Metabolite profiling of different solvent extracts of Moringa oleifera seeds and correlation with DPPH radical scavenging activity via 1H NMR-based metabolomics

In the present study, profiling of the Moringa oleifera seeds metabolome was carried out by employing proton nuclear magnetic resonance (1H NMR) spectroscopy combined with multivariate data analysis (MVDA) of 3 different solvent extracts. The principal component analysis (PCA) and partial least square-discriminant analysis (PLS-DA) score plot reveal that methanol extract was discriminated from ethyl acetate and hexane extracts by PC1 while ethyl acetate and hexane extracts were well separated from methanol extract by PC2. The PLS-DA loading plot highlighted the potential metabolites, which are responsible for the group separation observed in the score plot. Further detailed examination of the loading plot shows that methanol extract contains significantly higher amount of vitamins, sterols, amino acids and fatty acids compared to the other extracts. A total of 37 compounds were detected from the 3 different solvents upon which the methanolic extract was identified to contain more metabolites and in a wider range than the other organic solvent extracts. Based on PLS analysis, ergosterol, oleic acid, isoleucine, riboflavin, cholesterol, leucine, ascorbic acid, stigmasterol, tryptophan, choline, histidine and cysteine displayed strong correlation to 2,2-diphenyl-1-picrylhydrazyl (DPPH) free radical scavenging activity. Hence, this extract would be more appropriate in metabolite extraction for analysis and for therapeutical benefits. Therefore, NMR spectroscopy combined with MVDA in compliment with the right choice of solvent for extraction could be utilized by applicable industries to obtain maximum valued metabolites within a short period of time. Besides having high diversity of metabolites, M. oleifera seeds can serve as potential nutritional source to develop new functional foods, and even as a source of biodiesel.


INTRODUCTION
Plant metabolomics is an emerging technology which involves the comprehensive evaluation of all metabolites in a typical plant extract while offering a reasonable metabolic snapshot of a plant, aiding the development of natural product-derived therapeutic agents.Metabolite profiling of plant extracts provides information on their chemical composition.Therefore, it allows detection of chemically varied bioactive molecules and unknown compounds, as well as assessing the possibility of their isolation (Satheeshkumar et al., 2012).The fundamental goal in plant metabolomics is the ability to reliably detect and quantify every metabolite in a plant extract.To achieve this goal, nuclear magnetic resonance (NMR) spectroscopy has been recognized as a powerful platform for obtaining the metabolite profiles of plant extracts (Selegato et al., 2019).Its advantages are the universal detection of organic compounds with a high dynamic range, a good reproducibility, a relatively simple implementation for the screening and quantification of a range of major metabolites, and a provision of structural information for compound identification (Deborde et al., 2019).
The 1H NMR-based metabolomics is used for data acquisition.Simple 1D 1H NMR spectra are particularly useful for quantifying metabolites in the extract.Because of the large dynamic range of metabolite concentrations in bio-fluids, statistical analyses of one-dimensional (1D) proton NMR data tend to be biased toward selecting changes in more abundant metabolites (Que et al., 2008).To overcome the ubiquitous problem of overlapping peaks in 1D NMR spectra of complex mixtures, acquisition of two-dimensional (2D NMR) spectra allows a better separation between overlapped resonances while yielding accurate quantitative data when appropriate analytical protocols are implemented.
With the use of multivariate data analysis (MVDA) on the NMR data, the principal component analysis (PCA), an unsupervised multivariate data analytical method with the partial least square discriminative analysis (PLS-DA) and the partial least squares projection to latent structures (PLS), supervised multivariate data analytical tools will provide good categorization of metabolite according to the extraction solvents.The scores and loading plots displayed by the (MVDA) are used to extract significantly altered metabolites with pair wise comparison (Limin et al., 2016).Thus, this technique is a highly promising tool for quality control and standardization of natural products (Heyman and Meyer, 2012).
Moringa oleifera Lam (Moringaceae) is one of the world's most useful medicinal plants due to frequent usage of moringa tree for herbal medicine, nutrition and in the development of other pharmaceutical products.Traditional medicine and modern science now increasingly, have recognized the unique composition of this plant and its efficacy in the treatment of various diseases.In particular, the extracts of the seeds are known to possess antiproliferative, anti-inflammatory, antimicrobial, antioxidant, and osteoprotective effects (Basuny, 2016;Nadeem and Imran, 2016;Kou et al., 2018).M. oleifera seeds are a promising resource for food and non-food applications, due to their content of monounsaturated fatty acids with a high monounsaturated/saturated fatty acid (MUFA/SFA) ratio, sterols and tocopherols, as well as proteins rich in sulfated amino acids (Leone et al., 2016).
M. oleifera seeds can be an alternative to some leguminous seeds as a source of high-quality protein, oil and antioxidant compounds and a way to treat water in rural areas where appropriate water resources are not available (Ferreira et al., 2008).
In addition, monounsaturated fatty acids demonstrate cardio-protective effects by promoting a healthy lipid profile, mediating blood pressure, and improving insulin sensitivity and glucose levels (Gillingham et al., 2011).However, despite the recent advances in plant metabolomics research, only a few studies have been reported on the application of metabolomics for M. oleifera cultivars (Managa et al., 2021).Also, there is no data on the 1 H NMR-based metabolite profiling that discriminate between different solvent extracts of M. oleifera seeds using the metabolomics approach.
The extraction solvent and drying process used can also have a significant impact on the biological activities and metabolite contents of plant materials because distinct bioactive constituents with different chemical properties and polarities may have varying solubility in different solvents (Rafi et al., 2020;Sharma et al., 2021).
Therefore, the present study aims to establish the metabolite profile of the M. oleifera seeds extracted by three different polarity solvents; methanol (MeOH), ethyl acetate (EtOAc) and hexane (Hex) and to correlate these profiles to antioxidant activity using 1 H NMR-based metabolomics approach.
Furthermore, multivariate data analysis and relative quantification will be performed using binned NMR data.

Sample collection and preparation
Mature seeds of M. oleifera were collected from "Muda-Lawan" market, Bauchi city in northeast Nigeria and identified by a taxonomist in the Department of Biological Sciences, Abubakar Tafawa Balewa University, Bauchi, North East Nigeria.The seeds were cracked (by exerting manual pressure on them) and the shells were carefully removed.The seed kernels obtained were air dried for three weeks.The air dried seed kernels were ground into fine powder using a laboratory mortar and pestle, packed into air tight glass containers and kept in a dark, cool place until further use.

Chemicals and reagents
All chemicals and reagents used in this research were of the analytical grade and procured from Sigma-Aldrich (USA).Some of which include methanol, ethyl acetate, hexane, quercetin, 2,2-diphenyl-1-picrylhydrazyl, deuterated chloroform (CDCl 3 ) and tetramethylsilane (TMS).For calibration of chemical shifts, 0.1% of TMS was added to the final volume.

M. oleifera seeds solvent extraction for 1 H NMR-based metabolomics profiling
The following solvents were used in increasing polarity for solventextraction of the ground M. oleifera seed kernels; hexane (Hex), ethyl acetate (EtOAc) and methanol (MeOH).Extraction with each solvent was achieved using cold extraction method by soaking 100 g of the powdered seed kernel in 300 mL of solvent for 72 h in an enclosed glass jar and filtered.The procedure with each solvent was repeated, evaporated to dryness and stored at sterile laboratory conditions until further analysis.Each solvent extract was replicated four times for statistical analysis to be more valid.

Sample preparation for NMR analysis
Each of the solvent extract (10 mg) was mixed with NMR deuterated solvent system [0.7 mL; chloroform-d and methanol-d4 (2:8) containing 0.1% TMS], vortexed for 15 min, sonicated for 15 min, centrifuged for 15 min and 0.6 mL of the fine solution was dispensed into 5 mm NMR tubes for analysis.All samples were made in four replicates for multivariate analysis of the data.

NMR analysis
All 1H NMR and 2D; (J-resolved), 1H-Hetero-nuclear Single Quantum Coherence (HSQC) and Hetero-nuclear Multiple Bond Coherence (HMBC) NMR spectra were obtained on a 500 MHz Varian INOVA NMR spectrometer (Varian Inc., Palo Alto, California, USA), running at a frequency of 499.887MHz at room temperature (25°C).The solvent used was a combination of CD 3 OD 80% and CDCL 3 20% constituted with 0.1% tetramethylsilane (TMS) as internal standard.For each of the replica sample, the following parameters were used: set temperature of 24°C, acquisition time for 1H NMR spectra 8.49 min, spectral width was 14 ppm, number of scans was 28, and relaxation delay was 2 s.The pre-saturation (PRESAT) pulse sequence was applied to all the samples to reduce water (H 2 O) interrupting signals, which involved 64 scans.The 2D NMR experiments, that is, J-resolved, HSQC and HMBC spectroscopy were also carried out to assist biomarker identification.

NMR spectra preprocessing (Bucketing/Binning) and multivariate data analysis
The 1 H NMR spectra of all replicate samples were binned to ASCII files using Chenomx software (version 6.2, Edmonton, Alberta, Canada).The spectral region 0.30 to 14.00 ppm was bucketed into integrated regions with spectral width of δ 0.04.The residual signals for water, methanol and chloroform in the range of 4. 68 -4.88 ppm, 3.29 -3.32 ppm, and 7.50 -7.60 ppm, respectively, were excluded from the analysis.The standardized bucketed data were then pareto-scaled (Par) and subjected to principal component analysis (PCA), partial least square discriminant analysis (PLS-DA) and partial least squares projection to latent structures (PLS) using SIMCA-P+ software (version 14.1, Umetrics, Umea, Sweden).The 1D and 2D NMR spectra were viewed using MestRenova (version 6.02-5475, Mestrelab Research, Santiago de Compostella, Spain).

Famurewa et al. 3 2,2-Diphenyl-1-picrylhydrazyl (DPPH) Free Radical Assay
Determination of radical scavenging activity (RSA) of M. oleifera seed extracts was performed by the DPPH radical scavenging method.Different concentrations of each solvent extracts ranging from 5 to 30 mg/mL extract were prepared in methanol, 2 mL of each extract was vigorously shaken with 2 mL of methanolic solution of 0.1 mM 2,2-Diphenyl-1-picrylhydrazyl (DPPH).The mixture was stored in the dark for 30 min and absorbance of samples was measured on UV-Visible spectrophotometer at max 517 nm against blank which acts as control containing equal amount of DPPH and methanol.Quercetin was used as standard.
The free radical scavenging activity was calculated using the formula as described by Aryal et al., (2019).
% Scavenging of DPPH = where = Absorbance of control and = Absorbance of test extracts.

Statistical analysis
One-way analysis of variance (ANOVA) was performed using Graph Pad Prism 6.0 (Graph Pad Software, San Diego, CA, USA).Post-hoc analysis was carried out using Tukey's test, wherein values with were considered to be statistically significant.Values were expressed as mean ± standard deviation (SD).Relative quantification was done by manually entering the data into SPSS Data Analysis Version 22.0 (SPSS Inc., Chicago, IL, USA).

Assignment of metabolites by 1D and 2D NMR spectra of M. oleifera seed solvent extracts
The peak signals in the 1H NMR and 2D NMR spectra were assigned based on reference to Chenomx software database (Version 8.3, Alberta, Canada), previously reported literature and by comparison with several online databases (freely available), such as the Human Metabolome Database (HMDB) at http://www.hmdb.ca/, the Biological Magnetic Resonance (BMR) database at http://www.bmrb.wisc.edu/,PubChem at https://pubchem.ncbi.nlm.nih.gov/ and the Nuclear Magnetic Resonance database (NMRDB) at http://www.nmrdb.org/.
Figure 1 displays the representatives 1D 500 MHz 1H NMR spectra of M. oleifera seeds in Hex, EtOAc, and MeOH extracts.The spectra showed the presence of different classes of metabolites, including fatty acids, amino acids, vitamins and sterols which were successfully determined using NMR spectroscopy.These results verified the convenience of NMR in metabolomics profiling studies as almost all metabolites from different classes of compounds were detected in just one type of analysis (Emwas et al., 2013).

) in Table (1). (A higher resolution / colour version of this figure is available in the electronic copy of the article).
The identification of the metabolites was supported by 2D-NMR experiments in order to increase the metabolites specificity and to minimize the congestion of the signals.Figure S1 of supplementary materials provide 1H 2D-NMR (HSQC and HMBC) analysis for the representative VIP detected metabolites.
A total of thirty-seven (37) metabolites from hexane, ethyl acetate, and methanol extracts were putatively identified along with their chemical shift assignments corresponding to their multiplicities signals as summarized in Table 1.Overall, the methanol extract was identified to contain more metabolites followed by ethyl acetate while hexane extract gave the lowest number of metabolites.Thus suggesting that M. oleifera seeds contain more polar metabolites than the less and non-polar ones.Table (2) lists the yields of the individual solvent extract.
Marked differences within the different solvent extracts were observed in the up-field region of the spectrum (5.2-5.4 ppm), which were the characteristic chemical shift areas for monounsaturated fatty acids [MUFAs], polyunsaturated fatty acids [PUFAs], and sterol constituents.Variations were also detected in the highfield regions (2.7-4.0 ppm) which corresponded to the regions for amino acids.The signals for amino acids displayed characteristic amine and methyl protons which were detected only in the polar solvent extracts (7.08 -7.12 ppm) in the down-field region.However, a marked overlapping of signals was observed in the region (4.05-4.35ppm), which limits the identification of metabolites for the spectra.The identification was further supported by 2D experiments (J-resolved, heteronuclear multiplebond correlation, (HMBC) spectroscopy).Methanol, ethyl acetate and hexane extracts of M. oleifera seed were among the solvent extracts that showed these signal regions for the detected compounds in a decreasing order (Figure 1).
The vitamins have several distinct groups identified as methine, methyl, methylene, amine protons and hydroxylmethyl groups.The J-resolved spectrum showed a singlet signal at 3.23 ppm corresponding to three methyl substituents attached to an amino skeletal which signified the characteristic of choline.The upfield region of all spectra exhibited signals for both saturated and unsaturated fatty acids from the appearance of characteristic terminal methyl groups (-CH 3 ) at 0.86-0.89and methylene groups (-CH 2 -) at 1.25-1.27ppm of fatty acid chains.The allylic protons of the fatty acid chain, -CH 2 -CH= appeared at 2.00-2.02ppm.
The signals corresponding to unsaturated carbon (-CH=CH-) were deduced based on the presence of the olefinic protons (5.25 -5.37 ppm), protons attached to the bisallylic carbons (129 -132 ppm).The J-resolved spectrum showed multiplet signals of the olefinic protons at 5.25 -5.37 ppm and the HMBC spectrum revealed that these signals were correlated to methylene carbon signals at 25 -27 ppm which signified the characteristic Famurewa et al. 5 of PUFAs.The 1H NMR spectra indicate many aliphatic proton resonances namely methylenes and methyl protons which correspond to sterol signal in the chemical shift range of 0.86 -2.30 ppm.The proton signals at (3.47 -3.66 ppm) of a sterol moiety correspond to carbon (71 -72 ppm) attached to a hydroxyl group.The proton attached to olefinic linkage was observed at 5.25 to 5.37 ppm.

Classification of different solvent extracts of M. oleifera seed by PCA and PLS-DA
Assignment of metabolites in the 1H NMR and 2D NMR spectra of hexane, ethyl acetate and methanol extracts revealed notable variations in metabolite constituents.Consequently, PCA and PLS-DA multivariate data analysis were used to analyze the differences in metabolites present in the solvent extracts.The principal component analysis (PCA) score plot (Figure 2A) which is an unsupervised method was first used to identify trends, clusters and outliers among sampled data.
In contrast to unsupervised methods, the PLS-DA score plot (Figure 2B) which is a supervised method was used to get a clearer separation and to obtain improved class discrimination among the metabolite groups.The loading plot is complementary to the score plot (Bäckryd et al., 2018).The PLS-DA loading plot (Figure 3) was used to highlight the potential metabolites, which are responsible for the group separation observed in the score plot of the same model by PC1 and PC2 for the three solvent extracts.
It was evident from the score plot model (Figure 2) that the 12 samples were clustered into three groups.Satisfactory goodness of fit and high predictability was observed from the values of (R2X cum) and (Q2 cum) as 0.987 and 0.942 respectively, and which are close to 1 as described by Eriksson et al., (2006).In addition, no outliers were seen in the score plot.The first principal component (PC1) accounted for 51.5% of the variance, while the second principal component (PC2) accounted for 26.4% of the variance, cumulatively explaining the total variance of 77.9%.This implies that the metabolite profiles of the three solvent extracts were clearly different from each other.
The Orthogonal partial least squares discriminant analysis (OPLS-DA) was also generated to confirm the pattern of separation among the solvent extracts.The OPLS-DA 3D score scatter plot and loading plot (Figure S2), revealed a clear trend and widely scattered separation between the solvent extracts.
The PCA score plot (Figure 2A) shows that methanol extract was discriminated from ethyl acetate and hexane extracts by first principal component (PC1), while the PLS-DA score plot (Figure 2B) reveals that the ethyl      acetate and hexane extracts were well separated from methanol by second principal component (PC2).Further detailed examination of the PLS-DA loading plot (Figure 3) shows that methanol extract contains significantly higher amount of vitamins, sterols, amino acids and fatty acids compared to the other extracts.The variable importance for the projection (VIP) plot of the PLS-DA model (Figure 4) summarizes the importance of the metabolites in terms of content amounts.The plot is sorted to display the largest VIP to the left.In this regard, metabolites with VIP value greater than 1.0 were considered as most-ranked metabolites and are as shown in Figure 4.The top-ranked metabolites in the VIP are few sterols and many fatty acids, followed by few vitamins and many amino acids.

PLS-DA model validation
The statistical validation of the PLS-DA model was evaluated using 100-permutation test to nullify the risk of over-fitting.The plot shows, for a selected Y-variable, on the vertical axis, the values of R2 and Q2 for the original model (far to the right) and of the Y-permuted models further to the left.The horizontal axis shows the correlation between the permuted Y-variable and the original Y-variable for the selected Y.The original Y has a correlation 1.0 with itself, defining the high point on the horizontal axis.
The results from the permutations reveal that the PLS-DA model was clearly validated without over-fitting from the original model, since the Q2 regression lines had a negative intercept and the permuted Y-variable on the left was lower than the original Y-variable on the right (Figure 5).

Relative quantification using 1H NMR data
The levels of metabolites detected from 1H NMR spectra, which are responsible for the variations between different solvent extracts were relatively quantified based on the binned NMR data obtained from all three replicates of different solvent extracts to validate the results.
For the quantification, only those peaks that do not cluster but rather stand out in bold relief, distinct and well resolved were selected.Figure 6A to D shows bar chart with the metabolites grouped according to their class and There was a similar trend in all the solvent extracts with regard to the quantification of steroid content of M. oleifera seeds in the level of identified pigments with the more polar extracts, particularly the MeOH, followed by EtOAc and Hex (Figure 6A) and revealed that ergosterol (binned at δ 5.26 ppm) was found to be significantly higher in all solvent extracts followed by 7dehydrocholesterol (binned at δ 2.32 ppm), cholesterol (binned at δ 0.87 ppm) and stigmasterol (binned at δ 5.34 ppm), while β-sitosterol (binned at δ 2.00 ppm) was found  to be present in high amount in the ethyl acetate extract.
In the present study, the seed contained high amounts of not only ergosterol but also 7-dehydrocholesterol (7-DHC), with the highest concentrations.The results obtained for sterols profiles in the present research agree with recently published findings (Baur et al., 2016;Özcan, 2020).Vitamins (Figure 6B) also show the same consistent trend in the quantities present in all extracts as follows: vitamin E, -Tocopherol (binned at δ 4.17 ppm) was found to be significantly higher followed by vitamin B2 riboflavin (binned at δ 3.65 ppm), vitamin C, ascorbic acid (binned at δ 3.48 ppm) and vitamin B choline (binned at δ 3.23 ppm) was found to be present in the methanol extract.There is predominant high content of -Tocopherol commonly known as vitamin E which has good oxidative stability.This result is in agreement with the previous studies (Rahman et al., 2009;Gharsallah et al., 2021).In contrast, amino acid profile (Figure 6C) shows the alternating trend of their levels or amount in different solvent extracts; low levels of glycine (binned at δ 3.57 ppm), phenylalanine (binned at δ 3.97 ppm) and histidine (binned at δ 7.12 ppm) were detected in methanol extract.Low amount of methionine (binned at δ 2.01 ppm) was found to be present in ethyl acetate extract; and valine (binned at δ 2.30 ppm) was found to be high in hexane extract.
High levels of isoleucine (binned at δ 1.25 ppm), tryptophan (binned at δ 4.09 ppm) arginine at δ 3.27 ppm) and leucine (binned at δ 0.89 ppm) along with low levels of lysine (binned at δ 3.61 ppm) and cysteine (binned at δ 4.15 ppm) were found to be high in methanol extract.High amount of glutamic acid (binned at δ 2.28 ppm) and threonine (binned at δ 4.27 ppm) were detected in methanol and ethyl acetate extract.In addition, high level of proline (binned at δ 2.01 ppm) was found to be high in ethyl acetate and hexane extract.
The comparison of the inter-solvent extract in terms of amino acids yield therefore suggested that methionine was best extracted with ethyl acetate; valine was best extracted using hexane; proline was best extracted using ethyl acetate and hexane; glutamic acid and threonine were best extracted with methanol and ethyl acetate while glycine, isoleucine, phenylalanine, histidine, leucine, lysine, cysteine and arginine, tryptophan were best extracted using methanol.
Categorically, in the present study, nine essential amino acids (valine, leucine, isoleucine, methionine, lysine, phenylalanine, threonine, tryptophan and histidine) with five non-essential amino acids (glycine, proline, arginine, cysteine and glutamic acid) were putatively identified.Amino acid profile obtained in the present research agrees with recently published data on M. oleifera seeds (Aderinola et al., 2018;Igwilo et al., 2017).
Similarly, in the fatty acids profile (Figure 6D), the methanol extract was characterized by higher amounts of oleic acid (binned at δ 1.29 ppm), arachidic acid (binned at δ 1.26ppm), palmitic acid (binned at δ 0.86 ppm) and capric acid (binned at δ 1.29 ppm).Erucic acid (binned at δ 2.34 ppm) was found in high amount in all the solvent extracts.High levels of linoleic acid (binned at δ 2.35 ppm), palmitoleic acid (binned at δ 1.64 ppm), lauric acid (binned at δ 1.61ppm), and vaccenic acid (binned at δ 2.35 ppm) were detected in high amount in methanol and ethyl acetate extract.Caprylic acid (binned at δ 1.61 ppm) and myristic acid (binned at δ 1.29 ppm) were found to be high in ethyl acetate extract while stearic acid (binned at δ 1.26 ppm) was detected in high amount in hexane extract.
Also, linolenic acid (binned at δ 1.59 ppm) was present in high levels in ethyl acetate and hexane extracts while behenic acid (binned at δ 2.34 ppm) was found in high amount in methanol and hexane extract.However, the comparison between the solvent extracts in terms of fatty acids yield suggested that stearic acid was best extracted Famurewa et al. 15 with hexane; myristic acid and caprylic acid were best extracted using ethyl acetate; oleic acid, arachidic acid, capric acid and palmitic acid were best extracted using methanol.
The present fatty acid profile shows that M. oleifera seed contains high level of monounsaturated fatty acids (palmitoleic acid, oleic acid, erucic acid and vaccenic acid) compared to saturated fatty acids (stearic acid, arachidic acid, behenic acid, lauric acid, capric acid, myristic acid, caprylic acid and palmitic acid) and lower content of polyunsaturated fatty acids (linoleic acid and linolenic acid).These results agree with the previously published data on fatty acid content of M. oleifera seeds (Leone et al., 2016;Özcan, 2020).
Thus, results from PLS-DA and relative quantification established that polar and relatively polar solvent are the most suitable solvent for the recovery of high concentrations of metabolites from M. oleifera seeds as it is momentous to obtain a wide range of compounds in the highest intensity possible.

Effect of different solvent extractions on DPPH radical scavenging activity
The three different solvent system contributed variation to the DPPH free radical scavenging activities of M. oleifera seeds.Methanol extract exhibited high percentage of DPPH free radical scavenging activity followed by ethyl acetate and hexane extracts (Figure 7).The percentage inhibition for methanol extract was 47.65% whereas ethyl acetate and hexane extracts have 37.53 and 34.94% of inhibition, respectively compared to the percentage inhibition of quercetin, which is 50.75%.The IC 50 value for each extract was determined graphically by plotting inhibition percentage versus extract concentrations.The IC 50 values obtained for standard (quercetin), methanol, ethyl acetate and hexane extracts are 21.1, 28.71, 42.37, and 45.06 mg/mL, respectively.
The result for ANOVA showed that there were significant differences in all solvent systems.The results in the present study are in agreement with recently published findings where they found the methanol extracts exhibited better free radical scavenging activity in comparison to the other solvent systems (Tariq et al., 2022).

Correlation between identified metabolites and antioxidant activity in M. oleifera seed extracts
In order to comprehend the possible relationship between the metabolites and the bioactivity, the partial least squares projection to latent structures (PLS), a supervised multivariate data analysis was used to relate data of the independent variables (X-matrix, 1 H-NMR  spectral data) to data of the dependent variables (DPPH).Thus, the correlation reveals the metabolite that serves as potential biomarker for the studied bioactivity.PLS biplot (Figure 8) reveals three different clusters without outliers for the solvent extracts analyzed.The model showed the good fitness (R2Y) value of 0.982, whereas the predictive ability (Q2) was 0.951.
The Y-variables (DPPH activity) were clustered on the positive side of the plot, which are close to the methanol extracts.Vitamins such as riboflavin, ascorbic acid and choline; amino acids such as isoleucine, tryptophan, leucine, histidine and cysteine; sterols such as ergosterol, cholesterol and stigmasterol; that were concentrated in methanol extracts are the compounds responsible for influencing the bioactivity of Y-variables of M. oleifera seed samples.Oleic acid is also shown to be correlated.To test the credibility of the PLS model, permutation tests (200 permutations) were done.The permutation test result of DPPH free radical scavenging activity (Figure 9) showed the Y-intercepts of R2 and Q2 were 0.139 and -0.223, which demonstrated that the constructed model were valid and not over fitting.

Conclusion
In the present study, there is considerable variation of metabolites in M. oleifera seeds extracted by different organic solvents.By employing NMR-based metabolomics approach combined with multivariate data analysis (MVDA), the metabolite profile of M. oleifera seeds was characterized, and the comparison between different solvent extracts of M. oleifera was established.
A total of 37 compounds were detected from three different solvents upon which the methanolic extract was identified to contain more metabolites and in a wider range than the other organic solvents.For effective extraction of metabolite from plants sample, solvent selection is one of the crucial contributors as these factors contribute directly to the types of compounds that can be extracted, the extractable amount, and, hence, the biological activity of the extracts.The results indicated that ergosterol, oleic acid, isoleucine, riboflavin, cholesterol, leucine, ascorbic acid, stigmasterol, tryptophan, choline, histidine and cysteine are the metabolites that strongly correlated to DPPH free radical scavenging activity.
Furthermore, the methanolic extract contains several different classes of compounds in a wider range, including pigments and organic acids and hence would be more appropriate in metabolite extraction for analysis and for therapeutical benefits.Therefore, NMR spectroscopy combined with MVDA in compliment with the right choice of solvent for extraction could be utilized by applicable industries to obtain maximum valued metabolites within short period of time.,leucine;4,ascorbic acid;6,cysteine;7,histidine;8,ergosterol;11,12,oleic acid;13,glycine;20,isoleucine;24,riboflavin;26,caprylic acid;32,stigmasterol;33,methionine;34,erucic acid;35,linolenic acid;36,tryptophan;37,cholesterol.In addition to having high diversity of metabolites, M. oleifera seeds can serve as potential nutritional source to develop new functional foods, and even as a source of biofuel.

Fig. ( 2
Fig. (2).The PCA (A) and PLS-DA (B) score plots for the discrimination between different solvent extracts of Moringa oleifera seed.Key: EtOAc -ethyl acetate extract; Hex-hexane extract; MeOH-methanol extract.(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Figure 2 .
Figure 2. The PCA (A) and PLS-DA (B) score plots for the discrimination between different solvent extracts of Moringa oleifera seed.Key: EtOAc -ethyl acetate extract; Hex-hexane extract; MeOHmethanol extract.(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Fig. ( 6
Fig. (6).Relative quantification of 1H-NMR signals of identified metabolites in M. oleifera seeds.A; sterols B; vita mins, C; amino acids and D; fatty acids.Data presented are based on the mean of four replicates (n=4) from each of the solvent systems (methanol, ethyl acetate and hexane) ± standard deviation (SD).(Ahigher resolution / colour version of this figure is available in the electronic copy of the article).

Figure 6 .
Figure 6.Relative quantification of 1H-NMR signals of identified metabolites in M. oleifera seeds.A; sterols B; vitamins, C; amino acids and D; fatty acids.Data presented are based on the mean of four replicates (n=4) from each of the solvent systems (methanol, ethyl acetate and hexane) ± standard deviation (SD).(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Fig ( 7 )
Fig (7).Percentage inhibition of DPPH free radical scavenging activity for quercetin (standard) with the different solvent extracts of M. oleifera seed.The methanol extract is significantly higher than the other extracts.The level of significance difference was represented by alphabets.Values having different alphabets were statistically significant different .(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Figure 7 .
Figure 7. Percentage inhibition of DPPH free radical scavenging activity for quercetin (standard) with the different solvent extracts of M. oleifera seed.The methanol extract is significantly higher than the other extracts.The level of significance difference was represented by alphabets.Values having different alphabets were statistically significant different (p˂0.05).(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Fig. 9 .
Fig. 9. Validation of PLS model using permutation test (200 permutations) of DPPH free radical scavenging activity.(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Figure 9 .
Figure 9. Validation of PLS model using permutation test (200 permutations) of DPPH free radical scavenging activity.(A higher resolution / colour version of this figure is available in the electronic copy of the article).

Figure S2 .
Figure S2.(OPLS-DA) 3D score scatter plot scaled proportionally to R2X (A) showing variations between the solvent extracts.The corresponding (OPLS-DA) 3D loading scatter plot normalized to unit length; (B) show the relative contribution of spectral variables to clustering of the compared solvent extracts of M. oleifera seeds.

Table 2 .
Yield for each solvent extract.
a 18.407 ± 0.217 b 17.140 ± 0.182 c Data are expressed as mean ± SD.Values with the different superscript letters are significantly different as determined by the Turkey's test (p˂0.05).