Influence of bamboo fiber on the technological and nutritional qualities of semi-hard plant-based cheese alternative

Abstract

This study aimed to develop a plant-based cheese alternative with an improved nutritional profile and a texture similar to traditional Italian Caciotta cow’s cheese. The formulation combined a 1:1 blend of chickpea and oat concentrates (PCs), bamboo fiber (BF), and water. A simplex-centroid mixture design was employed to examine how varying these components (PCs: 10–20 g 100 g⁻¹, BF: 0–10 g 100 g⁻¹, Water: 60–70 g 100 g⁻¹) affected the textural and cutting properties of the product. Two formulations closest to Caciotta in texture were further tested for their nutritional and sensory attributes using the Just About Right scale. Results revealed BF and water levels significantly influenced the texture and cutting parameters. Both formulations exhibited high proteins (9.63–10.21 g 100 g⁻¹) and fibers (12.46–15.63 g 100 g⁻¹), alongside low fats (4.07–4.20 g 100 g⁻¹) and phytic acids (0.40 g/100 g). Despite their grainy and soluble nature and being less elastic and hard than Caciotta, one formulation, PCs: 16.7 g 100 g⁻¹, BF: 6.7 g 100 g⁻¹, and water: 66.7 g 100 g⁻¹, was favored in sensory evaluations. This study highlights bamboo fiber's role in achieving the desired firmness and fiber content, offering a promising path for developing functional foods that imitate the texture and nutritional profile of semi-hard cheeses.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13197-025-06206-1.

Introduction

Semi-hard cheeses, known for their versatility in consumption (be it cooked or not, standalone, or as a topping) present a significant opportunity for innovation in the plant-based food sector. Despite this potential, current plant-based (PB) cheese alternatives often fall short of consumer expectations, particularly in taste and texture; commonly described as gelatinous, mouthcoating, and rubbery (Falkeisen et al. 2022). These textures result from overusing of thickeners, saturated fats, and starches. This issue is reflected in the high failure rate of new PB foods, ranging from 50 to 75% (Rini et al. 2023). The market does offer affordable PB cheeses with a firm consistency. Still, these mostly fall into the category of ultra-processed food and contain high amounts of solid fats like shortening, palm, and coconut oils. The consumption of such plant-based rich in saturated fats has been associated with metabolic syndrome (Demmer et al. 2016), highlighting the need for healthier alternatives. This shift towards health-conscious food choices; in particular reduced fat, sodium, and synthetic additives, has ignited interest in exploring non-conventional ingredients (Mefleh et al. 2023). For instance, Moon et al. (2021) successfully utilized an oleogel made from canola oil and carnauba wax to replace palm oil in a PB hard cheese formulation.

In the present study, we aimed to formulate a low-fat, PB cheese with a balanced amino acid profile and a semi-hard texture. We utilized two dry-fractionated (DF) protein concentrates, chickpeas, and oats, with complementary amino acid compositions, alongside bamboo shoot dietary fiber as the primary ingredients. Bamboo fiber is known for its insolubility, tastelessness, and calorie-free nature and previous research has highlighted the significant benefits (on texture and as a means of fiber enrichment) of incorporating BF into a diverse range of foods, including dairy products (Zheng et al. 2017), meat and sausages (Li et al. 2020) and frozen dough (Zhang et al. 2017). Additionally, compared to soybean and rice bran fibers, bamboo fibers possess higher glucose retardation index and greater water and oil holding capacities (Zhang et al. 2017). These properties are crucial in maintaining the structure and consistency of PB products. For instance, adding BF to yogurt significantly reduced syneresis even after 21 days of storage at 4℃ (Zheng et al. 2017) and its incorporation into pork meat batter improved elasticity and viscosity (Li et al. 2020). Bamboo fiber was also shown to improve the gel's texture of wheat gluten, the dough microstructure, bread texture, and in vitro digestibility (Wang et al. 2017; Zhang et al. 2017).

This study addresses a notable gap in existing research by exploring the impact of bamboo fiber on the technological properties of PB cheese. We aimed to create an innovative semi-hard cheese alternative, utilizing chickpea and oat concentrates, with the texture of an Italian semi-hard cows’ cheese 'Caciotta' as our benchmark. To optimize the ingredient proportions for desired textural and cutting properties, we employed a simplex-centroid mixture design which enabled us to identify two formulations that most closely mirrored the characteristics of Caciotta. These selected formulations were then evaluated through nutritional and sensory assessments.

Materials and methods

Materials

The DF protein concentrate (dry-fractionated raw micronized flour) from dehulled Apulian black chickpeas (56% protein) and oats (55% protein) were sourced from Innovaprot SRL, Italy, and Proatein Lantmannen, Sweden, respectively. Among different cereals and pseudo-cereals tested (oats, wheat, barley), oats were selected based on preliminary sensory trials as the chickpea-oat combination provided optimal slurry homogenization and end-product cohesivity. Bamboo fiber (97% insoluble fiber) was acquired from Molino Bongiovanni S.r.l., Italy. This white, odorless powder had 0.30% carbohydrates, 0.10% protein, and 0.2% fat. Agar agar and lactic acid were bought from Pure Food Milano, Italy, and other ingredients like nutritional yeast flakes (deactivated yeast), salt, and sunflower oil from a local store in Bari, Italy. Nutritional yeast flakes and lactic acid were added for flavor enhancement. Reference 'Caciotta' cheese, used for texture comparison, was sourced locally, containing 54.2% moisture, 12.0% fat, and 18.0% protein.

D-optimal mixture design

The experimental design, using the software Design-Expert 11 (StatEase Inc., Minneapolis, USA), was planned to study the formulation by combining the three main ingredients i.e., the protein concentrates (PCs) (50:50 w/w of chickpea protein and oat protein concentrates), bamboo fiber (BF), and drinking water. In particular, a simplex centroid mixture design composed of seven experimental points was chosen according to a linear model, and to compute each response, the linear regression models were used according to the following equation: Y = β1 × 1 + β2 × 2 + β3 × 3, where Y represents the response variable, x is the components studied and β the model coefficients. The mixture design was arranged according to the following constraints (g/100 g): PCs (10 ≤ × 1 ≤ 20); bamboo fiber (0 ≤ × 2 ≤ 10); water (60 ≤ × 3 ≤ 70). The sum of the three ingredients was 90 g/100 g, whereas the other ingredients contributed to the remaining 10 g/100 g of the formulation. They were kept constant, and they consisted in agar agar, 1.5 g/100 g, sunflower oil, 2.5 g 100 g⁻¹, lactic acid, 1.0 g 100 g⁻¹, nutritional yeast, 4.0 g 100 g⁻¹, salt, 1.0 g 100 g⁻¹ (Supp Table 1).

Table 1.

Regression coefficients of the model; X1: PCs; X2: BF; X3: water, for TPA, and cut test parameters and water activity (a_w) measured on the formulated samples produced by the D-Optimal mixture design. p-value is significant when ≤ 0.05

	X1	X2	X3	R2	Adjusted R2	p-values
Firmness (N)	13.49	11.32	37.15	0.98	0.98	< 0.001
Springiness	− 0.56	− 0.84	0.73	0.21	− 0.02	0.44
Chewiness (N)	8.71	11.32	− 1.55	0.71	0.63	0.01
Cohesiveness	0.17	0.16	0.19	0.38	0.20	0.19
Shear Force (N/cm2)	6.95	3.93	14.64	0.99	0.99	< 0.001
aw	0.98	0.98	0.98	0.18	− 0.05	0.50

Methods of preparation

Half of the tap water quantity (X3) was used to hydrate the PCs (X1) while the other half was heated (40 ℃) and used to dissolve the agar agar and bamboo fiber (BF) (X2). The two compositions were mixed and stirred for 2 min with an Ultra-Turrax (IKA T 25 digital ULTRA-TURRAX®- Germany) at 20,000 rpm. During mixing lactic acid, sunflower oil, salt, and nutritional yeast were added. The homogenized slurry was transferred into sterile 300 mL glass jars which were then treated in a vertical Steristeam II autoclave at 121 ℃ for 15 min to stabilize the slurry from a microbiological point of view and to reduce the phytic acid content, considered an anti-nutritional factor (Mefleh et al. 2023), found in the chickpeas (0.78 g 100 g⁻¹) and oat (1.20 g 100 g⁻¹) concentrates. After cooling down at room temperature, samples were added to silicon square-shaped containers and stored at 4 ℃. The samples were analyzed after 24 h of refrigeration (Supp Fig. 1). The pH of all formulations was 4.40 ± 0.10.

Fig. 1.

Contour plots for firmness (N), chewiness (N), springiness, cohesiveness and shear force (N/cm²) produced by the D-Optimal mixture design

Water and oil absorption capacities

The water absorption of PCs and BF was measured using 3.0 g samples in 25 mL of water, vortexed and shaken for 30 min, then centrifuged at 3000 × g for 25 min. After discarding the supernatant, the pellets were dried at 50 °C for 25 min and reweighed. For oil absorption, 0.50 g samples were mixed with 6 mL of olive oil, vortexed, shaken for 30 min, centrifuged, and drained for 25 min before reweighing. Absorption capacities was calculated as grams of water/oil per gram of sample, with each sample tested twice.

Texture analyses

Texture profile analysis (TPA) and Cut test were determined on square-shaped samples (2 cm × 2 cm × 2 cm) directly after removal from the cell at 4 °C, using a Z1.0 TN texture analyzer (Zwick Roell, Ulm, Germany) with a 50 N load cell. Analyses were repeated 3 times on each formulation.

For the TPA analysis, a 36 mm diameter cylindrical probe was utilized. TPA conditions in the cyclic compression test were: 1 mm s⁻¹ compression; 50% sample deformation in both compressions; and 5 s pause before the second compression. The reported values represent the averages of six replicates. Data were acquired with TestXPert II version 3.41 software (Zwick Roell, Ulm, Germany).

For the Cut test, the blade of 50 mm length descended at a speed of 1 mm/s until the sample was completely cut. The shear force, i.e., the maximum force applied on the surface until the sample was completely cut (N/cm²), was measured. This is taken as an index of firmness, toughness or fibrousness. Samples were loaded on a plate and compressed, imitating the shear resistance force when the food was being cut by a knife.

Based on the results of the texture analysis, two samples were selected, based on their significant similarities with the Caciotta reference, and were subjected to nutritional and sensorial analyses.

Chemical analyses

Water activity was measured using Aqua Lab 4TE (Meter Group Inc., Pullman, WA, USA). pH was recorded using a pH meter (edge® HI2002, HANNA instruments, OH, USA). Moisture content was calculated by drying samples at 105 ℃ using MAC 110/NP RADWAG (Radom, Poland). Ash content followed the AACC Method 08-0.1.01 (Mefleh et al. 2024). Lipid content was determined by the AACC Soxhlet method 30-25.01 (Mefleh et al. 2024), and saturated fats were measured as per EEC 2568/91 (European Union Commission 1991). Total dietary fibers were quantified by AOAC method 991.43 (Association of Official Agricultural Chemists (AOAC) International, 2006). Protein content (N × 6.25) was determined with AACC Kjeldhal Method 46-13.01 (Mefleh et al. 2024). Total carbohydrate (g 100 g⁻¹ on dry matter) was calculated as 100—(ash + fat + protein), and total energy from proteins in kilocalories was estimated using Atwater factors.

Amino acids (AA) composition was analyzed post-acid hydrolysis using HCl 6N with 1% phenol (w/v) from Merck KGaA, Darmstadt, Germany, as described by De Angelis et al. (2022). This process involved a Pico-Tag workstation and glass reaction vials from Waters Corporation, Milford, USA. Briefly, a glass vial containing 10 mg of the sample was placed inside a glass reaction vial and 200 µL of HCl 6N with 1% phenol (w/v) were added to the bottom of the reaction vial. The entire system was sealed under nitrogen to prevent oxidation of the AA and the acid hydrolysis was carried out at 116 °C for 24 h.

Hydrolyzed AA were extracted with 5 mL of distilled water and analyzed using high-performance liquid chromatography (HPLC) (using an Agilent 1260 with Diode-Array-Detector and Poroshell 120, HPH-C18 column (Agilent Technologies, USA) after in-line derivatization with ortho-phthalaldehyde/9-fluorenyl methyl chloroformate (OPA/FMOC) reagents. The quantification of AA (expressed as mg/g of protein) was performed using external calibration curves prepared with AA standards at concentrations of 100, 250, and 1000 pmol. The analysis was conducted in duplicate.

The amino acid score (AAS) for adults was calculated as AA (mg g⁻¹ of protein) / Reference mg g⁻¹ for older children, adolescents, and adults (FAO/WHO/UNU 1991).

Sensorial analysis – JAR

To achieve a suitable texture, we compared experimental samples with fresh cows’ Caciotta cheese. Sensory analysis involved a trained, gender-balanced panel of 11 members aged 20–50, following the ISO 8586–1993 method. Training on Caciotta cheese was carried out before the sensory tests and helped the assessors to identify 12 different texture attributes (Supp Table 2). During testing, three experimental samples were weighed (10 g each) in labeled plastic dishes and were evaluated using a balanced design. The panelists used the JAR (Just-About-Right) scale to rate the texture attributes. The 5-point scale’s middle value represented the ideal texture, with higher values indicating excess and lower values indicating deficiency relative to Caciotta. Deviations from the JAR point signified texture differences from the target. Judges also rated overall liking on a 1–9 hedonistic scale. Triplicate analyses informed Penalty Analysis for texture improvement (Gere 2022).

Table 2.

Component Effects (Cox) (p ≤ 0.05) of X1: PCs; X2: BF; X3: Water, for the color, TPA, and cut test parameters measured on the formulates samples produced by the D-Optimal mixture design

	X1	X2	X3
Firmness (N)	14.92	15.82	− 19.43
Springiness	− 0.17	− 0.09	0.19
Chewiness (N)	0.71	0.98	− 0.91
Cohesiveness	0.01	0.02	− 0.02
Shear force (N/cm2)	4.70	6.88	− 6.84
aw	0.00	0.00	0.00

Statistical analysis

The results of the experimental design were modeled according to the postulated linear model. The coefficient of determination (R²), the adjusted coefficient of determination (R² adj) as well as the significance (p ≤ 0.05) of the regression model and the coefficients were calculated per each of the considered responses by the software Design-Expert 11 (StatEase Inc., Minneapolis, USA). All data were further subjected to one-way ANOVA followed by Tukey’s HSD (honestly significant difference) test at α = 0.05 using R software. The differences between the samples and the control were evaluated using Dunnett’s multiple comparisons test at α = 0.05. Data were expressed as means ± SD.

Regarding the sensorial test, penalty analysis and the percentages for the JAR levels of sensory analysis results were performed by Xlstat (Addinsoft, France).

Results and discussion

Mixture design results

The coefficients of the regression models calculated for the responses analyzed on the different formulations of the cheese alternatives are reported in Table 1, together with the calculated coefficients of determination and the significance of the ANOVA analysis of the regression. These results indicate that the proposed model effectively described most of the textural parameters. Particularly, the models for firmness, chewiness, and shear force were proven significant by ANOVA (p ≤ 0.05) (highlighted in bold in table 2), showing from acceptable to excellent values of adjusted R². The lack of significance for the other parameters can be related to the small variability observed in such responses, indicating a negligible effect of the three components. However, the sole model coefficients, in the mixture context, do not allow drawing a clear framework about the factor’s effect on the responses, and thus the observation of the contour plots (Fig. 1) is mandatory (De Angelis et al. 2022; Squeo et al. 2021). Together, complementary useful metrics to understand the effect of the components were evaluated by the calculation of the Cox’s effects of each component (De Angelis et al. 2022; Squeo et al. 2021) which are reported in Table 2.

An integrated analysis of the contour plots (Fig. 1) and the effects of the components (Table 2), reveals that both PCs and BF positively influenced the firmness, chewiness, and shear force of the cheese alternatives. This correlation is in line with expectations, particularly given BF’s superior water absorption capacity (3.73 ± 0.26 g g⁻¹ of sample) compared to oat (1.61 ± 0.01 g g⁻¹ of sample) and chickpea concentrate (0.63 ± 0.02 g g⁻¹ of sample).

A previous study by K. Li et al. (2020) has demonstrated BF’s ability to bind up to five times its weight in water, making it an effective ingredient for addressing moisture migration in high- and intermediate-moisture foods. As indicated in Table 2, BF’s impact is slightly more pronounced than that of PCs, specifically for chewiness and shear force. Conversely, water exhibited a notable negative effect on the structural properties of the cheese. Interestingly, the magnitude of this effect was found to be similar to that of BF for chewiness and shear force, as shown in Table 2.

Regarding springiness and cohesiveness, which reflect a food’s ability to return to its original form after deformation, our study found that none of the three components significantly affected these properties in the PB cheese, as indicated by minimal variability across trials (Fig. 1 and Table 2). This finding contrasts with Zheng et al. (2017) who reported that adding BF (2 g 100 g⁻¹) to milk pudding improved hardness but reduced cohesiveness. The difference in outcomes may be attributed to the initial viscosity of the products; the milk pudding in Zheng et al.’s study was liquid before the addition of BF. Therefore, it appears that BF is more effective in less viscous products.

Texture and shear force

Table 3 details the mean values, standard deviations, and significant differences in textural properties across the seven formulations we analyzed, along with their comparative analysis to the reference, that is Caciotta cheese (Supp Fig. 2). Among these, only formulations T5 (with PCs: 15 g 100 g⁻¹, BF: 10 g 100 g⁻¹, and water: 65 g 100 g⁻¹) and T7 (PCs: 16.7 g 100 g⁻¹, BF: 6.7 g 100 g⁻¹, Water: 66.7 g 100 g⁻¹) achieved a firmness level similar to the control. However, it’s notable that none of the formulations developed through the mixture design process successfully replicated the springiness, chewiness, cohesiveness, and shear force characteristic of Caciotta cheese. These textural attributes were significantly lower in our PB versions compared to the control. This outcome underscores the challenges in mimicking the semi-hard texture of traditional cheese using solely PB materials, especially without the inclusion of solid fats (Mefleh et al. 2023).

Table 3.

Textural parameters, as measured by the TPA and cut test and water activity (a_w) measured on the formulated samples and the commercial ‘caciotta’

	T1	T2	T3	T4	T5	T6	T7	CACIOTTA†
Firmness (N)	26.5 ± 1.57b	13.6 ± 2.74cd	37.3 ± 2.74a	13.8 ± 1.33cd	25.4 ± 3.2b†	10.9 ± 0.59d	18.0 ± 1.04c†	21.7 ± 1.17
Springiness	0.27 ± 0.02a	0.30 ± 0.03a	0.31 ± 0.01a	0.37 ± 0.06a	0.30 ± 0.03a	0.26 ± 0.05a	0.42 ± 0.15a	0.90 ± 0.07
Chewiness (N)	1.25 ± 0.23b	0.72 ± 0.05b	2.57 ± 0.28a	1.04 ± 0.06b	1.53 ± 0.24b	0.31 ± 0.07c	1.00 ± 0.04b	33.01 ± 3.09
Cohesiveness	0.19 ± 0.06a	0.15 ± 0.01ab	0.20 ± 0.02a	0.16 ± 0.01ab	0.17 ± 0.01ab	0.13 ± 0.02b	0.19 ± 0.00a	0.63 ± 0.03
Shear Force (N/cm2)	9.08 ± 1.45bc	7.00 ± 1.99cd	15.2 ± 2.07a	5.54 ± 1.00de	10.1 ± 0.44b	3.70 ± 0.43e	8.54 ± 0.68bcd	24.6 ± 2.84
aw	0.98 ± 0.00a	0.98 ± 0.00a	0.98 ± 0.00a	0.98 ± 0.00a	0.98 ± 0.00a	0.98 ± 0.00a	0.98 ± 0.00a	0.95 ± 0.00

Means with the same letter are not statistically different with Tukey’s test at p ≤ 0.05

^†Samples similar to the reference ‘caciotta’ at the Dunnett test for p ≤ 0.05

Fig. 2.

Upper graphes: Percentages for the JAR scale of sensory analysis results of T5 (PCs: 15, BF: 10, Water: 65) -the upper left- and T7 (PCs: 16.7, BF: 6.7, Water: 66.7) -the upper right. Lower graphes: Penalty analysis (Mean drops vs. % of judges) measuring the drop of liking score based on “too much” and “too little” parameters for T5 (PCs: 15, BF: 10, Water: 65) -the lower left- and T7 (PCs: 16.7, BF: 6.7, Water: 66.7)- the lower right

In our study, the springiness and cohesiveness of the seven formulated samples, as detailed in Table 3, were within the range of hard PB cheeses reported by Moon et al. (2021).

The analysis of the 7 samples revealed that T3, with the highest ratio of PCs plus fiber to water content, exhibited significantly higher firmness, chewiness, cohesiveness, and shear forces (p < 0.05). In contrast, T6, composed only of PCs and water, showed the lowest values for these textural parameters. This supports the findings of Zhang et al. (2017) who demonstrated that adding BF at even low levels (1–2 g 100 g⁻¹) significantly improves the rheological properties (viscoelasticity, extensibility, and plasticity) of frozen wheat dough. The beneficial impact of BF on texture is further substantiated by comparing T2 and T6: all textural values in T2 (PCs: 10 g 100 g⁻¹, BF: 10 g 100 g⁻¹, Water: 70 g 100 g⁻¹) were markedly reduced in T6 (PCs: 20 g 100 g⁻¹, BF: 0 g 100 g⁻¹, Water: 70 g 100 g⁻¹).

The cut force is a critical parameter in mimicking cheese texture, it reflects the resistance of the food to deformation and the force required to break it. With the domain of X used in this study, the samples did not reach the shear force level of Caciotta. Increasing the BF content beyond 10 g per 100 g might positively affect the shear forces; for instance, the shear force of T2 is twice that of T6. However, this adjustment could potentially lead to an excessive increase in the firmness of the PB cheese.

Hard and semi-hard cheeses typically exhibit a reduced susceptibility to microbial spoilage, primarily due to their relatively low moisture content (below 50 g 100 g⁻¹) and slightly acidic pH values (around 5.5). In our study, the pH of the PB cheese alternatives was deliberately kept below the threshold favorable for bacterial growth, at an acidic level of 4.5. This pH is notably lower than the typical range for traditional Caciotta cheese, which is 5.10 to 5.50 (Andersen et al. 2023). However, the moisture content (60–68 g 100 g⁻¹) and water activity (a_w) of 0.98 in our formulations were significantly higher than those in Caciotta, indicating an increased risk of microbial spoilage for the PB product.

In addition to its role in microbial growth, a_w affects the physical properties of food, such as texture and stability. A decrease in a_w generally leads to heightened friability and reduced softness in food products (Cauvain and Young 2008). These observations emphasize the critical role of a_w in food preservation and quality, highlighting the need for storage at low temperatures, to minimize spoilage risks and maintain desired textural properties.

Due to their close resemblance in firmness to traditional Caciotta cheese, formulations T5 and T7 were selected for further in-depth characterization.

Nutritional composition and food claims

The growing demand for high-protein PB food has spurred extensive scientific research into plant protein sources, their functionality, and potential food applications. The chemical analysis of the plant-based products T5 and T7 (Table 4) indicated that over 20% of their calories are from protein, making them eligible for the ‘high in protein’ claim (The European parliament and the council of the European 2006). All samples were low in saturated fat (SF < 1.5 g 100 g⁻¹) (The European parliament and the council of the European 2006). The total fat content in these samples, varying between 4.07 and 4.20 g 100 g⁻¹, was also substantially lower than that in Caciotta, which stands at 12 g 100 g⁻¹ (according to the nutrition facts reported in the label). Our formulations, T5 and T7, showcased a nutritional profile with higher protein and lower fat content compared to the semi-hard and hard PB cheeses developed by Grasso et al. (2021) and Moon et al. (2021). This notable difference in their nutritional composition underscores the positioning of the formulated products within the PB cheese market as alternatives that offer a higher protein content while maintaining a lower fat profile.

Table 4.

Proximate composition and phytic acid in g 100g⁻¹ of the two selected plant-based products T5 and T7

	T5	T7
Moisture	62.9 ± 0.01b	65.9 ± 0.06a
Ash	2.12 ± 0.01a	2.10 ± 0.03a
Carbohydrates	21.3 ± 0.06a	17.6 ± 0.01b
From which Total dietary fibers TDF	15.6 ± 0.85a	12.5 ± 0.96b
Phytic acid	0.40 ± 0.03a	0.40 ± 0.04a
Protein	9.63 ± 0.01b	10.2 ± 0.06a
Fat	4.07 ± 0.14a	4.20 ± 0.09a
From which SF	0.62 ± 0.02a	0.65 ± 0.04a
% of kCal from Proteins	24.0 ± 0.14b	27.4 ± 0.25a
% of kCal from UF	19.4 ± 0.78b	21.5 ± 0.64a

Means with the same letter are not statistically different with Tukey’s test at p ≤ 0.05

Notably, these formulations also boasted a high fiber content, surpassing the threshold of 6 g 100 g⁻¹, which allows them to be labeled as ‘high-fiber’ (The European parliament and the council of the European 2006). This aligns with the growing trend of incorporating fibers into foods, a movement gaining momentum for its health benefits (Brandl et al. 2022). Food rich in fibers are increasingly recognized as functional foods, attracting health-conscious consumers (Baker et al. 2022). Our nutritional results confirm the potential to develop an innovative and functional plant-based cheese using a mix of legume and cereal concentrates and BF.

The phytic acid content of the raw materials (0.78 g 100 g⁻¹ in chickpeas and 1.21 g 100 g⁻¹ in oat concentrates) was lower than the levels observed in dry seeds of soybeans (Glycine max), yellow shelled peas (Pisum sativum) and lentil (Lens esculentum) which ranged from 1.28 to 1.86 g 100 g⁻¹ as reported by Vojtíšková et al. (2010). Interestingly, the phytic acid levels in our samples T5 and T7, as shown in Table 4, were also lower than those found in a black chickpea-based beverage fermented with Lactococcus lactis (0.61 g 100 g⁻¹) (Mefleh et al. 2022). This reduction in phytic acid content could be attributed to the application of high heating temperatures (121 ℃) during processing, a method known to decrease phytic acid in legumes and cereals (Mefleh et al. 2023).

Blending legume and cereal proteins, as in our formulations, results in a comprehensive and high-quality essential amino acid profile, comparable to that of animal proteins (Mefleh et al. 2023). T7, with its higher protein content than T5, was selected for detailed amino acid analysis, as presented in Supplementary Table 3. This blend of chickpeas and oats effectively supplies methionine and lysine, typically deficient in legumes and cereals, respectively (Mefleh et al. 2023). However, a notable absence was cysteine, possibly a result of the high-temperature heating process triggering the Maillard reaction, which involves sulfur amino acids, such as methionine and cysteine, in flavor formation (Feiner 2006). The nutritional quality of plant proteins is often assessed by comparing their essential amino acid content against a reference standard set by the World Health Organization (FAO/WHO/UNU 1991). Our PB cheese contained all essential amino acids in proportions exceeding 100% of the AA score, with the exceptions of methionine + cysteine, and lysine, as detailed in supplementary Table 3. Considering that casein, the primary protein in milk is low in cysteine and has a moderate amount of methionine (Sarode et al. 2016), our developed cheese alternative can be considered a promising substitute for fresh cheese in terms of protein content (10 g 100 g⁻¹) and amino acid composition. Tryptophan analysis was omitted from this study due to a methodological issue with its hydrolysis, identified late in the study. As tryptophan is not typically deficient in chickpeas or oats, its exclusion did not affect the primary focus on assessing methionine, cysteine, and lysine deficiencies.

Just about right (JAR) sensorial analysis

Figure 2, upper graphs, displays the sensory panel’s evaluation of T5 and T7’s texture, with the x-axis showing texture attributes and the y-axis indicating percentage scores. Red, blue, and green colors represent ‘too much,’ ‘too little,’ and ‘Just About Right’ (JAR) scores, respectively, with JAR indicating similarity to reference Caciotta cheese. This method highlights significant deviations from the JAR value, guiding potential improvements.

Attributes like friability, adhesiveness, solubility, elasticity, deformability, and graininess often deviated from JAR. The smaller size of plant proteins, affecting aggregation and gelation, contributed to undesirable graininess sensation (Chen et al. 2019). Additionally, the use of bamboo, a known thickener in less firm products, has been shown to increase gumminess and hardness while reducing elasticity (Lopes et al. 2023). In our study, this was observed for the “hardness” of T5, which may be also attributed to its lower water content compared to T7.

T5 (Fig. 2, upper left graph) scored high in JAR for oiliness, chewiness, moisture, and graininess but differed significantly in graininess and deformability from the Caciotta reference, as perceived by all judges. Adhesiveness and solubility were mostly rated ‘too much’, while friability, lumpiness, and elasticity were ‘too little.’ T5’s hardness was comparable to Caciotta for 36% of panelists, aligning with the Texture Profile Analysis (TPA) results.

T7 (Fig. 2, upper right graph) scored well in oiliness, moisture, chewiness, and graininess. It needs improvements in graininess, deformability, solubility, and adhesiveness, with its high solubility reflecting the lower shear forces observed in the textural analysis. The addition of dietary fiber over 5 g 100 g⁻¹ has been linked to increased oil holding capacity (OHC), contributing to homogeneity and improved mouthfeel (Pinton et al. 2022). In our study, the OHC of BF (4.21 ± 0.31 g g⁻¹of sample) was significantly higher than that of oat (1.10 ± 0.06 g g⁻¹ of sample) and chickpea (0.99 ± 0.06 g g⁻¹ of sample) concentrates, and also higher than various legumes studied by Jamalullail et al. (2022). This led to the expectation that higher BF content would enhance the sensory experience, particularly in oiliness. However, T7 showed better responses than T5 in these aspects.

The penalty analysis (Fig. 2, lower graphs) assessed the impact on liking scores based on the difference between the Mean of Liking for Just-About-Right (JAR) and the combined mean for the other levels. If a level (too much, JAR, too little) wasn’t chosen by judges, no penalty was calculated. A 100% score in “too much” or “too little” meant the attribute didn’t match the reference. The T7 sample’s chewiness showed a significant penalty of 1.8 points for not meeting expectations (p < 0.05). Figure 2, lower graphs, provided further insights, illustrating the mean liking drop for non-JAR levels, calculated as the difference between the JAR level’s mean liking and the extreme levels “too much” or “too little”. The dotted line, perpendicular to the x-axis, indicates the 20% threshold above which values are significant, with “ + ” or “−” symbols denoting attributes greater or lesser than the reference (Gere 2022). The upper right quadrant presents the most penalizing attributes requiring improvement.

Over 90% of judges penalized T5 for higher adhesiveness (0.4) and solubility (0.4) (Fig. 2, lower left graph). Hardness was the most penalizing (1.35 mean drop), followed by graininess (1) and low chewiness (0.95).

Conversely, T7’s high solubility led to a 1-point drop for most judges (> 90%); it was softer and less lumpy than the reference (> 70%), with penalties of 0.8 and 0.5, respectively (Fig. 2, lower right graph). High graininess and low chewiness were also significant parameters for the judges, with a penalty of 0.66 and 0.28 points on the liking score. Also significantly impacted T7, with penalties of 0.66 and 0.28 on the liking score. T7 showed better potential for future texture refinement, with favorable scores in oiliness, chewiness, moisture, and graininess, but needing improvements in other areas.

Conclusion

This study represents a significant advancement in plant-based (PB) cheese alternatives, successfully developing semi-hard cheeses using chickpea, oat concentrates, and bamboo fiber (BF) that closely replicate the firmness of traditional ‘Caciotta’ cheese. The formulations not only meet European standards for “high in protein” and “high-fiber” claims but also maintain low saturated fat content, making them particularly appealing to health-conscious consumers. By achieving the desired firmness, the study effectively replicates a key textural characteristic of traditional cheese while delivering a balanced nutritional profile that aligns with modern health standards and the growing demand for sustainable, nutritious choices.

Despite these successes, the study acknowledges the ongoing challenges in fully replicating the sensory profile of dairy cheese, particularly in terms of springiness, chewiness, and cohesiveness. Sensory evaluations highlighted areas for improvement, such as graininess, adhesiveness, and solubility, suggesting potential avenues for refinement. Future work could explore the incorporation of alternative plant proteins, such as pea or rice, and innovative techniques like lactic acid bacteria fermentation, which has shown promise in enhancing texture and reducing graininess (Mefleh and Darwish 2024).

These findings lay a robust foundation for future innovations in PB cheese, offering valuable insights into ingredient and process refinements that could further enhance quality and consumer appeal in the rapidly evolving plant-based dairy market.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization and design: M.M., C.S, and M.F.; data acquisition, analysis, and interpretation: M.M., G.S., D.D., and G.N.; Drafting the manuscript: M.M., D.D., and G.N.; writing—review and editing: M.M., C.S., G.S., A.P., F.C., and M.F.; Supervision: M.M., C.S., and M.F.

Funding

Not applicable.

Data availability

Data will be available upon request.

Code availability

Not applicable.

Declarations

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

The sensory analysis was performed by a trained panel composed of 11 members (according to ISO 8586–1993 method).

Consent to participate

Not applicable.

Consent for publication

Not applicable.