Nutritional and Biochemical Diversity in Beans Accessions from Three Phaseolus Species Using Multiomics Characterization

Abstract

Comprehensive profiling of Phaseolus species revealed key biochemical differences relevant to nutrition and crop improvement. Genotyping confirmed species-level separation and two gene pools in P. vulgaris, with evidence of introgression. Proteomic analysis identified over 11100 proteins, with P. vulgaris showing the highest diversity. Functional proteins, storage proteins, digestive enzymes, and protease inhibitors were linked to amino acid composition and digestibility. All 20 standard and nine essential amino acids were detected, with P. lunatus displaying a distinct essentials profile. Metabolomics identified 6717 compounds, 57% shared across species, dominated by flavonoids, polyphenols, and terpenoids. P. lunatus had 592 unique metabolites, while P. acutifolius had 98, suggesting a conserved, drought-adapted metabolome. Although the total fat was low, P. vulgaris accessions had higher omega-3 content and favorable omega-6/omega-3. Ionomic profiling revealed a variation in Ca, Fe, and Zn, with one P. vulgaris line accumulating unsafe Pb and Cd. Multiomics enables identification of nutrient-rich, climate-resilient, food-safe accessions.

1. Introduction

Beans (Fabaceae) are among the most nutritionally and economically significant crops globally, serving as vital protein sources in regions with limited meat consumption and among populations adopting plant-based diets. , Their remarkable genetic diversity reflects broad adaptability and deep cultural importance. Phaseolus vulgaris (common bean) is the most extensively studied species, with comprehensive analyses of macronutrients, micronutrients, and antinutritional factors, as well as increasing molecular characterization. , Advances in multiomics have deepened our understanding of bean genomics, − transcriptomics, − proteomics , and metabolomics, shedding light on traits such as stress tolerance, seed development, and the presence of bioactive compounds like flavonoids and saponins. ,

Despite these advances, molecular data remains limited for many landraces and wild relatives, exacerbated by scarce metabolomic databases and a predominant research focus on P. vulgaris. Importantly, nutritional quality in Phaseolus is inherently multidimensional and cannot be adequately captured by single metrics such as total protein content. Broader studies are needed to unlock the full nutritional and functional potential of underutilized species. Accordingly, this study adopts a multiomics framework to characterize Phaseolus nutritional quality as a constellation of interacting traits rather than as a linear or hierarchical species ranking.

This study, conducted as part of the Periodic Table of Food Initiative (PTFI), characterizes the nutritional and molecular diversity of 44 genebank accessions, including P. vulgaris, P. lunatus (lima bean), P. acutifolius (tepary bean), and two commercial bean cultivars. By applying a multiomics approach, we aim to bridge existing knowledge gaps and support the sustainable utilization of diverse bean species.

2. Methods

2.1. Plant Material

A total of 46 bean accessions were selected based on morphological diversity and representing maximum variability in phaseolin, lectins, and alpha-amylase inhibitors as well as a range of seed coat colors. Seeds were obtained from the Future Seeds genebank collection (44 accessions) and two P. vulgaris cultivars chosen by CIAT Bean Program located in Palmira, Colombia. Accessions were distributed among three species as follows: 35 accessions of Phaseolus vulgaris (Common bean) from the core collection, 7 accessions of Phaseolus lunatus (Lima bean), and 4 accessions of Phaseolus acutifolius (Tepary bean). Between 30 and 50 g of seeds were obtained from each accession according to the genebank availability. Bean accessions metadata was extracted from the Genesys web portal (https://www.genesys-pgr.org/) and is included in the data repository for this article.

The plant material was obtained in compliance with all relevant international and national regulations, following the International Treaty on Plant Genetic Resources for Food and Agriculture, and transferred under a Standard Material Transfer Agreement (SMTA). For the three Colombian accessions (G26506, G4691, and G7309), we subscribed Addendum No. 17 to Contract No. 212 of 2018 under the Framework Agreement for Access to Genetic Resources and Their Derived Products, granted by the Colombian Ministry of Environment and Sustainable Development (MADS).

2.2. Sample Processing

Samples were lyophilized until fully dehydrated and then frozen at −80 °C. Lyophilized samples were then cryo-homogenized using a Tissue Homogenizer (SamplePrep MiniG), and 50 – 60 mg of the fine powder obtained was aliquoted and stored again at −80 °C until further analysis (Proteomics, metabolomics, and FAMEs).

2.3. Proximate Data Acquisition and Analysis

Proximate analysis was conducted to determine the macronutrient composition of the samples. Data acquisition was outsourced to Labconquim SAS (Cali, Valle del Cauca, Colombia) and performed according to the AOAC Official Methods of Analysis (2019), determining: moisture, protein, fat, and ash content. Carbohydrates were calculated by difference. One 5 g sample was sent for analysis for each accession.

2.4. DNA Extraction, Sequencing, Genotyping, and Data Analysis

Genomic DNA was extracted from pooled young leaf tissue of 25 individuals per accession following the protocol by Correa (2024). DNA from 43 of 46 accessions was processed, excluding accession G7309 and cultivars KAT B1 and ND Palomino. Samples were genotyped with DArTseq at medium sequencing density (∼1.25 million reads/sample). DNA was digested with PstI and MseI enzymes, and libraries were sequenced on an Illumina NovaSeq 6000. SNP calling was performed de novo using DArT’s DS14 software, aligning reads with a 2–3 nucleotide mismatch threshold. SNPs were assessed for read depth, call rate, reproducibility, and minor allele frequency (MAF). A pairwise genetic distance matrix was calculated from 57,871 SNPs using identity-by-state (1-IBS) in plink v1.0 and used for Principal Coordinate Analysis (PCoA) via the pcoa function in dartR. Population structure was assessed by Ward.D2 hierarchical clustering and by estimating the optimal number of ancestral populations (K) using the snmf function from LEA v3.2.0 across 32 P. vulgaris accessions.

2.5. Proteomic Data Acquisition and Analysis

Proteome profiling was outsourced to Innomics Inc. (Sunnyvale, CA, USA), where samples were prepared and analyzed by LC-MS/MS. Samples were homogenized, lysed in 5% SDS, 9 M urea, 50 mM TEAB buffer, reduced, alkylated, and digested using the STrap MS sample prep device (Protifi) following the manufacturer’s protocol. Peptides were eluted, dried by speedvac, and further cleaned using C18 stage tips to remove contaminants. Approximately 400 ng of each sample was reconstituted in mobile phase and analyzed using a 2-h gradient nano LC-MS/MS system. Postacquisition, MS data were searched against updated Uniprot databases at the species or genus level using Sequest analysis and AUC workflow for protein profiling.

With the raw data provided, and once identified, proteins were annotated using multiple public databases, including Uniprot, EMBL, PDB, and INSDC. This step involved assigning functional categories to each protein based on its known or predicted role in biological processes, molecular functions, and cellular components using Gene Ontology (GO) terms.

An initial analysis of essential amino acids used protein sequences to compare amino acid profiles across species. For each accession, we calculated the percentage contribution of each essential amino acid from the sequence and adjusted it by chromatographic abundance. Species-specific averages were then derived to generate overall profiles.

Additionally, we analyzed proteins linked to nutritional quality, crop performance, and stress response. Nutritional proteins, such as storage and digestive enzymes, were assessed for their contribution to bean composition. Proteins tied to nitrogen fixation, growth, and photosynthesis informed crop quality, while heat shock, antioxidant, and disease resistance proteins were examined for stress adaptation. Quantitative comparisons of protein abundance across accessions provided insight into their functional roles.

2.6. Metabolomic Data Acquisition and Analysis

We followed the Untargeted Metabolomics (SB-Aq) PTFI standard protocol for metabolite extraction. To each centrifuge tube previously filled with 50–60 mg of sample, we added 600 μL of methanol at 80%, vortexed for 10 min, followed by sample centrifugation at 14000 RCF for 10 min at room temperature. The supernatant was taken and combined with 10 μL of an internal retention time standard (IRTS) consisting of 34 well-characterized nonendogenous compounds. Then, the supernatant was filtered through Captiva EMR-Lipids previously conditioned following the manufacturer’s instructions. After lipid extraction with Captiva, the samples were centrifuged at 1000 RCF for 5 min at room temperature, dried using a nitrogen evaporator, then resuspended in 80 μL of methanol at 80% and vortexed for 10 min and centrifuged for 5 more minutes at 1000 RCF and room temperature. Finally, samples were transferred to 2 mL vials with glass inserts.

All LC-MS data were acquired using a Shimadzu LC-MS-9030 system (Japan), equipped with LC-ESI-Q-TOF and operated in both negative and positive ion modes using MS1 and data-dependent acquisition (DDA) modes. Chromatography employed a Zorbax SB-Aq RRHD column (2.1 × 100 mm, 1.8 μm) and an inline 0.3 μm, 2 mm ID filter, both from Agilent Technologies.

The mobile phases included Milli-Q water with 1% formic acid (A) and 1% formic acid in acetonitrile (B). The gradient began at 98.2% A and 1.8% B, held for 1 min postinjection, shifted to 90% B by 11 min, held for 1.8 min, then returned to initial conditions over 2.1 min at a flow rate of 600 μL/min.

Acquisition parameters included a nebulizing gas flow of 2.0 L/min (N₂), interface voltages of −4 kV (negative) and 3.5 kV (positive), heater temperature of 300 °C, desolvation line at 275 °C, DDA collision energy of 18–52 V, scan range 75–1700 m/z, and 2 μL injection volume. Each sample was analyzed in triplicate. Raw MS data were converted to.mzML format using MSConvert (ProteoWizard). Then, MzMine was used to process, filter, deconvolute, and align peaks in MS1 scan mode across the samples. The resulting peak matrix was normalized by dividing the intensity of each feature by the median of the IRTS molecules identified in the sample. By including the IRTS in each sample, we calculated a retention index (RI) for each detected metabolite based on linear regression analysis of the 34 IRTS peak retention times. Afterward, MetaboAnalyst 5.0 was used to perform PCA analysis.

The methodology used to compare relative metabolite levels across samples, including the Z-transformation, abundance shifting, and ecological diversity metrics (Shannon and Simpson indices), is detailed in the Supporting Information 1, and includes the equations and steps used to normalize, transform, and interpret metabolomic profiles in a scale-free, comparable framework.

For metabolite identification, DDA data were processed using MS-DIAL V17.0 and cross-referenced against public libraries as MoNA, FiehnLib, FooDB, GNPS, and NIST. A particularly stringent similarity cutoff of 90% with open public was established for metabolite identification. A total of 582 metabolites met this criterion and were functionally annotated by querying databases such as PubChem and KEGG , as well as incorporating manually curated and generated data. This annotation provided insights into the potential metabolic and biological roles of the identified compounds in our samples.

2.7. Fatty Acids Methyl Esters (FAMEs) Data Acquisition and Analysis

We followed the Fatty acid methyl esters (FAMEs) PTFI standard protocol described as follows. FAMEs were extracted, identified, and quantified using Gas Chromatography coupled with Flame Ionization Detection (GC-FID). The analysis was conducted on a Shimadzu 2010 Plus GC system equipped with a J&W DB-23 GC column (60 m, 0.25 mm, 0.25 μm, 7 in. cage, Agilent Technologies).

Food samples (10 ± 2 mg of plant material) were dissolved in methanolic HCl, spiked with 20 μL of internal standard (TAG 17:1, 2.5 mg/mL), vortexed, and incubated at 100 °C for 1 h. After cooling, water and hexane were added, followed by vortexing and centrifugation at 3,500 rpm for 5 min. The upper hexane layer was dried under nitrogen at 40 °C. A second hexane extraction was performed, and combined extracts were reconstituted in 200 μL hexane, vortexed, and transferred to GC vials. Samples were analyzed within 8 h.

Hydrogen served as the carrier gas at 43 cm/sec under constant pressure. Injections were made in split mode at 270 °C (split ratio 50:1), and the detector was set to 280 °C. The oven program held at 130 °C for 1 min, ramped to 170 °C at 6.5 °C/min, then to 215 °C at 2.75 °C/min (held 12 min), and finally to 230 °C at 40 °C/min (held 3 min). Samples were analyzed in triplicate.

Quantification used the Supelco 37 Component FAME Mix, with area normalization. Each batch included an extraction blank and two internal reference samples. Data were processed in the software Lab Solutions V5.84.

2.8. Ionomics (ICP) Data Acquisition and Analysis

Elemental analysis was outsourced to Agrosavia (Mosquera, Cundinamarca, Colombia). One 5 g sample was sent for analysis for each accession. Sample preparation involved acid digestion according to AOAC standard methodologies. Macro-minerals: nitrogen (N), phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), and sulfur (S), and microminerals: sodium (Na), iron (Fe), copper (Cu), manganese (Mn), zinc (Zn), and boron (B), were quantified using Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES). Heavy metals, including arsenic (As), mercury (Hg), cadmium (Cd), lead (Pb), and chromium (Cr), were determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS).

2.9. Statistical Analysis of Uneven Samples

For all data sets, nonparametric statistical methods were employed to robustly compare species despite unbalanced sample sizes (35 P. vulgaris, 7 P. lunatus, and 4 P. acutifolius accessions) and deviations from normality. Overall differences among species were initially assessed using the Kruskal–Wallis test for each measured parameter, including proximate nutrient concentrations, protein and amino acid profiles, metabolomic diversity indexes, fatty acid levels (total, individual, and omega ratios), and elemental composition (macro-minerals, microminerals, and heavy metals). For parameters with statistically significant overall differences (p < 0.05), pairwise comparisons were performed using Dunn’s test with Bonferroni correction to control for multiple testing; these adjusted p-values are referenced throughout the text as p_adj.

3. Results and Discussion

3.1. Selected Bean Diversity

The final list of 46 bean accessions included accessions from 19 different countries, with Mexico and Peru contributing the largest number of accessions (8 each) (Figure ). The wide diversity among these accessions is further highlighted by the range of elevations, which varied from 50 to 2,960 m above sea level, and by the weight of 100 seeds, which ranged from 14.5 to 146.2 g (metadata available in the data repository).

1.

Distribution map of 46 bean accessions. The 46 accessions from 3 species (P. vulgaris, P. lunatus, and P. acutifolius) selected for this study belong to 19 different countries.

P. vulgaris, P. lunatus, and P. acutifolius are widely consumed Phaseolus species, each originating in different regions of Central or South America and adapted to varied environments and latitudes. P. vulgaris includes two gene pools (Mesoamerican and Andean) distinguished by seed weight and phaseolin type. Its consumption is shaped by regional and cultural preferences, with landraces spanning diverse market classes (seed color, pattern, shape, size) and uses (dry, snap, toasted). Accessions exhibit determinate bush to indeterminate climbing growth, promoting broad climatic adaptability. Key threats include Empoasca and BCMV; resistant lines in this study include G3645, G8776, G9846, and G23621. KATB1 is a yellow Andean variety popular in East Africa, while ND Palomino is a slow-darkening pinto from the U.S. preferred by growers. P. lunatus, heat-tolerant, thrives in warm climates and diverse soils, with bush and climbing habits. P. acutifolius, adapted to arid zones, is highly drought-tolerant, with a prostrate, indeterminate habit and small, neutrally colored seeds.

3.2. Proximate Analysis

Proximate analysis revealed that carbohydrates dominate the composition of all three Phaseolus species, constituting 59–77% of the total, consistent with their role as staple foods providing essential dietary energy worldwide. Protein levels follow, ranging from 15% to 29%, underscoring the importance of these beans as a valuable plant-based protein source, particularly in regions with limited access to animal protein. In contrast, fat, moisture content, and ash contribute much smaller fractions, with moisture content ranging from 2–10%, ash consistently around 3–5%, and fat comprising less than 1% of the total. These values agreed with those reported on the Consensus Document on the Biology of common bean. Overall, as illustrated in Figure , the macronutrient profiles are quite similar across species.

2.

Proximate analysis. Macronutrient composition (percentage of fresh weight) of Phaseolus vulgaris, Phaseolus lunatus, and Phaseolus acutifolius accessions.

An omnibus Kruskal–Wallis test across all species showed no statistical differences (p-value > 0.05) for carbohydrates, protein, or moisture content, while significant differences emerged for ash (p-value = 0.0025) and fat (p-value = 0.0122). Subsequent post hoc Dunn’s tests confirmed differences between P. vulgaris and P. lunatus for both ash (p_adj = 0.0016) and fat (p_adj = 0.0099). Notably, P. lunatus exhibits a greater average ash content (5.0%) compared to P. acutifolius (4.4%) and P. vulgaris (4.3%). Conversely, P. vulgaris presented an increased average fat content (0.61%) relative to P. acutifolius (0.58%) and notably more than P. lunatus (0.31%).

Two P. vulgaris accessions, G747 and G7783, demonstrated protein content of 29% and 28%, respectively, coupled with reduced carbohydrate content of 59% and 60%. These accessions, two prostrate bean varieties from Guatemala, the first is black, while the latter is yellow, represented in the outermost rings of the concentric pie chart for this species (Figure ), represent promising targets for nutritional improvement efforts.

From a broader nutritional perspective, the high average carbohydrate and protein levels (68.5% and 20.1%, respectively) underscore the potential of these Phaseolus species to meet global energy and protein demands, while their low average fat content (<1%) supports their classification as low-fat legumes. However, although macronutrient profiles are broadly similar across species and differences in ash and fat content reveal exploitable variation, protein concentration alone is an incomplete indicator of nutritional quality, as proximate analysis does not capture differences in protein composition, digestibility, or inhibitory effects that vary substantially among accessions.

3.3. Genotypic Analysis

The genotypic analysis revealed clear species-level differentiation as previously reported, , with P. acutifolius, P. lunatus, and P. vulgaris accessions separating along the first three PCoA dimensions, which together explained over 98% of the genetic variation (Figure A). Pairwise genetic distance analysis revealed stronger differentiation between species than within species. P. vulgaris exhibited the highest mean interspecific genetic distance (0.733), followed by P. lunatus (0.712) and P. acutifolius (0.654), reflecting clear genomic divergence among species. In contrast, intraspecific diversity varied considerably: P. lunatus showed the lowest within-species genetic variation (mean = 0.015), followed by P. acutifolius (mean = 0.036), while P. vulgaris displayed substantially higher intraspecific diversity (mean = 0.160), suggesting greater genetic structuring within the species. This pattern is partially explained by differences in sample size, with P. vulgaris (n = 32) represented by a larger and more diverse core collection, compared to the smaller samples of P. acutifolius (n = 4) and P. lunatus (n = 7). The limited representation of P. lunatus and P. acutifolius may underestimate their true genetic diversity, especially given that both P. vulgaris and P. lunatus have undergone independent domestication events in Mesoamerica and South America. , Within P. vulgaris, hierarchical clustering and PCoA (dimension 3) revealed two major groups corresponding to the Andean and Mesoamerican gene pools (Figure A,B). This pattern was further supported by sNMF ancestry estimates at K = 2 (Figure C), consistent with previous studies on the species’ origin and evolution. , The mean genetic distance between the two gene pools was 0.200, substantially higher than within-group diversity, which was lower in the Andean pool (mean = 0.0412) than in the Mesoamerican group (mean = 0.0923), indicating greater genetic variability among Mesoamerican accessions. Although the P. vulgaris accessions G23814D and G14813 clustered with the Andean gene pool, they showed Mesoamerican ancestry proportions of 0.42 and 0.16, respectively, suggesting admixture or introgression between the two gene pools (Figure C). Within P. vulgaris, two popping bean accessions (G11786 and G23621), traditionally consumed by the Inca civilization and cultivated in high-altitude regions (above 2500 m) across Peru and Bolivia, respectively, clustered with the Andean gene pool, as expected.

3.

Genetic diversity and population structure of bean accessions. (A) PCoA of P. vulgaris, P. lunatus, and P. acutifolius based on genetic distances. Points are colored by species in the upper plot, and by species and gene pool in the middle and lower plots. (B) Dendrogram generated by hierarchical clustering of P. vulgaris accessions, illustrating their genetic relationships. (C) Ancestry proportions of P. vulgaris accessions for K = 2, illustrating population structure and gene pool differentiation. Accessions are ordered as in the dendrogram (B) and colored by gene pool: Andean (light green) and Mesoamerican (dark green).

3.4. Proteomic Analysis

3.4.1. Essential Amino Acid Comparison

As shown in Figure A, no apparent differences were observed when comparing the average percentage of essential amino acid content of the three species. The Kruskal–Wallis test used to evaluate interspecific variation, followed by the pairwise Dunn test, revealed statistically significant differences between species in the essential amino acid composition. For instance, Lysine levels in P. lunatus diverged from those in both P. vulgaris (p_adj = 0.00028) and P. acutifolius (p_adj = 0.036). In addition, the Dunn test for phenylalanine indicated that P. lunatus differed from P. vulgaris (p_adj = 0.00690), and Histidine levels in P. lunatus were distinct from those in P. vulgaris (p_adj = 0.000071). Methionine showed a difference exclusively between P. vulgaris and P. acutifolius (p_adj = 0.00616). Overall, these results demonstrate that the essential amino acid composition, expressed as percentages, varies among species, with P. lunatus exhibiting a particularly distinct profile compared to P. vulgaris, and P. acutifolius differing from P. vulgaris in specific cases. Such variation is crucial for understanding species-specific physiology and offers valuable insights for targeted breeding programs aimed at enhancing both nutritional quality and agronomic performance.

4.

(A) Comparison of essential amino acid distribution between P. vulgaris (Common bean), P. lunatus (Lima bean), and P. acutifolius (Tepary bean). Bars represent SE. (B) Venn diagram of proteins identified by species. (C) Bar plot indicating the number of unique proteins in each accession.

3.4.2. Comparative Analysis

An average of 3786 ± 447 proteins were identified across each of the 46 bean accessions, resulting in a total of 11105 unique proteins across all samples. These proteins, except 45, were functionally annotated, providing key information including protein names, associated gene names, molecular and biological function, among others.

Of the 11105 unique proteins identified, 9142 were detected in P. vulgaris accessions, 6794 in P. lunatus, and 6543 in P. acutifolius. Among these, 2861 proteins were exclusive to P. vulgaris, 920 to P. lunatus, and 687 to P. acutifolius (Figure B). It is important to note that the number of analyzed accessions from P. vulgaris was larger, likely influencing the total protein count for this species. Additionally, the genetic diversity within P. vulgaris may account for its increased number of unique proteins, reflecting broader phenotypic variability.

Despite the wider proteomic range observed in P. vulgaris, 4737 proteins were shared across the three species, indicating conserved metabolic and physiological functions likely tied to essential processes of cell survival and adaptation. This core set highlights the fundamental biological mechanisms maintained throughout the genus, even as P. vulgaris exhibits a notably broader proteomic landscape.

At the accession level, 1099 proteins were shared across all 46 accessions, while 2979 proteins were distributed uniquely among them. Notably, certain P. acutifolius accessions (i.e., G40084 and G40022 two landraces from Mexico and the USA, respectively) contained a higher number of unique proteins than many P. vulgaris and P. lunatus accessions. This suggests that P. acutifolius may possess more distinct proteomic profiles, potentially reflecting its adaptation to specific environmental conditions or its unique evolutionary history. Among P. vulgaris accessions, the number of unique proteins ranged from 111 (G13094, a bushy yellow commercial variety from Mexico) to 36 (G12119, a climbing pink landrace from Peru), indicating substantial within-species diversity (Figure C). This variation may stem from selective breeding or adaptation to particular ecological niches, resulting in specialized proteomic characteristics. For example, G13094 was bred for an atypical bean environment on the Pacific coast of Mexico.

3.4.3. Relative Abundance of Protein

Proteins with increased abundance levels were identified in each accession. In P. acutifolius accession G40084, the two most abundant proteins were identified as ferritin (EC 1.16.3.1) and diacylglycerol kinase (EC 2.7.1.107). Ferritin is a chloroplastic enzyme involved in intracellular iron homeostasis, playing a crucial role in protecting cells from oxidative stress by storing and releasing iron in a controlled manner. This function is particularly important for plants in environments with fluctuating iron availability, suggesting that G40084 might be adapted to such conditions. In addition, pulses store more than 90% of their iron in the form of ferritin. This makes them a good source of iron for vegetarians and people who eat less meat, or for iron-deficient people. Diacylglycerol kinase, on the other hand, is involved in lipid metabolism and plays a role in signal transduction pathways related to stress responses. This enzyme plays an essential role in plant biotic and abiotic stress response. , Its high abundance may indicate a robust capacity for responding to environmental stressors, making G40084 a candidate for further studies on stress resilience in P. acutifolius.

In contrast, P. vulgaris accession G747 showed a notable abundance of phaseolin, an important seed storage protein. The relative abundance of phaseolin in G747 is notably elevated in comparison to other proteins across the accessions analyzed, rendering it an outlier in terms of protein storage capacity. Phaseolin is a key indicator of nutritional quality due to its role in providing amino acids during seed germination. Over 40 phaseolin types or variants have been identified with differences in digestibility. Beans, including P. vulgaris, are acknowledged as a valuable protein source with additional benefits related to cholesterol levels, cardiovascular disease, obesity, diabetes, and cancer prevention. Thus, the marked phaseolin abundance can position accession G747 as a nutritionally valuable accession, particularly in breeding programs aimed at improving protein content in beans. However, it is important to evaluate the content of proteins related to digestive enzymes (amylases and proteases) and protease inhibitors (e.g., arcelins, lectins, and alpha-amylase and trypsin inhibitors). Additionally, the phaseolin type for G747, “S” (Sanilac), has been reported to have lower digestibility compared to other phaseolin types.

Similarly, in the P. vulgaris accession G23807, a landrace variety from Peru, the most abundant protein identified was formate dehydrogenase (EC:1.17.1.9). This enzyme catalyzes the NAD⁽⁺⁾-dependent oxidation of formate to carbon dioxide, playing a crucial role in cellular response to abiotic stress by participating in metabolic pathways that mitigate oxidative damage. The elevated formate dehydrogenase levels in G23807 suggest enhanced mechanisms for dealing with environmental stress, such as drought or salinity, conditions frequently associated with oxidative stress in plants.

Protein abundance varied across accessions, reflecting the genetic and environmental diversity within the bean species. As shown in Supporting Information 2, several proteins exhibited notably higher abundance in specific accessions (Figure S2.1) and the top five proteins exceeding abundance levels of 1 × 10⁸ are listed in Table S2.1. Proteins associated with stress response and storage functions stood out as potential targets for improving bean resilience and nutritional value through selective breeding.

This variability suggests differences in gene expression likely shaped by environmental conditions, evolutionary history, or breeding practices. For instance, P. acutifolius accession G40084 exhibited higher ferritin abundance than G40110, a Mexican landrace from 50 masl where diacylglycerol kinase was more prevalent. These patterns may reflect adaptations to nutrient stress or enhanced defense mechanisms.

To explore conserved protein functions across the 46 accessions, a coabundance network was constructed using proteins shared by all samples (Supporting Information 3). This network revealed that proteomic covariation in Phaseolus is structured around a limited set of functionally coherent hubs, rather than reflecting isolated accession-specific effects. The most highly connected nodes map to core cellular processes (including vesicle trafficking, redox/detox metabolism, central carbon metabolism, and water transport) supporting the view that primary metabolism and homeostasis are maintained by conserved modules that should be preserved during selection. In addition, smaller connected components provide actionable biological signals: Subnetwork A concentrates enzymes linked to energy production and biosynthetic capacity, whereas Subnetwork B groups proteins associated with cellular defense and stress adaptation Collectively, these patterns support a translational, systems-level framework in which conserved metabolic/homeostasis hubs act as stability anchors, while stress- and defense-related modules provide tractable biomarker candidates to guide selection for resilience and agronomic performance. These hub-centered signatures thus offer a concrete path from descriptive proteomics to selection criteria for developing stable, resilient, and nutritionally improved Phaseolus varieties.

Integrating network and functional analyses provided a detailed view of core proteomic architecture in beans. Identified hub proteins and subnetworks offer insights into conserved biological mechanisms that can be leveraged to enhance stress tolerance, nutritional quality, and overall productivity in Phaseolus species.

3.4.4. Targeted Protein Analysis

To explore the functional roles of key proteins that contribute to important agricultural traits, we aimed to identify and analyze a subset of proteins associated with three main categories: nutritional quality, crop quality, and stress response (Figure ). These categories are important for understanding how different proteins influence plant productivity, resilience to environmental stressors, and the nutritional value of crops. For each category, a list of specific proteins known to be involved in these processes was compiled and organized in subcategories as listed in Supporting Information 4.

5.

Cumulative abundance of proteins involved in functional categories: (A) Nutritional quality; (B) Crop quality; (C) Stress response.

3.4.4.1. Nutritional Quality

The proteomic analysis identified key proteins contributing to the nutritional value of beans, including storage proteins, digestive enzymes, and protease inhibitors, which influence protein content, amino acid composition, and digestibility. P. acutifolius exhibited the lowest abundance of nutritionally relevant proteins, likely due to its adaptation to extreme environments. P. lunatus displayed intermediate levels of storage proteins and a higher proportion of protease inhibitors, potentially affecting digestibility. P. vulgaris had the most diverse and abundant nutritional proteins, particularly phaseolin, a key source of essential amino acids. Differences in protease inhibitors across species suggest variability in amino acid bioavailability. , These findings highlight the broader proteomic diversity of P. vulgaris while also illustrating that protein quality, digestibility, and inhibitor balance vary independently across species and accessions. Consequently, nutritional value cannot be inferred from storage protein abundance alone and should be evaluated in a multidimensional framework.

3.4.4.2. Crop Quality

Proteins related to water use efficiency were most abundant in P. acutifolius, supporting its drought tolerance reported since 1912. Photosynthesis-related proteins were present across species but in lower proportions. Growth and development proteins, including expansins and hormone-signaling proteins, were more abundant in P. vulgaris, influencing cell expansion and productivity. Nitrogen fixation proteins were primarily found in P. vulgaris, suggesting their role in nutrient assimilation. Notably, aquaporins, crucial for water transport, were most prominent in P. acutifolius, reinforcing its adaptation to arid conditions, while P. vulgaris showed variable abundance, indicating differences in water regulation capacity among accessions.

3.4.4.3. Stress Response

Beans produce stress-response proteins to adapt to drought, temperature fluctuations, and pathogens. P. acutifolius exhibited the highest abundance of dehydration response proteins, reflecting strong drought tolerance. P. lunatus displayed a balanced distribution of stress-related proteins, while P. vulgaris displayed variable levels of antioxidant and signaling proteins, suggesting differential stress response across accessions. Resistance to pathogens was linked to glucanases and systemic acquired resistance (SAR) proteins. P. acutifolius relied on glucanase accumulation, P. lunatus maintained a balanced immune response, and P. vulgaris showed more variability, providing potential for selecting disease-resistant genotypes. Dehydrins, crucial for drought tolerance, were most abundant in P. acutifolius, moderate in P. lunatus, and highly variable in P. vulgaris, making it a key target for breeding drought-resistant varieties.

These findings reveal distinct adaptive strategies among species: P. acutifolius shows drought tolerance, P. lunatus integrates disease resistance and drought adaptation, and P. vulgaris exhibits genetic diversity, offering opportunities for breeding climate-resilient bean varieties.

3.4.5. Breeding Targets

Although the proteomic analyses are presented at high accession-level resolution, several consistent, breeding-relevant themes emerge across Phaseolus species. First, a large set of proteins shared across all accessions defines a conserved proteomic backbone dominated by primary metabolism, protein synthesis, redox regulation, and basal stress-response functions, indicating that these core pathways are largely invariant and unlikely targets for improvement. Instead, nutritionally and agronomically relevant variation arises from differences in the relative abundance and composition of peripheral protein classes.

Second, variation in seed storage proteins and protein-quality determinants, particularly phaseolin abundance and type, digestive enzymes, and protease inhibitors, emerges as a primary axis of nutritional differentiation within P. vulgaris. Accessions combining elevated storage protein abundance with balanced inhibitor profiles represent clear targets for improving protein yield and digestibility, while others illustrate trade-offs that must be managed explicitly rather than maximized.

Third, accession-specific enrichment of stress- and resilience-associated proteins (e.g., dehydrins, aquaporins, antioxidant enzymes, ferritin, formate dehydrogenase) converges into coherent molecular strategies for drought tolerance, oxidative stress mitigation, and nutrient limitation. These profiles suggest modular stress-response signatures that could be introgressed from resilient landraces (P. acutifolius) into nutritionally favorable genetic backgrounds.

Finally, proteins involved in mineral homeostasis and nutrient handling align closely with the ionomic patterns observed across accessions, reinforcing that mineral density and retention are under partial proteomic control and can be coselected alongside ionomic targets.

Taken together, the proteomic diversity observed here resolves into three actionable breeding targets: (i) optimization of storage protein composition for nutritional quality and digestibility, (ii) deployment of stress-response protein signatures as modular traits for climate resilience, and (iii) integration of proteomic markers of mineral handling to support micronutrient enhancement while avoiding undesirable accumulation.

3.5. Metabolomic Analysis

3.5.1. Comparative Analysis

A total of 6,717 metabolites were detected across all accessions of P. acutifolius, P. lunatus, and P. vulgaris. In P. acutifolius, the number of detected metabolites ranged from 3058 (G40110) to 3931 (G40200). P. lunatus ranged from 2490 (G26506) to 4316 (G26290), and P. vulgaris showed the widest span, from 2523 (G8776) to 4339 (G15326A). This wider distribution in P. vulgaris points to a higher level of metabolomic diversity, potentially attributable to genetic variability, environmental adaptation, or unique secondary metabolite biosynthesis pathways.

Across all accessions, 3820 metabolites were common to the three species (Figure A), suggesting a conserved metabolomic core within the Phaseolus genus, likely involved in essential primary functions such as energy production, carbohydrate metabolism, lipid processing, and amino acid biosynthesis. From these shared metabolites, 82 were functionally annotated and assigned to major classes, with flavonoids (32%) and glycosides (20%) being the most prevalent. Fatty acids and purine/pyrimidine nucleosides and derivatives each comprised 11%, followed by terpenoids and alpha/proteinogenic amino acids at 8% apiece. Triterpenoid saponins, hydroxycinnamic acids and derivatives, and carboxylic acids each constituted 3%, while polyphenols and peptides made up 2%. This finding is expected as the Phaseolus genus is monophyletic.

6.

(A) Overview of detected metabolites and major functional classes of the shared annotated compounds. (B) Principal component analysis (PCA) of detected metabolites. (C) Boxplots of diversity indexes: (i) Shannon index; (ii) Simpson index.

Flavonoids and glycosides are especially abundant, highlighting their fundamental roles in plant defense, UV protection, and symbiotic interactions. Polyphenols and hydroxycinnamic acid derivatives likewise stand out for their strong antioxidant capacity and potential health benefits, including anti-inflammatory and cardioprotective effects. These metabolites play a key role in oxidative stress management and lignin biosynthesis, crucial for structural integrity and pathogen resistance. Alpha and proteinogenic amino acids further emphasize active protein metabolism and nitrogen utilization, suggesting species-specific differences in primary metabolism and signaling pathways for stress responses and plant development. Terpenoids and triterpenoid saponins also feature prominently, contributing to defense against herbivores and pathogens through volatile signaling and antimicrobial activities. The presence of such secondary metabolites, especially flavonoids, polyphenols, and terpenoids, likely confers an evolutionary advantage for stress adaptation, defense, and plant-microbe interactions.

Beyond the core metabolites shared by all three species, notable differences emerge in the pairwise overlaps. P. vulgaris and P. lunatus share 1023 metabolites, indicating a closer metabolomic relationship, whereas P. vulgaris and P. acutifolius share 474, reflecting a more limited overlap. P. lunatus and P. acutifolius share only 166 metabolites, suggesting more divergent metabolomic behaviors. Each species also harbors unique metabolites, likely tied to distinct ecological and genetic adaptations. P. lunatus has the highest number of unique metabolites (592), followed by P. vulgaris (544), and P. acutifolius (98). This relatively small set in P. acutifolius points to a streamlined metabolomic profile, presumably shaped by extreme environmental adaptation and narrower ecological niches.

3.5.2. Principal Component Analysis (PCA)

We conducted a PCA on the detected metabolites in bean accessions to visualize metabolomic variability and assess clustering patterns based on metabolomic profiles. As shown in Figure B, the first three principal components, PC1 (12.1%), PC2 (7.8%), and PC3 (7.2%), together explain 35.5% of the total variance. The analysis revealed three distinct, minimally overlapping clusters corresponding to P. vulgaris, P. lunatus, and P. acutifolius, highlighting inherent metabolomic differences. P. vulgaris displays the greatest dispersion (average distance to the centroid = 28.4884), indicating an important metabolomic heterogeneity, which is consistent with its larger number of identified metabolites (Figure A). In contrast, P. acutifolius forms a more compact cluster (average distance to the centroid = 26.9392), suggesting a homogeneous metabolomic profile, while P. acutifolius exhibits the tightest clustering (average distance to the centroid = 12.5629), reflective of a more specialized, conserved metabolism. The clear separation among species suggests that genetic factors and ecological adaptations significantly shape their metabolomic profiles. A Kruskal–Wallis test demonstrated a significant difference in Euclidean median distances to the centroid among the Phaseolus species (P. vulgaris, P. lunatus, and P. acutifolius). The Dunn’s test (Bonferroni corrected) confirmed that the highly distinct metabolomic profile of P. acutifolius is significantly different from those of P. lunatus (p = 6.02e-05) and P. vulgaris (p = 5.40e-06), while P. lunatus and P. vulgaris show comparable levels of metabolomic dispersion. To further quantify interspecies separation, we also computed the Euclidean distances between centroids in the full three-dimensional score space. The greatest distance (86.92) occurs between P. acutifolius and P. lunatus, confirming their strong metabolomic divergence. P. vulgaris lies roughly midway, 64.89 units from P. acutifolius and 68.52 units from P. lunatus, indicating that it shares metabolomic features with both species.

3.5.3. Diversity Analysis

The diversity analysis complements the PCA results by quantifying chemical variability within the Phaseolus genus. We used ecological metrics, Shannon and Simpson diversity indices, to assess metabolomic diversity among the three species. The Shannon index emphasizes richness and evenness of metabolite distribution, while the Simpson index is more sensitive to dominance by highly abundant metabolites. These indices are valuable for evaluating how species adapt to their environments and for measuring metabolomic diversity across accessions. In untargeted metabolomics, each metabolite is treated as a unique entity, with relative intensity serving as a proxy for species abundance.

Figure C­(i) shows Shannon index values, with P. vulgaris exhibiting the greatest diversity, indicating a broader and more balanced metabolite distribution. P. lunatus and P. acutifolius showed lower values, suggesting less richness and evenness. The Shannon index is more influenced by rare metabolites. Figure C­(ii) shows Simpson index values, also highest in P. vulgaris, further highlighting its metabolic diversity. This index reflects dominance of a few high-abundance metabolites, underscoring the dynamic metabolic profile of P. vulgaris.

To statistically validate these findings, Kruskal–Wallis tests were applied to both indices, followed by Dunn’s post hoc tests. For the Shannon index, the overall p-value was 0.0003, with significant differences between P. vulgaris and both P. lunatus (p_adj = 0.0018) and P. acutifolius (p_adj = 0.0375). The Simpson index yielded similar results: overall p = 0.0003, with differences between P. vulgaris and P. lunatus (p_adj = 0.0016), and P. acutifolius (p_adj = 0.0332). No significant differences were found between P. lunatus and P. acutifolius. These results confirm that P. vulgaris possesses the most diverse metabolomic profile, reflecting higher adaptability and nutritional potential.

Figure presents a correlation heatmap of metabolomic features across all 46 accessions. Each cell represents a pairwise similarity, with yellow indicating high overlap (∼3750 shared features) and dark blue/purple indicating lower similarity (∼2000). Dendrograms show clustering of accessions with similar profiles, highlighting intraspecies consistency and interspecies divergence.

7.

Correlation heatmap of metabolomic features. Heatmap showing the correlation of shared metabolites among the 46 accessions, based on hierarchical clustering. Yellow indicates a high number of shared features, while dark blue represents low similarity. Accessions are color-coded by species: Phaseolus vulgaris (green), P. lunatus (orange), and P. acutifolius (blue).

The classification shows that P. vulgaris accessions cluster closely, displaying high intraspecies similarity, evident in the yellow-toned blocks within the green-labeled group. In contrast, P. lunatus and P. acutifolius, notably G26506 (“Machete,” a Colombian landrace) and G40110, form separate clusters with lower similarity to P. vulgaris, reflecting distinct species-specific metabolomic profiles. This divergence may arise from differences in primary and secondary metabolite biosynthesis, the landrace or wild status of certain P. lunatus accessions, or environmental variability where samples were collected or grown.

The accompanying bar chart displays the number of metabolite features detected per accession, ranging from under 2500 to over 4300, indicating differences in metabolic richness or detection sensitivity. G26506, for instance, had the lowest feature count at 2490.

Altogether, the heatmap underscores clear species-level distinctions in metabolomic profiles while also illustrating notable diversity within P. vulgaris accessions, supporting its broad adaptability and biochemical complexity.

3.6. Fatty Acids Methyl Esters (FAMEs)

Comparison of total fatty acids revealed differences among species (Kruskal–Wallis p = 0.00157). Dunn’s test confirmed that P. vulgaris differs from P. lunatus (p_adj = 0.00098), with P. vulgaris presenting an average total fatty acid content of 15782 mg/kg. As illustrated in Figure A, the P. vulgaris accession G15131, a black bean landrace from Cameroon, recorded an elevated total FA content compared to the other species accessions, suggesting that genetic or environmental factors may promote lipid accumulation, which is a rarity in beans typically characterized by low fat levels. Increased lipid content can enhance bean palatability and culinary appeal, making this accession an attractive target for further study. Although this accession does not have the highest fat content according to the proximate analysis, this can be explained by the fact that the fatty acid fraction excludes unsaponifiable lipids such as sterols, tocopherols, and waxes.

8.

Fatty acids content. (A) Average total FA (saturated and unsaturated) content for each accession. (B) Individual fatty acids content by species. (C) Average omega-3, -6, and -9 FA content for each accession. (D) Average omega-6/3 FA ratio by species. Bars represent SE.

Accession G15131 also exhibits increased levels of both saturated and unsaturated fatty acids, contributing to its elevated total FA content. Unsaturated fatty acids (UFAs) are consistently more abundant than saturated fatty acids (SFAs) across all accessions (Figure A), aligning with the general lipid profile of legumes. Notably, P. vulgaris exhibited a notable unsaturated-to-saturated FA ratio (3:1), a value that differs from P. acutifolius (2:1, p_adj = 0.002114) and P. lunatus (2:1, p_adj = 0.000163). This higher ratio in P. vulgaris suggests a more favorable lipid profile that may enhance its nutritional quality and functional properties.

Comparison of individual FAs revealed distinct profiles among species. Out of 36 fatty acids detected, only 19 exhibited nonzero content (mg/kg) across all samples, which were arranged in descending order by their global average values, as shown in Figure B. In agreement with previous reports, C18:1, C18:2, C18:3, and C16:0 ranked among the top five. Linoleic acid (C18:2, omega-6) emerged as the predominant polyunsaturated fatty acid, consistent with literature highlighting linoleic acid’s dominance in legume lipid fractions. For this FA, P. vulgaris showed the lowest average compared to P. acutifolius (p_adj = 2.35 × 10^–8) and P. lunatus (p_adj = 1.20 × 10^–5). Similarly, α-linolenic acid (C18:3, omega-3) varied among species, with P. vulgaris displaying the highest levels relative to P. lunatus (p_adj = 3.16 × 10^–14) and P. acutifolius (p_adj = 2.00 × 10^–6). Additionally, long-chain fatty acids (C20:0, C22:0, C24:0), though present in lower concentrations, contribute to the structural lipid components of the seeds and underscore their role in maintaining cellular integrity. Overall, the comparison between P. vulgaris and P. acutifolius indicated differences in 12 of the 20 FA peaks, suggesting that P. vulgaris possesses a particularly distinct fatty acid composition. These findings highlight potential targets for nutritional enhancement and breeding strategies.

Omega’s FA content varies among accessions, as shown in Figure C. While omega-9 levels were comparable across species (global average 1060 mg/kg), statistically significant differences were observed for omega-3 (Kruskal–Wallis p-value = 3.94 × 10^–6) and omega-6 (Kruskal–Wallis p-value = 0.0003187). For both, only P. acutifolius and P. lunatus did not differ (p_adj = 1 for omega-3; p_adj = 0.89 for omega-6). P. vulgaris exhibited a notable average omega-3 content (7265 mg/kg), while its omega-6 content was the lowest (3762 mg/kg) relative to P. lunatus (5048 mg/kg) and P. acutifolius (5959 mg/kg). Analysis of the omega-6/3 ratio (Figure D) revealed that P. vulgaris has a markedly lower ratio (0.53) compared to P. acutifolius (2.26) and P. lunatus (2.44). This elevated omega-3 content in P. vulgaris, reflected by its lower omega-6/3 ratio, suggests a nutritionally advantageous- lipid profile, given the benefits of omega-3 fatty acids in reducing inflammation and cardiovascular risk.

3.7. Ionomics Analysis

Elemental profiling was grouped into macro-minerals, microminerals, and heavy metals. Macro-minerals, needed in high amounts, are vital for plant growth, development, and metabolism. Microminerals, though required in smaller quantities, are essential for enzymatic activity, redox balance, and physiological processes. Heavy metals were assessed for potential toxicity, as excessive accumulation can threaten plant health and food safety.

Among macrominerals, nitrogen was most abundant, followed by potassium and phosphorus (Figure A). Nitrogen and phosphorus levels did not differ significantly among species, but potassium (p = 0.0001) and calcium (p = 0.0002) showed species-level differences. Dunn’s post hoc tests confirmed these for all pairs except P. acutifolius and P. vulgaris. P. lunatus had notably high potassium (18636 ppm), while P. acutifolius showed elevated calcium (1657 ppm).

9.

Ionomics. (A) Macrominerals content in ppm: N, P, K, Ca, Mg, and S. (B) Microminerals content in ppm: Na, Fe, Cu, Mn, Zn, and B. (C) Heavy metals content in ppb: As, Hg, Cd, Pb, and Cr, excluding accession G3645 from P. vulgaris. (D) Lead content in ppb by species and accession.

Magnesium also varied (p = 0.0015), mainly between P. lunatus and P. acutifolius (p_adj = 0.0009), with P. lunatus reaching 1808 ppm. Sulfur levels differed across species (p = 0.0047); P. acutifolius had the highest concentrations (3072 ppm). Dunn’s test revealed differences among all species pairs except between P. lunatus and P. vulgaris. These variations may influence both nutritional quality and sensory attributes such as flavor.

Several accessions warrant special mention. G9846 (P. vulgaris) and G40110 (P. acutifolius) both had exceptionally high calcium content (2430 ppm), surpassing their species’ averages (1231 ppm and 1658 ppm, respectively). G25657 and G26542 (P. lunatus) exhibited the highest magnesium concentrations among all accessions (2120 ppm and 2080 ppm). For sulfur, P. acutifolius accessions G40022, G40084, and G40110 had the highest levels (2270–3280 ppm).

Among microminerals, iron had the highest overall average, followed by sodium, zinc, manganese, boron, and copper (Figure B). Only manganese showed a significant interspecies difference (p = 0.00076), driven by the contrast between P. acutifolius and P. vulgaris (p_adj = 0.0035). This general uniformity aligns with shared physiological functions across species. Nonetheless, individual accessions showed distinct profiles: G40084 (P. acutifolius) had 100 ppm sodium (vs a 40 ppm average); G8776 (P. vulgaris) had 99 ppm iron (species average: 64 – 69 ppm); and G3645 (P. vulgaris) had elevated copper (13.4 ppm; average: 5.6–7.3 ppm). G40022 (P. acutifolius) had 28 ppm manganese, higher than the typical 15–22 ppm. P. vulgaris accessions G4472 and G15131 had zinc levels of 39 ppm, exceeding species averages (24–27 ppm). These outliers represent promising candidates for breeding to address micronutrient deficiencies in diets.

Overall, heavy metal concentrations were within Codex Alimentarius safety limits, except for one outlier: G3645 (“Jamapa”, P. vulgaris) from Veracruz, Mexico, showed 3311 ppb lead and 134 ppb cadmium (Figure C), both exceeding the 100 ppb threshold. This accession represents a clear ionomic outlier and should be interpreted cautiously. The elevated Pb and Cd levels are most plausibly explained by localized soil contamination at the site of cultivation or regeneration, although genotype-dependent metal accumulation or postharvest contamination cannot be excluded based on the present data. Though other heavy metals were within limits, these levels call for follow-up studies to investigate hyperaccumulation and local soil conditions. From a breeding and dissemination perspective, accessions exhibiting unsafe metal concentrations should not be advanced for food-oriented deployment. However, such outliers remain scientifically valuable as contrast lines or negative selectors to identify genetic architectures associated with metal exclusion, supporting the development of safer cultivars.

Excluding G3645, no significant species-level differences were found for lead, chromium, or mercury (Figure D). However, cadmium (p = 0.0166) and arsenic (p = 0.0097) differed significantly. P. acutifolius had the highest average cadmium (36 ppb vs 11–12 ppb) and arsenic (6 ppb vs 1.3–1.8 ppb). Despite these differences, levels remain below thresholds for human health risk.

Mineral density likewise varied independently of species boundaries and other nutritional traits, with Fe, Zn, Ca, and Mg enrichment occurring at the accession level rather than following a consistent species hierarchy. At the same time, rare but critical heavy-metal outliers highlight the importance of integrating food safety constraints into any assessment of nutritional quality.

Altogether, the integration of ionomic, lipidomic, metabolomic, proteomic, and genotypic data sets reveals clear biochemical differentiation and complementary nutritional traits among Phaseolus species. P. vulgaris exhibited the highest overall biochemical and molecular diversity, including superior omega-3 fatty acid content, a favorable omega-6/3 ratio, and rich profiles of essential amino acids and functional proteins such as phaseolin. P. lunatus stood out for its unique metabolite composition and elevated potassium and magnesium levels, while P. acutifolius displayed distinctive drought-adaptive signatures, such as high calcium and sulfur content and abundant ferritin and aquaporin proteins, consistent with its arid-zone origins. These multilayered differences underscore species-specific nutritional and physiological strategies that can be leveraged to enhance food quality, resilience, and safety.

Taken together, this work provides one of the most comprehensive comparative multiomics data sets for Phaseolus to date and demonstrates that nutritional quality is best understood as a multidimensional trait space rather than a linear or hierarchical ranking of species. Protein composition and digestibility, lipid balance, mineral density, bioactive metabolite richness, antinutrient burden, and food safety constraints vary independently, defining complementary nutritional and adaptive profiles across species and accessions. By identifying nutrient-rich and resilient accessions while flagging potential safety risks, such as heavy-metal accumulation, these results underscore the value of multiomics approaches for precision breeding and support the integration of underutilized species, including P. lunatus and P. acutifolius, to advance sustainable, climate-resilient, and nutritionally diverse food systems.

The metadata of the bean accessions and the raw data used for all of the multiomic analysis except genomics is available in the following repository: Javeriana Dataverse 10.60790/DZWXCO. Genomics raw data is available in the following repository: 10.7910/DVN/DDX28G.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnutrsci.5c00061.

• S1: Ecological diversity metrics: Shannon and Simpson indexes; S2: Most abundant proteins; S3: Co-abundance network analysis of shared proteins; S4: Functional protein subcategories (PDF)

J.C. and G.E.V. contributed to investigation, data curation, formal analysis, and writing of the original draft. D.C.C.B. and C.R.R. contributed to data curation, formal analysis, and writing of the original draft. L.M.G. contributed to data curation and writing – review and editing. S.B., J.G.-F., J.W., G.K., J.T., S.A., and J.d.l.P. contributed to writing – review and editing. M.S., P.W., and D.D. were responsible for the selection of accessions and writing – review and editing. M.C.-Y., M.C.A., and J.A.O. conducted investigation, formal analysis, and data curation of genotypic information, and contributed to writing – review and editing. S.B. also contributed to formal analysis. J.W. contributed to investigation and methodology. G.K., J.T., S.A., and J.d.l.P. contributed to conceptualization, funding acquisition, writing – review and editing. A.J.B. led the conceptualization, funding acquisition, formal analysis, investigation, project administration, supervision, and validation, and contributed to writing – original draft and review and editing.

This work was partially funded by the Periodic Table of Food Initiative (PTFI), under contract with the American Heart Association (Contract #200341-A02). We are grateful for the support of the Rockefeller Foundation, as part of The Periodic Table of Food Initiative (PTFI), the Foundation for Food and Agriculture Research, the Seerave Foundation, the 4-fold Foundation, Atria Health Collaborative, and the Bill and Melinda Gates Foundation. The genotyping work was supported by The Crop Trust and the CGIAR Genebank Initiative. The content, findings, and conclusions presented are those of the authors and do not necessarily reflect the official views, positions, or policies of the funding institutions or the author’s affiliated institutions.

The authors declare no competing financial interest.