Discrimination of Phytosterol and Tocopherol Profiles in Soybean Cultivars Using Independent Component Analysis

Abstract

Soybean (Glycine max (L.) Merrill) is a major oilseed crop rich in phytosterols and tocopherols, compounds associated with functional and nutritional properties of vegetable oils. This study aimed to apply, for the first time, Independent Component Analysis (ICA) to discriminate the composition of phytosterols (β-sitosterol, campesterol, stigmasterol) and tocopherols (α, β, γ, δ) in 20 soybean genotypes—14 non-transgenic and six transgenic—cultivated in two major producing regions of Paraná state, Brazil (Londrina and Ponta Grossa). Lipophilic compounds were extracted from soybean seeds, quantified via gas chromatography and HPLC, and statistically analyzed using ICA with the JADE algorithm. The extracted independent components successfully differentiated soybean varieties based on phytochemical profiles. Notably, transgenic cultivars from Ponta Grossa exhibited higher levels of total tocopherols, including α- and β-tocopherol, while conventional cultivars from both regions showed elevated phytosterol content, particularly campesterol and stigmasterol. ICA proved to be a powerful unsupervised method for visualizing patterns in complex compositional data. These findings highlight the significant influence of genotype and growing region on the nutraceutical potential of soybean, and support the use of multivariate analysis as a strategic tool for cultivar selection aimed at enhancing functional quality in food applications.

1. Introduction

Soybean (Glycine max (L.) Merr.) is the second most cultivated oilseed crop worldwide, primarily used for the production of refined oil [1,2,3]. Historically, soybean breeding programs have focused on increasing oil content and developing herbicide-resistant cultivars [4]. Soybean oil is rich in polyunsaturated fatty acids (PUFAs), such as linoleic and α-linolenic acids, which are susceptible to oxidation during frying and other food processing applications [5]. Tocopherols act as endogenous antioxidants, protecting PUFAs from peroxyl radicals generated during lipid oxidation [3]. In addition, tocopherols exhibit anti-inflammatory properties in vivo and help prevent lipid peroxidation [6]. Phytosterols are also present in soybean oil and have been associated with a range of biological activities [7]. Consequently, there is growing interest in increasing phytosterol and tocopherol levels in soybean grains [8].

Phytosterols are bioactive compounds found in all plant-derived foods, with their primary sources being the unsaponifiable fractions of oils from soybean, canola, and sunflower seeds [9]. Over 200 phytosterol types have been identified in plants, with β-sitosterol (24-α-ethylcholesterol), campesterol (24-α-methylcholesterol), and stigmasterol (∆22,24-α-ethylcholesterol) being the most abundant [7,10,11]. These compounds have demonstrated potential to lower low-density lipoprotein (LDL) and total cholesterol levels in human serum, making them valuable for the formulation of phytosterol-enriched functional foods [11,12,13].

Tocopherols, which constitute vitamin E, occur in four isomeric forms (α-, β-, γ-, δ-). Among these, α-tocopherol is considered the most biologically active and health-promoting due to its superior antioxidant capacity [14,15]. Nevertheless, γ- and δ-tocopherol exhibit greater oxidative stability at high temperatures, such as those encountered during frying processes [3]. As natural antioxidants, tocopherols help prevent oxidative damage to human tissues caused by free radicals [16].

Therefore, compositional data on soybean cultivars are essential in various fields, including plant physiology, nutrition, food science, and functional food development. The accuracy of such data is crucial for estimating nutrient intake and for supporting the commercialization of novel oilseed varieties. Previous studies have demonstrated that the levels of phytosterols and tocopherols in soybeans are influenced by seasonal variation, genetic background, and climatic conditions [8,17].

However, interpreting the phytosterol and tocopherol composition of transgenic and non-transgenic soybean grains cultivated across different regions often yields complex datasets. In this context, multivariate statistical techniques offer a valuable means of reducing data dimensionality and improving interpretability [18,19]. Among these, Independent Component Analysis (ICA), first introduced in 1986, has shown strong capabilities for blind-source separation and has been widely adopted in diverse scientific disciplines. In analytical chemistry, ICA is frequently used for data pre-processing, exploratory analysis, classification, regression, and resolution of overlapping signals [20].

Brazil is a major soybean-producing country, with cultivation zones spanning a wide latitudinal range (from 5° N to 32° S). In this context, evaluating the phytochemical composition of soybean cultivars is highly relevant. Thus, in this study, ICA was applied for the first time to assess the variation in phytosterol and tocopherol contents across 20 Brazilian soybean genotypes—14 non-transgenic and six transgenic—grown in two distinct regions of southern Brazil.

2. Materials and Methods

2.1. Chemicals and Standards

HPLC-grade solvents including hexane, isopropanol, chloroform, methanol, tetrahydrofuran, methyl tert-butyl ether (MTBE), ethanol, and n-heptane were obtained from J.T. Baker (Phillipsburg, NJ, USA). Standards of β-cholestanol (purity > 98%), campesterol, stigmasterol, β-sitosterol, (+)-δ-tocopherol (90%), and γ-tocopherol (98%) were purchased from Sigma-Aldrich (St. Louis, MO, USA), while α-tocopherol (98%) was supplied by Merck (Darmstadt, Germany). Analytical-grade reagents included potassium hydroxide (KOH; Ecibra, São Paulo, Brazil) and anhydrous sodium sulfate (Na₂SO₄; Nuclear, São Paulo, Brazil).

2.2. Sampling

A total of twenty soybean (Glycine max (L.) Merrill) samples representing newly developed genotypes—14 non-transgenic (EMBRAPA 48, BRS184, BRS213, BRS232, BRS233, BRS257, BRS258, BRS259, BRS260, BRS261, BRS262, BRS267, BRS268, and BRS282) and 6 transgenic cultivars (BRS242RR, BRS244RR, BRS245RR, BRS246RR, BRS255RR, BRS256RR)—were used in this study. These cultivars were provided by the Brazilian Agricultural Research Corporation (EMBRAPA) and cultivated at two experimental farms: Site I: 23°11′37″ S, 51°11′03″ W; 630 m altitude (Londrina, PR, Brazil); Site II: 25°09′31″ S, 50°04′32″ W; 884 m altitude (Ponta Grossa, PR, Brazil). Both regions are located in southern Brazil but differ in elevation, climate, and soil characteristics. Londrina (Site I) is situated at 630 m altitude and has a humid subtropical climate (Cfa) with an average annual temperature of 20.9 °C and annual rainfall of 1429 mm. The predominant soil type is Rhodic Ferralsol, characterized by high clay content and good drainage. Ponta Grossa (Site II), at 884 m altitude, presents a temperate oceanic climate (Cfb), with a lower average temperature of 17.5 °C and precipitation of 1495 mm. The soil in this region is classified as dystrophic Red Latosol, also rich in clay but with higher organic matter content. These environmental differences may influence the biosynthesis of lipophilic compounds in soybean grains. Experimental conditions, including soil correction, batch separation, grain collection, sample handling, and storage procedures, followed the protocol described by Galão et al. [21].

2.3. Total Lipids, Phytosterol, and Tocopherol Analysis

Total lipids were extracted using a chloroform/methanol/water mixture (2:2:1.8, v/v/v) according to the method of Bligh and Dyer [22]. Phytosterol analysis was performed by saponifying the oil extracted from ground soybean samples. An internal standard (100 µL of β-cholestanol at 3 mg·L⁻¹ in MTBE) was added prior to saponification [23]. Samples were analyzed via gas chromatography (Shimadzu GC-14B) equipped with a flame ionization detector (FID) set at 320 °C, an autosampler, and a cool-on-column injector. A CP-Sil-5CB capillary column (10 m × 0.20 µm × 0.32 mm i.d., Varian CP7730) with a pre-column (2.5 m × 0.53 µm, Varian CP8009) was used. Hydrogen was employed as the carrier gas (flow rate: 45.6 cm·s⁻¹). The oven program was: initial temperature 60 °C (1 min), ramped to 300 °C at 10 °C·min⁻¹, held for 10 min.

For tocopherol analysis, 0.4 g of oil was mixed with 4 mL of an isopropanol:chloroform solution (75:25, v/v) in a glass flask. The HPLC system (Varian) consisted of a ternary pump model 9012, an autosampler model 9100, a DuPont Zorbax ODS column (15 cm × 0.46 mm i.d., 5 µm particle size), and a DAD detector model 9065. Isocratic elution was performed using a mobile phase composed of methanol:tetrahydrofuran:water (67:27:6, v/v/v) at a flow rate of 0.8 mL·min⁻¹, with detection at 292 nm [24].

The oil extracts used for phytosterol and tocopherol quantification were prepared shortly after seed harvesting and stored under appropriate long-term conservation protocols at Embrapa. These included low-temperature storage, exclusion of oxygen and light, and sealed amber glass containers. Such conditions are known to minimize oxidative degradation and preserve the integrity of the unsaponifiable fraction, including tocopherols, for extended periods. These procedures are standard in lipid research and ensure the reliability of the analytical results even after prolonged storage.

2.4. Statistical and Chemometric Analysis

All statistical analyses were conducted using Statistica 8.0 software. Each measurement was performed in triplicate, and results are expressed as mean values. Comparisons among samples within and between regions were assessed by Tukey’s test, with statistical significance set at p > 0.05. Identical letters indicate non-significant differences within the same region, and identical numbers indicate non-significant differences between regions.

Independent Component Analysis (ICA) was employed as an unsupervised pattern recognition method using the Joint Approximate Diagonalization of Eigenmatrices (JADE) algorithm [25] implemented in Matlab R2007b. ICA is a blind-source separation technique designed to extract statistically independent components from a mixture of signals. It operates via a linear transformation that maximizes statistical independence among components [26]. The JADE algorithm is particularly effective at extracting non-Gaussian sources from noisy signal mixtures [27]. The mathematical formulation of ICA using the JADE algorithm has been previously described in detail [27,28].

3. Results and Discussion

The total phytosterol content in transgenic and conventional soybean cultivars from the Londrina region (

Table 1. Phytosterol contents (mg 100 g⁻¹ of total lipid) of soybean seeds from Londrina’s region.

Conventional Cultivar	Campesterol	Stigmasterol	β-Sitosterol	Total Phytosterol
1 Embrapa-48	68.55 ± 0.70 1	67.26 ± 0.90 d1	160.87 ± 0.56 c	296.68 ± 1.27 1
2 BRS 184	57.29 ± 0.19 ef	57.08 ± 0.23 f3	145.14 ± 0.45 f2	259.50 ± 0.54 f
3 BRS 213	59.36 ± 0.26 de3	82.12 ± 0.42	145.22 ± 1.07 f2	286.70 ± 1.17 c2
4 BRS 232	85.84 ± 0.17	92.35 ± 0.31	177.47 ± 0.65 1	355.67 ± 0.74
5 BRS 233	61.50 ± 0.73 cd2	83.48 ± 0.30	225.32 ± 0.77	370.30 ± 1.10
6 BRS 257	56.71 ± 0.46 f	67.60 ± 0.55 d1	122.94 ± 0.68 6	247.24 ± 0.98 h3
7 BRS 258	29.72 ± 0.62	40.30 ± 0.17 m	100.76 ± 0.45 7	170.78 ± 0.78 o6
8 BRS 259	63.00 ± 2.41 c2	67.08 ± 0.63 d1	154.55 ± 0.74 d	284.63 ± 2.59 c2
9 BRS 260	33.67 ± 0.78 m12	37.93 ± 0.46 7	97.77 ± 0.57 m8	128.70 ± 1.07
10 BRS 261	35.42 ± 0,78 m12	41.47 ± 0.76 lm6	93.36 ± 0.74 9	170.24 ± 1.31 o6
11 BRS 262	39.33 ± 0.57 l,10,11	45.86 ± 0.31	132.54 ± 0.70 4	217.73 ± 0.95
12 BRS 267	44.86 ± 0.17 j5,6	58.98 ± 0.20 f	151.65 ± 0.53	255.49 ± 0.59 g
13 BRS 268	44.13 ± 0.31 jk5,6,7	32.63 ± 0.15	97.90 ± 0.36 m8	174.67 ± 0.49
14 BRS 282	57.28 ± 0.47 ef	61.42 ± 0.37 2	135.88 ± 1.22 h3	254.58 ± 1.35 g
Transgenic cultivar	Campesterol	Stigmasterol	β-Sitosterol	Total phytosterol
15 BRS 242RR	50.00 ± 0.36 i4	51.72 ± 0.16	135.83 ± 0.32 h3	237.55 ± 0.5 4
16 BRS 244RR	53.49 ± 0.20 g	49.91 ± 0.19 j4	144.98 ± 0.19 f2	248.38 ± 0.33 h3
17 BRS 245RR	37.88 ± 0.29 l,11	41.90 ± 0.46 l6	126.52 ± 0.26 5	206.30 ± 0.60 5
18 BRS 246RR	42.06 ± 0.31 k8,9	49.91 ± 0.19 j4	141.22 ± 0.16	233.19 ± 0.39
19 BRS 255RR	52.70 ± 0.20 gh	57.72 ± 0.16 fg3	159.90 ± 0.20 c	270.32 ± 0.32
20 BRS 256RR	50.89 ± 0.28 hi4	55.56 ± 0.31	156.13 ± 0.16 d	262.57 ± 0.44 f

The number in brackets is the standard deviation of each measure. Means followed by the same letters do not differ by Tukey test (p < 0.05). Numbers after letters refer to the averages do not differ region by region by Tukey test (p < 0.05).

) ranged from 128.70 to 370.33 mg·100 g⁻¹ of oil, whereas samples from the Ponta Grossa region (

Table 2. Phytosterol contents (mg 100 g⁻¹ of total lipid) of soybean seeds from Ponta Grossa’s region.

Conventional Cultivar	Campesterol	Stigmasterol	β-Sitosterol	Total Phytosterol
21 Embrapa-48	68.48 ± 0.59 jk1	61.89 ± 0.31 g2	113.18 ± 1.08	243.56 ± 1.27
22 BRS 184	44.00 ± 0.20 jk5,6,7,8	48.15 ± 0.22 jk	107.07 ± 0.15 k	199.23 ± 0.64 m
23 BRS 213	62.68 ± 0.22 f2	64.74 ± 0.42 f	98.10 ± 0.43 8	225.52 ± 0.64 j
24 BRS 232	61.86 ± 0.32 2	64.71 ± 0.18 f	122.56 ± 0.36 h6	249.13 ± 0.29 3
25 BRS 233	73.21 ± 0.09 d1	77.44 ± 0.25	176.50 ± 0.36 1	327.15 ± 0.51
26 BRS 257	88.59 ± 0.12	67.22 ± 0.44 1	186.33 ± 0.24	412.14 ± 0.51
27 BRS 258	43.01 ± 0.74 k6,7,8	49.11 ± 0.18 ij4	89.14 ± 0.66	181.26 ± 1.07 n
28 BRS 259	59.35 ± 0.28 3	61.36 ± 0.32 g2	125.36 ± 0.42 5	246.07 ± 0.59 3
29 BRS 260	37.70 ± 0.20 m11	37.20 ± 0.30 m7	93.76 ± 0.26 9	168.66 ± 0.44 6
30 BRS 261	37.63 ± 0.25 m11	41.98 ± 0.19 6	100.90 ± 0.20 7	180.52 ± 0.37 n
31 BRS 262	34.03 ± 0.71 12	36.17 ± 0.31 m7	96.20 ± 0.60 8	166.40 ± 0.98
32 BRS 267	40.37 ± 0.81 k9,10	48.30 ± 0.36 jk	133.50 ± 0.46 f4	222.17 ± 0.99 k
33 BRS 268	42.70 ± 0.20 k7,8	47.58 ± 0.30 k	108.38 ± 0.28 jk	198.67 ± 0.45 m
34 BRS 282	81.13 ± 0.58	69.96 ± 0.22	144.87 ± 0.15 2	295.95 ± 0.63 1
Transgenic cultivar	Campesterol	Stigmasterol	β-Sitosterol	Total phytosterol
35 BRS 242RR	73.56 ± 0.48 d	75.62 ± 0.63	163.98 ± 1.07	313.17 ± 1.33
36 BRS 244RR	72.51 ± 0.26 d	65.62 ± 0.80 f	147.39 ± 0.45	285.52 ± 0.95 2
37 BRS 245RR	78.20 ± 0.56	72.36 ± 0.82	189.27 ± 0.42	339.83 ± 1.07
38 BRS 246RR	45.99 ± 0.20 5	57.07 ± 0.21 3	134.85 ± 0.91 f3	237.91 ± 0.95 4
39 BRS 255RR	44.43 ± 0.25 j5,6,7	49.62 ± 0.53 i4	109.00 ± 0.20 j	203.05 ± 0.62 5
40 BRS 256RR	50.61 ± 0.51 4	49.18 ± 0.40 ij4	123.72 ± 0.86 h6	223.51 ± 1.08 jk

The number in brackets is the standard deviation of each measure. Means followed by same letters do not differ according to Tukey’s test (p < 0.05). Numbers after letters refer to averages that do not differ region by region according to Tukey’s test P(p < 0.05).

) ranged from 166.40 to 412.14 mg·100 g⁻¹ of oil. These values fall within the typical range observed for vegetable oils (100–500 mg·100 g⁻¹ of oil), with a reference value for soybean oil at approximately 327 mg·100 g⁻¹ of oil [29,30].

All soybean cultivars analyzed contained campesterol, stigmasterol, and β-sitosterol. As shown in Table 1 and Table 2, β-sitosterol was the predominant phytosterol across all samples. In Ponta Grossa, β-sitosterol content ranged from 89.14 to 186.33 mg·100 g⁻¹ of oil in conventional cultivars and from 109.00 to 189.27 mg·100 g⁻¹ in transgenic cultivars. In the Londrina region, the values varied from 93.36 to 225.32 mg·100 g⁻¹ for conventional soybeans and from 126.52 to 159.90 mg·100 g⁻¹ for transgenic ones. The highest β-sitosterol concentration observed was 225.32 mg·100 g⁻¹ in the BRS233 conventional cultivar grown in Londrina. When cultivated in Ponta Grossa, the same variety presented a lower concentration (176.60 mg·100 g⁻¹). For the transgenic variety BRS255RR, β-sitosterol levels were 159.90 mg·100 g⁻¹ in Londrina and 109.00 mg·100 g⁻¹ in Ponta Grossa. Similar trends were observed for campesterol and stigmasterol content across all samples.

For comparison, β-sitosterol concentrations in Japanese and non-Japanese soybean seeds have been reported to range from 10.1 to 41.2 mg·100 g⁻¹ of seed, stigmasterol from 3.1 to 16.1 mg·100 g⁻¹, and campesterol from 4.1 to 18.1 mg·100 g⁻¹. In wheat, stigmasterol was found to be more abundant than campesterol and β-sitosterol, suggesting that genetic and geographic factors can influence phytosterol profiles [31]. Similarly, in six Tunisian olive cultivars, stigmasterol was the predominant compound, followed by β-sitosterol [32]. Conversely, a study on Turkish apricot kernel oil identified β-sitosterol as the principal compound, with stigmasterol being the least abundant [33]. These findings reinforce that phytosterol composition in vegetable oils can vary significantly depending on the food matrix.

The total tocopherol content in transgenic and conventional soybean cultivars from the Londrina region (

Table 3. Tocopherol contents (mg 100 g⁻¹ of total lipid) of soybean seeds from Londrina’s region.

Conventional Cultivar	α-tocopherol	β-tocopherol	γ-tocopherol	δ-tocopherol	Total Tocopherols
1 Embrapa-48	11.19 ± 0.25 ef	18.90 ± 0.59 bcd1	46.67 ± 0.23 bcde1	14.73 ± 0.81 defg	91.49 ± 1.06 defg
2 BRS 184	10.78 ± 0.59 fg1	6.80 ± 0.40 g2	45.08 ± 0.11 defg	15.74 ± 0.65 cde1	78.41 ± 0.97 kl2
3 BRS 213	10.32 ± 0.14 g	18.70 ± 0.70 bcd	44.37 ± 0.07 efgh	16.38 ± 2.03 bcde2	89.77 ± 2.15 fgh3
4 BRS 232	12.53 ± 0.25 c	8.10 ± 0.20 g	48.09 ± 1.10 abcd	12.97 ± 0.59 ghi3	81.69 ± 1.29 ijk
5 BRS 233	11.35 ± 0.23 ef2	11.90 ± 0.6 1efg	41.95 ± 0.90 gh	15.48 ± 0.96 efg	80.68 ± 1.47 jkl
6 BRS 257	12.91 ± 0.06 bcd	9.50 ± 0.11 fg	45.08 ± 0.23 defg	16.12 ± 0.63 bcde4	83.61 ± 0.68 hij
7 BRS 258	10.96 ± 0.07 efg	16.71 ± 0.20 bcde3	42.64 ± 0.60 gh	19.00 ± 0.56 a	89.31 ± 0.85 fgh
8 BRS 259	10.91 ± 0.11 efg3	20.50 ± 0.50 bc	50.76 ± 2.12 a	16.77 ± 0.75 bc	98.94 ± 2.3 1bc
9 BRS 260	12.71 ± 0.03 cd	20.92 ± 0.90 bc	47.53 ± 0.70 abcde	13.71 ± 0.50 fghi	94.87 ± 1.25 cdef
10 BRS 261	12.78 ± 0.07 cd4	23.03 ± 0.30 b	34.81 ± 0.26 i	16.68 ± 0.16 bcd	87.30 ± 0.43 ghi4
11 BRS 262	13.67 ± 0.61 ab1,2	15.34 ± 0.30 cdef	47.99 ± 1.12 abcd	12.35 ± 0.22 i	89.35 ± 1.33 fgh
12 BRS 267	14.09 ± 0.11 a1	33.80 ± 0.80 a	46.31 ± 0.19 cdef2	13.37 ± 0.08 fghi	107.57 ± 0.83 a
13 BRS 268	10.79 ± 0.08 fg15,1	12.33 ± 0.30 defg	42.95 ± 3.01 fgh	10.27 ± 0.24 i	76.34 ± 3.04 l
14 BRS 282	13.20 ± 0.07 bcd1,2	7.34 ± 0.20 g	41.22 ± 1.23 h	15.32 ± 0.05 cdef	77.08 ± 1.25 kl
Transgenic cultivar	α-tocopherol	β-tocopherol	γ-tocopherol	δ-tocopherol	Total tocopherols
15 BRS 242RR	12.51 ± 0.07 d	19.32 ± 0.30 bc	42.41 ± 0.7 gh	16.33 ± 0.12 bcde5	90.57 ± 0.77 efg
16 BRS 244RR	13.11 ± 0.07 bcd	22.51 ± 0.50 b	45.14 ± 0.98 defg	14.50 ± 0.12 efgh	95.26 ± 0.77 cde
17 BRS 245RR	13.47 ± 0.06 abc	8.01 ± 0.20 g	50.04 ± 1.25 ab3	17.82 ± 0.48 ab6	89.34 ± 1.11 fgh
18 BRS 246RR	13.43 ± 0.23 abc	7.03 ± 0.20 g	36.01 ± 0.98 i	12.54 ± 0.27 hi	69.01 ± 1.36 m
19 BRS 255RR	11.61 ± 0.22 e	21.4 ± 0.40 b	49.64 ± 0.61 abc	14.43 ± 0.12 efgh	97.08 ± 1.06 bcd
20 BRS 256RR	11.13 ± 0.40 ef5	32.21 ± 2.50 a	41.64 ± 0.53 h	16.89 ± 0.07 bc	101.87 ± 0.77 b

The number in brackets is the standard deviation of each measure. Means followed by the same letters do not differ according to Tukey’s test (p < 0.05). Numbers after letters refer to averages that do not differ region by region according to Tukey’s test (p < 0.05).

) ranged from 76.34 to 107.57 mg·100 g⁻¹ of oil. In the Ponta Grossa region, the same cultivars exhibited a wider range of tocopherol concentrations, from 69.52 to 115.23 mg·100 g⁻¹ of oil.

Table 3 and

Table 4. Tocopherol contents (mg 100 g⁻¹ of total lipid) of soybean seeds from Ponta Grossa’s region.

Conventional Cultivar	α-tocopherol	β-tocopherol	γ-tocopherol	δ-tocopherol	Total Tocopherols
21 Embrapa-48	13.58 ± 0.56 a	18.70 ± 0.32 bcd1	36.56 ± 0.50 jk	17.18 ± 0.15 bcdef4	94.60 ± 0.86 c1
22 BRS 184	11.10 ± 0.40 efg1	7.41 ± 0.40 fg2	38.27 ± 0.13 ij	15.24 ± 0.11 h	83.93 ± 1.28 ghi
23 BRS 213	12.17 ± 0.19 bcd	22.62 ± 0.12 abc	47.78 ± 0.56 abc	18.53 ± 0.97 ab	88.66 ± 0.74 de3
24 BRS 232	11.15 ± 0.61 defg	10.80 ± 0.60 efg	39.52 ± 0.65 hi	15.66 ± 0.22 gh	69.52 ± 1.33 j
25 BRS 233	11.26 ± 0.12 defg2	23.70 ± 0.40 ab	42.71 ± 0.51 ef	18.34 ± 0.58 abc	93.84 ± 0.53 c
26 BRS 257	11.88 ± 0.44 bcde	6.75 ± 0.70 fg	44.16 ± 1.12 de	16.29 ± 0.06 fgh	72.37 ± 0.98 j
27 BRS 258	12.43 ± 0.11 b	14.33 ± 1.10 def3	45.65 ± 0.02 cd	16.99 ± 0.04 cdefg	80.27 ± 1.12 i
28 BRS 259	11.84 ± 0.26 bcde	14.27 ± 1.70 def	40.36 ± 0.06 ghi	16.44 ± 0.09 defgh	92.42 ± 2.05 cd
29 BRS 260	10.99 ± 0.26 fg7	6.22 ± 0.30 g	48.57 ± 0.61 a	17.75 ± 0.18 abcde	72.39 ± 0.79 j
30 BRS 261	12.66 ± 0.11 ab4	10.54 ± 1.01 efg	36.56 ± 0.50 jk	17.18 ± 0.15 bcdef4	84.25 ± 1.28 fgh
31 BRS 262	11.87 ± 0.41 bcde	8.62 ± 0.60 efg	38.27 ± 0.13 ij	15.24 ± 0.11 h	80.94 ± 1.34 hi
32 BRS 267	12.46 ± 0.12 b	18.82 ± 0.40 bcd	47.78 ± 0.56 abc	18.53 ± 0.97 ab	93.92 ± 0.42 c
33 BRS 268	12.67 ± 0.05 ab	15.87 ± 0.80 cde	39.52 ± 0.65 hi	15.66 ± 0.22 gh	85.34 ± 0.81 efg
34 BRS 282	12.11 ± 0.10 bcde	10.62 ± 0.80 efg	42.71 ± 0.51 ef	18.34 ± 0.58 abc	89.05 ± 1.03 de
Transgenic cultivar	α-tocopherol	β-tocopherol	γ-tocopherol	δ-tocopherol	Total tocopherols
35 BRS 242RR	19.91 ± 0.37 fg	29.64 ± 1.20 a	47.81 ± 0.84 abc	17.87 ± 0.56 abcd5	115.23 ± 1.61 a
36 BRS 244RR	12.07 ± 0.02 bcde	8.34 ± 0.40 efg	41.95 ± 0.80 fg	11.04 ± 0.18 ij	73.40 ± 0.91 j
37 BRS 245RR	12.36 ± 0.05 bc	22.45 ± 0.50 ab	48.85 ± 1.25 a3	17.32 ± 0.21 abcdef6	101.01 ± 1.36 b
38 BRS 246RR	11.66 ± 0.11 bcdef	14.03 ± 0.40 de	45.56 ± 0.43 d	16.25 ± 0.18 fgh	92.52 ± 0.62 cd
39 BRS 255RR	10.50 ± 0.85 g	12.26 ± 0.20 de	40.85 ± 0.79 fgh	18.56 ± 0.92 ab	82.19 ± 2.18 ghi
40 BRS 256RR	11.31 ± 0.16 cdefg5	15.44 ± 1.60 cd	42.63 ± 0.60 ef	18.74 ± 1.26	88.14 ± 2.13 ef

The number in brackets is the standard deviation of each measure. Means followed by the same letters do not differ according to Tukey’s test (p < 0.05). Numbers after letters refer to averages that do not differ region by region according to Tukey’s test (p < 0.05).

present the individual tocopherol composition for all samples. Quantification of the four main tocopherol isomers—α-, β-, γ-, and δ-tocopherol—was achieved for each cultivar. Among these, γ-tocopherol was the predominant compound across all samples. The highest γ-tocopherol content was observed in the non-transgenic cultivar BRS259 grown in Londrina (50.76 mg·100 g⁻¹ of oil), while the same variety grown in Ponta Grossa contained 40.36 mg·100 g⁻¹. A similar pattern was observed for the transgenic variety BRS245RR, as well as for α-, β-, and δ-tocopherol contents. This trend has also been reported in Plukenetia volubilis (Sacha inchi), where γ-tocopherol was the predominant isomer [34]. San Andrés et al. also identified γ-tocopherol as the major form in soybean oil using a novel HPLC method for tocopherol analysis [35]. Furthermore, a study evaluating 20 lentil cultivars confirmed the predominance of γ-tocopherol in all samples [36].

Pearson correlation analyses were conducted to evaluate associations among phytosterol and tocopherol contents within the same region, as well as between different regions. Pearson’s r values range from +1 (perfect positive correlation) to −1 (perfect negative correlation), with 0 indicating no linear relationship. In the Londrina and Ponta Grossa regions, the correlation coefficients between total phytosterol and tocopherol contents were 0.0 and 0.2, respectively, indicating no meaningful association. These findings suggest that the biosynthetic pathways of phytosterols and tocopherols are likely independent. This hypothesis is supported by Yamaya et al., who reported distinct regulatory mechanisms for the synthesis of each compound class [30].

In comparisons across regions, correlation values for phytosterol and tocopherol contents were 0.3 and 0.0, reinforcing the role of environmental factors in modulating phytochemical profiles. During the soybean growing season, Londrina presented an average temperature of 22.8 °C and total precipitation of 726.6 mm, while Ponta Grossa exhibited slightly cooler and drier conditions, with 20.9 °C and 588.8 mm of rainfall [21,37,38,39]. These values, obtained during the actual soybean development period rather than annual means, offer a more accurate representation of the environmental influence on lipid biosynthesis. Such differences likely contributed to the variability observed in phytosterol and tocopherol content among cultivars grown in the two regions.

The phytochemical composition of soybean oil plays an important role in its industrial applications. Phytosterols, due to their physicochemical properties, enhance membrane fluidity and permeability in plant cells [40]. Furthermore, tocopherol profiles are directly related to the oxidative and flavor stability of vegetable oils. According to Warner, the stability of soybean and sunflower oils under oxidative conditions is strongly associated with their tocopherol composition, particularly the concentration of α-tocopherol [41].

In this study, α-tocopherol levels across all transgenic and conventional samples ranged from 10.32 to 19.91 mg·100 g⁻¹ of oil. The highest α-tocopherol concentration was observed in the BRS242RR variety cultivated in Ponta Grossa, followed by BRS267 from Londrina. However, most cultivars exhibited α-tocopherol levels below 13.7 mg·100 g⁻¹. Structurally, vitamin E compounds consist of a chromanol ring linked to a saturated (tocopherol) or unsaturated (tocotrienol) side chain. Among the tocopherol isomers, α-tocopherol is the most biologically active form due to its high free radical scavenging capacity [42].

Vitamin E activity (expressed as α-tocopherol equivalents) was calculated using conversion factors of 1.0, 0.5, 0.1, and 0.03 for α-, β-, γ-, and δ-tocopherol, respectively [43,44,45]. The resulting values ranged from 19.06 to 40.05 mg·100 g⁻¹ of oil, which were substantially higher than those previously reported for standard soybean oils [45]. The highest vitamin E activity values were observed in BRS256RR and BRS267 cultivars from Londrina, and BRS244RR from Ponta Grossa (31.91, 36.02, and 40.05 mg·100 g⁻¹ of oil, respectively). In contrast, BRS184 (19.16 mg·100 g⁻¹) and BRS260 (19.49 mg·100 g⁻¹), grown in Londrina and Ponta Grossa, respectively, exhibited the lowest activity values.

Given the complexity of the soybean matrix and the number of variables involved (phytosterols, tocopherols, cultivars, and geographic origin), Independent Component Analysis (ICA) was employed to facilitate data interpretation and reveal underlying patterns. ICA was selected over Principal Component Analysis (PCA) due to its ability to identify statistically independent and non-Gaussian sources, which better reflects the biological system under study. Phytosterol and tocopherol biosynthesis are governed by distinct metabolic pathways and regulatory gene sets—such as the mevalonate pathway for phytosterols and the shikimate pathway for tocopherols—implying biological independence. In addition, these compound classes respond differently to environmental factors such as temperature, rainfall, and photoperiod. Such biological and environmental dissociation supports the core assumptions of ICA and justifies its use over PCA, which relies on variance maximization but does not assume independence among sources. Thus, ICA provides a more biologically coherent and statistically suitable multivariate framework for this dataset.

An ICA model with nine independent components (ICs) was applied to the dataset. As illustrated in Figure 1, each IC corresponds to an extracted source signal and its respective score evolution, allowing for the discrimination of soybean cultivars based on phytochemical content.

The extracted source signals (Figure 1A) indicate that the independent components (ICs) one through nine correspond to total phytosterol, β-sitosterol, total tocopherol, stigmasterol, campesterol, γ-tocopherol, β-tocopherol, δ-tocopherol, and α-tocopherol, respectively.

The score plots (Figure 1B) revealed clear discrimination among soybean cultivars. IC1 was associated with the conventional cultivar BRS257 from Ponta Grossa, which exhibited the highest total phytosterol content. IC2 highlighted BRS233 from Londrina as the main contributor to β-sitosterol concentration. IC3 showed that the transgenic cultivar BRS242RR from Ponta Grossa had the highest total tocopherol content. IC4 identified BRS232 from Londrina as presenting the greatest stigmasterol content. IC5 indicated that both BRS232 from Londrina and BRS257 from Ponta Grossa shared similar levels of campesterol. In IC7, BRS267 from Londrina, BRS256RR from Londrina, and BRS242RR from Ponta Grossa displayed elevated β-tocopherol levels. IC9 revealed that BRS242RR from Ponta Grossa also exhibited the highest α-tocopherol concentration.

These results suggest biochemical similarities among cultivars from different regions. For example, BRS232 and BRS257 exhibited comparable campesterol profiles, while the transgenic varieties BRS256RR and BRS242RR showed β-tocopherol contents similar to that of the conventional cultivar BRS267. The transgenic cultivar BRS242RR from Ponta Grossa stood out for its high levels of total tocopherol, β-tocopherol, and α-tocopherol. The conventional cultivar BRS257 from Ponta Grossa was notable for its elevated total phytosterol and campesterol content, whereas BRS232 from Londrina showed a predominance of stigmasterol and campesterol.

In contrast, IC6 and IC8 identified cultivars with lower levels of γ- and δ-tocopherol, respectively. Specifically, BRS261 from Londrina exhibited the lowest γ-tocopherol content, while BRS262 from Londrina and BRS244RR from Ponta Grossa had the lowest δ-tocopherol concentrations.

4. Conclusions

This study demonstrates the feasibility of using Independent Component Analysis (ICA) to discriminate soybean cultivars based on their phytosterol and tocopherol profiles. The extracted independent components effectively distinguished varieties with higher concentrations of total phytosterol, β-sitosterol, total tocopherol, stigmasterol, campesterol, β-tocopherol, and α-tocopherol. The results highlight that transgenic cultivars from the Ponta Grossa region exhibited elevated levels of total tocopherol, β-tocopherol, and α-tocopherol. In contrast, conventional cultivars from both the Ponta Grossa and Londrina regions were characterized by higher phytosterol content. Overall, ICA proved to be a powerful tool for revealing underlying biochemical variability among soybean cultivars and offers valuable insight for targeted selection based on functional lipid composition.

It is important to note that this study was conducted using soybean samples harvested in a single agricultural year. Therefore, while the Independent Component Analysis (ICA) provided clear discrimination of cultivars based on phytochemical profiles, long-term assessments across multiple growing seasons would be valuable to validate the consistency of these patterns. Future studies may also explore the application of ICA in conjunction with other multivariate techniques such as supervised learning algorithms, aiming to enhance predictive power and robustness in cultivar discrimination models.