Total nitrogen content was estimated using the Kjeldahl method (AOAC 984.13), with some modifications in the digestion method. The organic compounds present in the sample were digested overnight at room temperature by adding the digestion mixture (mixture of sulfuric acid, selenium catalyst, lithium sulfate, and hydrogen peroxide) (Michałowski et al., 2013). Partially digested samples were heated at 400°C for 20 min to complete the digestion and get a clear and transparent solution. A Foss Kjeltec Nitrogen Autoanalyzer (Foss Tecator 2300 Kjeltec Analyzer Distiller Unit) estimated the nitrogen content in the samples and converted it to protein percentage using Jone's factor 6.25.

About 100 mg of homogenized sample was extracted twice using 80% ethanol at 80°C for 30 min. After each extraction cycle, the samples were centrifuged at 10,000 rpm for 10 min, supernatants were pooled, and the residue thus obtained was used to estimate the total starch content. Total starch content was estimated by Megazyme total starch assay kit as per AOAC 996.11. It initially uses α-amylase to hydrolyze starch into maltodextrin, which is followed by the action of amyloglucosidase to hydrolyze maltodextrin into D-glucose. Glucose was oxidized causing the liberation of hydrogen peroxide, which was measured using a GOPOD reagent that gave pink color. The absorbance was recorded at 510 nm using a UV–Vis spectrophotometer. The results were expressed as g100 g⁻¹.

Amylose was determined by the modified KI–I² iodometric assay (Perez and Julian, 1978) based on the binding capacity of the amylose–iodine complex. The absorbance was recorded at 620 nm using a UV–Vis spectrophotometer, and the results were expressed as g100 g⁻¹. A standard calibration curve was also developed using potato amylose obtained from Sigma Aldrich.

The TDF content was estimated in duplicates using a commercial assay kit purchased from Megazyme International, Wicklow, Ireland (AOAC 985.29). The method included the use of α-amylase, protease, and amyloglucosidase, followed by precipitation with 95% ethanol. After vacuum filtration, one of the duplicates was estimated for total ash content using a muffle furnace. In contrast, the other one was used for the estimation of total protein content using the Kjeldahl method, and the result was expressed as g100 g⁻¹.

The extract obtained by the procedure mentioned in section starch was used to estimate the TSS content using the anthrone reagent method (Dubois et al., 1956), which involves the dehydration of sugars by sulfuric acid to form furfural. The furfural–anthrone complex (blue-green) was read for absorbance at 630 nm using a UV–Vis spectrophotometer, while D-glucose was used as a standard to develop a calibration curve. This method measures the content of total soluble sugars, including monosaccharides, disaccharides, and oligosaccharides. The results were expressed as glucose equivalent in g 100 g⁻¹.

The extract obtained by the procedure mentioned in section starch was used to estimate the total phenolic content using an FCR assay (Singleton et al., 1999), which includes oxidation/reduction reaction. FCR is a mixture of molybdates and tungstates that get reduced by phenolic compounds, which is marked by the development of a blue complex. The absorbance was read at 650 nm using a UV–Vis spectrophotometer, and the results were expressed as GAE in g 100 g⁻¹.

Phytic acid content was estimated by enzymatic method using a Phytate assay kit (K-PHYT) from Megazyme International, Wicklow, Ireland (AOAC 986.11). Total and free phosphorous content was estimated separately using an ascorbic acid-molybdate color reagent. The absorbance was recorded after incubation at 655 nm using a UV–Vis spectrophotometer, and the result was expressed as g 100 g⁻¹.

There was a significant difference between the accessions where the protein content ranged from 19.3% (EC724681) to 27.9% (EC390248) with a mean value of 24% (Table 1), which is considerably higher than that found in white and brown cowpea seeds (20.4 and 21.3%, respectively; Longvah et al., 2017). Our values agree with the previously reported values in cowpea which ranged from 17.4 to 31.7% (Antova et al., 2014). Cowpea is known for its high protein content in seeds, but it can vary depending on the cultivar or edaphic factors. The protein content values higher than those obtained in our study (26.5–29%) have been reported in three cowpea varieties from Ghana (Appiah et al., 2011). Owing to the high protein present in cowpea, it will fulfill SDG 2, i.e., achieving zero hunger by tackling malnutrition diseases like kwashiorkor in poor children. It can be used to make protein-rich food and food additives that could be beneficial for infants in particular and as high protein feed for animals. Moreover, a good crop improvement program could be conducted by selecting promising accessions, including EC390248 (27.9%), EC169879 (27.7%), and EC240667 (26.6%), from our study.

Significant variation was observed in the starch content between accessions and ranged from 27.5 g 100 g⁻¹ (IC214752) to 42.7 g 100 g⁻¹ (EC724681) with a mean value of 32.6 g 100 g⁻¹ (Table 1), which is considerably lower than the starch values reported for brown (47.1 g100 g⁻¹) and white (47.5 g 100 g⁻¹) cowpea grains (Longvah et al., 2017). The difference could be due to the presence of highly hydrated fine fiber fraction derived from the cell wall enclosing starch granules or high content of insoluble and closely associated starch granules (Hall et al., 2017). The values obtained in our study were in the range of 28.3–36.2 g 100 g⁻¹ (Antova et al., 2014), which are lesser than the values reported by the studies done on 15 Nigerian cowpea cultivars (60 g 100 g⁻¹; Gonçalves et al., 2016). Moreover, compared to other legumes, cowpea has a higher starch content and in vitro starch hydrolysis rate (Nkhata et al., 2018). The starch yield of isolated cowpeas was low when compared to the yield of other legumes like black gram (45 g 100 g⁻¹) and red bean (46 g100 g⁻¹; Hoover et al., 2010); however, the starch yield was found to be high when compared to beach pea (12.3 g100 g⁻¹), grass pea (26 g100 g⁻¹), green pea (30 g100 g⁻¹), and adzuki bean (21.5 g100 g⁻¹; Ren et al., 2021). These results confirmed that cowpea starch could be a valuable source of legume starch. It is considered to be beneficial for human health, as the legume starch is digested more slowly than the starch obtained from cereals. High starch-containing accessions can also be used to make processed products like moin moin and akara (Phillips et al., 2022) because of their superior processing properties, which will significantly contribute to SDG 2 and 3.

Amylose content of cowpea differed significantly between accessions and ranged from 9.7 g 100 g⁻¹ (IC249141) to 21.7 g100 g⁻¹ (EC472284) with a mean value of 15.8 g 100 g⁻¹ (Table 1). Higher amylose content has been reported in the starch of cowpea and mungbean in the range of 36–39 g100 g⁻¹ (Kim et al., 2018). The amylose content affects the physicochemical properties of starch, such as viscosity, swelling power, gelatinization capacity, retrogradation, and starch crystallinity (Petroni et al., 2017). High-amylose starch is prone to retrogradation by nature and is resistant to digestion. Although legume starches have high amylose content (Zia-ud-Din et al., 2017), the starch and protein contents are inversely related, and the high protein content in our samples explains the low amylose content. The reported range of amylose in cowpea cultivars is 13.9–15.5 g 100 g⁻¹ and 17.0–18.7 g 100 g⁻¹ for whole-grain and decorticated flour respectively (Adebooye and Singh, 2008) and 16.7–19.75 g 100 g⁻¹ (Naiker et al., 2019) are similar to our values. Accessions having high amylose content, i.e., EC472284 (21.7 g 100 g⁻¹), IC128727 (21.4 g 100 g⁻¹), and EC244077 (20.5 g100 g⁻¹), would have a low GI, which slows down the process of digestion, thus preventing the increase of blood sugar levels right after a meal and thereby helping diabetic patients. Low-amylose accessions, such as IC249141 (9.7 g 100 g⁻¹), EC240900A (10 g100 g⁻¹), and IC214752 (10.8 g 100 g⁻¹) have high swelling power, which gradually softens the grains and facilitates easy cooking of the grains (Wani et al., 2012). Hence, accessions containing low and high amylose content have their significance depending on the user's interests and requirements.

The content of total dietary fiber ranged from 13.7 g 100 g⁻¹ (EC724690) to 21.1 g 100 g⁻¹ (EC244138) with an average value of 17.3 g 100 g⁻¹ (Table 1), which is greater than the values of brown and white cowpea grains (11.5 g 100 g⁻¹ and 11.7 g 100 g⁻¹, respectively; Longvah et al., 2017), but close to the previously reported value of 15 g 100 g⁻¹ (Akissoé et al., 2021) and less than the reported values of 27.4 g 100 g⁻¹ (Ghavidel and Prakash, 2007) and 27 g 100 g⁻¹ (Kalogeropoulos et al., 2010) for raw cowpea. Genetic and environmental factors might attribute to the variation of TDF content. Accessions with high TDF (EC724033 and EC244138) could be a preferred dietary fiber source for people suffering from constipation, cardiovascular diseases, and obesity, which will facilitate achieving SDG 3, i.e., good health and wellbeing of individuals. WHO recommends an intake of >25 g of fiber, which could be contributed by cowpea along with cereals, fruits, and vegetables (Joint and Consultation, 2018). Foods derived from cowpea accessions having high dietary fiber could prove to be beneficial for maintaining a regular bowel function, assisting in weight loss, and lowering the risk of heart diseases and type-II diabetes.

The TSS content varied from 1.30 g 100 g⁻¹ (EC724385) to 8.73 g 100 g⁻¹ (EC149303) with a mean value of 5.58 g 100 g⁻¹ (Table 1), which is comparatively higher than the reported range of 1.11–4.07 g 100 g⁻¹ with a mean value of 2.23 g 100 g⁻¹ (Nassourou et al., 2017). The experimental values are similar to the reported range of 3.26–8.60 g 100 g⁻¹ with a mean value of 5.45 g 100 g⁻¹ (Weng et al., 2018). High TSS content imparts tolerance to abiotic stress factors and improves the storability of seeds by increasing desiccation tolerance. TSS contributes to the desirable taste and availability of fermentable sugars. Fermented pulse-based food products in different forms (masyaura, papad, sepubari, daal vada, kanji vada, idly, dosa, and uttapam) are commonly consumed in India and other developing countries (Tamang et al., 2016). Pulses are a rich source of vitamins and minerals compared to cereals, but they also contain a high level of phytates. The fermentation process increases the content and bioavailability of thiamine, niacin, phosphorus, energy, protein, and minerals (Kapravelou et al., 2020). Thus, accessions with high fermentability are desired not only for making local dishes but also to meet the nutritional needs in a better manner. In our study, we have identified high TSS content with a possibility of finding oligosaccharide-rich accessions. Oligosaccharides have a probiotic activity, help regulate gut microflora, and aid in synthesizing short-chain fatty acids and vitamins of the B group. Not much work has been done so far to study the TSS of cowpea, and this aspect needs to be explored in breeding programs and genetic improvement of the cowpea accessions, which will generally meet the needs of the market and consumer.

The total phenolic content ranged from 0.026 g100 g⁻¹ (EC724674) to 0.883 g 100 g⁻¹ (EC109112) with a mean value of 0.272 g 100 g⁻¹ (Table 1). The amount of polyphenol tends to be higher in brown and cream-colored (0.192–0.196 g 100 g⁻¹) rather than white-colored seeds (0.099 g 100 g⁻¹; Gonçalves et al., 2016) and is present mostly in seed coats. Polyphenolics determine the antioxidant capacity of legumes (Zhang et al., 2015) and differ across the cultivars in legumes, which might be due to genotypic variations and geographical conditions. Higher values observed in our samples are similar to the reported values of 0.544–0.844 g 100 g⁻¹, while the lower values agree with 0.035–0.377 g 100 g⁻¹ (Cai et al., 2003). Polyphenols cause a reduction in protein and starch digestibility and also act as a metal chelator (Hall et al., 2017). However, cowpea polyphenols have the most active DPPH radical scavenging activity, thus acting as hydrogen peroxide scavengers and strong antioxidants. Therefore, low and high phenol content has unique relevance from a nutritional perspective.

The content of phytic acid varied significantly across the different accessions and ranged from 0.690 g 100 g⁻¹ (EC723824) to 1.88 g 100 g⁻¹ (EC101970) with a mean value of 1.10 g 100 g⁻¹ (Table 1), which is considerably higher than the phytic acid value of brown and white cowpea grains (0.550 g 100 g⁻¹ and 0.573 g 100 g⁻¹, respectively; Longvah et al., 2017). The values are quite similar to the results reported for Brazilian cowpea (0.870–1.26 g 100 g⁻¹; Almeida et al., 2008) but considerably lower than the values reported in soy (3.17–3.90 g 100 g⁻¹; Kumar et al., 2006). Phosphorus is mainly stored as phytic acid in the seeds, which is widely distributed in the grains of cereals and legumes. Humans lack digestive enzymes to hydrolyze phytate, as phytate binds to the minerals, thus limiting their bioavailability. Phytate is also known to inhibit the activities of food digestive enzymes, such as proteases and amylases (Pramitha et al., 2021); hence, low phytate-containing germplasm is much in demand among the crop breeders. The results show that cowpea has high phytate content, which can be reduced by cooking and long germination time (Jayathilake et al., 2018). In contrast, low phytate-containing accessions, such as EC723824 (0.690 g 100 g⁻¹), EC240697 (0.797 g 100 g⁻¹), and EC004218 (0.810 g 100 g⁻¹), could be consumed regularly along with the staple food.

Correlation studies are extremely important for finding positive or negative values which indicate significant or non-significant relationships between the parameters. The substantial relationships between the nutritional indicators were determined using Pearson's correlation test. Pearson's r > 0 indicates a positive correlation, whereas r < 0 indicates a negative correlation with p < 0.05. Boost in a particular parameter leads to changes in the characteristics of other related parameters. The content of phytate, protein, TDF, phenols, TSS, starch, and amylose was evaluated to determine their correlation coefficients.

Table 2 shows the correspondence between seven biochemical parameters of 120 cowpea accessions, along with the exact values. It can be inferred that a highly significant and negative correlation was found between starch and protein content (r = −0.63, p < 0.001), where protein increases due to the inhibition of starch synthesis (Yu et al., 2017). The catalytic activity of starch synthases varies depending on the protein levels, thus resulting in changes in starch content. Such correlations could occur due to the nature of the starch–protein matrix.

A significant negative correlation was found between phenol and starch content (r = −0.326, p < 0.001), since phenolics interact with enzymes such as amylases, which is followed by a decrease in starch digestion in several legumes (Zhu, 2015). At the same time, phytic acid has a positive correlation with protein (r = 0.22, p < 0.05) and TDF (r = 0.238, p < 0.01). The positive correlation with protein could be attributed to the anionic phytic acid groups, which attach tightly to the cationic groups of protein, i.e., the ε-NH2 of lysine, the α-NH2 terminal group of the protein chain, the guanidyl group of arginine, and the imidazole group of histidine (Nkhata et al., 2018). TDF also tends to bind mineral ions, which might decrease the bioavailability of some mineral elements, similar to phytic acid, which explains the positive correlation between them (Nkhata et al., 2018). Phytic acid showed a negative and significant correlation with amylose (r = −0.226, p < 0.05), which is similar to the results reported by Sivakumaran et al. (2018) in legumes. The relationship between the traits indicates that it is possible to find accessions with low starch content and higher dietary fiber, protein, phytate, and phenol levels. High phenolic, tannin, and phytate levels are associated with enzyme inhibition and prevent the interaction of amylases with the starch molecule. Accessions with low starch and high dietary fiber, protein, phytate, and phenol content are likely to have low GI and impair the release of glucose from other food components consumed with it.

The PCA was used to correlate the findings obtained for different parameters, such as protein content, starch content, amylose content, TDF, TSS, phenolic profile, and phytic acid. Loadings were studied to identify the underlying link between the data structure after representative PCs were established based on sample grouping/differentiation and variance explained. An Eigenvalue is assigned to each PC that accounts for a fraction of the variation in the dataset. Each Eigenvalue has an eigenvector linked to it, which defines the variation within the main components. A Scree plot was used to plot the eigenvalues of factors or principal components. The scree plot containing the principal components based on eigenvalues (>1) is shown in Figure 1A. Three components have been identified (Table 3), where phytic acid (FL2: 0.708), TSS (FL3: 0.712), phenol (FL3: 0.677), protein (FL1: 0.878), and TDF (FL2: 0.636) have large positive factor loading values, which indicate that the specific values will be higher on the positive axis (Figure 1B). Correspondingly, starch (FL1: −0.886) and amylose (FL2: −0.662; FL3: −0.454) have negative factor loading values, indicating that the values are higher on the negative axis (Figure 1B).

Nearly, all the useful data are present in the first few PCs, while noise is mostly present in the remaining PCs (Chaudhry et al., 2018). As in our study, 62.7% of the overall variation was accounted for by the first three main components (PC1, PC2, and PC3; eigenvalues >1; Figure 1A), which is in agreement with the studies conducted in maize hybrids where the remaining principal components accounted for a minimal variability (Ali et al., 2015). Protein (0.878) is the major contributor to the variability, and on the negative axis, starch (−0.886) contributes to the maximum variation in PC1, accounting for 27.05% of the total variation (Table 3). Phytic acid (0.708) followed by TDF (0.636) contributed to the maximum variation, whereas amylose (−0.662) contributed negatively in PC2 (Table 3), accounting for 20.68% of the total variation. The third principal component (PC3) accounted for 14.98% of the total variation, where maximum variability was contributed by TSS (0.712), followed by phenols (0.677), and negatively contributed by amylose (−0.454) (Table 3). The first six principal components explained 95.5% of the variability within the observations, where PC1, PC2, PC3, PC4, PC5, and PC6 showed 27.05, 20.68, 14.98, 12.00, 11.30, 9.47, and 4.52% of the total variation, respectively, as given in Supplementary Table 2.

The dominant factors showed high component loadings along with low uniqueness values. Uniqueness defines the fraction of variation that is unique to a single variable and cannot be explained by PC, while loading establishes the contribution of each variable to PC. Thus, the larger the contribution of a given variable to PCA, the lower the uniqueness value. According to Figure 1B, the lower angle between the two vectors represents the positive correlation between the traits. When the vectors make 90°, they are less likely to be correlated, and when the vectors make 180°, the traits are negatively correlated. Thus, accessions could be grouped based on the relationship between variables and the uniqueness of dominant factors.

Hierarchical clustering analysis was employed to evaluate the multivariate association between biochemical characteristics of accessions being assessed. The accessions were divided into six clusters using Ward's method at a square Euclidian distance of 5 and are shown as a dendrogram in Supplementary Figure 3. Table 4 depicts the accessions that are divided into six clusters, along with the mean values of their nutritional properties.

Figure 2 illustrates the grouping of accessions into six clusters, where cluster I contains 22 accessions characterized by a low amount of TSS (4.73 g 100 g⁻¹), a high amount of TDF (18.8 g 100 g⁻¹), and a moderate level of other constituents. Cluster II contains 28 accessions characterized by a low amount of amylose (12.4 g 100 g⁻¹) and a moderate level of other constituents. Cluster III contains 11 accessions characterized by a low amount of phytate (1.02 g 100 g⁻¹) and protein (21.8%), a higher amount of starch (36.4 g 100 g⁻¹), and a moderate level of other constituents. Cluster IV contains 28 accessions characterized by a low amount of phytic acid (1.06 g 100 g⁻¹), a higher amount of TSS (6.07 g 100 g⁻¹) and phenol (0.303 GAE g 100 g⁻¹), and a moderate level of other constituents. Cluster V contains 18 accessions characterized by a higher amount of phytic acid (1.19 g 100 g⁻¹), protein (26.2%), TSS (5.89 g 100 g⁻¹), and phenol (0.302 GAE g 100 g⁻¹) and a lower amount of starch (29.7 g 100 g⁻¹). Cluster VI contains 13 accessions characterized by a higher amount of amylose (19.5 g 100 g⁻¹), a lower amount of phenol (0.170 GAE g 100 g⁻¹), and a moderate level of other constituents.

Germplasm of clusters IV and V had higher TSS content, and cluster V contained a high amount of protein; the accessions of these clusters could be used as a source for designing high protein and palatable food, suitable for use as a protein supplement, for making protein isolates, and useful for combating protein-energy malnutrition in the developing countries. Accessions of cluster III contain high starch content, and cluster VI contains higher amylose content, and these characteristics contribute to the best physico-functional features. Therefore, these clusters might be useful in making processed products, confectionery items, and legume starch in the food processing industry. In comparison, owing to the high TDF and amylose content in clusters I and VI, respectively, they could be helpful in making low GI foods. Accessions from both of these clusters are suitable for persons with cardiovascular ailments and diabetes. Accessions of cluster II have low amylose content and high swelling power, and hence are preferred for ease in cooking. Clusters III and VI contain a lower amount of phytic acid and phenol, respectively, which can be used as functional ingredients and for food diversification. This information could help in the identification of eligible parents in the cowpea improvement programs for improving the nutritional properties.

The studied cowpea accessions have good nutritional properties which could be exploited for nutrition and food formulations. Some nutri-dense accessions have been found that can be used in different areas of the food industry and crop improvement programs. EC169879 and IC201086 (cluster V) have high protein, TSS, amylose, and TDF with low starch content. These accessions can be used to make high-protein foods and feeds and can be used for people with obesity, cardiovascular diseases, and type-II diabetes, because of their high amylose and TDF content. Several authors have discussed the use of legume flours as components in diverse culinary products, such as bread, biscuits, pasta, tortillas, and doughnuts (Naiker et al., 2019). Keeping this information in mind, EC724681, EC240663, and EC472284 (cluster III) can be used for this purpose owing to their high starch content. Except for EC240663, the other two accessions also have high amylose and high TDF content. The food products prepared from these accessions will have a low glycemic index, ultimately helping diabetic and obese patients. Accessions like IC214752, IC249141, and EC 367678 (cluster II) have low amylose and high TDF and protein content, which will contribute to higher swelling capacity and solubility index (Ashogbon and Akintayo, 2012), and lower syneresis rate (Naiker et al., 2019). These properties can further help to increase the shelf life, texture, and frozen quality of processed food products.