An Optimal Probiotic Carrier: Multiple Steps Toward Selection and Application in Kombucha

Abstract

Kombucha is widely recognized as a functional beverage with potential probiotic effects, yet maintaining probiotic viability remains a challenge due to the harsh conditions of fermentation. This study focuses on optimizing probiotic retention by identifying the most effective carrier for Lactobacillus rhamnosus using a multi-criteria decision-making approach. Five carrier materials—pea protein, whey protein, maltodextrin, inulin, and pectin—were assessed through three critical phases: evaluating encapsulated probiotic survival in different pH solutions, examining the impact of carriers on kombucha fermentation, and assessing probiotic stability during storage. The findings indicate that whey protein serves as the most effective carrier, offering superior bacterial protection and enhancing fermentation efficiency. Kinetic modeling further demonstrated a significant correlation between probiotic survival, pH, and titratable acidity, while artificial neural network models achieved high predictive accuracy (r² > 0.9). Functional analysis revealed that kombucha enriched with probiotic whey protein encapsulates exhibited improved bioactivity, including enhanced antidiabetic properties through α-glucosidase and α-amylase inhibition, antihypertensive effects via ACE inhibition, and antihypercholesterolemic activity through HMGCR inhibition. These findings suggest that probiotic fortification contributes to the beverage’s overall health-promoting potential. Sensory evaluation highlighted that while enriched kombucha exhibited slight modifications in texture and acidity, overall consumer acceptability remained high. The study underscores whey protein’s role as an optimal probiotic carrier, significantly enhancing kombucha’s probiotic stability and bio functional properties. The results contribute to advancements in functional beverage formulation, paving the way for the development of probiotic-enriched kombucha with improved stability, bioactivity, and consumer appeal.

1. Introduction

Kombucha is a fermented functional beverage that rapidly gained popularity because of its therapeutic properties. Although knowledge of its exact origin has been lost over the years, it is believed that fermented tea was first used in northeastern China (Manchuria region) in 220 BC, where it was adopted for its detoxifying and energetic properties [1]. With the expansion of trade routes, kombucha was introduced to western countries as well. Since then, its popularity has grown, and today, kombucha can be purchased in retail stores around the world, as well as tea fungus for its preparation [2]. This traditional beverage was originally made by fermenting sweetened black or green tea (Camellia sinensis), but other teas [3,4] and different raw materials such as coffee [5], fruit juices [6,7], and by-products [8,9] can also be used. The fermentation, which lasts from a few days to two weeks, is performed by a symbiotic colony of bacteria and yeasts embedded in a cellulose film called a Symbiotic Consortium of Bacteria and Yeast (SCOBY) [2]. Sucrose is often used as a primary source of carbon during fermentation, and the presence of tea in the kombucha preparation medium provides microorganisms with nitrogen, which is necessary for their metabolism [10]. The yeast component of this culture is usually made up of osmophilic yeast species, while the bacterial component includes acetic acid bacteria (AAB) [11]. The chemical compounds present in kombucha are either products of the metabolism of SCOBY microorganisms or originate from the substrate itself, whereby their structure can be modified and transformed into new compounds during fermentation. Kombucha consists of a wide range of chemical compounds, including polyphenols, sugars, organic acids, ethanol, caffeine, amino acids, biogenic amines, proteins, anions, minerals, vitamins, carbon dioxide, and hydrolytic enzymes, as well as other lesser-known metabolites of yeasts and bacteria [12]. Numerous health benefits of kombucha have been noticed based on personal experiences and testimonies, but recently more scientific evidence has been provided to explain these therapeutic effects [11]. There have been several reports of the antimicrobial [13,14], anticancer [15,16,17], antidiabetic [18,19], antioxidant [20,21], and detoxifying and hepatoprotective [22] properties of kombucha. However, it should be noted that most of these benefits have only been studied in experimental models, with the absence of scientific evidence in humans [2].

In addition, kombucha is known worldwide as a probiotic drink, since it contains live microorganisms in its composition. However, most of these statements are unfounded or at least unproven [23]. Although lactic acid bacteria (LAB) are present in some studies both in SCOBY and in kombucha itself [24,25], their numbers are often inconsistent and there is no evidence of the probiotic potential of kombucha. The probiotic strains are usually absent or remain in low concentrations, mainly after storage [23]. Harrison et al. [26] examined 39 retail soft, soft-aberrant, and hard retail kombucha samples for their probiotic potential and concluded that only 6.3% of the soft, 10% of the soft-aberrant, and none of the hard kombucha products sampled exceeded 10⁶ CFU/mL, the limit that would deliver at least a billion cells in one package and provide an adequate probiotic dose. An alternative strategy for obtaining kombucha with pronounced probiotic properties may be to add well-known and studied probiotic strains to the fermentation process or finalized product [27]. However, probiotic bacteria usually do not remain viable during the sometimes harsh conditions of fermentation, manufacturing, and storage, such as low pH, oxygen levels, temperature, etc. [28]. Several studies showed that LAB’s survivability in the conditions of kombucha fermentation is low [29,30].

LAB are a heterogeneous group of phylogenetically related microorganisms that produce lactic acid as the main or sole product of carbohydrate fermentation [31]. These bacteria are defined as rod-shaped, Gram-positive, immobile, non-spore-forming, and anaerobic or aerotolerant species [32,33]. LAB play an important role in the food and agricultural industries and are also used in the clinical sector [34]. Among this group, Lactobacillus create organic substances that contribute to the taste, texture, and aroma of the product and prevent spoilage. In addition, these bacteria can also boost therapeutic and health benefits for consumers. These health-promoting properties are related to the probiotic characteristics of LAB [35,36]. According to the World Health Organization (WHO/FAO), probiotics are formally defined as “living microorganisms that, when used in adequate amounts, confer health benefits on the host” [37]. When the microbial community in the human intestinal tract is in dysbiosis, an imbalance occurs, and various diseases can arise in individuals [38]. One of the main therapies that has been implemented is the use of beneficial microorganisms, such as probiotics, in a functional diet [23,39]. Probiotics are widely used in the pharmaceutical industry because of their ability to improve health, including mental state, and some secondary functions such as vitamin synthesis, reduce cholesterol levels and lower the risk of colorectal cancer [40], stimulate the immune system, enhance bowel motility, and prevent diarrhea [41], diabetes, and cardiovascular diseases [40]. Public awareness is slowly increasing regarding the health benefits of fermented food and beverages, so numerous functional foods with probiotics are emerging [35]. To achieve potential health effects, probiotic food and beverages are required to contain a recommended minimum level of 6–7 log CFU/mL or g of probiotic bacteria at the time of consumption [42].

Microencapsulation emerges as one of the main technologies used for protecting and enhancing the survivability and viability of probiotics, ensuring product stability, and achieving a targeted release in the gut in an adequate amount [39,43]. During the encapsulation process, small quantities of core materials containing nutrients or therapeutic compounds and probiotic bacteria are entrapped within the wall material to form capsules [44]. The survival of encapsulated probiotic cells will depend on several factors, including the used strain, the carrier material and concentration, the number of initial cells, the microcapsule size, the encapsulation method, etc. The most often used materials for microencapsulation of probiotic microorganisms include various biopolymers such as polysaccharides (gum Arabic [45], pectin [46], maltodextrin [47], alginate, inulin [48], chitosan, xanthan [38]), proteins (whey protein, gelatin, legume proteins [49]), and dietary fibers (resistant starch, maize starch [38]). Freeze drying, also known as lyophilization, is a widely used dehydration method for the preservation and encapsulation of probiotics, pharmaceuticals, and food [50]. This process combines the critical freezing and sublimation steps by first freezing the liquid and then converting the frozen liquid into vapor by sublimation under reduced pressure [38]. Although very convenient and effective, the freeze-drying method has some disadvantages, such as being expensive and time-consuming. Also, it can cause damage to the cellular membrane, surface proteins, and cell wall, thus lowering the survivability of probiotics [50]. Encapsulated probiotic bacteria have not yet been explored as a possible addition to kombucha fermentation, but they have successfully been employed in different fruit juices [51,52] and milk products, such as yogurt [53] and ice cream [54].

Since manipulation in kombucha production, probiotic enrichment, and encapsulation protocols can be laborious and time- and cost-consuming, a different mathematical strategy might be a solution for obtaining better validation for all these steps without additional resources. For example, the classical machine learning models, including artificial neural networks (ANNs), random forest regression (RFR), support vector machines (SVMs), extreme learning machines (ELMs), K-nearest neighbors (KNN)s, and decision trees (DTs), are widely applied across various scientific fields [55]. SVM, grounded in statistical learning theory, is recognized for its strong generalization capability, effectively balancing model complexity and training error. ELM utilizes a single-layer feedforward network with randomly assigned input weights and biases. Advanced ensemble models such as XGBoost, LightGBM, and CatBoost are particularly well suited for sequence data [56]. XGBoost provides high prediction accuracy and interpretability, LightGBM efficiently processes large datasets, and CatBoost enhances forecasting by effectively integrating categorical attributes [57].

This paper explores the influence of different carriers on probiotic survival (Step 1), the optimization of kombucha fermentation parameters (Step 2), and the stability of LAB during storage (Step 3). The optimization process employed advanced mathematical tools—including cluster analysis (CA), Principal Component Analysis (PCA), kinetic modeling, ANN modeling, and multi-objective optimization—to systematically evaluate and refine key parameters related to probiotic survival, kombucha fermentation, and LAB stability. Additionally, mathematical verification was performed to validate the kinetic model- and ANN-predicted values. Lastly, kombucha enriched with probiotic bacteria encapsulated in the optimal carrier was assessed in terms of antidiabetic, antihypertensive, and antihypercholesterolemic potential, as well as sensory characteristics.

2. Materials and Methods

The study is divided into three key experimental segments, each assessing a different aspect of probiotic stability (Figure 1). The stepwise selection process is based on mathematical models and statistical evaluation, leading to the identification of the best-performing encapsulation material for probiotic cells for obtaining enriched kombucha with probiotic properties. As the final step, kombucha enriched with the optimal carrier for Lactobacillus cells was evaluated in terms of health-promoting parameters as well as sensory attributes.

2.1. Preparation of Probiotic Culture

Lactobacillus rhamnosus ATCC 53103 was stored at −80 °C in an ultra-low-temperature deep freezer (Snijders Labs, Tilburg, Norway) in Lactobacillus MRS Broth (Himedia, Mumbai, India), supplemented with 200 g/L of glycerol, at the Laboratory of Microbiology, Faculty of Technology, University of Novi Sad, Serbia. Prior to use, the strain was recultured by evenly spreading 25 µL of strain mixture on the surface of several previously poured Lactobacillus MRS Agar (Himedia, Mumbai, India) plates. The plates were stored anaerobically by using Anaerocult A^® (Merck, Darmstadt, Germany) at 30 °C for 48–72 h. Then, a suspension in saline peptone solution (0.85% NaCl, 0.1% pepton) was made. The concentration of L. rhamnosus ATCC 53103 was estimated at 3 × 10⁹ CFU/mL by comparing with McFarland Standards.

In each experiment step, L. rhamnosus concentration was determined in the same way. Namely, this parameter was determined using Lactobacillus MRS Agar (Himedia, Mumbai, India), and the incubation was performed under anaerobic conditions by using Anaerocult A^® (Merck, Darmstadt, Germany) at 30 °C for 48–72 h. All samples were tested in three repetitions, and the results are expressed as log CFU/mL. Since the L. rhamnosus is lactic acid bacteria, the abbreviation LAB is used in the following text to refer to using this probiotic strain.

2.2. Encapsulation of LAB Cells

Three polysaccharides, pectin (Sigma-Aldrich, St. Louis, MO, USA), maltodextrin (Battery Nutrition Limited, London, UK), and inulin (purchased from a local shop), and two proteins, whey (BioTech USA, Szada, Hungary) and pea protein (purchased from a local shop), were used as carrier materials. An amount of 8 g of carrier material was dissolved in 40 mL sterile distilled water (20% w/w). After that, 20 mL of LAB cell suspension, 4.4 mL of sunflower oil, and 0.4 g of Tween 80 were added to the mixture. The solutions were then stirred for 1 h using a magnetic stirrer at 25 °C and 200 rpm and stored in an ultra-low-temperature deep freezer (Snijders Labs, Tilburg, Norway) at −80 °C overnight. The lyophilization process consisted of primary freezing of all samples at −40 °C for 2 h in the Martin Crist Alpha 2–4 freeze-drier (Osterode, Germany), while the main drying process was performed at 0.01 bar at −40 to 20 °C for 48 h, and the final drying step lasted for 4 h at 0.005 mbar at 20 to 30 °C. The samples were kept in the freezer until further use.

2.3. Stability of LAB Cells in Different Carriers at Different pH Solutions

Saline peptone solution was adjusted to four different pH values, 2.5, 3, 4, and 4.5, using 1 mol/L acetic acid. Then, 0.1 g of carrier–LAB encapsulates were separately added to 1 mL of four different pH solutions and kept at 28 °C. Samples were taken after 1 h and every day for the first 4 days and on days 7 and 10. The pH values were chosen to resemble typical kombucha fermentation changes. After sampling, a series of dilutions were made, and the number of LAB was determined.

2.4. Kombucha Preparation and Fermentation

Fermentation was performed by using the local household tea fungus culture (Medusomyces gisevii, Manchurian mushroom, SCOBY), which has traditionally been used over a long period for all research at the Laboratory of Microbiology at the Faculty of Technology Novi Sad, Serbia (Supplementary Figure S1). The authors have previously reported that the fungus culture contains at least two yeast strains and two acetic acid bacteria strains [58]. The substrate for fermentation was prepared by adding 70 g/L of sucrose into tap water. After the boiling point, 3 g/L of black tea leaves (Camellia sinensis L.) (“Fructus”, Bačka Palanka, Serbia) was added and left to steep for 15 min. The tea leaves were then removed by filtration through a sterile filter paper. After cooling down to room temperature, the tea was inoculated with 100 mL/L of the kombucha from previous fermentation obtained under the same conditions (Supplementary Figure S2). Sterile glass jars (volume 0.72 L, diameter 8 cm) were filled with 0.33 L of the mixture, and 1.5 g of carrier–LAB encapsulates was added separately. The control sample was prepared without the addition of carrier–LAB encapsulates. The glass jars were covered with cheesecloth, and the fermentation at 28 ± 0.2 °C was monitored for 4 days. Four glass jars per fermentation were prepared and each was sampled only once to avoid potential contamination. Sampling was performed after 1 h of the addition of carrier–LAB encapsulates and for the next 4 days every day for microbiological (number of yeasts, AAB, and LAB) and chemical (pH and titratable acidity (TA)) analysis.

Microbiological (number of yeasts and AAB) and chemical analyses (pH and TA) were performed using methods previously described by Ranitovic et al. [58]. The number of LAB was determined as described previously.

2.5. Stability of L. rhamnosus in Kombucha During Refrigarated Storage

After four days of fermentation, samples of kombucha were filtered, bottled, and stored in the refrigerator at +4 °C to test LAB viability during storage. During 21-day storage, samples were taken on days 2, 4, 6, 8, 10, and 21.

2.6. In Vitro Testing of Health-Promoting Potential

In vitro testing of selected health-promoting potentials (antidiabetic, antihypertension, and antihypercholesteromic) was performed using the protocols described in the study by Vučetić et al. [59] with differences in the preparation of protein-rich samples.

Namely, to ensure that the whey protein carrier did not interfere with enzyme inhibition assays, a series of purification steps was applied to both control and enriched kombucha samples. These included centrifugation, filtration, protein removal (for the enriched sample), and ethanol extraction to obtain a standardized and bioactive compound-rich extract for enzymatic analysis. In the beginning, each kombucha sample was subjected to centrifugation at 5000 rpm for 10 min at 4 °C to separate microbial cells and suspended particles. The resulting supernatant was then passed through a 0.45 µm membrane filter (Whatman No.1 or PVDF membrane, Maidstone, UK) to ensure a clear, cell-free fraction suitable for subsequent assays. Since the enriched kombucha contained protein, an additional trichloroacetic acid (TCA) precipitation step was applied to remove protein content that could interfere with enzymatic activity. To achieve this, 10% (w/v) TCA solution was added to the enriched kombucha (1:10 ratio), resulting in a final concentration of 1% TCA. The mixture was incubated at 4 °C for 30 min to facilitate protein aggregation, followed by centrifugation at 10,000 rpm for 10 min at 4 °C. The clear supernatant was carefully collected and neutralized with 1 M NaOH (pH 7.4) to restore compatibility with enzymatic assays. The neutralized sample was then stored at 4 °C for further processing. To concentrate bioactive compounds, ethanol extraction was performed on both control and enriched kombucha samples. A 1:1 (v/v) ratio of kombucha filtrate to ethanol (99%) was mixed and stirred at room temperature (25 °C) for 30 min to facilitate compound dissolution. The suspension was then subjected to centrifugation at 4000 rpm for 10 min, and the supernatant was collected. To remove ethanol, the supernatant was evaporated under reduced pressure using a rotary evaporator. The resulting dried extract was dissolved in phosphate buffer specific to each enzyme assay (pH 6.8 for α-amylase and α-glucosidase, pH 8.3 for ACE, and pH 7.4 for HMGCR) at a final concentration of 1 mg/mL. The stock solutions were stored at −20 °C until enzymatic inhibition assays were conducted.

To describe the method briefly, the α-amylase inhibition assay used porcine pancreatic α-amylase (Sigma-Aldrich, USA) following the 3,5-dinitrosalicylic acid method. A mixture of 250 µL kombucha extract and 250 µL α-amylase (0.5 mg/mL in phosphate buffer, pH 6.8) was pre-incubated at 37 °C for 10 min. Then, 250 µL of 1% starch solution was added and incubated for 30 min at the same temperature. The reaction was stopped with 500 µL reagent, boiled at 100 °C for 5 min, and cooled, and then absorbance was read at 540 nm (Shimadzu UV-Vis). α-Glucosidase inhibition was assessed using yeast α-glucosidase (EC 3.2.1.20). An amount of 50 µL kombucha extract was mixed with 100 µL enzyme solution (1 U/mL, pH 6.8) and pre-incubated at 37 °C for 10 min. After adding 50 µL of 5 mM pNPG substrate, incubation continued for 30 min. The reaction was halted with 100 µL of 0.2 M Na₂CO₃, and absorbance was measured at 405 nm (Bio-Rad reader). For ACE inhibition, the assay followed the Hippuryl-Histidyl-Leucine (HHL) method. A mix of 50 µL kombucha extract, 50 µL ACE (2.5 mU/mL, pH 8.3), and 100 µL HHL (5 mM) was incubated at 37 °C for 60 min. The reaction was stopped with 200 µL of 1 M HCl, followed by 500 µL ethyl acetate extraction. The organic layer was dried and then redissolved in water, and absorbance was recorded at 228 nm. The HMG-CoA reductase assay used a commercial kit (Sigma-Aldrich, USA). The reaction contained 50 µL kombucha extract, 50 µL HMG-CoA (0.2 mM), and 50 µL NADPH (0.2 mM), with initiation by 50 µL of HMGCR enzyme (1 U/mL, pH 7.4). After 10 min at 37 °C, NADPH absorbance was measured at 340 nm using a microplate reader.

2.7. Sensory Analysis

The sensory analysis was conducted in the laboratory for sensory analyses at the Faculty of Technology, University of Novi Sad, Serbia, after approval by the Ethics Committee of Faculty of Technology, Novi Sad, Serbia (Ref. No. 020-8/26). Since there is no standardized approach for the sensory evaluation of kombucha, evaluation techniques and methodologies from the works of Everitt [60], Kim [61], and Ranitović [58] were utilized. All participants approved and signed the informed consent form and evaluated the kombucha samples as trained sensory panelists. Following the average structure of kombucha consumers [62], participation included 4 men (35 to 55 years old) and 12 women (25 to 45 years old). The sensory evaluation was carried out as a blind test—trained participants knew that both served samples were kombucha beverages, but not which one was enriched with whey protein–LAB encapsulates. In brief, eight aroma descriptors (odor, taste, acidity, sweetness, bitterness, color tone, visual viscosity, and general impression) were selected for descriptive analysis, enabling rating from 1 to 5 for each descriptor. A 5-point scale was used to assess the complete evaluation using the following scheme:

Odor: mild vinegar (1); strong vinegar (2); yeast (3); tea (4); other (5).

Taste: mild vinegar (1); strong vinegar (2); yeast (3); tea (4); other (5).

Acidity: imperceptibly sour (1); slightly sour (2); medium sour (3); noticeably sour (4); too sour (5).

Bitterness: imperceptibly bitter (1); slightly bitter (2); average bitter (3); noticeably bitter (4); too bitter (5).

Sweetness: imperceptibly sweet (1); slightly sweet (2); average sweet (3); noticeably sweet (4); too sweet (5).

Color tone: pale (transparent) (1); light yellow (2); golden-yellow (3); light brown (4); dark brown (5).

Visual viscosity: like milk (1); as clear juice (2); as thick juice (3); as thick yogurt (4); like honey (5).

General impression: completely unacceptable (1); partially unacceptable (2); neutral (3); partially acceptable (4); completely acceptable (5).

Each kombucha sample (50 mL) was placed in 150 mL odorless sensory plastic cups at room temperature and marked with a random letter. The panelists were asked to assess the kombucha samples and record impressions of each sensory descriptor. Each panelist was asked to have a rest for sensory recovery between different samples.

2.8. Statistical Analysis

The normality of data distribution was assessed using the Shapiro–Wilk and Anderson–Darling tests. The results indicated that most variables significantly deviated from a normal distribution (p < 0.05). Consequently, the Kruskal–Wallis test, a non-parametric method, was applied to ensure methodological rigor and provide a more robust statistical perspective. Results are presented as mean values (n = 3) with standard deviations. Tukey’s HSD test was used to evaluate differences between sample means. All data underwent comprehensive statistical analysis, including descriptive statistics and Pearson’s correlation analysis, using the STATISTICA 10.0 software package (StatSoft Inc., Tulsa, OK, USA).

To evaluate probiotic survival (Step 1), optimize kombucha fermentation parameters (Step 2)—including pH, TA, yeast count, and AAB and LAB levels—and assess LAB stability during storage (Step 3), kinetic modeling was performed. The temporal kinetics of these parameters, tested with different carriers (pea protein, whey protein, maltodextrin, inulin, and pectin), was described using a four-parameter sigmoidal mathematical model (Equation (1)), which is highly suitable for microbiological systems [58,63].

(1)$y(t)=d+{\displaystyle \frac{a-d}{1+{\left(\frac{t}{c}\right)}^b}}$

In Equation (1), probiotic survival (log CFU/mL) (Step 1), and fermentation parameters (Step 2) are represented as y(t), whereas the regression coefficients are denoted as follows: a—minimum of the experimentally obtained values (at t = 0); d—the maximally acquired value; c—the inflection point (the point between a and d); and b—Hill’s slope (the steepness of the inflection point c).

CA and PCA were used as multivariate statistical techniques for data exploration and pattern recognition. CA groups objects based on their similarities, helping to identify underlying structures in complex datasets. PCA reduces dimensionality by transforming correlated variables into a smaller set of uncorrelated principal components, preserving the most significant variance in the data while enhancing interpretability.

The sensory evaluation followed a balanced factorial design, with the order of sample presentation determined by the experimental design for sensory analysis using XLSTAT-MX (XLSTAT 2018.7, Addinsoft, http://www.xlstat.com/ accessed on 10 January 2025).

A multi-layer perceptron (MLP) with three layers (input, hidden, and output) was used to develop ANN models for predicting probiotic survival (ANN 1), kombucha fermentation parameter optimization (ANN 2), and LAB stability during storage (ANN 3). ANN 1 estimated probiotic survival using pH and the application of different carriers (such as pea protein, whey protein, maltodextrin, inulin, and pectin). ANN 2 predicted kombucha fermentation parameters (such as pH, TA, yeast, and AAB and LAB count), based on time and the used carrier (pea protein, whey protein, maltodextrin, inulin, and pectin). ANN 3 predicted LAB stability during storage.

ANN models, widely recognized for addressing nonlinear problems [64], were standardized to improve accuracy. Model training was optimized using the BFGS algorithm [65]. Data were split into 70% for training and 30% for testing. A total of 100,000 MLP configurations were evaluated by varying the number of hidden-layer neurons (5–10), weight initialization methods, and activation functions (tanh, logistic, exponential, identity). Successful training was achieved by minimizing the squared error.

Hidden-layer coefficients (weights and biases) were stored in matrices W₁ and B₁, while output-layer coefficients were in W₂ and B₂. The ANN models were mathematically represented using matrix notation, with output Y and input X corresponding to accident-related variables (Equation (2)) [66].

(2)$Y=f_1\left(W_2\cdot f_2\left(W_1\cdot X+B_1\right)+B_2\right)$

Weight coefficients in the ANN models (elements of matrices W₁ and W₂ and vectors B₁ and B₂) were defined through determination of the ANN models [67]. The widely applied BFGS algorithm was utilized to consolidate the convergence in resolving the solution of the nonlinear problem.

In terms of error analysis, the accuracy of the developed models was evaluated through several key metrics: coefficient of determination (r²), reduced chi-square (χ²) (Equation (3)), root mean square error (RMSE) (Equation (4)), mean bias error (MBE) (Equation (5)), mean percentage error (MPE) (Equation (6)), the sum of squared errors (SSE) (Equation (7)), and average absolute relative deviation (AARD) (Equation (8)). These widely used parameters were employed to assess the validity of the models, as follows:

(3)${\chi }^2={\displaystyle \frac{{\displaystyle \sum _{i=1}^N{(x_{\exp ,i}-x_{pre,i})}^2}}{N-n}}$

(4)$RMSE={\left[{\displaystyle \frac{1}{N}}\cdot {\displaystyle \sum _{i=1}^N{(x_{pre,i}-x_{\exp ,i})}^2}\right]}^{1/2}$

(5)$MBE={\displaystyle \frac{1}{N}}\cdot {\displaystyle \sum _{i=1}^N(x_{pre,i}-x_{\exp ,i})}$

(6)$MPE={\displaystyle \frac{100}{N}}\cdot {\displaystyle \sum _{i=1}^N(\frac{\left|x_{pre,i}-x_{\exp ,i}\right|}{x_{\exp ,i}})}$

(7)$SSE={\displaystyle \sum _{i=1}^N{(x_{pre,i}-x_{\exp ,i})}^2}$

(8)$AARD={\displaystyle \frac{1}{N}}\cdot {\displaystyle \sum _{i=1}^N\left|\frac{x_{exr,i}-x_{pre,i}}{x_{exr,i}}\right|}$

where x_exp,i stands for the experimental values and x_pre,i are the predicted values obtained by calculating from the model for these measurements. N and n are the numbers of observations and constants, respectively.

3. Results and Discussion

Selecting an optimal probiotic carrier for kombucha fermentation plays a key role in enhancing probiotic viability while preserving the beverage’s functional properties. This study systematically investigates probiotic survival, fermentation dynamics, and stability during storage through a multi-criteria decision-making framework. The research unfolds in three critical stages: (1) evaluating probiotic survival across different carriers, (2) examining how carrier selection affects kombucha fermentation parameters, and (3) assessing probiotic stability over storage time. To ensure data-driven optimization, CA, PCA, and advanced mathematical models, including kinetic modeling and ANN, were applied. The findings offer a detailed perspective on how various carriers influence probiotic retention, fermentation kinetics, and overall kombucha quality. The following sections provide an in-depth analysis, reinforced by statistical validation and predictive modeling.

3.1. Step 1: Probiotic Survival in Different Carriers

The probiotic survival varies significantly depending on the carrier type, pH level, and storage time, with statistically significant differences observed (p < 0.05) between samples based on Tukey’s HSD test (database is provided as Supplementary Table S1). The pea protein and whey protein carriers provide better protection for probiotics than maltodextrin, inulin, and pectin carriers, particularly under the acidic conditions and pH values that typically change during kombucha fermentation, where degradation is most pronounced for probiotic survival in maltodextrin, inulin, and pectin. Pea protein and whey protein carriers provide the highest probiotic stability, with initial values ranging from 8.09 to 8.185 log CFU/mL and 8.499 to 8.835 log CFU/mL, respectively, across all pH levels. Over time, the pea protein carrier shows a moderate decline in probiotic survival, particularly at pH 2.5, where it decreases from 8.124 at day 0 to 7.730 at day 10, corresponding to a 0.394 log reduction, while probiotic survival in the whey protein carrier remains stable under neutral conditions but declines significantly at pH 2.5 (from 8.499 at day 0 to 7.033 at day 10, representing a 1.466 log reduction). The probiotic survival in maltodextrin, inulin, and pectin carriers exhibits greater degradation under acidic conditions. The probiotic survival in maltodextrin declines rapidly at pH 2.5, from 8.029 (day 0) to 4.657 (day 1), corresponding to a 3.372 log reduction, and is undetectable by day 10. The probiotic survival in inulin follows a similar trend, dropping from 6.866 (day 0) to 1.617 (day 2), corresponding to a 5.249 log reduction and becoming undetectable by day 7 at pH 2.5, but remaining stable at higher pH levels. The probiotic survival in pectin is the most sensitive to acidity. After 1 h of addition, survivability decreased to 4.344 (time 0, pH 2.5) and continued to further drop, reaching 0.000 by day 4. In neutral and higher pH environments (pH 4–4.5), probiotic survival demonstrates greater stability for all carriers, with survival values remaining above 5.0 log CFU/mL for most carriers up to day 10.

The diagrams in Figure 2 depict probiotic survival (log CFU/mL) across different carriers (pea protein, whey protein, maltodextrin, inulin, and pectin), with data points representing sample ages from 0 to 10 days. The results indicate a consistent trend across all samples, showing a gradual changes in probiotic survival over time.

Previous studies indicate that whey [68] and pea proteins [49] are commonly used carriers for probiotic bacteria since they have specific physicochemical properties, such as good emulsification, gelation, and fill-forming and water binding capacities [49,69]. Sun et al. [28] successfully encapsulated the Lactobacillus plantarum strain using whey protein, thus promoting its survivability and stability in apple juice during refrigerated storage. Polysaccharides are also commonly used; for example, Tarifa et al. [46] encapsulated two Lactobacillus sp. strains in pectin and pectin inulin microgels and their shelf life was extended but the results of survivability under simulated gastric and intestinal conditions were not as high as expected. This degradability under lower pH values of pectin and inulin is in correlation with the results of the current study. The stability of encapsulated cells in pectin, inulin, and maltodextrin could be further improved by coating the capsules with additional materials, such as chitosan or gelatine [70].

Analysis of the coefficient values (a, b, c, and d) from the kinetics study shown in Figure 2 using a four-parameter sigmoidal mathematical model (Table 1) reveals that coefficient a remains relatively consistent across most carriers and pH levels (ranging from 6.9 to 8.9), while b shows greater variability with values often close to 0 but exceeding 15 in some cases (e.g., for whey protein at pH 4.5 with b = 18.0). Coefficient c exhibits high variability, with values close to 1 in some instances and reaching 115.0 for pectin at pH 4.5, whereas d shows considerable variability, with notable decreases to 0 for inulin and pectin carriers at higher pH values. Coefficient a represent a baseline property, b is sensitive to specific protein–pH interactions, and c and d capture processes strongly influenced by both carrier type and pH value.

The verification kinetic model results (Table 2) presented for various carriers across different pH levels demonstrates a generally strong model performance. Specifically, high r² values are observed in most instances, with values ranging from 0.644 for whey protein at pH 3 to 0.998 for maltodextrin at pH 4.5. Correspondingly, low RMSE values are also prevalent, ranging from 0.040 for maltodextrin at pH 4.5 to 0.431 for maltodextrin at pH 2.5. However, certain conditions exhibit comparatively lower accuracy. For example, the maltodextrin carrier at pH 2.5 shows an r² of 0.970 and an RMSE of 0.431, while the inulin carrier at pH 4 presents an r² of 0.024 and an RMSE of 0.193. These cases indicate reduced model accuracy in specific scenarios. Also, MBE and MPE values, while generally close to zero, show some deviations, such as an MBE of −0.051 and an MPE of 7.309 for maltodextrin at pH 2.5, and an MBE of −0.130 and an MPE of 5.817 for pectin at pH 2.5. These deviations suggest minor systematic biases in certain cases. Overall, the model demonstrates good predictive capability for most carrier–pH combinations, as evidenced by the predominantly high r² values and low RMSE values. However, further investigation and potential model refinement may be necessary to address the less accurate predictions observed in specific conditions, particularly those with higher RMSE and lower r² values, as well as those exhibiting larger deviations in MBE and MPE.

The five carriers were ranked by comparing their averages to extreme values, favoring higher values, as described by Brlek et al. [71]. Maximizing the score identified whey protein as the optimal carrier, while pea protein also had a very high score (Table 3). All other types of carriers obtained scores below 0.5, which was crucial for the selection of a protein-based carrier for further experiment steps. Although polysaccharide carriers have various advantages in the encapsulation of probiotics, such as high stability and availability, they also have reactive functional groups and are brittle and sensitive to moisture. On the other hand, protein encapsulation adds to the nutritional value of the product and due to their amphiphilic nature, they are suitable candidates for the encapsulation of probiotics. Whey protein, in particular, has shown the ability to enhance the resilience of probiotics [72].

The correlation analysis (Table 4) shows varying degrees of associations between the different carriers, with some statistically significant relationships. The correlation between pea protein and whey protein carriers (r = 0.472, p = 0.011) is moderate and statistically significant, indicating a positive relationship. The correlation between whey protein and maltodextrin carriers (r = 0.658, p = 0.000) is also moderate and statistically significant. Likewise, the correlations between whey protein and inulin carriers (r = 0.485, p = 0.009) and whey protein and pectin carriers (r = 0.602, p = 0.001) are moderate and statistically significant. The correlation between maltodextrin and inulin carriers (r = 0.902, p = 0.000) is very strong and statistically significant, as is the correlation between maltodextrin and pectin carriers (r = 0.937, p = 0.000), indicating strong positive relationships. Similarly, the correlation between inulin and pectin carriers (r = 0.918, p = 0.000) is also very strong and statistically significant.

Based on the dendrogram resulting from a hierarchical cluster analysis using complete linkage and city-block (Manhattan) distances (Figure 3), a clear separation of samples into two primary clusters is observed. The first cluster is exclusively composed of samples with a pH of 2.5, while the second cluster encompasses all remaining samples with pH values of 3.0–4.5. This primary division suggests that pH is the dominant factor influencing sample differentiation. Within the second cluster, further sub-structuring is evident. It is noted that samples with pH 4.0 and 4.5 are closely intermixed, indicating a high degree of similarity. In contrast, a temporal trend is discerned within the pH 3.0 subgroup, where samples from earlier time points appear to cluster separately from those at later time points. This suggests that while pH is the primary determinant of sample clustering, temporal changes also contribute to sample variability.

PCA was used to explore the relationships between variables during probiotic survival experiments in different carriers and during the kombucha fermentation process. In the PCA plot (Figure 4), the proximity of spots signifies similarity in patterns [73]. The direction of vectors in the factor space indicates variable trends, while their length represents the strength of the correlation between the fitting value and the variable [74]. Smaller angles between variables denote stronger correlations. The first two principal components demonstrated 89.10% of the total variance in the recorded data. The first PC explained 68.56%, and the second explained 20.54% of the total variance between the collected data (eigenvalues were 3.428 and 1.027 for the first two PCs). The projection of the variables in the factor plane indicated that whey protein (−17.01%, based on correlation), maltodextrin (−27.56%), inulin (−23.68%), and pectin (−26.09%) contributed most negatively to the first principal component (PC1). Pea protein (68.36%, based on correlation) and whey protein (12.93%) contributed positively to the second principal component (PC2), while inulin (−11.64%) contributed negatively to PC2.

The optimal neural network models (Figure 5) for predicting probiotic survival in different carriers demonstrated strong generalization capabilities for the collected data and can accurately predict the observed outputs based on input parameters such as pH value and carrier selection (Table 5).

The number of neurons in the ANN models ranged from 3 to 9, which yielded the highest r² values: 0.755, 0.868, 0.995, 0.943, and 0.980 for training, and 0.592, 0.927, 0.995, 0.744, and 0.931 during the testing cycle. Training accuracy improved as the number of training cycles increased, reaching a nearly constant value around the 40th to 50th epoch. Training beyond 50–60 epochs could potentially lead to overfitting, with 60 epochs being sufficient to achieve high model accuracy without the risk of overfitting.

The verification of the ANN models (Table 6), applied to various carriers, reveals a generally good predictive performance, albeit with some variation across different carriers. High r² values, ranging from 0.694 for pea protein to 0.994 for maltodextrin, indicate a strong correlation between predicted and observed values for most carriers. However, pea protein exhibits a comparatively lower r² of 0.694, suggesting a somewhat reduced accuracy of prediction for this carrier. Correspondingly, the RMSE values are relatively low, ranging from 0.136 for pea protein to 0.718 for inulin, further supporting the generally good model performance. However, inulin shows a notably higher RMSE, indicating a larger average prediction error for this carrier. The MBE and MPE values, while generally close to zero, show some deviations, particularly for pea protein with an MBE of −0.018 and an MPE of 1.422, suggesting a slight systematic underestimation in this case. The ANN model demonstrates good predictive capability for most carriers, as evidenced by the predominantly high r² values and low RMSE values. The lower accuracy observed for pea protein and inulin, as well as the minor biases indicated by MBE and MPE values, suggest potential areas for model refinement and further investigation to enhance the overall prediction accuracy and robustness across all carriers.

3.2. Step 2: Kombucha Fermentation Parameters

The results of pH value, TA, number of yeasts, AAB, and LAB during kombucha fermentation (Supplementary Table S2) demonstrate that the type of carrier significantly influences the fermentation process. The database includes control kombucha fermentation as well as the fermentation of kombuchas enriched with a whey protein–LAB or pea protein–LAB encapsulate system. The initial pH of uninoculated tea was 7.30 and decreased significantly in pH, ranging from 4.817 ± 0.012 for pea protein–LAB-enriched kombucha to 4.153 ± 0.015 for the control kombucha after inoculation (day 0). Over time, all kombucha samples exhibited a decrease in pH value, with whey protein–LAB-enriched kombucha and control kombucha showing the most pronounced acidification by day 4 (2.570 ± 0.010 for whey protein and 2.807 ± 0.006 for control), while pea protein–LAB-enriched kombucha maintained a higher pH value, reaching 3.100 ± 0.010 by day 4. TA increased across all treatments, with whey protein–LAB-enriched kombucha showing the highest final TA (6.090 ± 0.030), followed by the control kombucha (5.140 ± 0.035) and pea protein–LAB-enriched kombucha (4.210 ± 0.017). During kombucha fermentation, yeasts hydrolyze sucrose to glucose and fructose, thus producing ethanol. At the same time, AAB use glucose to produce gluconic acid and ethanol to produce acetic acid [2]. Acetic acid is the most dominant acid in fermentation liquid and is responsible for the sourness of kombucha [10]. In addition to acetic acid, through the metabolism of SCOBY microorganisms several other organic acids are produced, including lactic acid by LAB in enriched kombucha samples [75]. The accumulation of acids over time affects TA, which linearly increases. This in turn influences the pH value that decreases gradually and stabilizes as the fermentation time increases. These results are in accordance with the standard pattern of kombucha fermentation [76,77,78,79].

The number of yeasts was 5.343, 5.347, and 5.614 log CFU/mL for the control sample, and the pea protein–LAB-enriched and whey protein–LAB-enriched kombucha samples, respectively, on day 0 and increased for about 1 log unit until the next day. Yeast counts then peaked between days 1 and 3, with the control and pea protein–LAB-enriched samples maintaining higher counts, 6.527 ± 0.196 and 6.441 ± 0.053, respectively, at day 2, while the whey protein–LAB-enriched kombucha showed a slight decline by day 2 (6.114 ± 0.163). Yeast count showed a decline on day 4 for all the samples, corresponding to 0.04, 0.44, and 1.17 log reduction for pea protein, whey protein, and control, respectively. A similar result was observed by Majid et al. [80] where the maximum number of yeasts was noticed on the fourth day of fermentation, after which a slow decrease appeared. Cvetković et al. [81] also noticed an increase in the number of yeasts on the first day, after which the number was relatively stable until the end of kombucha fermentation. Since there were no significant differences in yeast numbers in the enriched kombucha samples compared to the control sample, it could be concluded that the addition of carrier–LAB encapsulates did not have an influence on yeasts during the fermentation. On the other hand, more noticeable changes were observed in AAB numbers. In the control sample there was an increase in AAB numbers until day 2, after which a slight decrease appeared, reaching 5.967 log CFU/mL on day 4 of kombucha fermentation. Similarly, in pea protein–LAB-enriched kombucha an increase was observed until day 3, with a slight decrease on day 4, but still a higher number of AAB of 6.309 log CFU/mL was noticed compared to the control sample. The highest concentration of AAB was observed in whey protein–LAB-enriched kombucha, where it increased from 6.883 to 7.576 log CFU/mL on day 1, after which it remained mainly stable, reaching a maximum on day 4 of 7.832 log CFU/mL. Yang et al. [82] noted that the presence of LAB during fermentation could support the growth and reproduction of other present bacteria, including AAB, which could be the reason for the higher numbers of AAB in enriched kombucha samples compared to the control sample. A similar pattern was noticed in the studies of Majid et al. [80] and Cvetković et al. [78], where the concentration of AAB was higher by the end of fermentation in the samples enriched with free LAB cells in comparison to the control. Also, the highest number of AAB in whey protein–LAB-enriched kombucha is in accordance with acetic acid production, since TA values were also the highest in this sample by the end of the 4-day fermentation. On the other hand, Wang et al. [83] reported that the production of acetic acid was the most pronounced in the control sample compared to LAB-enriched kombucha samples and concluded that the addition of LAB strains limited the production of acids. The differences in the results of acetic acid production could be due to the different LAB strains used between the studies.

LAB growth was the highest in the whey protein–LAB-enriched kombucha (7.706 ± 0.029 at day 4), with the pea protein–LAB-enriched sample supporting moderate growth (6.550 ± 0.072 at day 3), while LAB was absent in the control sample. Since the optimal pH range for the L. rhamnosus strain is between 6.4 and 4.5 [84] and pH values during kombucha fermentation often decrease to below 3, it could be concluded that free cells would not survive these conditions [27]. There is not much research available that has successfully produced a kombucha beverage that meets the requirements of a probiotic product, due to the drop in the concentration of LAB during fermentation and storage. In a previous study by Cvetković et al. [27], the same strain of L. rhamnosus as in the current study and an additional L. plantarum strain were used and free cells were added at the beginning of kombucha fermentation, which lasted for 5 days. The number of both probiotic strains drastically decreased until the end of fermentation with less than 2.0 log CFU/mL on day 5, supporting the mentioned conclusion. Similarly, Bromley and Perry [29] found that Lactobacillus sp. final populations decreased and ranged from 1.0 to 3.0 log CFU/mL at the end of kombucha fermentation. Fu et al. [30] concluded that the survival rate of LAB was only 0.98% on the eighth day of refrigerated storage of this beverage. On the other hand, Majid et al. [80] managed to successfully produce blue pea tea (Clitoria ternatea L.) kombucha with probiotic properties by adding the strain of Lactiplantibacillus plantarum subsp. plantarum Dad-13 before 8-day fermentation, and their number after fermentation and 28-day storage was 6.26 log CFU/mL. This discrepancy in results may be due to the use of different LAB strains, as well as different tea substrates, for the preparation of kombucha.

The diagrams in Figure 6 show the kombucha fermentation parameters across different carriers (pea protein, whey protein, and control), with data points corresponding to sample ages from 0 to 4 days. The results reveal a clear trend: pH decreases while TA increases as the material ages. Meanwhile, yeast count and AAB and LAB levels remain relatively stable throughout the fermentation process.

Kinetic analysis (Table 7) using a four-parameter model reveals that pea protein–LAB and whey protein–LAB systems differentially influence fermentation kinetics compared to the control, with the pea protein–LAB system often exhibiting slower rates of change (lower b coefficient values, e.g., 2.4 for pH) and the whey protein–LAB system potentially promoting faster LAB growth (higher b value of 10.4 for LAB) and higher final acidity (higher c value of 6.4 for TA). The pea protein–LAB encapsulate sample shows a slightly higher initial pH (a = 4.8) compared to the whey protein–LAB encapsulate sample (a = 4.3) and the control (a = 4.2). The high c values for AAB with the pea protein–LAB system (159.1) and the whey protein–LAB system (25.7) suggest greater growth compared to the control. The whey protein–LAB system also exhibits a higher ‘c’ value for TA (6.4) compared to the pea protein–LAB system (1.3) and the control (1.8). Variations in ‘d’ values indicate protein-specific effects on lag phases, with the pea protein–LAB system having a longer lag phase for yeast growth (‘d’ = 9.1) compared to the whey protein–LAB system (‘d’ = 6.5) and the control (‘d’ = 2.8). However, further analysis and domain expertise are needed to interpret these findings within the context of the specific model used and to elucidate the underlying mechanisms driving these observed differences.

Verification of kinetic models for various parameters (pH, TA, yeasts, AAB, LAB) across different kombucha samples (pea protein, whey protein, control) during fermentation reveals a range of model performances with varying accuracy (Table 8). For the pea protein–LAB-enriched kombucha, pH and TA exhibit exceptional accuracy with χ², RMSE, and MBE values all at 0.000 and an r² of 1.000. Yeasts show slightly lower accuracy, with an r² of 0.989, χ² of 0.004, RMSE of 0.053, and MBE of 0.000. AAB and LAB present lower accuracy, with r² values of 0.066 and 0.806, respectively. Specifically, AAB have χ² of 0.128, RMSE of 0.320, and MBE of 0.000, while LAB show χ² of 0.025, RMSE of 0.142, and MBE of −0.010. For the whey protein–LAB-enriched kombucha, the pH value maintains high accuracy with χ² of 0.003, RMSE of 0.050, MBE of 0.000, and r² of 0.994. TA also shows good accuracy with χ² of 0.040, RMSE of 0.178, MBE of 0.000, and r² of 0.993. However, yeasts and AAB exhibit lower accuracy with r² values of 0.438 and 0.285, respectively. For yeasts, χ² is 0.055, RMSE is 0.210, and MBE is 0.000, while for AAB, χ² is 0.131, RMSE is 0.324, and MBE is 0.000. LAB in whey protein shows good accuracy with χ² of 0.009, RMSE of 0.085, MBE of 0.000, and r² of 0.943.

The control kombucha demonstrates high accuracy for pH and TA with χ², RMSE, and MBE values all at 0.000 and r² of 1.000. LAB also shows perfect accuracy with all metrics at 0.000. However, yeasts and AAB exhibit lower accuracy, with r² values of 0.477 and 0.072, respectively. For yeasts, χ² is 0.196, RMSE is 0.396, and MBE is 0.000, while for AAB, χ² is 0.160, RMSE is 0.358, and MBE is 0.000.

The verification results highlight varying model performance across different parameters and kombucha samples. While pH and TA consistently demonstrate high accuracy, AAB and yeasts often exhibit lower accuracy, particularly in enriched kombuchas. The control sample shows high accuracy for pH, TA, and LAB, but lower accuracy for AAB and yeasts. These findings emphasize the need for further investigation and potential model refinement to improve prediction accuracy for specific parameters and carriers, ultimately enhancing the model’s overall robustness and reliability.

The control and enriched kombuchas were ranked by comparing their averages to extreme values, favoring lower values for pH value and higher values for TA, yeasts, AAB and LAB, as described by Brlek et al. [71]. Maximizing the score identified the whey protein–LAB-enriched kombucha as the better option in view of probiotic stability for kombucha (Table 9).

According to correlation analysis for kombucha samples (Table 10), a strong negative correlation between pH value and TA (r = −0.895, p = 0.000) and a moderate, statistically significant negative correlation between pH value and AAB (r = −0.515, p = 0.049) suggest that lower pH values are associated with higher AAB concentrations. A strong positive correlation between AAB and LAB (r = 0.744, p = 0.001) suggests that higher AAB concentrations are associated with higher LAB counts [82].

Additionally, hierarchical cluster analysis (Figure 7) using complete linkage and city-block distances reveals distinct clustering based on protein source, indicating that protein type carrier is a primary determinant of sample characteristics. Sub-clustering within each protein group highlights temporal variations, suggesting that protein properties evolve over time, potentially due to degradation, aggregation, or other time-dependent changes. Pea protein exhibits greater stability compared to the whey protein carrier, which shows more variability over time. The use of Manhattan distance emphasizes the magnitude of changes in measured variables, further supporting the observation of differential temporal stability among the protein sources.

As shown in Figure 8, the first two principal components accounted for 77.55% of the total variance in the data, with PC1 explaining 49.24% and PC2 accounting for 28.31% (eigenvalues: 2.462 and 1.416, respectively). The factor plane projection indicated that TA (−26.82%, based on correlation), AAB (−24.99%), and LAB parameters (−9.45%) contributed most negatively to PC1, while pH value contributed positively (32.25%). In PC2, pH value (11.03%), AAB (20.99%), and LAB parameters (47.84%) had positive contributions, whereas TA (−12.32%) and yeast count (−7.81%) contributed negatively.

The optimal neural network models (Figure 9, Table 11) for predicting kombucha parameters demonstrated strong generalization capabilities for the collected data and can accurately predict the observed outputs based on input parameters such as time and carrier selection. The model with six neurons yielded the highest r² values, with 0.903 for training and 0.946 during the testing cycle.

The training performance for the variables was as follows: 0.974 for pH value, 0.976 for TA, 0.641 for yeast count, 0.931 for AAB, and 0.995 for LAB. During testing, the r² values reached 0.979 for pH value, 0.993 for TA, 0.763 for yeast count, 0.993 for AAB, and 1.000 for LAB. Training accuracy improved as the number of cycles increased, stabilizing around the 40–50th epoch. Exceeding 50–60 epochs could lead to overfitting, so 60 epochs were sufficient to achieve high model accuracy without the risk of overfitting.

The verification analysis (Table 12), conducted across various fermentation parameters (pH, TA, yeasts, AAB, LAB), reveals a range of model performances with varying accuracy. Generally, the model demonstrates good predictive capabilities for most parameters, as evidenced by high r² values and relatively low RMSE values. Specifically, pH and TA exhibit strong agreement between predicted and observed values, with r² values of 0.946 and 0.956, respectively. LAB also shows a high r² value of 0.993, indicating excellent model performance for this parameter. However, the model exhibits comparatively lower accuracy for yeasts, with an r² value of 0.217 and a higher RMSE of 0.473. Similarly, AAB shows a moderate r² value of 0.890 and a higher RMSE of 0.276, suggesting some discrepancies between predicted and observed values for this parameter. The MBE and MPE values, while generally close to 0, indicate minor systematic biases for certain parameters. For instance, pH shows a slight underestimation with an MBE of −0.043, while yeasts exhibit a slight overestimation with an MBE of 0.208. Overall, the verification results demonstrate robust model performance for most fermentation parameters, particularly pH, TA, and LAB.

3.3. Step 3: LAB Stability During Storage

In the final step of the experimental and mathematical evaluation of the optimal carrier for LAB encapsulation and application in kombucha, whey protein was defined as the best choice according to the previously defined results. Therefore, LAB stability in this enriched kombucha was additionally evaluated during cold storage. The results obtained during cold storage of the whey protein–LAB-enriched kombucha (Supplementary Table S3) indicate that LAB stability slightly decreases over time during storage. Initially, LAB levels remain relatively stable between 0 and 4 days (p > 0.05), showing minor fluctuations. However, a slight decline is observed from day 6 onward (p < 0.05), as indicated by the change in superscript letters. Between days 6 and 21, LAB levels continue to decrease, but the differences among these time points are not statistically significant. This suggests that while LAB viability gradually declines over storage period, the most pronounced reduction occurs between days 4 and 6. On the other hand, the total decreasing rate is lower than 0.3 log CFU/mL, which is microbiologically acceptable and presents minor fluctuations in LAB population in the sample. As previously mentioned, the defined minimum of probiotic concentration for a beverage to obtain therapeutic and health-improving effects is 6–7 log CFU/mL of LAB cultures [42]. Enriched kombucha beverage in this study presented with LAB concentrations above the mentioned minimum for probiotic effect after fermentation and for all storage periods, and it can be concluded that the finalized beverage presents the targeted properties.

The diagrams in Figure 10 show LAB stability during storage in kombucha with whey protein–LAB encapsulates, with data points representing sample ages from 0 to 21 days. The results reveal a clear trend, with LAB levels generally decreasing as the material ages. A similar decreasing trend was also noticed by Majid et al. [80] during 28-day storage of a kombucha beverage with the addition of free L. plantarum cells.

Kinetic analysis using a four-parameter sigmoidal model indicates that the LAB stability of whey protein–LAB-enriched kombucha during storage showed slow rates of change, as reflected by the relatively low c and b coefficients (5.1 and 3.4, respectively). The parameters a and d reached similar values (7.7 and 7.4, respectively), suggesting a potentially stable LAB value during storage.

The results in Table 13 show that the kinetics model for LAB stability during storage provides an excellent fit to the data. The χ² value of 0.000 indicates no significant discrepancy between the observed and predicted values. With an RMSE of 0.017 and MBE of 0.000, the model’s predictions are accurate and unbiased. The low MPE (0.173) and SSE (0.002) further suggest minimal error in the model, while the AARD of 0.173 confirms the small average deviation between the predicted and observed values. The high r² value of 0.978 indicates that the model explains 97.8% of the variance in the data, demonstrating strong predictive power and reliability.

The optimal neural network models (Figure 11, Table 14) for predicting LAB stability during storage demonstrated strong generalization capabilities for the collected data and can accurately predict the observed outputs based on input parameters such as time and carrier selection. The model with eight neurons yielded the highest r² values, with 0.945 for training and 1.000 during the testing cycle.

The results in Table 15 indicate that the ANN model for predicting LAB stability during kombucha storage provides a good fit to the data. The χ² value of 0.005 suggests minimal discrepancy between the observed and predicted values. The RMSE of 0.066 and MBE of 0.040 demonstrate that the model’s predictions are both accurate and slightly biased, but within an acceptable range. The MPE of 0.687 indicates the average percentage error is relatively high, suggesting some variability in the predictions. The SSE of 0.030 further reflects low overall error in the model. The AARD of 0.687 confirms that the average deviation between the predicted and observed values is consistent. Finally, the r² value of 0.892 indicates that the model explains 89.2% of the variance in the LAB stability data, demonstrating strong predictive power.

3.4. Evaluation of Enriched Kombucha in Terms of Health-Promoting Potential and Sensory Attributes

As final step in this study, two crucial segments of the final enriched beverage were tested—their health-improving parameters and their sensory attributes. Namely, this study explores the impact of whey protein–LAB encapsulate enrichment on kombucha’s biological activities, with a particular focus on its antidiabetic, antihypertensive, and antihypercholesterolemic effects, and probiotic viability (Figure 12). All results are compared to the control kombucha sample.

Diabetes mellitus is the most common endocrine disease that affects many people worldwide and continues to grow yearly [85]. It is characterized by chronic hyperglycemia, which occurs due to impaired insulin secretion and/or action and is closely linked to the inhibition of digestive enzymes like α-glucosidase and α-amylase, which are crucial for carbohydrate digestion and glucose absorption [86]. The findings from this study reveal that enriched kombucha exhibits significantly higher inhibition of these enzymes compared to its control counterpart, suggesting an improved ability to regulate postprandial glucose levels. The results highlight a significant enhancement in α-glucosidase inhibition in enriched kombucha compared to the control, suggesting an improved ability to slow carbohydrate digestion and reduce postprandial blood glucose spikes. This increase in inhibitory potential is likely due to the higher polyphenolic content and the presence of bioactive peptides produced through probiotic fermentation. Similarly, α-amylase inhibition follows the same trend, demonstrating a stronger regulatory effect on starch breakdown. These results are aligned with research by Azadikhah et al. [87], where it was found that probiotic fermentation enhances bioactive peptide and polyphenol production, leading to increased inhibition of α-glucosidase and α-amylase. Similarly, previous studies have shown that Lactobacillus-fermented products can support glucose metabolism and improve insulin sensitivity, further reinforcing the antidiabetic potential of enriched kombucha [88,89].

Hypertension affects more than 1 billion adults worldwide and is associated with increased risk of various cardiovascular diseases that could ultimately lead to heart failure, stroke, and death [90]. The analysis of antihypertensive activity reveals that enriched kombucha exhibits a significantly higher degree of ACE (angiotensin-converting enzyme) inhibition compared to control kombucha. Hypertension is often managed by inhibiting the angiotensin-converting enzyme (ACE), which plays a role in blood pressure regulation [91]. This study demonstrates that enriched kombucha has a greater ACE inhibitory effect than the control sample, highlighting its potential to help manage hypertension. Additionally, tea polyphenols, naturally present in kombucha, have been linked to ACE inhibition and vascular relaxation, further contributing to its antihypertensive benefits [92]. Given the crucial role of ACE in blood pressure regulation by converting angiotensin I into angiotensin II, a potent vasoconstrictor, this increased inhibition suggests that enriched kombucha may contribute to blood pressure modulation. The observed enhancement in ACE inhibition is likely attributed to bioactive peptides derived from whey protein hydrolysates and probiotic fermentation, reinforcing prior research on the cardiovascular benefits of probiotic-fermented beverages [93].

In addition to its impact on glucose metabolism and blood pressure regulation, enriched kombucha demonstrated superior inhibition of HMG-CoA reductase (HMGCR), the key enzyme responsible for cholesterol biosynthesis. The increased inhibitory activity suggests greater potential for cholesterol-lowering effects, which could be linked to probiotic-induced modulation of lipid metabolism and enhanced bile salt hydrolase activity. By facilitating the reduction in cholesterol absorption, enriched kombucha may contribute to lowering LDL cholesterol levels, further supporting its role in cardiovascular health. Prior research supports this observation, with Begley et al. [94] demonstrating that probiotics can influence cholesterol metabolism by increasing bile salt hydrolase activity, thereby reducing cholesterol absorption. Additionally, whey protein hydrolysates have been associated with cholesterol-lowering effects, further validating the superior efficacy of enriched kombucha [95].

Another critical finding of the study is the probiotic presence in kombucha, which adds another dimension of kombucha’s functionality. The protective effects of whey protein encapsulation appear to enhance the survival of probiotic bacteria, ensuring greater stability in acidic environments. Since probiotic viability is essential for gut microbiota modulation, immune support, and overall metabolic health [40], the increased microbial stability in enriched kombucha suggests superior functional benefits. The results indicate that enriched kombucha has significant probiotic viability, confirming that whey protein encapsulation effectively preserves LAB stability. This aligns with findings from Sompach et al. [96], who reported that whey protein carriers enhance probiotic survival in acidic environments. Moreover, probiotic strains such as Lactobacillus have been shown to improve kombucha fermentation, leading to increased production of organic acids, mainly glucuronic and bioactive compounds, thus enhancing the antibacterial and antioxidant activities of kombucha [97]. These results suggest that enriched kombucha may offer greater probiotic benefits for gut health compared to conventional formulations.

It can be summarized that the enriched kombucha consistently outperforms the control kombucha across all measured health-promoting parameters. This enhanced bioactivity can be attributed to the synergistic effects of probiotic fermentation, whey protein encapsulation, and the increased bioavailability of polyphenols and bioactive peptides. The combined impact of these factors suggests that enriched kombucha offers a promising alternative for metabolic disease prevention, particularly in relation to diabetes, hypertension, and hypercholesterolemia. Future research should focus on evaluating the long-term stability of probiotics in enriched kombucha, as well as conducting clinical trials to validate its metabolic effects in human subjects.

Additionally, an analysis of sensory attributes and consumer acceptance was performed as one of the essential steps for completely assessing the market potential of enriched kombucha as a functional beverage with significant health benefits. Figure 13 presents a radar chart comparison of sensory attributes between the control kombucha (blue, dotted line) and the whey protein–LAB-encapsulate-enriched kombucha (orange, dotted line). The analysis covers key sensory descriptors: odor, taste, acidity, bitterness, sweetness, color tone, visual viscosity, and general impression.

Currently, no specific standards or regulations govern the sensory analysis of kombucha beverages. This lack of standardization largely stems from the vast diversity in microbial consortia, the wide range of production methods used both domestically and industrially, and the varying physicochemical properties of the ingredients used in kombucha preparation [61]. Given these complexities, sensory evaluations across different studies exhibit considerable variation. In this study, 16 trained panelists evaluated the mentioned descriptors (Figure 13). A comparative sensory analysis between the control and enriched kombucha reveals notable differences in key attributes, including odor, taste, acidity, bitterness, sweetness, color, and viscosity. The results suggest that the incorporation of whey protein–LAB encapsulate may have contributed to subtle but meaningful enhancements in the overall sensory profile. In terms of odor and taste, the enriched kombucha scored slightly higher, indicating a more favorable aroma and flavor profile. This improvement may be attributed to the presence of the LAB strain, probiotic fermentation, and the production of bioactive metabolites, which are known to contribute to the taste, texture, and aroma of the product, resulting in unique organoleptic characteristics [36]. Previous studies have demonstrated that microbial activity during fermentation can enhance aroma complexity and modify flavor compounds, potentially explaining the improved perception of the enriched formulation [98]. Bacterial growth in probiotic beverages leads to the consumption and/or production of compounds that change the aroma and flavor [99]. Wang et al. [83] concluded that the addition of different LAB strains to kombucha fermentation positively affected the production of volatile compounds that contributed to the flavor quality of kombucha and lactic acid that generally has softer taste.

Regarding acidity and bitterness, both samples exhibited comparable acidity levels, with only minor differences in sensory perception. However, the control kombucha appeared slightly more bitter than its enriched counterpart. This suggests that whey protein encapsulation may have contributed to a milder taste profile, potentially by modifying the organic acid composition or masking harsher flavors. Differences in sweetness and color tone were also observed. The control kombucha demonstrated slightly higher sweetness levels compared to the enriched version, likely due to variations in residual sugar content and fermentation kinetics. Probiotic-enriched formulations tend to exhibit higher sugar consumption rates, which could explain the reduced sweetness perception in the enriched kombucha. Additionally, subtle variations in color tone were noted, potentially influenced by protein interactions or changes in pigment stability resulting from encapsulation. Visual viscosity was another distinguishing factor between the two formulations. The enriched kombucha exhibited a higher viscosity, indicating a slightly thicker and more stable texture. This outcome aligns with the known effects of whey protein encapsulation, which can improve the overall sensory experience by increasing viscosity [99,100]. The general impression of the enriched kombucha was slightly more favorable, suggesting that the sensory modifications introduced by probiotic encapsulation were positively received. This finding is consistent with research indicating that probiotic-fortified fermented beverages often achieve improved consumer ratings due to their enhanced texture, balanced taste, and perceived health benefits [101].

Tenea et al.’s [101] results demonstrated that the addition of LAB strains to maracuya/coconut juice significantly influenced the sensory properties of the final product, with enriched juice being preferred for its taste and flavor over the control sample. Sensory analysis of enriched kombucha in the study of Majid et al. [80] is in accordance with the current study. Kombucha with added probiotic, the L. plantarum strain, was better accepted by panelists with higher ratings for aftertaste, taste, and overall impression. The color of enriched kombucha in their study seems to be generally the same as the control sample, which was not the case in the current study. This could be due to the tea used for preparation, but also due to the presence of whey protein as a carrier in the current study.

From a market potential perspective, the results indicate that whey protein–LAB encapsulate enrichment maintains or enhances the sensory characteristics of kombucha, making it a viable candidate for commercial development. However, consumer preferences can vary based on several factors, including sweetness levels, texture, and viscosity. While a thicker beverage may be associated with improved quality, excessive viscosity could be undesirable for certain consumers. Similarly, variations in odor and taste acceptance may necessitate further sensory optimization to align with broader market expectations. To ensure commercial success, future studies should incorporate a larger consumer panel and employ quantitative sensory evaluation techniques, including hedonic testing, to validate broader consumer acceptance. Additionally, targeted formulation adjustments may be required to optimize the balance between acidity, sweetness, and mouthfeel to cater to varying consumer preferences.

Additionally, a panel analysis was conducted using a Mixed Models—Type III Sum of Squares approach to evaluate the significance of various sources of variation, including products, assessors, and their interactions. ANOVA was performed on all dependent variables based on the following model:

(9)$Y=\mu +P+A+P\times A$

where P represents product (a fixed factor), while A (assessor) and P × A are random factors. The results of the panel analysis are summarized in Table 16.

The panel analysis conducted using Mixed Models—Type III Sum of Squares evaluated the significance of variation sources, specifically the effects of product and assessor, across multiple sensory descriptors. The product effect was found to be significant for color tone (F = 49.906, p < 0.001) and visual viscosity (F = 5.788, p = 0.029), indicating that these attributes differed significantly among the tested samples. No significant differences among products were observed for odor (F = 0.063, p = 0.806), taste (F = 2.612, p = 0.127), acidity (F = 3.030, p = 0.102), bitterness (F = 1.154, p = 0.300), sweetness (F = 0.319, p = 0.581), or general impression (F = 0.513, p = 0.485), suggesting that these descriptors did not effectively distinguish between the samples. The assessor effect was significant for odor (F = 5.251, p = 0.001), bitterness (F = 4.596, p = 0.003), color tone (F = 2.610, p = 0.036), visual viscosity (F = 8.622, p < 0.001), and general impression (F = 4.072, p = 0.005), indicating variability in individual evaluations for these attributes. Conversely, taste (F = 1.495, p = 0.223), acidity (F = 0.386, p = 0.963), and sweetness (F = 1.851, p = 0.122) did not exhibit significant variation among assessors, suggesting consistent evaluation of these descriptors. Sensory attributes that did not show significant differentiation across products, namely odor, taste, acidity, bitterness, sweetness, and general impression, were removed from further analysis. The results suggest that color tone and visual viscosity serve as the most discriminatory attributes among the examined samples, while the variability observed among assessors underscores the need for refined panel training to enhance the reliability of sensory evaluations.

The distance to consensus across descriptors was computed for each assessor concerning the two evaluated products (control kombucha and whey protein–LAB-enriched kombucha), providing insights into the level of agreement among panelists (Figure 14). Lower values indicate a closer alignment with the panel consensus, whereas higher values suggest greater deviation from the group assessment. For control kombucha, the distance to consensus ranged from 0.265 (panelists 6, 9, and 16) to 1.856 (panelist 3), indicating notable variability in individual assessments. Several assessors exhibited a strong alignment with the panel’s consensus, with multiple values clustering around 0.834 and 0.972, suggesting moderate agreement among the majority of panelists. For the whey protein–LAB-enriched kombucha, the distance to consensus varied between 0.337 (panelists 9, 12, and 14) and 1.597 (panelist 3). The overall variability appeared slightly higher compared to the control kombucha, with some assessors (e.g., panelists 2 and 3) showing a greater deviation from the consensus. Notably, panelist 3 exhibited the highest deviation for both products (1.856 for control kombucha and 1.597 for whey protein–LAB-enriched kombucha), suggesting possible inconsistencies in their evaluations. Conversely, panelists 6 and 9 demonstrated the lowest distance to consensus, implying a high level of agreement with the panel. The analysis suggests that control kombucha exhibited slightly greater variability in panelist agreement, while the whey protein–LAB-enriched kombucha showed a more consistent evaluation pattern. The presence of high deviations in certain assessors highlights the potential need for panel training to improve alignment in sensory assessments.

4. Conclusions

This research outlines a methodical, multi-stage approach to identifying the most effective probiotic carrier for Lactobacillus rhamnosus within kombucha. Five potential probiotic carriers were assessed, polysaccharides pectin, inulin, and maltodextrin, and proteins pea and whey, regarding their stability in several pH level solutions, with pectin, inulin, and maltodextrin exhibiting weaker stability at lower pH levels compared to pea and whey proteins. During the next step, kombucha fermentation, pea and whey protein carriers were monitored, with whey protein showing greater stability regarding the protection of LAB cells, as well as better influence on the fermentation process. Overall, whey protein stood out, offering notable protection to probiotic cells across a range of pH conditions, accelerating fermentation processes, and preserving microbial viability throughout cold storage. The kombucha enriched with this encapsulated probiotic not only retained a higher concentration of live bacteria but also showed improved health-promoting attributes, including antidiabetic, antihypertensive, and cholesterol-lowering effects. Comprehensive validation—through advanced statistical analysis, kinetic modeling, and predictions using artificial neural networks—supported the reliability of these results. Furthermore, sensory analysis revealed that consumer acceptance remained high, with minimal alterations in taste and texture. Overall, whey protein encapsulation presents a compelling strategy for creating probiotic-rich kombucha that balances stability, functionality, and consumer appeal. Future investigations may focus on scaling up production and assessing long-term benefits through clinical studies.

Author Contributions

Conceptualization, T.B. and A.R.; methodology, A.R.; software, L.P.; validation, O.Š., L.P. and T.B.; formal analysis, T.B., T.C. and A.V.; investigation, A.R.; resources, D.C.; data curation, T.B. and L.P.; writing—original draft preparation, T.B.; writing—review and editing, O.Š. and A.R.; visualization, L.P.; supervision, D.C. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

The study was approved by the Ethics Committee of the Faculty of Technology Novi Sad, University of Novi Sad, Serbia (Ref. No. 020-8/26).

Informed Consent Statement

Not applicable.

Data Availability Statement

The original contributions presented in the study are included in the article and Supplement Material; further inquiries can be directed to the corresponding authors.

Conflicts of Interest

The authors declare no conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1 Research setup for multi-criteria decision-making for optimal probiotic carrier selection in kombucha. Statistically significant correlations are marked with asterisks: * p < 0.001; ** p < 0.01; *** p < 0.05.

Figure 2 Kinetic modeling of probiotic survival (log CFU/mL) in different carriers: (a) pea protein, (b) whey protein, (c) maltodextrin, (d) inulin, and (e) pectin (points indicate calculated results, while lines present experimentally obtained results).

Figure 3 Cluster analysis for probiotic survival experiment in different carriers.

Figure 4 The PCA biplot diagram, depicting the relationships between probiotic survival experiments in different carriers.

Figure 5 ANN models for probiotic survival experiments in different carriers, (a) whey protein, (b) pea protein, (c) maltodextrin, (d) inulin, (e) pectin.

Figure 6 Kinetic modeling of fermentation parameters for different carriers: (a) pH, (b) TA, (c) yeast count, (d) AAB, (e) LAB (points indicate calculated results, while lines present experimentally obtained results).

Figure 7 Cluster analysis of fermentation parameters.

Figure 8 The PCA biplot diagram, depicting the relationships among pH, TA, yeasts, AAB, and LAB parameters during fermentation.

Figure 9 ANN models for pH value, TA, yeasts, AAB, and LAB parameters during kombucha fermentation, (a) pH, (b) TA, (c) Yeast, (d) acetic acid, (e) LAB.

Figure 10 Kinetic modeling of LAB stability during storage in kombucha with whey protein–LAB encapsulates (points indicate calculated results, while lines present experimentally obtained results).

Figure 11 ANN model for LAB stability during storage prediction.

Figure 12 Differences in health-promoting parameters between control kombucha and enriched kombucha with whey protein–LAB encapsulate.

Figure 13 Sensory analysis of control kombucha and kombucha enriched with whey protein–LAB encapsulate.

Figure 14 Distance to consensus computed across descriptors.

Table 1

Kinetics model parameters (a, b, c, and d) for a four-parameter sigmoidal mathematical model of probiotic survival in different carriers.

Kinetic Parameter	Type of Carrier
	Pea Protein				Whey Protein				Maltodextrin				Inulin				Pectin
	2.5	3	4	4.5	2.5	3	4	4.5	2.5	3	4	4.5	2.5	3	4	4.5	2.5	3	4	4.5
d	7.3	7.7	4.5	7.7	6.8	7.7	6.2	0.0	0.0	6.3	6.5	6.6	0.0	8.0	8.0	8.2	0.0	1.8	0.0	0.0
a	8.1	8.1	8.2	8.1	8.3	8.9	8.6	8.6	7.9	8.7	8.8	8.7	6.9	7.6	8.0	7.9	4.2	6.4	6.5	6.7
c	28.4	4.0	83.1	1.0	7.6	32.5	9.8	21.3	1.6	2.4	2.3	2.6	1.2	4.2	5.0	28.4	2.2	10.2	115	25.8
b	0.0	1.7	0.5	18.0	5.5	0.6	5.9	2.7	1.2	6.1	7.4	4.6	2.2	15.9	0.0	0.0	2.5	0.6	0.5	1.1

Table 2

Verification of kinetics model of probiotic survival in different carriers.

Type of Carrier	pH Value	Verification Parameters
		χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
Pea protein	2.5	0.011	0.099	0.000	0.790	0.069	0.790	0.665
	3	0.024	0.143	0.000	1.507	0.144	1.507	0.410
	4	0.035	0.173	0.001	1.713	0.210	1.713	0.716
	4.5	0.010	0.091	0.000	0.897	0.058	0.897	0.723
Whey protein	2.5	0.010	0.091	0.000	0.788	0.058	0.788	0.963
	3	0.008	0.084	0.002	0.670	0.049	0.670	0.644
	4	0.018	0.124	0.000	1.009	0.107	1.009	0.926
	4.5	0.011	0.099	0.000	1.010	0.069	1.010	0.921
Maltodextrin	2.5	0.217	0.431	−0.051	7.309	1.300	7.309	0.970
	3	0.010	0.094	0.000	1.134	0.062	1.134	0.992
	4	0.006	0.069	0.000	0.744	0.033	0.744	0.996
	4.5	0.002	0.040	0.000	0.439	0.011	0.439	0.998
Inulin	2.5	0.106	0.302	0.000	27.535	0.638	27.535	0.984
	3	0.021	0.135	0.000	1.088	0.128	1.088	0.599
	4	0.044	0.193	0.000	1.845	0.261	1.845	0.024
	4.5	0.004	0.055	0.000	0.535	0.021	0.535	0.355
Pectin	2.5	0.156	0.366	−0.130	5.817	0.936	5.817	0.968
	3	0.064	0.233	0.000	3.670	0.381	3.670	0.901
	4	0.157	0.367	0.000	4.757	0.943	4.757	0.646
	4.5	0.080	0.261	−0.001	4.074	0.479	4.074	0.834

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 3

Score table for average probiotic survival in different carriers.

Time (Days)	Type of Carrier
	Pea Protein	Whey Protein	Maltodextrin	Inulin	Pectin
0	8.129	8.678	8.567	7.609	6.074
1	7.959	8.596	7.660	6.966	5.207
2	7.724	8.376	7.128	6.306	4.932
3	7.744	8.621	5.799	5.941	4.538
4	7.896	8.451	5.530	6.250	3.938
7	7.455	8.198	5.178	5.983	4.025
10	7.701	7.621	4.780	6.022	3.287
Average	7.801	8.363	6.377	6.440	4.572
Score	0.852	1.000	0.476	0.493	0.000

Table 4

Correlation matrix for probiotic survival in different carriers.

Carrier	Whey Protein	Maltodextrin	Inulin	Pectin
Pea protein	0.472 *	0.361 n	0.162 n	0.225 n
Whey protein		0.658 ***	0.485 **	0.602 **
Maltodextrin			0.902 ***	0.937 ***
Inulin				0.918 ***

*** Statistically significant correlation at p < 0.001 level; ** statistically significant correlation at p < 0.01 level; * statistically significant correlation at p < 0.05 level; ⁿ statistically insignificant correlation.

Table 5

ANN model performance for probiotic survival prediction in different carriers.

Type of Carrier	Network Name	Training Performace	Test Performance	Training Error	Test Error	Training Algorithm	Error Function	Hidden Activation	Output Activation
Pea protein	MLP 2-8-1	0.755	0.592	0.007	0.016	BFGS 71	SOS	Exponential	Identity
Whey protein	MLP 2-7-1	0.868	0.927	0.015	0.009	BFGS 36	SOS	Logistic	Exponential
Maltodextrin	MLP 2-5-1	0.995	0.995	0.018	0.014	BFGS 118	SOS	Tanh	Exponential
Inulin	MLP 2-9-1	0.943	0.744	0.355	0.014	BFGS 29	SOS	Logistic	Identity
Pectin	MLP 2-3-1	0.980	0.931	0.049	0.023	BFGS 58	SOS	Tanh	Identity

Table 6

Verification of ANN models of probiotic survival in different carriers.

Type of Carrier	Verification Parameters
	χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
Pea protein	0.019	0.136	−0.018	1.422	0.522	1.422	0.694
Whey protein	0.028	0.164	0.016	1.609	0.756	1.609	0.860
Maltodextrin	0.035	0.183	0.030	2.545	0.940	2.545	0.994
Inulin	0.535	0.718	0.040	32.162	14.434	32.162	0.948
Pectin	0.087	0.290	0.030	4.372	2.348	4.372	0.979

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 7

Kinetics parameters (a, b, c, and d) for a four-parameter sigmoidal mathematical model of kombuchas.

Sample	Kinetic Parameters	Kombucha Parameter
		pH	TA	Yeasts	AAB	LAB
Pea protein-LAB enriched kombucha	d	3.0	5.2	9.1	2.6	8.5
	a	4.8	0.3	5.3	6.2	5.8
	c	1.4	2.5	16.5	159.1	9.4
	b	2.4	3.0	0.4	4.2	0.9
Whey protein-LAB enriched kombucha	d	2.4	20.3	6.5	3.0	9.6
	a	4.3	0.3	5.6	7.5	6.8
	c	1.3	6.4	12.6	25.7	8.4
	b	2.1	1.9	0.0	10.4	0.8
Control kombucha	d	2.8	5.9	7.2	3.3	0.0
	a	4.2	0.5	5.3	5.8	0.0
	c	1.8	2.6	16.6	23.1	13.3
	b	4.0	4.2	0.0	10.8	1.4

Table 8

Verification of kinetics model of control and enriched kombuchas.

Sample	Kombucha Parameter	Verification Paraemeters
		χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
Pea protein–LAB-enriched kombucha	pH	0.000	0.010	0.000	0.233	0.000	0.233	1.000
	TA	0.000	0.013	0.000	1.456	0.001	1.456	1.000
	Yeasts	0.004	0.053	0.000	0.615	0.014	0.615	0.989
	AAB	0.128	0.320	0.000	4.094	0.510	4.094	0.066
	LAB	0.025	0.142	−0.010	1.974	0.101	1.974	0.806
Whey protein–LAB-enriched kombucha	pH	0.003	0.050	0.000	1.394	0.012	1.394	0.994
	TA	0.040	0.178	0.000	15.590	0.158	15.590	0.993
	Yeasts	0.055	0.210	0.000	2.493	0.220	2.493	0.438
	AAB	0.131	0.324	0.000	3.377	0.525	3.377	0.285
	LAB	0.009	0.085	0.000	0.901	0.036	0.901	0.943
Control	pH	0.000	0.005	0.000	0.139	0.000	0.139	1.000
	TA	0.001	0.027	0.000	3.261	0.004	3.261	1.000
	Yeasts	0.196	0.396	0.000	5.189	0.783	5.189	0.477
	AAB	0.160	0.358	0.000	5.600	0.639	5.600	0.072
	LAB	0.000	0.000	0.000	0.000	0.000	0.000	0.000

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 9

Score table for average probiotic survival in kombucha.

Pea protein–LAB-enriched kombucha					Score
pH	TA	Yeasts	AAB	LAB
0.000	0.000	1.000	0.267	0.852	0.424
Whey protein–LAB-enriched kombucha
pH	TA	Yeasts	AAB	LAB
1.000	1.000	0.000	1.000	1.000	0.800
Control kombucha
pH	TA	Yeasts	AAB	LAB
0.636	0.570	0.365	0.000	0.000	0.314

Table 10

Correlation matrix for kombucha fermentation process.

Parameters	TA	Yeasts	AAB	LAB
pH	−0.895 ***	−0.358 n	−0.515 *	−0.058 n
TA		0.203 n	0.353 n	0.067 n
Yeasts			0.105 n	0.038 n
AAB				0.744 **

*** Statistically significant correlation at p < 0.001 level; ** statistically significant correlation at p < 0.01 level; * statistically significant correlation at p < 0.05 level; ⁿ statistically insignificant correlation.

Table 11

ANN model performance for kombucha fermentation parameter prediction.

Network Name	Training Performance	Test Performance	Training Error	Test Error
MLP 4-6-5	0.903	0.946	0.318	0.423
	Training algorithm	Error function	Hidden activation	Output activation
	BFGS 8	SOS	Identity	Logistic

Table 12

Verification of ANN models of kombucha fermentation parameter prediction.

Kombucha Parameter	Verification Parameter
	χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
pH	0.035	0.180	−0.043	4.280	0.484	4.280	0.946
TA	0.218	0.451	−0.042	32.581	3.052	32.581	0.956
Yeasts	0.240	0.473	0.208	6.673	3.359	6.673	0.217
AAB	0.082	0.276	0.052	3.687	1.146	3.687	0.890
LAB	0.169	0.396	−0.011	3.274	2.346	3.274	0.993

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 13

Verification of kinetics model of LAB stability during storage.

Storage Parameter	Verification Parameters
	χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
LAB	0.000	0.017	0.000	0.173	0.002	0.173	0.978

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 14

ANN model performance in prediction of LAB stability during kombucha storage.

NetworkName	Training Performance	Test Performance	TrainingError	TestError
MLP 1-8-1	0.945	1.000	0.002	0.001
	Training algorithm	Error function	Hidden activation	Output activation
	BFGS 3	SOS	Tanh	Tanh

Table 15

Verification of ANN model predictions of LAB stability during kombucha storage.

Storage Parameter	Verification Parameter
	χ 2	RMSE	MBE	MPE	SSE	AARD	r 2
LAB	0.005	0.066	0.040	0.687	0.030	0.687	0.892

Legend: χ²—reduced chi-square; RMSE—root mean square error; MBE—mean bias error; MPE—mean percentage error; SSE—sum of squared errors; AARD—average absolute relative deviation; r²—coefficient of determination.

Table 16

Panel analysis for kombucha samples.

	Product		Assessor
Descriptors	F	Pr > F	F	Pr > F
Odor	0.063	0.806	5.251	0.001
Taste	2.612	0.127	1.495	0.223
Acidity	3.030	0.102	0.386	0.963
Bitterness	1.154	0.300	4.596	0.003
Sweetness	0.319	0.581	1.851	0.122
Color tone	49.906	<0.001	2.610	0.036
Visual viscosity	5.788	0.029	8.622	<0.001
General impression	0.513	0.485	4.072	0.005

Values displayed in bold correspond to descriptors that were removed.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/fermentation11050256/s1: Figure S1. Tea fungus used for kombucha preparation; Figure S2. Kombucha beverage obtained at the Laboratory of Microbiology at the Faculty of Technology Novi Sad, Serbia; Table S1: Probiotic survival (log CFU/mL) in different carriers; Table S2: Kombucha fermentation (parameters: pH, TA, number of yeasts, acetic acid bacteria and lactic acid bacteria); Table S3: LAB stability during storage.

References

1. Selvaraj, S.; Kalaichelvan, G. An overview of probiotic health booster-kombucha tea. Chin. Herb. Med.; 2023; 15, pp. 27-32. [DOI: https://dx.doi.org/10.1016/j.chmed.2022.06.010] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36875441]

2. Coelho, R.M.D.; de Almeida, A.L.; do Amaral, R.Q.G.; da Mota, R.N.; de Sousa, P.H.M. Kombucha. Int. J. Gastron. Food Sci.; 2020; 22, 100272. [DOI: https://dx.doi.org/10.1016/j.ijgfs.2020.100272]

3. Vitas, J.S.; Cvetanović, A.D.; Mašković, P.Z.; Švarc-Gajić, J.V.; Malbaša, R.V. Chemical composition and biological activity of novel types of kombucha beverages with yarrow. J. Funct. Foods; 2018; 44, pp. 95-102. [DOI: https://dx.doi.org/10.1016/j.jff.2018.02.019]

4. Battikh, H.; Bakhrouf, A.; Ammar, E. Antimicrobial effect of Kombucha analogues. LWT Food Sci. Technol.; 2012; 47, pp. 71-77. [DOI: https://dx.doi.org/10.1016/j.lwt.2011.12.033]

5. Bueno, F.; Chouljenko, A.; Sathivel, S. Development of coffee kombucha containing Lactobacillus rhamnosus and Lactobacillus casei: Gastrointestinal simulations and DNA microbial analysis. LWT; 2021; 142, 110980. [DOI: https://dx.doi.org/10.1016/j.lwt.2021.110980]

6. Sharifudin, S.A.; Ho, W.Y.; Yeap, S.K.; Abdullah, R.; Koh, S.P. Fermentation and characterisation of potential kombucha cultures on papaya-based substrates. LWT; 2021; 151, 112060. [DOI: https://dx.doi.org/10.1016/j.lwt.2021.112060]

7. Liamkaew, R.; Chattrawanit, J.; Danvirutai, P. Kombucha production by combinations of black tea and apple juice. Prog. Appl. Sci. Technol.; 2016; 6, pp. 139-146.

8. Vukmanović, S.; Vitas, J.; Malbaša, R. Valorization of winery effluent using kombucha culture. J. Food Process. Preserv.; 2020; 44, e14627. [DOI: https://dx.doi.org/10.1111/jfpp.14627]

9. Phung, L.T.; Kitwetcharoen, H.; Chamnipa, N.; Boonchot, N.; Thanonkeo, S.; Tippayawat, P.; Klanrit, P.; Yamada, M.; Thanonkeo, P. Changes in the chemical compositions and biological properties of kombucha beverages made from black teas and pineapple peels and cores. Sci. Rep.; 2023; 13, 7859. [DOI: https://dx.doi.org/10.1038/s41598-023-34954-7]

10. Bishop, P.; Pitts, E.R.; Budner, D.; Thompson-Witrick, K.A. Chemical composition of kombucha. Beverages; 2022; 8, 45. [DOI: https://dx.doi.org/10.3390/beverages8030045]

11. Watawana, M.I.; Jayawardena, N.; Gunawardhana, C.B.; Waisundara, V.Y. Health, wellness, and safety aspects of the consumption of kombucha. J. Chem.; 2015; 1, 591869. [DOI: https://dx.doi.org/10.1155/2015/591869]

12. de Miranda, J.F.; Ruiz, L.F.; Silva, C.B.; Uekane, T.M.; Silva, K.A.; Gonzalez, A.G.M.; Fernandes, F.F.; Lima, A.R. Kombucha: A review of substrates, regulations, composition, and biological properties. J. Food Sci.; 2022; 87, pp. 503-527. [DOI: https://dx.doi.org/10.1111/1750-3841.16029] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35029317]

13. Al-Mohammadi, A.R.; Ismaiel, A.A.; Ibrahim, R.A.; Moustafa, A.H.; Abou Zeid, A.; Enan, G. Chemical constitution and antimicrobial activity of kombucha fermented beverage. Molecules; 2021; 26, 5026. [DOI: https://dx.doi.org/10.3390/molecules26165026] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34443614]

14. Mizuta, A.G.; de Menezes, J.L.; Dutra, T.V.; Ferreira, T.V.; Castro, J.C.; da Silva, C.A.J.; Pilau, E.J.; Junior, M.M.; de Abreu Filho, B.A. Evaluation of antimicrobial activity of green tea kombucha at two fermentation time points against Alicyclobacillus spp. LWT; 2020; 130, 109641. [DOI: https://dx.doi.org/10.1016/j.lwt.2020.109641]

15. Monia, D.; Chriaa, J.; Battikh, H.; Abid, K.; Bakhrouf, A. Antiproliferative and antimicrobial activities of kombucha tea. Afr. J. Microbiol. Res.; 2013; 7, pp. 3466-3470.

16. Rasouli, L.; Aryaeian, N.; Gorjian, M.; Nourbakhsh, M.; Amiri, F. Evaluation of cytotoxicity and anticancer activity of kombucha and doxorubicin combination therapy on colorectal cancer cell line HCT-116. J. Educ. Health Promot.; 2021; 10, 376. [DOI: https://dx.doi.org/10.4103/jehp.jehp_1456_20]

17. Villarreal-Soto, S.A.; Beaufort, S.; Bouajila, J.; Souchard, J.P.; Renard, T.; Rollan, S.; Taillandier, P. Impact of fermentation conditions on the production of bioactive compounds with anticancer, anti-inflammatory and antioxidant properties in kombucha tea extracts. Process Biochem.; 2019; 83, pp. 44-54. [DOI: https://dx.doi.org/10.1016/j.procbio.2019.05.004]

18. Zubaidah, E.; Afgani, C.A.; Kalsum, U.; Srianta, I.; Blanc, P.J. Comparison of in vivo antidiabetes activity of snake fruit Kombucha, black tea Kombucha and metformin. Biocatal. Agric. Biotechnol.; 2019; 17, pp. 465-469. [DOI: https://dx.doi.org/10.1016/j.bcab.2018.12.026]

19. Xu, S.; Wang, Y.; Wang, J.; Geng, W. Kombucha reduces hyperglycemia in type 2 diabetes of mice by regulating gut microbiota and its metabolites. Foods; 2022; 11, 754. [DOI: https://dx.doi.org/10.3390/foods11050754]

20. Jakubczyk, K.; Kałduńska, J.; Kochman, J.; Janda, K. Chemical profile and antioxidant activity of the kombucha beverage derived from white, green, black and red tea. Antioxidants; 2020; 9, 447. [DOI: https://dx.doi.org/10.3390/antiox9050447]

21. Nurikasari, M.; Puspitasari, Y.; Siwi, R.P.Y. Characterization and analysis kombucha tea antioxidant activity based on long fermentation as a beverage functional. J. Glob. Res. Public Health; 2017; 2, pp. 48-52.

22. Abaci, N.; Deniz, F.S.S.; Orhan, I.E. Kombucha–An ancient fermented beverage with desired bioactivities: A narrowed review. Food Chem.; 2022; 14, 100302. [DOI: https://dx.doi.org/10.1016/j.fochx.2022.100302] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35434600]

23. Vargas, B.K.; Fabricio, M.F.; Ayub, M.A.Z. Health effects and probiotic and prebiotic potential of Kombucha: A bibliometric and systematic review. Food Biosci.; 2021; 44, 101332. [DOI: https://dx.doi.org/10.1016/j.fbio.2021.101332]

24. Antolak, H.; Piechota, D.; Kucharska, A. Kombucha tea—A double power of bioactive compounds from tea and symbiotic culture of bacteria and yeasts (SCOBY). Antioxidants; 2021; 10, 1541. [DOI: https://dx.doi.org/10.3390/antiox10101541]

25. Marsh, A.J.; O’Sullivan, O.; Hill, C.; Ross, R.P.; Cotter, P.D. Sequence-based analysis of the bacterial and fungal compositions of multiple kombucha (tea fungus) samples. Food Microbiol.; 2014; 38, pp. 171-178. [DOI: https://dx.doi.org/10.1016/j.fm.2013.09.003]

26. Harrison, K.; Navarro, R.; Jensen, K.; Cayler, W.; Nielsen, T.; Curtin, C. Live, Probiotic, or Neither? Microbial Composition of Retail-Available Kombucha and “Hard” Kombucha in the Pacific Northwest of the United States. Beverages; 2023; 9, 59. [DOI: https://dx.doi.org/10.3390/beverages9030059]

27. Cvetković, D.; Ranitović, A.; Budimac, T.; Šovljanski, O. Examination of Lactobacillus plantarum and Lactobacillus rhamnosus during kombucha fermentation. Proceedings of the 2nd International Symposium on Biotechnology; Čačak, Republic of Serbia, 14–15 March 2024; pp. 409-414.

28. Sun, W.; Nguyen, Q.D.; Sipiczki, G.; Ziane, S.R.; Hristovski, K.; Friedrich, L.; Visy, A.; Hitka, G.; Gere, A.; Bujna, E. Microencapsulation of Lactobacillus plantarum 299v Strain with Whey Proteins by Lyophilization and Its Application in Production of Probiotic Apple Juices. Appl. Sci.; 2023; 13, 318. [DOI: https://dx.doi.org/10.3390/app13010318]

29. Bromley, A.; Perry, J. Survival of Probiotic Lactobacillus spp. During Kombucha Fermentation. Curr. Dev. Nutr.; 2022; 6, 507. [DOI: https://dx.doi.org/10.1093/cdn/nzac077.010]

30. Fu, C.; Yan, F.; Cao, Z.; Xie, F.; Lin, J. Antioxidant activities of kombucha prepared from three different substrates and changes in content of probiotics during storage. Food Sci. Technol.; 2014; 34, pp. 123-126. [DOI: https://dx.doi.org/10.1590/S0101-20612014005000012]

31. Mozzi, F. Lactic Acid Bacteria. Encyclopedia of Food and Health; Caballero, B.; Finglas, P.M.; Toldrá, F. Elsevier: Amsterdam, The Netherlands, 2016; pp. 501-508.

32. Munoz, R.; Moreno-Arribas, M.V.; Rivas, B. Lactic Acid Bacteria; Elsevier: Amsterdam, The Netherlands, 2011; Volume 1, pp. 191-226.

33. Dempsey, E.; Corr, S.C. Lactobacillus spp. for gastrointestinal health: Current and future perspectives. Front. Immunol.; 2022; 13, 840245. [DOI: https://dx.doi.org/10.3389/fimmu.2022.840245]

34. Bintsis, T. Lactic acid bacteria: Their applications in foods. J. Bacteriol. Mycol.; 2018; 6, pp. 89-94.

35. Minj, J.; Chandra, P.; Paul, C.; Sharma, R.K. Bio-functional properties of probiotic Lactobacillus: Current applications and research perspectives. Crit. Rev. Food Sci. Nutr.; 2021; 61, pp. 2207-2224. [DOI: https://dx.doi.org/10.1080/10408398.2020.1774496] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32519883]

36. Ayivi, R.D.; Gyawali, R.; Krastanov, A.; Aljaloud, S.O.; Worku, M.; Tahergorabi, R.; Silva, R.C.D.; Ibrahim, S.A. Lactic acid bacteria: Food safety and human health applications. Dairy; 2020; 1, pp. 202-232. [DOI: https://dx.doi.org/10.3390/dairy1030015]

37. FAO/WHO. The Food and Agriculture Organization of the United Nations and the World Health Organization Joint FAO/WHO Expert Consultation on the Evaluation of Health and Nutritional Properties of Probiotics in Food Including Powder Milk with Live Lactic Acid Bacteria FAO/WHO Report No. 10-1-2001; 2001. Available online: https://openknowledge.fao.org/items/db384295-64d9-47e2-b65b-3c918efc5140 (accessed on 10 January 2025).

38. Rajam, R.; Subramanian, P. Encapsulation of probiotics: Past, present and future. Beni Suef Univ. J. Basic Appl. Sci.; 2022; 11, 46. [DOI: https://dx.doi.org/10.1186/s43088-022-00228-w]

39. Misra, S.; Pooja, P.; Hari, N.M. Novel approaches for co-encapsulation of probiotic bacteria with bioactive compounds, their health benefits and functional food product development: A review. Trends Food Sci. Technol.; 2021; 109, pp. 340-351. [DOI: https://dx.doi.org/10.1016/j.tifs.2021.01.039]

40. Kowalska, E.; Ziarno, M.; Ekielski, A.; Żelaziński, T. Materials Used for the Microencapsulation of Probiotic Bacteria in the Food Industry. Molecules; 2022; 27, 3321. [DOI: https://dx.doi.org/10.3390/molecules27103321]

41. Mabrouk, A.M.; Salama, H.H.; El Sayed, H.S.; El Sayed, S.M. Preparation of symbiotic whey protein gel as a carrier of free and encapsulated probiotic bacteria. J. Food Process. Preserv.; 2021; 45, 15612. [DOI: https://dx.doi.org/10.1111/jfpp.15612]

42. Zhu, W.; Lyu, F.; Naumovski, N.; Ajlouni, S.; Ranadheera, C.S. Functional Efficacy of Probiotic Lactobacillus sanfranciscensis in Apple, Orange and Tomato Juices with Special Reference to Storage Stability and In Vitro Gastrointestinal Survival. Beverages; 2020; 6, 13. [DOI: https://dx.doi.org/10.3390/beverages6010013]

43. Krunić, T.Ž.; Obradović, N.S.; Rakin, M.B. Application of whey protein and whey protein hydrolysate as protein based carrier for probiotic starter culture. Food Chem.; 2019; 293, pp. 74-82. [DOI: https://dx.doi.org/10.1016/j.foodchem.2019.04.062]

44. Rama, G.R.; Führ, A.J.; da Silva, J.A.B.S.; Gennari, A.; Giroldi, M.; Goettert, M.I.; Volken de Souza, C.F. Encapsulation of Lactobacillus spp. using bovine and buffalo cheese whey and their application in orange juice. 3 Biotech; 2020; 10, 263. [DOI: https://dx.doi.org/10.1007/s13205-020-02255-9]

45. Gul, O.; Atalar, I. Different stress tolerance of spray and freeze dried Lactobacillus casei Shirota microcapsules with different encapsulating agents. Food Sci. Biotechnol.; 2019; 28, pp. 807-816. [DOI: https://dx.doi.org/10.1007/s10068-018-0507-x] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31093438]

46. Tarifa, M.C.; Piqueras, C.M.; Genovese, D.B.; Brugnoni, L.I. Microencapsulation of Lactobacillus casei and Lactobacillus rhamnosus in pectin and pectin-inulin microgel particles: Effect on bacterial survival under storage conditions. Int. J. Biol. Macromol.; 2021; 179, pp. 457-465. [DOI: https://dx.doi.org/10.1016/j.ijbiomac.2021.03.038] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33711368]

47. Barbosa, J.; Brandão, T.R.; Teixeira, P. Spray drying conditions for orange juice incorporated with lactic acid bacteria. Int. J. Food Sci. Technol.; 2017; 52, pp. 1951-1958. [DOI: https://dx.doi.org/10.1111/ijfs.13472]

48. Frakolaki, G.; Kekes, T.; Lympaki, F.; Giannou, V.; Tzia, C. Use of encapsulated Bifidobacterium animalis subsp. lactis through extrusion or emulsification for the production of probiotic yogurt. J. Food Process Eng.; 2022; 45, e13792.

49. Gharibzahedi, S.M.T.; Smith, B. Legume proteins are smart carriers to encapsulate hydrophilic and hydrophobic bioactive compounds and probiotic bacteria: A review. Compr. Rev. Food Sci. Food Saf.; 2021; 20, pp. 1250-1279. [DOI: https://dx.doi.org/10.1111/1541-4337.12699]

50. Nguyen, T.H.; Kim, Y.; Kim, J.S.; Jeong, Y.; Park, H.M.; Kim, J.W.; Kim, J.E.; Kim, H.; Paek, N.S.; Kang, C.H. Evaluating the cryoprotective encapsulation of the lactic acid bacteria in simulated gastrointestinal conditions. Biotechnol. Bioprocess Eng.; 2020; 25, pp. 287-292. [DOI: https://dx.doi.org/10.1007/s12257-019-0406-x]

51. Varela-Pérez, A.; Romero-Chapol, O.O.; Castillo-Olmos, A.G.; García, H.S.; Suárez-Quiroz, M.L.; Singh, J.; Figueroa-Hernández, C.Y.; Viveros-Contreras, R.; Cano-Sarmiento, C. Encapsulation of Lactobacillus gasseri: Characterization, Probiotic Survival, In Vitro Evaluation and Viability in Apple Juice. Foods; 2022; 11, 740. [DOI: https://dx.doi.org/10.3390/foods11050740]

52. Afzaal, M.; Saeed, F.; Saeed, M.; Ahmed, A.; Ateeq, H.; Nadeem, M.T.; Tufail, T. Survival and stability of free and encapsulated probiotic bacteria under simulated gastrointestinal conditions and in pasteurized grape juice. J. Food Process. Preserv.; 2020; 44, e14346. [DOI: https://dx.doi.org/10.1111/jfpp.14346]

53. Romero-Chapol, O.O.; Varela-Pérez, A.; Castillo-Olmos, A.G.; García, H.S.; Singh, J.; García-Ramírez, P.J.; Viveros-Contreras, R.; Figueroa-Hernández, C.Y.; Cano-Sarmiento, C. Encapsulation of Lacticaseibacillus rhamnosus GG: Probiotic Survival, In Vitro Digestion and Viability in Apple Juice and Yogurt. Appl. Sci.; 2022; 12, 2141. [DOI: https://dx.doi.org/10.3390/app12042141]

54. Afzaal, M.; Saeed, F.; Arshad, M.U.; Nadeem, M.T.; Saeed, M.; Tufail, T. The effect of encapsulation on the stability of probiotic bacteria in ice cream and simulated gastrointestinal conditions. Probiotics Antimicrob. Proteins; 2019; 11, pp. 1348-1354. [DOI: https://dx.doi.org/10.1007/s12602-018-9485-9]

55. Pan, G.; Wang, F.; Shang, C.; Wu, H.; Wu, G.; Gao, J.; Wang, S.; Gao, Z.; Zhou, X.; Mao, X. Advances in machine learning-and artificial intelligence-assisted material design of steels. Int. J. Miner. Metall. Mater.; 2023; 30, pp. 1003-1024. [DOI: https://dx.doi.org/10.1007/s12613-022-2595-0]

56. Chen, R.; Wang, Q.; Javanmardi, A. A Review of the Application of Machine Learning for Pipeline Integrity Predictive Analysis in Water Distribution Networks. Arch. Comput. Methods Eng.; 2025; pp. 1-29. [DOI: https://dx.doi.org/10.1007/s11831-025-10251-6]

57. Ahn, J.M.; Kim, J.; Kim, K. Ensemble Machine Learning of Gradient Boosting (XGBoost, LightGBM, CatBoost) and Attention-Based CNN-LSTM for Harmful Algal Blooms Forecasting. Toxins; 2023; 15, 608. [DOI: https://dx.doi.org/10.3390/toxins15100608] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37888638]

58. Ranitović, A.; Šovljanski, O.; Aćimović, M.; Pezo, L.; Tomić, A.; Travičić, V.; Saveljić, A.; Cvetković, D.; Ćetković, G.; Vulić, J. . Biological Potential of Alternative Kombucha Beverages Fermented on Essential Oil Distillation By-Products. Fermentation; 2022; 8, 625. [DOI: https://dx.doi.org/10.3390/fermentation8110625]

59. Vučetić, A.; Šovljanski, O.; Pezo, L.; Gligorijević, N.; Kostić, S.; Vulić, J.; Čanadanović-Brunet, J. A Comprehensive Antioxidant and Nutritional Profiling of Brassicaceae Microgreens. Antioxidants; 2025; 14, 191. [DOI: https://dx.doi.org/10.3390/antiox14020191]

60. Everitt, M. Consumer-targeted sensory quality. Global Issues in Food Science and Technology; Academic Press: Cambridge, MA, USA, 2009; pp. 117-128.

61. Kim, J.; Adhikari, K. Current Trends in Kombucha: Marketing Perspectives and the Need for Improved Sensory Research. Beverages; 2020; 6, 15. [DOI: https://dx.doi.org/10.3390/beverages6010015]

62. Mintel. Tea and RTD Teas—US—August 2019. Reasons for Drinking Kombucha.; 2019; Available online: http://www.reports.mintel.com (accessed on 3 March 2025).

63. Tomić, A.; Šovljanski, O.; Nikolić, V.; Pezo, L.; Aćimović, M.; Cvetković, M.; Stanojev, J.; Kuzmanović, N.; Markov, S. Screening of Antifungal Activity of Essential Oils in Controlling Biocontamination of Historical Papers in Archives. Antibiotics; 2023; 12, 103. [DOI: https://dx.doi.org/10.3390/antibiotics12010103]

64. Ewim, D.R.E.; Okwu, M.O.; Onyiriuka, E.J.; Abiodun, A.S.; Abolarin, S.M.; Kaood, A. A quick review of the applications of artificial neural networks (ANN) in the modelling of thermal systems. Eng. Appl. Sci. Res.; 2021; 9, pp. 444-458.

65. Niu, Y.; Fabian, Z.; Lee, S.; Soltanolkotabi, M.; Avestimehr, S. Ml-bfgs: A momentum-based l-bfgs for distributed large-scale neural network optimization. arXiv; 2023; arXiv: 2307.13744

66. Kramar, V.; Kramar, O.; Kabanov, A. An Artificial Neural Network Approach for Solving Inverse Kinematics Problem for an Anthropomorphic Manipulator of Robot SAR-401. Machines; 2022; 10, 241. [DOI: https://dx.doi.org/10.3390/machines10040241]

67. Šovljanski, O.; Lončar, B.; Pezo, L.; Saveljić, A.; Tomić, A.; Brunet, S.; Filipović, V.; Filipović, J.; Čanadanović-Brunet, J.; Ćetković, G. . Unlocking the Potential of the ANN Optimization in Sweet Potato Varieties Drying Processes. Foods; 2024; 13, 134. [DOI: https://dx.doi.org/10.3390/foods13010134] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38201161]

68. Doherty, S.B.; Gee, V.L.; Ross, R.P.; Stanton, C.; Fitzgerald, G.F.; Brodkorb, A. Development and characterisation of whey protein micro-beads as potential matrices for probiotic protection. Food Hydrocoll.; 2011; 25, pp. 1604-1617. [DOI: https://dx.doi.org/10.1016/j.foodhyd.2010.12.012]

69. Khem, S.; Small, D.M.; May, B.K. The behaviour of whey protein isolate in protecting Lactobacillus plantarum. Food Chem.; 2016; 190, pp. 717-723. [DOI: https://dx.doi.org/10.1016/j.foodchem.2015.06.020]

70. Nualkaekul, S.; Cook, M.T.; Khutoryanskiy, V.V.; Charalampopoulos, D. Influence of encapsulation and coating materials on the survival of Lactobacillus plantarum and Bifidobacterium longum in fruit juices. Food Res. Int.; 2012; 53, pp. 304-311. [DOI: https://dx.doi.org/10.1016/j.foodres.2013.04.019]

71. Brlek, T.; Pezo, L.; Voća, N.; Krička, T.; Vukmirović, Đ.; Čolović, R.; Bodroža-Solarov, M. Chemometric approach for assessing the quality of olive cake pellets. Fuel Process. Technol.; 2013; 116, pp. 250-256. [DOI: https://dx.doi.org/10.1016/j.fuproc.2013.07.006]

72. Sun, Q.; Yin, S.; He, Y.; Cao, Y.; Jiang, C. Biomaterials and Encapsulation Techniques for Probiotics: Current Status and Future Prospects in Biomedical Applications. Nanomaterials; 2023; 13, 2185. [DOI: https://dx.doi.org/10.3390/nano13152185]

73. Abid, A.; Zhang, M.J.; Bagaria, V.K.; Zou, J. Exploring patterns enriched in a dataset with contrastive principal component analysis. Nat. Commun.; 2018; 9, 2134. [DOI: https://dx.doi.org/10.1038/s41467-018-04608-8]

74. McBratney, A.B.; Santos, M.M.; Minasny, B. On digital soil mapping. Geoderma; 2003; 117, pp. 3-52. [DOI: https://dx.doi.org/10.1016/S0016-7061(03)00223-4]

75. Wang, Y.; Wu, J.; Lv, M.; Shao, Z.; Hungwe, M.; Wang, J.; Bai, X.; Xie, J.; Wang, Y.; Geng, W. Metabolism characteristics of lactic acid bacteria and the expanding applications in food industry. Front. Bioeng. Biotechnol.; 2021; 9, 612285. [DOI: https://dx.doi.org/10.3389/fbioe.2021.612285]

76. Miranda, B.; Lawton, N.M.; Tachibana, S.R.; Swartz, N.A.; Hall, W.P. Titration and HPLC characterization of kombucha fermentation: A laboratory experiment in food analysis. J. Chem. Educ.; 2016; 93, pp. 1770-1775. [DOI: https://dx.doi.org/10.1021/acs.jchemed.6b00329]

77. Dwiputri, M.C.; Feroniasanti, Y.L. Effect of fermentation to total titrable acids, flavonoid and antioxidant activity of butterfly pea kombucha. J. Phys. Conf. Ser.; 2019; 1241, 012014. [DOI: https://dx.doi.org/10.1088/1742-6596/1241/1/012014]

78. Chakravorty, S.; Bhattacharya, S.; Chatzinotas, A.; Chakraborty, W.; Bhattacharya, D.; Gachhui, R. Kombucha tea fermentation: Microbial and biochemical dynamics. Int. J. Food Microbiol.; 2016; 220, pp. 63-72. [DOI: https://dx.doi.org/10.1016/j.ijfoodmicro.2015.12.015] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26796581]

79. Barbosa, C.D.; Uetanabaro, A.P.T.; Santos, W.C.R.; Caetano, R.G.; Albano, H.; Kato, R.; Cosenza, G.P.; Azeredo, A.; Góes-Neto, A.; Rosa, C.A. . Microbial–physicochemical integrated analysis of kombucha fermentation. LWT; 2021; 148, 111788. [DOI: https://dx.doi.org/10.1016/j.lwt.2021.111788]

80. Majid, A.A.; Suroto, D.A.; Utami, T.; Rahayu, E.S. Probiotic potential of kombucha drink from butterfly pea (Clitoria ternatea L.) flower with the addition of Lactiplantibacillus plantarum subsp. plantarum Dad-13. Biocatal. Agric. Biotechnol.; 2023; 51, 102776. [DOI: https://dx.doi.org/10.1016/j.bcab.2023.102776]

81. Cvetković, D.; Ranitović, A.; Savić, D.; Joković, N.; Tomić, A.; Pezo, L.; Markov, S. Survival of wild strains of lactobacilli during Kombucha fermentation and their contribution to functional characteristics of beverage. Pol. J. Food Nutr. Sci.; 2019; 69, pp. 407-415. [DOI: https://dx.doi.org/10.31883/pjfns/112276]

82. Yang, Z.; Zhou, F.; Ji, B.; Li, B.; Luo, Y.; Yang, L.; Li, T. Symbiosis between microorganisms from kombucha and kefir: Potential significance to the enhancement of kombucha function. Appl. Biochem. Biotechnol.; 2010; 160, pp. 446-455. [DOI: https://dx.doi.org/10.1007/s12010-008-8361-6]

83. Wang, S.; Li, C.; Xu, Q.; Wang, Y.; Wang, S.; Zou, Y.; Yang, Z.; Yuan, L. Addition of lactic acid bacteria modulates microbial community and promotes the flavor profiles of Kombucha. Food Biosci.; 2024; 60, 104340. [DOI: https://dx.doi.org/10.1016/j.fbio.2024.104340]

84. Liptáková, D.; Valík, L.; Medveďová, A. Characterization of the growth of Lactobacillus rhamnosus GG in milk at suboptimal temperatures. J. Food Nutr. Res.; 2008; 47, pp. 60-67.

85. Nolasco-Rodríguez, G.; Ordaz-Castro, L.E.; Albarrán-Rodríguez, E.; García-Delgado, L.; Bañuelos-Pineda, J.; Rosales-Cortés, M. Diabetes, Allopathic and Alternative Methods (Kombucha) for Its Treatment. J. Diabetes Mellit.; 2022; 12, pp. 187-203. [DOI: https://dx.doi.org/10.4236/jdm.2022.124015]

86. Alqahtani, A.S.; Hidayathulla, S.; Rehman, M.T.; ElGamal, A.A.; Al-Massarani, S.; Razmovski-Naumovski, V.; Alqahtani, M.S.; El Dib, R.A.; AlAjmi, M.F. Alpha-Amylase and Alpha-Glucosidase Enzyme Inhibition and Antioxidant Potential of 3-Oxolupenal and Katononic Acid Isolated from Nuxia oppositifolia. Biomolecules; 2019; 10, 61. [DOI: https://dx.doi.org/10.3390/biom10010061]

87. Azadikhah, F.; Khalesi, M.; Shad, E.; Majdinasab, M.; Razmjooei, M.; Eskandari, M.H. Inhibition of α-amylase and lipase by probiotic fermented milk-derived bioactive peptides: A comprehensive in vitro investigation. LWT; 2025; 215, 117208. [DOI: https://dx.doi.org/10.1016/j.lwt.2024.117208]

88. Chen, P.; Zhang, Q.; Dang, H.; Liu, X.; Tian, F.; Zhao, J.; Chen, Y.; Zhang, H.; Chen, W. Antidiabetic effect of Lactobacillus casei CCFM0412 on mice with type 2 diabetes induced by a high-fat diet and streptozotocin. Nutrition; 2014; 30, pp. 1061-1068. [DOI: https://dx.doi.org/10.1016/j.nut.2014.03.022] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25102821]

89. Wang, X.; Qi, Y.; Zheng, H. Dietary polyphenol, gut microbiota, and health benefits. Antioxidants; 2022; 11, 1212. [DOI: https://dx.doi.org/10.3390/antiox11061212]

90. Carey, R.M.; Moran, A.E.; Whelton, P.K. Treatment of hypertension: A review. Jama; 2022; 328, pp. 1849-1861. [DOI: https://dx.doi.org/10.1001/jama.2022.19590]

91. Ahmad, H.; Khan, H.; Haque, S.; Ahmad, S.; Srivastava, N.; Khan, A. Angiotensin-converting enzyme and hypertension: A systemic analysis of various ACE inhibitors, their side effects, and bioactive peptides as a putative therapy for hypertension. J. Renin Angiotensin Aldosterone Syst.; 2023; 2023, 7890188. [DOI: https://dx.doi.org/10.1155/2023/7890188]

92. Amalia, R.I.; Fahrudin, F.I.; Alvi, A.S.; Intipunya, P. Myths and Potential Benefits of Kombucha as a Functional Food: A Review. Int. J. Res. Innov. Appl. Sci.; 2024; 9, pp. 484-491. [DOI: https://dx.doi.org/10.51584/IJRIAS.2024.906043]

93. Saubenova, M.; Oleinikova, Y.; Rapoport, A.; Maksimovich, S.; Yermekbay, Z.; Khamedova, E. Bioactive Peptides Derived from Whey Proteins for Health and Functional Beverages. Fermentation; 2024; 10, 359. [DOI: https://dx.doi.org/10.3390/fermentation10070359]

94. Begley, M.; Hill, C.; Gahan, C.G. Bile salt hydrolase activity in probiotics. Appl. Environ. Microbiol.; 2006; 72, pp. 1729-1738. [DOI: https://dx.doi.org/10.1128/AEM.72.3.1729-1738.2006]

95. Mei, L.; Fu, Q.; Guo, T.; Ji, Q.; Zhou, Y. Structural changes and cholesterol-lowering in denatured whey protein isolate: Malic acid combined enzymolysis. Food Hydrocoll.; 2022; 127, 107502. [DOI: https://dx.doi.org/10.1016/j.foodhyd.2022.107502]

96. Sompach, G.; Rodklongtan, A.; Nitisinprasert, S.; Chitprasert, P. Microencapsulating role of whey protein isolate and sucrose in protecting the cell membrane and enhancing survival of probiotic lactobacilli strains during spray drying, storage, and simulated gastrointestinal passage. Food Res. Int.; 2022; 159, 111651. [DOI: https://dx.doi.org/10.1016/j.foodres.2022.111651]

97. Nguyen, N.K.; Dong, N.T.N.; Nguyen, H.T.; Le, P.H. Lactic acid bacteria: Promising supplements for enhancing the biological activities of kombucha. SpringerPlus; 2015; 4, pp. 1-6. [DOI: https://dx.doi.org/10.1186/s40064-015-0872-3]

98. de Souza Neves Ellendersen, L.; Granato, D.; Bigetti Guergoletto, K.; Wosiacki, G. Development and sensory profile of a probiotic beverage from apple fermented with Lactobacillus casei. Eng. Life Sci.; 2012; 12, pp. 475-485. [DOI: https://dx.doi.org/10.1002/elsc.201100136]

99. Danesh, E.; Goudarzi, M.; Jooyandeh, H. Effect of whey protein addition and transglutaminase treatment on the physical and sensory properties of reduced-fat ice cream. J. Dairy Sci.; 2017; 100, pp. 5206-5211. [DOI: https://dx.doi.org/10.3168/jds.2016-12537] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28527808]

100. Tenea, G.N.; Perugachi, E. Innovative functional juice enriched with native probiotics for enhanced nutrition and antimicrobial properties. Front. Nutr.; 2025; 12, 1552745. [DOI: https://dx.doi.org/10.3389/fnut.2025.1552745] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/40046763]

101. Bulatović, M.L.; Krunić, T.Ž.; Vukašinović-Sekulić, M.S.; Zarić, D.B.; Rakin, M.B. Quality attributes of a fermented whey-based beverage enriched with milk and a probiotic strain. RSC Adv.; 2014; 4, pp. 55503-55510. [DOI: https://dx.doi.org/10.1039/C4RA08905G]