronAh14_5692509Upstream gene variantAh145692509AG1.41E−0714.8565.38FarmCPU, BLINKArahy.NX1XSTDihydroxy-acid dehydrataseS6IronAh03_87593234Intergenic regionAh0387593234TG9.09E−105.5151.21BLINK, MLMM, FarmCPUArahy.HNJF4G26S proteasome regulatory subunit like isoformArahy.9RX38MDCD (Development and Cell Death) domain proteinS3ZincAh05_35417862Intergenic regionAh0535417862TC5.09E−08− 1.8955.01MLMMArahy.AN6HNQUncharacterized proteinArahy.GRNJ5WZinc finger, RING/FYVE/PHD-typeS3ZincAh05_27023896Intergenic regionAh0527023896TC1.04E−10− 1.1167.42BLINKArahy.F1RJAKFormin, FH2 domainArahy.G26LFAPhosphorelay response regulatorS2ZincAh03_28583444Downstream gene variantAh0328583444CT5.29E−131.4113.25BLINK, FarmCPUArahy.9BBP6DL-lactate/malate dehydrogenaseS2ZincAh05_15332507Intergenic regionAh0515332507CT7.09E−091.7512.08BLINKArahy.TLT0RJZinc finger, CCHC-typeArahy.29A50MMyc-type, basic helix-loop-helix (bHLH) domainS2ZincAh06_100584371Intergenic regionAh06100584371CT7.27E-−11− 1.5010.72BLINKArahy.NWPU6YUncharacterized proteinArahy.12HUMHNAC domain proteinS2ZincAh07_51562128Upstream gene variantAh0751562128AG1.17E−093.2316.89MLMMArahy.6YW6LEAldo/keto reductase family oxidoreductase; NADP-dependent oxidoreductase domainS2ZincAh09_7062575Intergenic regionAh097062575CA1.74E−10− 1.901.65FarmCPUArahy.C4WT98P-loop containing nucleoside triphosphate hydrolaseArahy.KAVD32Ethylene insensitive 3 family proteinS2ZincAh10_86959972Intron variantAh1086959972GA2.35E−10− 1.9729.06BLINKArahy.4HXN36Myb/SANT-like DNA-binding domain proteinS2ZincAh15_103380195Intergenic regionAh15103380195AG2.23E−092.038.40BLINKArahy.QXL28EGDSL-like Lipase/Acylhydrolase superfamily proteinArahy.YMBN2YGDSL-like Lipase/Acylhydrolase superfamily proteinS2ZincAh19_84532565Intergenic regionAh1984532565GA5.55E−091.316.61BLINKArahy.N0IL4QPhosphoribosylaminoimidazole carboxylaseArahy.6PZ0D3DNA mismatch repair protein, UbiB domainS4ZincAh13_120211775Intergenic regionAh13120211775GA1.69E−09− 3.5238.05BLINKArahy.292F09CLPTM1-like membrane proteinArahy.H71TASCalcium-dependent protein kinaseS5ZincAh02_51505705Intergenic regionAh0251505705GA4.82E−30− 29.2299.12BLINK, MLMM, CMLM, FarmCPUArahy.U6R2AIUncharacterized proteinArahy.J9ASUVMicronuclear linker histone polyproteinS5ZincAh02_46495272Intergenic regionAh0246495272CA9.91E−1216.212.66MLMMArahy.7X6UW5Uncharacterized proteinArahy.J1W7N5Unknown proteinS1 L2ProteinAh03_125540699Intergenic regionAh03125540699TG5.43E−113.2825.69FarmCPU, GLMArahy.VMLN7LProtein serine/threonine kinasesArahy.B3BEZ6Protein serine/threonine kinasesS1 L2ProteinAh13_18601962Intergenic regionAh1318601962GT2.93E−081.885.54FarmCPUArahy.NG0HJGUnknown proteinArahy.H2FU9ZUnknown proteinS1 L2ProteinAh17_49411056Intergenic regionAh1749411056AG2.46E−08− 2.341.07FarmCPUArahy.P38DRGUnknown proteinArahy.DD6N9KProtein tyrosine/serine/threonine phosphataseS1 L2ProteinAh17_133010553Intron variantAh17133010553CT3.47E−09− 3.078.53FarmCPUArahy.RFW94SSyntaxin, putative;S2 L2ProteinAh03_125540699Intergenic regionAh03125540699TG8.41E−124.4010.95GLM, FarmCPUArahy.VMLN7LProtein serine/threonine kinasesArahy.B3BEZ6Protein serine/threonine kinasesS2 L2ProteinAh03_107178229Intergenic regionAh03107178229CT5.43E−10− 3.970.50MLMMArahy.D5URA7Novel plant SNAREArahy.9A4AY7MYB transcription factorS2 L2ProteinAh12_116861708Upstream gene variantAh12116861708GA3.90E−084.1918.70MLMMArahy.SITW69FAD-binding Berberine family protein, FAD-binding, UDP-N-acetylmuramate dehydrogenase activity, flavin adenine dinucleotide bindingS2 L2ProteinAh18_126543814Upstream gene variantAh18126543814AT1.49E–−091.642.60FarmCPUArahy.19AV4IP-loop containing nucleoside triphosphate hydrolaseS1 L1ProteinAh03_125540699Intergenic regionAh03125540699TG1.20E−072.457.82GLMArahy.VMLN7LProtein serine/threonine kinasesArahy.B3BEZ6Protein serine/threonine kinasesS1 L1ProteinAh03_68769471Upstream gene variantAh0368769471GA1.59E−152.423.50FarmCPU, GLM, MLMMArahy.GBY2EBPutative pectinesterase/pectinesterase inhibitorS1 L1ProteinAh15_49814667Intergenic regionAh1549814667GT6.05E−10− 1.756.03FarmCPUArahy.0RJM1QUDP-glucose 6-dehydrogenase family proteinArahy.I0PDCZCinnamyl alcohol dehydrogenaseS1 L1ProteinAh19_6490479Intergenic regionAh196490479CT9.35E−150.0317.59FarmCPUArahy.7NK30AHeat shock proteinArahy.KJ6PSZHeat shock proteinS2 L1ProteinAh03_27122996Upstream gene variantAh0327122996TC4.17E−09− 2.145.24FarmCPUArahy.ZJ4Z2JProtein serine/threonine kinaseS2 L1ProteinAh03_125540699Intergenic regionAh03125540699TG7.81E−124.444.23FarmCPU, GLMArahy.VMLN7LProtein serine/threonine kinasesArahy.B3BEZ6Protein serine/threonine kinasesS2 L1ProteinAh03_107178229Intergenic regionAh03107178229CT1.80E−11− 3.9754.30MLMMArahy.D5URA7Novel plant SNAREArahy.9A4AY7MYB transcription factorS2 L1ProteinAh12_116861708Upstream gene variantAh12116861708GA3.63E−094.1120.01MLMMArahy.SITW69FAD-binding Berberine family protein, FAD-binding, UDP-N-acetylmuramate dehydrogenase activity, flavin adenine dinucleotide bindingSeasons: Rainy seasons 2008 (S1); Rainy seasons 2009 (S2); post-rainy 2009–10 (S3), post rainy 2010–11 (S4), rainy 2020 (S5) and rainy 2021 (S6), Locations: ICRISAT, Patancheru, (L1); Dharwad (L2). List of MTAs detected for Fe, Zn and PC using different models for season-wise data. Seasons: Rainy seasons 2008 (S1); Rainy seasons 2009 (S2); post-rainy 2009–10 (S3), post rainy 2010–11 (S4), rainy 2020 (S5) and rainy 2021 (S6), Locations: ICRISAT, Patancheru, (L1); Dharwad (L2). Genomic associations identified for Fe For pooled season data, 4 significant MTAs were identified for Fe on chromosomes Ah02 (Ah02_5010027), Ah13 (Ah13_17399102), Ah19 (Ah19_27281401) and Ah20 (Ah20_80483965) (Fig. 2A). Their p-values ranged from 1.20 × 10–8 (Ah20_80483965) to 1.80 × 10–9 (Ah19_27281401), explaining 14.04% to 18.18% of phenotypic variation. The SNPs Ah13_17399102 and Ah20_80483965 were uniquely identified by BLINK model. Notably, MTA (Ah02_5010027) associated with Fe, identified on chromosome Ah02 was detected by the models (BLINK, MLMM and SUPER), with the highest phenotypic variance of 26.57% and ‘p’ value of 1.23 × 10–9, whereas Ah13_17399102 was found to have the lowest phenotypic variance of 11.27% with a ‘p’ value of 3.96 × 10–8. Manhattan plots and optimal Q-Q plots for (A) Fe (B) Zn and (C) PC for pooled season data. The horizontal dashed lines represent the significant threshold (p = 1 × 10−7, Bonferroni correction) and the vertical dashed lines represent the candidate genes at the significant MTAs associated with (A) Fe on chromosomes- Ah02, Ah13, Ah19 and Ah20 (B) Zn on chromosomes- Ah02, Ah16, Ah17 and Ah18 (C) PC on chromosomes- Ah03, Ah11, Ah12, Ah13, Ah17 and Ah19. For seasonal analysis, 20 significant MTAs were identified for Fe on different chromosomes Ah03, Ah04, Ah05, Ah09, Ah10, Ah11, Ah13, Ah17, Ah18, Ah19 and Ah20. The p-values ranged from 1.10 × 10–7 (Ah18_122103781) to 1.47 × 10–13 (Ah03_28583444), explaining 14.63% to 15.37% of phenotypic variation. Three MTAs on chromosome Ah03 (Ah03_28583444, Ah03_105594763 and Ah03_87593234) were consistently identified by at least three models during 2009 rainy and 2021 rainy (Supplementary Fig. 1 and 2). The highest phenotypic variance (65.38%) was observed for Ah14_5692509 with a ‘p’ value of 1.41 × 10–7 followed by Ah20_80483965 (53.43%) with a ‘p’ value of 1.53 × 10–10. Genomic associations identified for Zn For pooled season data, 4 significant MTAs were identified for Zn on chromosomes Ah02 (Ah02_80149889), Ah16 (Ah16_114034439), Ah17 (Ah17_102733285), and Ah18 (Ah18_127902116) (Fig. 2B). All the MTAs associated with the Zn were detected by the models- FarmCPU, GLM, MLM and CMLM. The MTA Ah16_114034439 was observed with highest phenotypic variance of 20.49% and a p-value of 1.19 × 10–14, whereas Ah17_102733285 was observed with lowest phenotypic variance of 0.71% and a p-value of 3.28 × 10–9. The other MTAs, Ah02_80149889 and Ah18_127902116, located on chromosomes Ah02 and Ah18 were identified with phenotypic variance values of 14.83% and 12.48%, and ‘p’ values of 1.38 × 10–7 and 2.40 × 10–10, respectively. Seasonal analysis identified 13 significant MTAs across 10 chromosomes. The p-values for these significant MTAs ranged from 5.09 × 10–8 (Ah05_35417862) to 4.82 × 10–30 (Ah02_51505705), explaining 55.01% to 99.12% of phenotypic variance. The MTA Ah02_51505705 on chromosome Ah02 was detected by the four models (BLINK, CMLM, FarmCPU, MLMM) during 2020 rainy (Supplementary Fig. 3 and 4). The highest phenotypic variance (99.12%) was observed for Ah02_51505705 with a ‘p’ value of 4.82 × 10–30 followed by Ah05_27023896 (67.42%) with a ‘p’ value of 1.04 × 10–10. Genomic associations identified for PC For PC, seven significant MTAs were identified for pooled season data across six chromosomes like Ah03 (Ah03_2532408), Ah11 (Ah11_48679516), Ah12 (Ah12_56160933), Ah13 (Ah13_58948078), Ah17 (Ah17_10878383 and Ah17_110193095), Ah19 (Ah19_28004976) (Fig. 2C). All the MTAs associated with PC were detected by FarmCPU, GLM, MLM and CMLM models. The p-values for these MTAs ranged from 1.99 × 10–8 (Ah11_48679516) to 5.72 × 10–10 (Ah19_28004976), explaining 5.42% to 4.17% of phenotypic variance. Interestingly, two MTAs coding for the PC were found on the chromosome Ah17 at two different positions. The highest phenotypic variance of 16.05% was observed for Ah03_125540699 with a ‘p’ value of 4.70 × 10–12 and the lowest phenotypic variance of 1.84% was observed for Ah13_58948078 with a ‘p’ value of 2.67 × 10–8. For season-wise analysis, 11 significant MTAs were identified on seven chromosomes (Ah03, Ah12, Ah13, Ah15, Ah17, Ah18 and Ah19). The p-values for these significant MTAs ranged from 1.20 × 10–7 (Ah03_125540699) to 9.35 × 10–15 (Ah19_6490479), explaining 7.82% to 17.59% of phenotypic variance. The MTA Ah03_68769471 on chromosome Ah03 was consistently detected by the three models (FarmCPU, GLM, MLMM) (Supplementary Fig. 5 and 6). Identification of candidate genes and their expression analysis A total of 28 candidate genes were identified for pooled season data, including 8 each for Fe and Zn while 12 for PC. The seasonal data analysis identified 62 genes, including 30 for Fe, 18 for Zn and 14 for PC, associated with the MTAs. These candidate genes were involved in different molecular, cellular, and biological functions related to the traits. These candidate genes encode various important proteins like MYB transcription factor, Zn finger MYM type protein, RING finger MYM type protein, ATP binding/ threonine kinase/ protein serine/ protein kinase, NAC domain protein, and Glycine rich cell wall structural protein. Several other genes like Heavy metal transport/detoxification superfamily protein, ATP phosphoribosyl transferase, NADPH-cytochrome family reductase, Nudix hydrolase, Probable protein phosphatase, Calcium-dependent protein kinase, Heat shock protein, E3 ubiquitin-protein ligase, 2-oxoglutarate (2OG) and Fe (II)-dependent oxygenase superfamily protein, Phosphomevalonate kinase, aspartic proteinase, 3-ketoacyl-CoA synthase, Dolichyl-diphosphooligosaccharide-protein and Jasmonic acid carboxyl methyltransferase were identified as candidate genes related to the trait functions and biological processes. The domains for the proteins encoded by these genes were either directly or indirectly related to the Fe, Zn and PC. The SNPs, Ah02_5010027 associated with NAC domain protein, Ah02_80149889, Ah18_127902116, Ah13_17399102, Ah17_7608976, Ah18_122103781, Ah05_35417862, Ah05_15332507 associated with RING and Zn finger protein, Ah16_114034439, Ah18_27281401, Ah19_121223186, Ah10_86959972 associated with MYB transcription were found linked to Fe and Zn homeostasis pathway. Among the total candidate genes identified, 17 genes showed expression during different developmental stages based on the Arachis hypogaea gene expression atlas (AhGEA)38 (Fig. 3). Gene expression studies using AhGEA show that NAC domain protein (Arahy.LV3APC), Heavy metal transport/detoxification superfamily protein (Arahy.CT4NDI), 2-oxoglutarate (2 OG) and Fe (II)-dependent oxygenase (Arahy.KX8RC4) and Heat shock protein (Arahy.7NK30A) were found to show higher expression in majority of the plant tissues studied, whereas Glycine rich cell wall structural protein (Arahy.LJ25B0), MYB transcription factor (Arahy.QI0PHV) and Phosphomevalonate kinase (Arahy.QVG73I) showed higher expression in some specific plant tissues only. Tissue-specific expression study of candidate genes during different developmental stages based on AhGEA data. The expression of 17 candidate genes in 16 different tissues (Pre-soaked seeds, emerging radicle, cotyledons, embryo, root seedling, leaves at vegetative stage, root at vegetative stage, shoot at vegetative stage, shoot seedling, immature bud, seeds at 5 DAM, seeds at 15 DAM, seeds at 25 DAM, immature pod wall, mature pod wall, nodules) are plotted in the heatmap. Allele mining for significantly associated SNPs In the current study, significant MTAs for all three traits and their associated candidate genes were identified on different chromosomes. To utilize these MTAs and genes, SNP polymorphism was examined in a diverse mini-core collection. Based on MTA positions associated with the three traits, allele mining was conducted to identify the alleles associated with high and low trait values in the diverse mini-core lines (Supplementary Fig. 7). The diverse set included 16 genotypes for Fe, 22 genotypes for Zn and 18 genotypes for PC. Among the 16 Fe genotypes, high Fe genotypes ranged from 28.7 to 52.5 ppm and low Fe genotypes ranged from 15.5 to 19.4 ppm. Among the 22 Zn genotypes, high Zn genotypes ranged from 40.1 to 49.2 ppm, while low Zn genotypes ranged from 19.8 to 27.9 ppm. Similarly, for 18 PC genotypes, high PC genotypes ranged between 26.8 to 28.6%, while low PC genotypes ranged from 18.7 to 19.8%. Of the four MTAs associated with Fe, three MTAs -Ah02_5010027, Ah13_17399102 and Ah20_80483965 showed clear polymorphism between high and low Fe genotypes. Among the four MTAs identified for Zn, two MTAs -Ah16_114034439 and Ah18_127902116 showed clear polymorphism between the high and low Zn content lines. Similarly, among the seven MTAs identified for PC, two MTAs Ah03_125540699 and Ah19_28004976 showed clear polymorphism between high and low PC lines. Interestingly, genotype ICG 9249 with mean values of Fe-30.1 ppm, Zn-40.3 ppm and PC-27.3% carried alleles associated with high Fe, Zn and PC. In contrast, four genotypes, ICG 2777 (Fe- 19.4 ppm, Zn- 27.9 ppm), ICG 6813 (Fe- 19 ppm, Zn- 26.3 ppm), ICG 5016 (Fe- 18.4 ppm, Zn- 25.7 ppm) and ICG 14834 (Fe- 15.5 ppm, Zn- 19.8 ppm) carried alternate alleles for low Fe and Zn. Likewise, in-silico validation of the MTAs identified from the season-wise analysis was done using diverse set of 40 genotypes for both Fe and Zn (Supplementary Fig. 8). Among the 20 MTAs associated with Fe, 8 MTAs- Ah05_10145205, Ah18_122103781, Ah20_70265096, Ah18_55925435, Ah19_135254456, Ah19_145627456, Ah14_5692509, Ah03_87593234 showed clear polymorphism between high and low Fe genotypes. Similarly, among the 13 MTAs associated with Zn, three MTAs- Ah05_35417862, Ah05_27023896 and Ah02_46495272 showed clear polymorphism between high and low Zn lines. Seven genotypes—ICG 14118 (Fe = 42.9 ppm, Zn = 53.3 ppm), ICG 15309 (Fe = 37.6 ppm, Zn = 52.9 ppm), ICG 36 (Fe = 36.4 ppm, Zn = 56.6 ppm), ICG 4911 (Fe = 35.2 ppm, Zn = 53.9 ppm), ICG 7963 (Fe = 34.7 ppm, Zn = 47.7 ppm), ICG 10092 (Fe = 34 ppm, Zn = 48.4 ppm), and ICG 1274 (Fe = 31.1 ppm, Zn = 47.7 ppm)—carried alleles associated with high Fe and Zn. In contrast, genotypes such as ICG 8760 (Fe = 20.9 ppm, Zn = 25.7 ppm), ICG 4412 (Fe = 20.6 ppm, Zn = 13.7 ppm), ICG 4156 (Fe = 20.3 ppm, Zn = 15.9 ppm), ICG 4389 (Fe = 19.7 ppm, Zn = 16.4 ppm), ICG 76 (Fe = 18.9 ppm, Zn = 14.2 ppm), ICG 4527 (Fe = 17.7 ppm, Zn = 15.3 ppm), ICG 13723 (Fe = 10.9 ppm, Zn = 14.8 ppm), and ICG 5016 (Fe = 9.2 ppm, Zn = 24.4 ppm) carried alleles associated with low Fe and Zn. These findings suggest that these identified SNPs can be effectively deployed for allele-specific marker development for marker-assisted selection (MAS) to improve nutritional quality in groundnut. Development and validation of KASP markers Based on the identified SNP polymorphisms, 9 MTAs (2 MTAs on Ah05, 2 MTAs on Ah18, 2 MTAs on Ah19, 1 MTA each on Ah03, Ah14 and Ah20) were used to develop KASP markers. These markers were successfully developed and validated in a panel of 50 genotypes, representing the ICRISAT reference set and mini-core collection (Fig. 4). The validation panel comprised of wide range of Fe (7.6 to 42.9 ppm) and Zn (10.9 to 62.4 ppm) contents. Of the nine validated KASPs (eight for Fe and one for Zn), three KASPs (snpAH00636—intergenic SNP between Arahy.HNJF4G and Arahy.9RX38M on chromosome Ah03, snpAH00641 -intergenic SNP between Arahy.2U0JEX and Arahy.999YA8 on chromosome Ah19, snpAH00644—intergenic SNP between Arahy.AN6HNQ and Arahy.GRNJ5W on chromosome Ah05) showed clear polymorphism in the diverse germplasm. These 3 KASP markers can be used as diagnostic markers for their effective utilization in marker-assisted breeding for enhancing Fe and Zn in groundnut. Validation of Kompetitive allele specific polymerase chain reaction (KASP) markers for Fe and Zn. KASP markers were validated in a panel consisting of 50 genotypes of both high and low Fe, Zn content lines with Fe ranging from 7.6 to 42.9 ppm and Zn ranging from 10.9 to 62.4 ppm. Three KASP markers (snpAH00636, snpAH00641 and snpAH00644) were successfully validated and showed polymorphism between high and low Fe, Zn content genotypes. SNPs show clear homozygous clusters- snpAH00636 for genes Arahy.HNJF4G (26S proteasome regulatory subunit), Arahy.9RX38M (Development and Cell Death domain protein); snpAH00641 for genes Arahy.2U0JEX (UDP-Glycosyltransferase superfamily protein), Arahy.999YA8 (calcium-dependent lipid-binding family protein); snpAH00644 for genes Arahy.AN6HNQ (uncharacterized protein), Arahy.GRNJ5W (Zinc finger, RING/FYVE/PHD-type). Haplo-pheno analysis for the identification of superior haplotypes for Fe, Zn and PC Haplo-pheno analysis with haplotype and phenotype data on 178 accessions was used to identify the superior haplotypes. The performance of a haplogroup (a set of genotypes with the same haplotype sequence) was compared to other haplogroups in order to identify a superior haplotype for the MTAs associated with Fe, Zn and PC. The 9 superior haplotypes for Fe were identified as H3 for Ah03_8759323, H4 for Ah18_55925435, H1 for Ah13_58751943, H1 for Ah17_12433761, H2 for Ah19_145627456, H4 for Ah13_17399102, H1 for Ah20_80483965, H4 for Ah09_10736323 and H1 for Ah19_13525456 (Supplementary Fig. 9 and 10). The 3 superior haplotypes for Zn were found to be H4 for Ah16_114034439, H2 for Ah05_15332507 and H4 for Ah05_35417862 (Supplementary Fig. 11). The superior haplotypes for PC were observed as H2 for Ah15_49814667 and H4 for Ah13_18601962 (Supplementary Fig. 12). The superior haplotypes for the two traits- Fe and Zn majorly belong to the subspecies fastigiata, and agronomic type, Valencia bunch/Spanish bunch, whereas the inferior haplotypes of these traits belong to the subspecies hypogaea, and agronomic type, Virginia bunch/Virginia Runner. Similarly, the superior haplotypes related to the PC were observed in the accessions that belong to Spanish Bunch and Valencia Bunch, however the common accessions for the haplotypes with lowest mean values of PC were identified in all the agronomic types, Spanish Bunch/Valencia Bunch/Virginia Bunch/Virginia runner. Therefore, the haplotype diversity for PC did not show subspecies relationship with the inferior haplotype groups. The genotypes that had superior haplotype combinations showed an increase in Fe and Zn levels in the accessions (ICG 14118, ICG 5494), compared to the accessions carrying inferior haplotype combinations (ICG 721, ICG 6813) which exhibited reduced levels of Fe and Zn. The genotypes ICG 14118, ICG 11249, ICG 9507, ICG 7906 and ICG 8567 exhibited increased Fe, Zn and PC levels with different combinations of superior haplotypes for the three traits (Supplementary Table 6). Discussion Malnutrition, or “hidden hunger,” remains a significant global health challenge arising from inadequate nutrient intake. Biofortification offers a sustainable solution by improving essential nutritional content of crops, thus directly linking agriculture with nutrition and health41. Several crop varieties have been enriched with essential nutrients to combat malnutrition41, including groundnut, which provides essential nutrients and amino acids necessary for good health. Further enhancement of its protein and micronutrient content could make groundnut an important food crop in addressing malnutrition. The identification and characterization of genomic regions and candidate genes associated with these traits will advance biofortification efforts in groundnut. In this study, significant phenotypic variability was observed for the three traits, consistent with earlier reports for Fe, Zn and PC in groundnut42, and other crops, such as chickpea43 and wheat44. Breeding strategies, such as multiple trait pyramiding, can be used to improve correlated traits like Fe and Zn5. A significant positive correlation was observed between Fe, Zn and PC across locations and seasons in this study. Similar results were previously reported in groundnut (r = 0.31)5, (r = 0.535)42, (r = 0.714)26 and in other legumes, including soybean (r = 0.29)45, mung bean (r = 0.469)46, and common bean (r = 0.416)47. These findings indicate that several common pathways are associated with the homeostasis of Fe and Zn. PCA further supported this, as Fe and Zn contributed positively to PCA1, while PC contributed to both PCA1 and PCA2. Similar results were observed in chickpea, where Fe and Zn loaded on PCA148 and PC contributed to both PCA1 and PCA249. This study highlights the genetic potential for enhancing Fe, Zn, and protein content in groundnut; however, the lack of multi-environment phenotypic data limited the ability to assess genotype stability. Micronutrient traits, including Fe and Zn, were significantly influenced by genotype-by-environment (G × E) interactions, and exhibited significant variability across different environments. Therefore, future studies should incorporate multi-environmental trials to validate these findings to ensure accurately capture environmental variance. GWAS analysis detected a stable MTA (Ah03_28583444) on chromosome Ah03 for both Fe and Zn in S2. Similarly, a previous study identified a co-localized QTL linked with Fe and Zn on Ah035. The genetic marker Ah03_28583444, closely associated with the MTA on chromosome Ah03, showed potential for further validation across diverse populations and could be used for early-generation selection in breeding programmes50. Previous studies reported significant QTLs for Fe, Zn on chromosomes Ah03, Ah05 and Ah135, and for PC on chromosomes Ah03, Ah11, Ah12, and Ah1851, which are consistent with the present results. The highest number of MTAs for the three traits were identified on chromosomes Ah03, Ah17, Ah05 and Ah19. These genomic regions can be considered as hotspot regions for simultaneous trait improvement52. Another stable MTA (Ah03_125540699) associated with PC, was identified on Ah03 in four different seasons (S1, S2 at L1 and S1, S2 at L2), while two stable MTAs (Ah03_107178229, Ah03_116861708) on chromosome Ah03 were consistently detected in two different seasons (S2 at L2, S2 at L1). Earlier studies suggest that significant MTAs are less influenced by environmental variation53. This study identified key candidate genes associated with the significant SNPs that regulate the Fe and Zn homeostasis pathways. The genes encoding the proteins and transcription factors associated with Fe and Zn were found to suppress or inhibit the plant’s response to the deficiency conditions and also help in activation of a series of response mechanisms to enhance Fe and Zn uptake and homeostasis54,55 (Fig. 5). These genes facilitated Fe+3 and Zn+2 ions uptake from the rhizosphere into the root tissues and also involve in the mobilization of these ions from root to the shoot and seed54,55. The genes, RING/ Zn finger domains (Arahy.7P97F6, Arahy.9R964H and Arahy.I3B88T), which act as E3 ubiquitin ligases were identified to negatively regulate the Fe deficiency response56,57. Similarly, MYB 10 and MYB 72 that belong to the family, MYB transcription factor (Arahy.QI0PHV, Arahy.1I6ZSS), induce the NAS genes under the Fe and Zn deficiencies that increase the levels of Fe and Zn in the shoot and the seed58. Another transcription factor, IDEF2 that belongs to the NAC domain protein family (Arahy.LV3APC), binds to cis-regulatory element, IDE2 present at the promoter site of Fe-deficiency response genes and controls the homeostasis of Fe by modulating the expression of these genes59. These genes serve as a valuable resource for the development of molecular markers and for further genetic and functional characterization60. The genes MYB transcription factor (Arahy.QI0PHV, Arahy.1I6ZSS), and NAC domain (Arahy.LV3APC) were earlier reported to show their expression in groundnut kernel61,62. Molecular mechanism of Fe and Zn homeostasis in groundnut root cell and kernel. The figure depicts the Fe and Zn deficiency response in the groundnut root cell and the uptake, storage of both Fe and Zn in the groundnut kernel. bHLH basic helix-loop helix; FIT FER-like iron deficiency induced transcription factor; FRO ferric reduction oxidase; IRT iron-regulated transporter; IRO iron reduction oxidase; MYB myeoloblastosis; PYE POPEYE transcription factor; NAS nicotianamine adenine synthase; HRZ Haemerythrin motif-containing Really Interesting New Gene (RING)-and Zinc-finger protein; IDEF iron deficiency responsive element binding factor; NA nicotianamine adenine; YSL yellow stripe like; VIT vacuolar iron transporter; NRAMP natural resistance associated macrophage protein; FRD ferric reductase defective; FPN ferriportin; IDE iron deficiency responsive element; SAM S-adenosyl L-methionine. This study also reported several other candidate genes that are either directly or indirectly associated with the functional regulation of Fe, Zn and PC in the plants. Previous studies have shown that P-loop containing nucleoside triphosphate hydrolases (Arahy.BEB4AA) play a key role in the nuclear and cytosolic compartmentalisation, storage, and export of Fe in the forms of ferritin and ferroportin, and are involved in Zn-ion binding57. The gene, ATP-binding/protein serine/threonine kinase (Arahy.11RBMI, Arahy.VMLN7L, Arahy.B3BEZ6, and Arahy.0IU8HI), is involved in the homeostasis of Fe in Arabidopsis and the transcriptional activity regulation63. The autophagy-related protein (Arahy.M50AVM) is involved in the intracellular utilisation of Zn and inhibits the accumulation of reactive oxygen species (ROS) caused by Zn deficiency64. Calcium-dependent protein kinase family helps in the stress responsive regulation of Zn+2 ions65. The gene, E3 ubiquitin-protein ligase (Arahy.PE3CF6), identifies the target proteins and has crucial role in protein degradation and is a major component of ubiquitination cascade during plant response to stress66. Heat shock proteins (HSPs) play an important role in protein folding, assembly, translocation, and degradation under stress and in several regular cellular activities67. The gene expression analysis based on the AhGEA data showed that NAC transcription factor (Arahy.LV3APC) showed higher expression in multiple plant tissues, including seeds, roots, and nodules. Whereas MYB transcription factor (Arahy.QI0PHV) exhibited higher expression in seeds, cotyledon and embryo, Earlier studies revealed the higher expression of NAC genes in root tissues, nodules, and developing seed9,62, and MYB transcription factor expressed highly in developing seed and shell13,61. Overall, this study highlights the crucial role of key candidate genes in Fe, Zn, and protein homeostasis. Identification and cloning of these genes will improve understanding of genetic and biological functioning of the traits. Furthermore, haplotype-based approaches using diverse sequencing data can aid in developing diagnostic markers to accelerate nutritional biofortification in groundnut. Diagnostic markers enhance breeding efficiency by enabling early-generation selection through MAS68. The validated KASP markers for various traits, such as LLS, high oleic acid and leaf rust resistance, were widely utilized in groundnut breeding programs globally, contributing to the development of advanced breeding lines, some of which are available for commercial cultivation69. To identify and confirm the diagnostic markers associated with Fe and Zn, 9 SNPs were selected for the development of KASP markers, Among the 9 KASPs, 3 were successfully validated, and these validated markers can accelerate biofortification breeding programs through MAS. LD based haplotype identification offers advantage, focusing on identifying comprehensive historical recombination events within a population and allowing for easy visualization70. Haplotypes offer a more comprehensive view by combining SNPs and other associated polymorphisms into inherited units71. Likewise, significant haplotype diversity was observed for the MTAs associated with the Fe, Zn and PC in the current study. Previous studies also reported superior haplotypes for Fe, Zn and PC in rice, wheat, sorghum and pigeonpea72–74. The different combinations of haplotypes showed variations in Fe and Zn contents, with superior haplotype combination showing higher levels of Fe and Zn. Therefore, pyramiding of superior haplotype blocks associated with micronutrient accumulation, using HBB, could enhance the concentrations of essential micronutrient elements, in the grains, including Zn75,76. In the present study, the accessions with superior haplotypes identified for Fe, Zn and PC showed better phenotypic performance compared to the accessions with inferior haplotypes (Supplementary Fig. 13). Overall, incorporation of these superior haplotypes in groundnut breeding programs has the potential for developing groundnut cultivars with improved kernel Fe, Zn, and PC (Supplementary Fig. 14). Conclusion This study presents a comprehensive GWAS analysis utilizing WGRS data and multi-season phenotypic data from a mini-core collection to identify significant MTAs and candidate genes associated with Fe, Zn and PC in groundnut. A total of 15 MTAs and 28 candidate genes were identified from pooled season data, while 44 MTAs and 62 candidate genes were detected from seasonal data for Fe, Zn and PC. Among the nine KASP markers, three KASP markers (snpAH00636—intergenic SNP between Arahy.HNJF4G and Arahy.9RX38M on chromosome Ah03, snpAH00641—intergenic SNP between Arahy.2U0JEX and Arahy.999YA8 on chromosome Ah19, snpAH00644—intergenic SNP between Arahy.AN6HNQ and Arahy.GRNJ5W on chromosome Ah05) showed clear polymorphism between high and low content genotypes in the diverse germplasm. Key candidate genes including MYB transcription factor (Arahy.QI0PHV, Arahy.1I6ZSS), Zn finger MYM type protein and RING finger MYM type protein (Arahy.7P97F6, Arahy.9R964H and Arahy.I3B88T), and NAC domain protein (Arahy.LV3APC) were found to be involved in the regulation of Fe and Zn homeostasis pathways. Haplotype analysis identified 9 superior haplotypes for Fe, 3 for Zn and 2 for PC. The haplo-pheno analysis revealed that superior haplotypes majorly found in the subspecies fastigiata, and agronomic type, Valencia bunch/Spanish bunch, whereas the haplotypes with the lowest mean values of these traits belong to the subspecies hypogaea, and agronomic type, Virginia bunch/Virginia Runner. The accessions ICG 14118, ICG 11249, ICG 9507, ICG 7906 and ICG 8567 showed higher contents of Fe, Zn and PC with different combinations of superior haplotypes. The superior haplotypes identified in the groundnut germplasm collection will serve as a potential resource for HBB for enhancing nutritional quality in groundnut. Supplementary Information Supplementary Information 1. Supplementary Information 2. Supplementary Information 3. Acknowledgements The authors are thankful to GeneBank, ICRISAT for their support in providing seed material and assistance in phenotyping work. Thanks also extended to Dr. Ragavendran Abbai for haplotype analysis. UNS is grateful to ICRISAT for providing the facilities to conduct the work. Author contributions MKP conceived the idea and supervised and finalized the manuscript. SP, RSB, AKP, VS, KS phenotyped the mini-core collection. ND developed the hapmap file and mined gene sequences for marker development. UNS, SSG, and VS performed the analysis. UNS wrote the manuscript. UNS, VS, and DKM contributed in improvising figures. VS, KJ, MDS, PSR, RKV, MKP contributed to reviewing and improving the manuscript. All authors have read and agreed to the published version of the manuscript. Funding This research is partially funded by Bill & Melinda Gates Foundation (BMGF) through Tropical Legumes III (TL III) project, MARS, USA and ICAR-ICRISAT collaboration project, India. Data availability The data used in this study correspond to previously reported NCBI bio-projects ID (PRJNA1002116, PRJNA490835 and PRJNA490832). Competing interests The authors declare no competing interests. Selection of handling editor and peer reviewers carried out independently by the journal and authors did not have any role in review process to avoid conflict of interest Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data used in this study correspond to previously reported NCBI bio-projects ID (PRJNA1002116, PRJNA490835 and PRJNA490832).