Introduction

Bambara groundnut (BGN) (Vigna subterranea L. Verdc) is a sustainable and affordable source of carbohydrates, protein, unsaturated fatty acids, and essential minerals (magnesium, iron, zinc, and potassium)1. Often cultivated by women, this legume crop is primarily grown to nourish their families2. Bambara groundnut seeds contain approximately 64% carbohydrates, 23.6% protein, 6.5% fat, and 5.5% fibre3. Tan, Azam-Ali1 reported BGN to be rich in Fe (4.9–40 mg 100 g− 1), Ca (95.5–99 mg 100 g− 1), K (11.44–19.35 mg 100 g− 1), and Na (2.9–12 mg 100 g− 1). Apart from their nutritional value, BGN has been found to contain polyphenolic compounds, suggesting their potential as nutraceuticals4,5.

Pregnant and lactating women, along with young children, are particularly vulnerable to micronutrient deficiencies due to high nutrient demands for reproduction and growth6. Legumes like BGN can play a pivotal role in bridging this nutritional gap, especially in low-income settings. Despite its nutritional richness, BGN remains underutilised in Africa due to undesirable cooking quality traits, notably the hard-to-cook (HTC) trait7.

Cooking quality is increasingly considered a critical parameter in legume improvement programs, influencing consumer acceptance and energy requirements during preparation8,9. Key metrics for assessing cooking quality include cooking time, cooking loss, water absorption index, swelling index, and textural properties10, yet many studies assess these traits in isolation or using limited genotypic diversity. Breeding programs in the Global South must prioritise cooking quality improvement to promote greater adoption and utilisation of climate-resilient crops like BGN.

Cooking quality of legume seeds has garnered significant attention due to its implications on nutrition and energy consumption, making it a focal point for consumers, manufacturers, and researchers11,12. The cooking process involves applying heat to food items, making grain legumes more tender and palatable enhancing their nutritional quality13,14. Most grain legumes are soaked before cooking to ensure uniformity during the expansion of the seed coat and to reduce cooking time, thereby influencing texture14,15.

However, hard-to-cook (HTC) legumes pose a significant obstacle to consumption, prolonging cooking time and requiring more energy16. Water absorption, an indirect criterion used to determine cooking quality, is linked to seed hydration capacity and the HTC phenomenon17,18,19. Additionally, morphological traits such as seed coat thickness and flatness correlate with cooking time, indicating a genetic influence on cooking quality20. Instrumental texture analysis has emerged as a valuable method for evaluating legume hardness, aiding in high-throughput phenotyping21. Cooking time, a heritable trait, varies among genotypes, underlining its importance in breeding programs to improve cooking and nutritional quality22,23.

While cooking improves nutrient availability, prolonged cooking may decrease nutrient retention, highlighting the need for optimal cooking times8. Despite the importance of cooking quality traits, they are relatively overlooked in crop improvement programs, creating a gap in the management of HTC16. Short-cooking varieties of BGN hold promise to increase consumption, reduce energy usage, and address food and nutrition insecurity, particularly in rural areas11,12. Collaborative efforts between plant breeders and nutrition experts are crucial for developing BGN varieties that are nutritionally superior and possess cooking characteristics consumers prefer11,12.

Hence, this study assessed the cooking quality and nutritional diversity among Bambara groundnut recombinant inbred lines. Integrating cooking quality traits with nutritional profiling across genetically diverse RILs presents a novel approach to inform breeding decisions. These findings will inform the selection of superior genetic stock for breeding programs targeting the improvement of HTC traits.

Materials and methods

Plant material

This study used 135 BGN F6 lines and 21 landraces from the University of Nottingham Malaysia to capture a broad genetic spectrum. The F6 lines were derived from two crosses, S19/Ankpa 4 and IITA686/LunT (Fig. 1). Variations are observed in seed coat colour, size, and lustre, ranging from dark purple and black to cream, red, and speckled brown. A millimetre scale beneath each set of seeds provides a visual reference for size comparison. The BGN seeds were manually harvested and sun-dried at Ukulinga Research Farm, Pietermaritzburg, South Africa. The seeds were stored for one month at 25 °C before conducting experiments. Figure 2 summarizes all the experimental procedures and collected data for the study.

Fig. 1
figure 1

A visual sample of Bambara groundnut seed material used in the study, including IITA686, LunT, Ankpa and S19 with multiple morphotypes.

Full size image
Fig. 2
figure 2

Flow chart illustrating experimental methodology from seed storage to nutritional analysis.

Full size image

Seed physicochemical properties

The seed size, volume and density were determined as described by24,25,26, respectively. Seed hydration and swelling properties were measured according to Yadav, Singh27. Fifty seeds per genotype and 100 ml of deionised H2O were added to the Erlenmeyer flask and left for 16 h at 25 °C. Seed weight and volume were measured before and after soaking to calculate hydration and swelling properties. Leached electrolytes and pH were determined using the method described by28. Hunter Lab colorimeter (Hunter Associates Laboratory Inc, Reston, VA) was used to measure the seed coat colour of raw BGN seeds. The seed colour was measured as L*, a*, b*, according to an international standard Commission Internationale d’Eldarage (CIE) adopted. Seeds were enclosed in a glass cup topped with a metal cup for colour measurement and shaken to minimize sampling error caused by colour variation among seeds. The process was repeated in triplicate in each sample.

Preparation of Bambara groundnut samples for cooking

Each sample of BGN seeds was measured using a 25 ml glass cup separately, and the quantity of the BGN seeds depended on the seed size. The BGN seeds were placed in a pot containing 200 ml deionised water and boiled on a Defy Thermofan Stove (Model 731 MF) at heat setting 5. Cooking time for each BGN was recorded. Cooking time was from boiling until 90% of the seeds were cooked. The visual determination of gelatinisation and finger-pressing methods were used to determine the cooking level. The cooked BGN seeds were pressed between the thumb, and the forefinger and seed were recorded as cooked when cotyledons disintegrated upon pressing29. The samples were withdrawn from the pot without interrupting boiling, and the degree of cooking was tested at five-minute intervals. This procedure was repeated until all samples of BGN were classified as cooked. Seeds were considered cooked when five consecutive tested seeds were cooked, and at this time, total cooking time was recorded.

Texture analysis of cooked seeds

Once the samples were cooked, the texture of the BGN was determined using the Texture Analyser TA-TX2 (Stable Microsystem, Surrey, England). A return-to-start method was used to measure force under compression using a Warner-Bratzler shear cell. A 2 mm probe was used to record the maximum peak force. The BGN seeds were axially compressed to 75% of their original height. The force-time curves were recorded at a speed of 2 mm/s, and the results were expressed in Newton (N).

Nutritional analysis

Ten samples with the shortest and longest cooking times were selected for nutritional analysis. The raw and cooked samples were freeze-dried and ground to a fine powder using a coffee grinder model CBM4-B5. Five grams of fine powder per sample were taken to the Analytic Services Department at Cedara College of Agriculture for nutritional analysis (moisture, ash, fat, fibre, crude protein, and total mineral content).

Elemental analysis (Ca, Cu, Fe, K, Mg, Mn, and Zn) was done using Fast Sequential Absorption Spectrometer (Varian AA280FS) interfaced with SpectrAA version 5.1 PRO software. The AAS was calibrated using an ICP multi-element standard solution IV, prepared within the optimal working range (0–100 ppm) for FAAES, following the protocol outlined by Mandizvo, Odindo30. Flame conditions were standardized by adjusting the airflow to 12 L/min and acetylene gas flow to 8 L/min. A hollow cathode lamp specific to each element was stabilized for 5 min and optimized for maximum emission intensity. Instrumental zero was established by aspirating a 5% HNO3 blank into the flame. Sensitivity was fine-tuned using working standard solutions to achieve full-scale deflection of the galvanometer. Each sample solution was aspirated in triplicate, and the mean absorbance values were recorded. Elemental concentrations were determined by interpolating absorbance values from standard calibration curves. Final concentrations in the original samples were calculated following Kurilenko and Kostyreva31, based on the mass of the sample and the volume of the digest solution, as illustrated in Eq. (1).

$$\:\text{F}\text{i}\text{n}\text{a}\text{l}\:\text{c}\text{o}\text{n}\text{c}.\:(\text{m}\text{g}/\text{l})=\text{A}\text{v}\text{e}\text{r}\text{a}\text{g}\text{e}\times\:\frac{\text{v}\text{o}\text{l}\text{u}\text{m}\text{e}\:\left(25\:\text{m}\text{l}\right)}{\text{m}\text{a}\text{s}\text{s}\:\left(0.2\:\text{g}\right)}$$
(1)

Statistical analysis

Data collected were subjected to analysis of variance (ANOVA) using GenStat statistical analysis software 23rd edition (VSN International, Hempstead, UK). Means were separated using Fisher’s protected least significant difference (LSD) when treatments showed significant differences at 5% level of probability. Correlation analysis was performed to determine associations among Bambara groundnut physicochemical properties using the corrplot package in R version 4.0 (R Core Team, 2023). Principal component analysis (PCA) was performed using the correlation matrix using XLSTAT (XLSTAT 2023.5.1.1075). A paired t-test was performed to identify statistical significance between the raw and cooked nutritional composition of BGN. Agglomerative Hierarchical Clustering (AHC) was performed using cooking time and texture according to Ward’s method using squared Euclidean distance to measure similarity using ORIGINPRO ® 2024 software.

Results

Physicochemical properties

The ANOVA revealed significant differences in physicochemical properties among the 156 Bambara groundnut accessions (P < 0.001), indicating highly significant variation (Table 1). The ANOVA test suggests substantial differences in the measured physicochemical properties across the diverse array of Bambara groundnut accessions studied, highlighting the inherent variability within this species.

Table 1 Analysis of variance showing mean squares and significant tests for seed physicochemical properties of 156 Bambara groundnut accessions.
Full size table

For seed size (SS), accessions UNISWA, S19/Ankpa4-65-56, S19/Ankpa4-136-116, IITA686/LunT-419-324 and S19/Ankpa4-199-169 had highest SS (≥ 0.750 g). Conversely, accessions such as S19/Ankpa4-106-92, ANKPA 4, IITA686/LunT-370-288 and S19/Ankpa4-112-96 had lowest SS (≤ 0.330 g). Highest seed volume (SV) values (≥ 0.630 ml) were recorded in S19/Ankpa4-136-116, UNISWA, S19/Ankpa4-199-169 and S19/Ankpa4-21-19. Lowest SV values (≤ 0.270 ml) were recorded in IITA686/LunT-387-301, IITA686/LunT-403-314, NAV 4, S19/Ankpa4-96-84 and S19/Ankpa4-112-96 (Table 2).

Seed density (ρ) was highest in S19/Ankpa4-130-112, S19/Ankpa4-18-17 and S19/Ankpa4-67-58 (≥ 1.670 g ml− 1). Lowest ρ (≤ 1.050 g ml− 1) was recorded in S19/Ankpa4-247-208, S19/Ankpa4-139-119, IITA686/LunT-365-284 and S19/Ankpa4-49-42. The hydration capacity (HC) was highest (≥ 0.370 ml g− 1) in BURKINA, S19/Ankpa4-92-79, DodR, and TIGD, and lowest (≤ 0.010 ml g− 1) in IITA686/LunT-301-240, IITA686/LunT-365-284, IITA686/LunT-420-325, S19/Ankpa4-125-107, S19/Ankpa4-188-159, S19/Ankpa4-253-213, and S19/Ankpa4-73-64 (Table 2).

Highest values (≥ 0.840) for hydration index (HI) were recorded in BURKINA, NAV 4, TIGD and ANKPA 4 while lowest HI (≤ 0.020) was recorded in IITA686/LunT-301-240, IITA686/LunT-420-325, S19/Ankpa4-125-107, S19/Ankpa4-188-159, S19/Ankpa4-253-213 and S19/Ankpa4-73-64. Swelling capacity (SC) ranged from 0.020 to 0.470; the highest SC was recorded in BURKINA (0.470), and lowest SC (≤ 0.020) was recorded in S19/Ankpa4-235-198, S19/Ankpa4-253-213, S19/Ankpa4-88-76 and UNISWA (Table 2).

Swelling index (SI) was highest in IITA686/LunT-403-314, BURKINA, TIGD and NAV 4 (≥ 1.110). Lowest SI was recorded in UNISWA (0.030). Longest cooking time (CT) (≥ 120 min) was recorded in eight accessions namely, S19/Ankpa4-106-92, SONGKHLA, S19/Ankpa4-89-76, UNISWA, S19/Ankpa4-171-143, GHC37105, DIP-C and MOQ-4. Shortest (CT) (≤ 60 min) was recorded in eleven accessions namely, IITA686/LunT-427-330, DodR, ANKPA 4, BURKINA, S19/Ankpa4-141-121, S19/Ankpa4-234-197, IITA686/LunT-403-314, TIGD, S19/Ankpa4-92-79, NAV 4 and S19/Ankpa4-100-87 (Table 2).

S19/Ankpa4-106-92, SONGKHLA, S19/Ankpa4-89-76, GHC37105 and UNISWA recorded highest electrical conductivity (EC) (≥ 9.280 kΩ cm− 1). Lowest EC (≤ 5.380 kΩ cm− 1) was recorded in IITA686/LunT-403-314, TIGD, S19/Ankpa4-92-79, NAV 4, IITA686/LunT-337-264 and S19/Ankpa4-100-87. Highest pH values (≥ 8.140) were recorded in S19/Ankpa4-100-87, NAV 4, S19/Ankpa4-92-79, TIGD, S19/Ankpa4-130-112 and BURKINA. Lowest pH values (≤ 5.580) were recorded in IITA686/LunT-334-262, MOQ-4, S19/Ankpa4-89-76, S19/Ankpa4-171-143, UNISWA, SONGKHLA and S19/Ankpa4-106-92 (Table 2).

Seed texture ranged from 5.40 to 32.70 N, with highest texture values (≥ 28 N) recorded in S19/Ankpa4-106-92, SONGKHLA, S19/Ankpa4-171-143, S19/Ankpa4-89-76, UNISWA and GHC37105. Lowest texture values (≤ 8 N) were recorded in S19/Ankpa4-141-121, IITA686/LunT-403-314, S19/Ankpa4-92-79, TIGD, NAV 4 and S19/Ankpa4-100-87.

Table 2 Means for seed properties across 156 Bambara groundnut accessions.
Full size table

Data visualisation

Figure 3 provides insights into the underlying distribution of the data observed. Figure 3a,h,l depict symmetrical violin plots representing SS, CT, and texture data. The density is balanced on both median sides, creating a mirror-like appearance. In these violin plots, the density of the data is evenly distributed around the median, leading to a balanced, mirror-like shape on both sides of the plot. This symmetry indicates that the data points are spread out equally on either side of the median value, suggesting that SS, CT, and texture distributions are not skewed in any direction. Instead, they exhibit a central tendency where the data is clustered around the median. By observing the balanced density in these violin plots, we can infer that the central measure (median) is a good representative of the dataset for SS, CT, and texture. This symmetry also suggests that any statistical analyses or modelling approaches applied to these datasets will likely yield reliable and meaningful results, given the normal-like distribution of the data.

Figure 3c,d,e,f,g depict skewed violin plots representing ρ, HC, HI, SC and SI, respectively. There is a noticeable imbalance in the density of data points on one side compared to the other, suggesting that the data distribution is skewed, with more data points concentrated on one side of the median. In contrast to the symmetrical violin plots seen in Fig. 3a,h,l, the violin plots in Fig. 3c–g display an evident imbalance in the density of data points between the two sides of the median. This disparity suggests that the data distribution is skewed, with a greater concentration of data points observed on one side of the median than the other.

Figure 3b,i,j depict bimodal violin plots representing SV, EC and pH, respectively. There are two distinct peaks or modes, indicating the presence of two separate groups or clusters within the dataset. This suggests that the data may comprise two distinct populations or categories. Identifying bimodal distributions is crucial as it implies heterogeneity within the dataset and may necessitate separate analyses or modelling approaches for each subgroup. Texture data (Fig. 3k) is depicted using a constricted violin plot, indicating narrow data distribution and low variability.

Fig. 3
figure 3

Data visualisation with violin plots combining box plots and kernel density plots to represent the distribution and shape of recorded data among 156 Bambara groundnut accessions.

Full size image

Principal component analysis (PCA)

Table 3 shows the PCA with factor loadings, eigenvalues, and percent variance for the evaluated physicochemical traits. PC1 accounted for 44.187% of the total variation and was positively correlated with HC, SC, SI and pH. PC2 accounted for 18.437% of the total variation and was positively correlated with CT, EC, L*, a*, b* and texture. Seed size and SV were positively correlated with PC3, accounting for 12.158% of the total variation.

Table 3 Summary of factor loadings, eigenvalues, and percent variation for cooking quality parameters assessed for 156 Bambara groundnut genotypes.
Full size table

Based on the evaluated physicochemical properties, the PC biplots based on PCA analysis were used to picture the relationship among Bambara groundnut genotypes (Fig. 4). Traits depicted by parallel vectors or those positioned closely to each other indicate a strong positive association, whereas traits situated nearly opposite each other (at 180 °) demonstrate a highly negative association. Vectors directed towards the sides indicate a weaker relationship. BURKINA, IITA686/LunT-403-314, NAV4, S19/Ankpa4-92-79, TIGD, ANKPA 4, IITA686/LunT-261-220, IITA686/LunT-427-330 and KENYA CAPSTONE are grouped based on high HI, HC, SI and SC.

The genotypes IITA686/LunT-387-301, S19/Ankpa4-254-215, IITA686/LunT-305-243, S19/Ankpa4-5-4, S19/Ankpa4-339-266, S19/Ankpa4-49-42, and IITA686/LunT-360-280 are clustered together due to their high values of a*, b*, and L* parameters. This grouping suggests a similarity in lightness attributes among these genotypes, indicating potential common traits or characteristics in their colour profiles (Fig. 4).

The genotypes UNISWA, GHC37105, KANO2, S19/Ankpa4-89-76, S19/Ankpa4-171-143, S19/Ankpa4-161-136, S19/Ankpa4-179-150, and IITA686/LunT-334-262 have been clustered together due to their shared characteristics of high cooking time, electrical conductivity, and texture (Fig. 4).

S19/Ankpa4-17-16, IITA686/LunT-284-230, IITA686/LunT-366-285, IITA686/LunT-295-235, IITA686/LunT-415-321, IITA686/LunT-390-304, S19/Ankpa4-224-189, IITA686/LunT-419-324, S19/Ankpa4-232-195, IITA686/LunT-398-310, S19/Ankpa4-65-56, S19/Ankpa4-21-19, S19/Ankpa4-86-74, S19/Ankpa4-47-40 and S19/Ankpa4-247-208 are grouped based on high SS and SV (Fig. 4).

IITA686/LunT-337-264, IITA686/LunT-362-281, S19/Ankpa4-100-87, S19/Ankpa4-141-121, S19/Ankpa4-50-43, IITA686/LunT-420-325, IITA686/LunT-332-260, S19/Ankpa4-13-12, S19/Ankpa4-238-201, S19/Ankpa4-133-114, DodR, S19/Ankpa4-12-11 and IITA686/LunT-426-329 are grouped on high ρ and pH.

Fig. 4
figure 4

Principal component (PC) biplot of PC 1 vs. PC 2 demonstrating the relationships among physicochemical properties among 156 Bambara groundnut genotypes.

Full size image

Pearson correlation analysis

Pearson correlation coefficients showing relationships among physicochemical traits evaluated among Bambara groundnut genotypes are presented in (Fig. 5a). Significant and positive correlations were observed between SS and SV (r = 0.88; p = 0.028). Seed hydration capacity was positively correlated with HI (r = 0.94; p = 0.028), SC (r = 0.83; p = 0.035) and SI (r = 0.76; p = 0.037). Cooking time was positively correlated with texture (r = 0.96; p = 0.040) and EC (r = 0.95; p = < 0.05).

Significant and negative correlations were observed between SC and CT (r = − 0.55; p = 0.046), EC (r = − 0.56; p = 0.041) and texture (r = − 0.56; p = 0.039). Significant inverse associations were also noted between pH and CT (r = − 0.96; p < 0.05), EC (r = − 0.90; p = 0.010) and texture (r = − 0.92; p = 0.037) (Fig. 5a).

The circular chord diagram (Fig. 5b) illustrates the interrelationships among Bambara groundnut’s various physicochemical, hydration, and cooking quality traits, with each segment around the circle representing a specific trait. The coloured ribbons, or chords, connecting these segments denote the degree of association between the traits—where thicker chords suggest stronger correlations and thinner ones indicate weaker associations.

Figure 5b reflects BGN’s complex phenotypic integration among hydration, physical, and cooking-related traits. These insights are valuable for breeding programs, highlighting key trait combinations that can be targeted to develop fast-cooking, nutritionally rich, and consumer-acceptable BGN varieties. By visualizing these relationships, the chord diagram facilitates a holistic understanding of how different seed characteristics influence cooking quality, enabling more informed and efficient selection in breeding pipelines.

Fig. 5
figure 5

Pairwise correlation matrix and chord diagram illustrating the relationships among evaluated BGN seed physicochemical properties: (a) heatmap of Pearson correlation coefficients between size (SS), seed volume (SV), density (ρ), hydration capacity (HC), swelling capacity (SC), swelling index (SI), cooking time (CT), electrical conductivity (EC), potential of hydrogen (pH), lightness from black to white on a scale of zero to 100 (L*), chromaticity with no specific numeric limits (a* and b*) and (b) circular chord diagram visualizing the strength and direction of relationships among traits. Thicker ribbons indicate stronger correlations, with the width and colour of the connections highlighting the degree of association.

Full size image

Agglomerative hierarchical clustering (AHC)

Based on cooking time and texture, BGN genotypes were clustered into five groups (Fig. 6). Group A (soft-cooking genotypes) comprised of 14 genotypes including DodR, BURKINA, ANKPA 4, TIGD, NAV 4, S19/Ankpa4-100-87 and IITA686/LunT-292-233. Group B (slightly soft-cooking genotypes) consisted of 27 genotypes, among which were PONG-CR, IITA686/LunT-420-325, S19/Ankpa4-89-102, KANO2, S19/Ankpa4-238-201, IITA686/LunT-314-249, IITA686/LunT-261-220, and S19/Ankpa4-5-4.

Group C (slightly hard-cooking genotypes) comprised of 37 genotypes including IITA686, S19/Ankpa4-122-104, S19/Ankpa4-73-64, S19/Ankpa4-241-203, IITA686/LunT-415-321, IITA686/LunT-419-324 and IITA686/LunT-262-221. Group D (moderately hard-cooking genotypes) comprised of 61 genotypes including S19/Ankpa4-61-52, S19/Ankpa4-190-160, S19/Ankpa4-108-93, UKZN-1, LUNT, S19, IITA686/LunT-444-346, IITA686/LunT-365-284 and IITA686/LunT-284-230. Group E (hard-cooking genotypes) comprised of 17 genotypes including DIP-C, MOQ-4, GHC37105, PONG-BR, SONGKHLA, UNISWA, S19/Ankpa4-206-175 and IITA686/LunT-334-262 (Fig. 6).

Fig. 6
figure 6

Dendrogram showing agglomerative hierarchical clustering of 156 Bambara groundnut genotypes cooking quality.

Full size image

Proximate and mineral nutrient compositions

The proximate composition of raw and cooked BGN is shown in Tables 4 and 5. The results show that the concentrations of all proximate nutrients analysed were significantly different using a paired t-test. The protein content of raw BGN seeds ranged from 15.16 to 29.32 g/100 g− 1. The mean protein content of cooked BGN samples was significantly higher (p = 0.035) than the raw samples. The genotype DodR showed the highest protein recovery after cooking (19.09–25.39 g/100 g− 1). The fat content of raw BGN samples (3.16–5.55 g/100 g− 1) was lower than that of cooked BGN samples (5.95–8.16 g/100 g− 1) and was significantly different (p < 0.001). After cooking, the genotype Nav 4 showed the highest fat content recovery (3.16–6.77 g/100 g− 1). The result indicates that BGN mainly comprises Neutral Detergent Fibre (NDF) with a maximum of 46.8 and 64.09 g/100 g− 1 for raw and cooked samples, respectively. The total mineral content (ash) of raw BGN was higher (4.18–5.23 g/100 g− 1) and significantly different (p < 0.001) from that of cooked BGN samples (2.5–4.85 g/100 g− 1).

The mineral concentration of raw and cooked BGN samples is shown in Tables 4 and 5. All the selected nutrients analysed exhibited significant differences except for P (p = 0.309) and Mg (p = 0.22). Iron was not detected in some cooked samples. An increase in the number of nutrients after cooking was observed in zinc (2.4–3.2 mg 100 g− 1) to (2.5–3.7 mg 100 g− 1), Na (0.00–0.0002 mg 100 g− 1) to (0.002–0.007 mg 100 g− 1), Cu (0.6–2.1 mg 100 g− 1) to (0.9–6.3 mg 100 g− 1) and Ca (0.0004–0.001 mg 100 g− 1) to (0.008–0.018 mg 100 g− 1) after cooking. Cooking also caused a reduction in the number of nutrients as observed in Fe from (1.7–9.6 mg 100 g− 1) to (0.00–2.9 mg 100 g− 1) and K (0.117–0.176 mg 100 g− 1) to (0.046–0.149 mg 100 g− 1). Iron and Zn were predominant minerals in raw BGN genotypes ranging from 1.7 to 9.6 mg 100 g− 1 and 2.4–3.2 mg 100 g− 1, respectively.

Table 4 Proximate, macro and micro-nutrient composition of 10 Bambara groundnut genotypes analysed in their Raw state, categorised as fastest and slowest cooking.
Full size table
Table 5 Proximate, macro and micro-nutrient composition of 10 Bambara groundnut genotypes analysed in their cooked state, categorised as fastest and slowest cooking.
Full size table

Discussion

Seed size showed a significant and positive correlation with seed volume. Geethanjali, Rani32 reported comparable findings for Cicer arietinum L, and Singh, Kaur33 observed similar results for Phaseolus mungo L. Swelling capacity and swelling index were significantly positively corrected to the degree of lightness (L) r = 0.367 and r = 0.389 respectively. These results showed that light-coloured genotypes swelled more than dark-coloured genotypes. Light-coloured genotypes that swell more could cook faster because they imbibe water more efficiently, leading to quicker softening of tissues28.

Hydration capacity was positively correlated to the hydration index, swelling capacity and swelling index (Fig. 4a). The findings of this study align with those of28, indicating that seed coat colour impacts the structural and imbibitional characteristics of BGN seeds. Both hydration and swelling properties were significantly correlated to cooking time (Fig. 5a). Seeds with greater hydration and swelling properties cook fast due to (i) enhanced water absorption, (ii) improved heat transfer, (iii) reduced structural integrity and (iv) faster gelatinisation.

Seed size was also significantly correlated with cooking time; similar findings of a positive correlation between cooking time and seed size were reported. Smaller seeds have a less dense internal structure compared to larger seeds. A less dense structure allows water and heat to penetrate more efficiently throughout the seed, promoting faster cooking. Cooking time was significantly positively correlated with texture (r = 0.96). These results indicated that BGN genotypes that took longer to cook had a firmer texture. This implies that fast-cooking genotypes also give a better texture.

Seed size showed a significant positive correlation with cooking time, corroborating earlier findings by Santos, Carbas34, who observed that larger seeds generally required longer cooking durations. This relationship may be attributed to size and internal structural density, as smaller seeds tend to possess less compact microstructure, facilitating more efficient water and heat penetration during thermal processing30,35. However, while these studies support the trend observed in our results, contrasting findings have been reported in other legumes, where cooking time was more strongly influenced by seed coat thickness and composition than size alone36. The relative influence of seed size may vary depending on species-specific characteristics or genetic variation within the crop.

Cooking time was also positively correlated with post-cooking texture (r = 0.96), indicating that genotypes requiring longer cooking times also exhibited firmer textures. This strong association may reflect slower starch gelatinization and delayed solubilization of cell wall components in these genotypes, which could be linked to genetic factors and differential seed maturity at harvest36. Furthermore, fast-cooking genotypes yielded softer textures, which are generally preferred in consumer acceptability studies37. Mubaiwa, Fogliano22 highlighted that differences in texture and cooking performance among genotypes may also be influenced by agroecological conditions (temperature, soil nutrient availability) during seed development, which have been shown to affect hydration dynamics and seed hardness.

In our study, uniform cooking conditions were applied to all genotypes, minimizing environmental and procedural variability; hence, observed differences can be primarily attributed to inherent physicochemical properties and genetic background. This strengthens the reliability of the correlations and underscores the need to integrate seed structural traits into breeding programs that target reduced cooking time and improved sensory quality in BGN.

Variations in the nutritional composition of BGN genotypes amount from (i) environmental factors affecting plant growth38, (ii) genetic diversity39 and (iii) techniques used for nutrition analysis40,41. The protein concentration of all BGN genotypes increased after soaking and cooking, and these findings were in line with the report for common beans and chickpeas by42. The increase in protein content after cooking is attributed to the denaturing of some anti-nutritional factors, resulting in the release of nutrients9,43.

Protein-energy malnutrition (PEM) is a major problem in developing countries44,45,46. This condition arises from a severe deficiency in protein and caloric intake, leading to health problems (stunted growth, weakened immunity, and increased susceptibility to infections). According to the World Health Organization (WHO), PEM affects millions of children and adults, contributing significantly to morbidity and mortality rates in low-income regions47. In this context, the study findings highlight the potential of BGN as a valuable nutritional resource. Bambara groundnut seeds are rich in protein, containing approximately 18–25% protein by dry weight, comparable to other legumes such as chickpeas and lentils48. This high protein content makes BGN an excellent dietary option to combat protein deficiencies, particularly in regions where protein-rich foods (meat and dairy) are economically inaccessible.

Fats play a crucial role in maintaining good health by providing a dense energy source and facilitating the absorption of fat-soluble vitamins and the synthesis of hormones49. In evaluating the cooking quality of BGN, the study observed a significant increase in fat content following soaking and cooking (p < 0.001). This enhancement in fat content post-cooking is consistent with the findings of Mazahib, Nuha50, who reported similar outcomes for other legumes. Notably, despite the significant rise in fat levels, the overall fat content of cooked BGN remains within the recommended dietary allowance (RDA) range of 3–17 g 100 g− 1, ensuring its nutritional adequacy for consumption. These results are pivotal for breeding programs that optimise BGN for nutritional quality. The ability to increase fat content through cooking while maintaining it within safe and beneficial dietary limits highlights the potential of BGN as a functional food. This information is valuable for developing BGN varieties with enhanced cooking quality and nutritional profiles, aligning with improving food security and nutritional health in various populations.

The moisture content of dried legumes is crucial for determining their shelf life, product quality, and processing techniques51. The moisture content of the BGN genotypes decreased after cooking. These findings contradict Shanono and Muhammad52, who observed a significant increase in moisture content after cooking. Although the moisture content of BGN samples increased after cooking, it was in the same range as recommended values (3–8%) for the extended storage of legumes. Significant disparities in the availability of selected mineral elements were observed between soaking and cooking, except for Mg and P. Specifically, there was a notable increase in Ca, Cu, Zn, and Na levels. In contrast, Fe levels experienced a significant decrease.

The disparities in mineral element availability between soaking and cooking can be attributed to distinct physicochemical changes during these processes53. Soaking leads to the leaching of soluble minerals, while cooking enhances the availability of some minerals by breaking down complexes and reducing anti-nutritional factors54. However, minerals like Fe may decrease due to leaching or loss in the cooking water. Magnesium and P levels remain relatively stable due to their chemical properties and interactions within the food matrix55.

Conclusion

The Bambara groundnut genotypes evaluated in this study exhibited significant variation in physicochemical properties, enabling classification into five distinct cooking categories: A (soft-cooking), B (slightly soft-cooking), C (slightly hard-cooking), D (moderately hard-cooking), and E (hard-cooking). Genotypes such as DodR, BURKINA, ANKPA 4, TIGD, NAV 4, and several S19/Ankpa4 and IITA686/LunT lines were consistently classified as soft-cooking. In contrast, genotypes including UNISWA, SONGKHLA, DIP-C, EXSOCOTO, and others exhibited hard-cooking characteristics. These findings have clear implications for breeding programs. The soft-cooking genotypes represent promising parental lines for developing improved Bambara groundnut varieties with enhanced cooking quality—traits desirable for household consumption and agro-processing. Conversely, hard-cooking genotypes may offer utility in breeding for industrial applications requiring structural resilience, such as flour production or extrusion-based processing. Future research should focus on identifying and mapping key genomic regions or candidate genes controlling cooking time and texture through molecular and genomic approaches. Additionally, exploring gene–trait associations and the underlying biochemical mechanisms will be critical for marker-assisted selection. Optimizing cooking protocols for different genotype categories may also support product development and value chain expansion.