Optimization of a stabilizer mixture for stirred yogurt formulation using native starch, acetylated distarch phosphate, and gelatin Authors: Saqi Fakhraei, Roghayeh Amini Sarteshnizi, Bahareh Borhani, Ronak Gholami, Mehdi Jafari Asl, Arash Soltani Published: 2026 ABSTRACT In this study, we have optimized a stabilizer formulation using native corn starch (A: 0–2%), acetylated distarch phosphate (B: 0–2%), and gelatin (C: 0.1–0.4%) via a D-optimal mixture design approach. The study aimed to quantify main and interaction effects and identify an optimum formulation under predefined quality targets. Physicochemical, textural, and sensory characteristics were monitored during storage at 4 °C. Syneresis after 28 days was predominantly influenced by ingredient type and concentration; gelatin significantly reduced syneresis, whereas native and modified starches contributed to increased syneresis. Gelatin showed a softening effect on texture immediately post-production but induced hardness during storage. Both starch types increased hardness throughout storage. Gelatin improved overall acceptability and a significant positive interaction was observed among all three ingredients (ABC). Conversely, binary interactions (AB and AC) negatively impacted sensory scores. Two optimized formulations with highest desirability were optimum 1 (0.998% A, 0.872% B, and 0.129% C) and optimum 2 (1.116% A, 0.526% B, and 0.358% C). After the experiments, observed values for optimum 1 closely matched the predicted values, confirming model accuracy and it was recommended for stirred yogurt formulation.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-51376-3. Introduction There is a high demand for yogurt, principally fermented by lactic acid bacteria, due to its unique organoleptic properties, health and nutritional benefits. According to the production method it can be categorized as set or stirred yogurt1–3. Stirred yogurt, produced by the mild homogenization of gel formed during fermentation, currently is the most popular yogurt type in the market4. Yogurt gelation occurs near the isoelectric point, where casein micelles interact with whey proteins and κ-casein via non-covalent interactions2,5. Moreover, fat globules contribute to the final yogurt texture through linking proteins and enhancing creaminess and firmness. However, the yogurt gel is inherently unstable and prone to syneresis. Throughout the storage, ongoing rearrangement of casein particles leads to the whey proteins separation from the casein network and syneresis6. Various parameters influence stirred yogurt rheology and texture including the composition of the milk and processing conditions such as homogenization and thermal treatments, fermentation variables and post-fermentation treatments. To improve the stirred yogurt characteristics, hydrocolloid polymers are frequently applied in yogurt formulations4. Hydrocolloids are usually applied as stabilizers in stirred yogurt to control syneresis, and improve viscosity and sensory properties3,6–9. Gelatin is one of the most frequently used stabilizers, gelling agents, or thickeners in various types of stirred yogurt, related to its remarkable hydrophilic properties6,8,10–12. Gelatin’s compatibility with milk proteins and its unique ability to melt at mouth temperature contributes to improved palatability by mimicking the sensory effects of fat6,12. Gelatin with the bloom of 225 and 250 is a typical choice for stabilizing stirred yogurt6,10,12. The widespread application of starch in stirred yogurt is contributed to its ease of application and cost effectiveness3. Maize starch is extensively used in yogurt as thickener and fat replacer. During gelatinization, upon heating beyond a critical temperature, starch granules undergo significant water absorption and swelling, a phenomenon referred as gelatinization and pasting. This process results in a pronounced volume increase, consequently enhancing the viscosity of the system. A starch phase volume slightly above the close packing point, achieved with low starch swelling volumes, yield stirred yogurt with optimal textural and sensory attributes, while excessive starch leads to phase separation, and a paste like texture3. The application of native starches is limited because of their susceptibility to low pH, high shear forces and retrogradation13. To overcome these limitations, chemical modifications are often employed. Acetylated distarch phosphate (ADP), a starch modified through phosphate cross-linking and acetic anhydride esterification, is commonly used in yogurt production13. While market preference for stirred yogurt is a smooth, spoonable and lump-free texture, using gelatin alone may result in a jelly-like texture that exhibits lumpiness during agitation10. Moreover, gelatin is an animal-based product which needs innovative and expensive technologies for production and has limited resources for Muslim’s populations. Some studies have investigated the effect of plant-based ingredients in yogurt. Aljobair (2024) investigated the influence of clove extract (CE) on the viscoelastic properties, syneresis, textural attributes and sensory acceptance of set yogurt, demonstrating that CE reduced whey separation and elastic modulus without significantly affecting firmness14. Similarly, high molecular weight polysaccharides including basil seed gum and carrageenan, have been reported to limit water separation in dairy systems through the formation of cross-linked networks arising from polysaccharide-proteins interactions15. However, application of starch from different botanical sources is more acceptable by dairy industry due to ease of application and cost effectiveness than hydrocolloids3. Some previous studies have investigated the individual effects of native starch, modified starch, and gelatin in yogurt13,16–18. Wong et al. (2020) reported that thickened starch can be effectively utilized as a stabilizer in yogurt, yielding products with acceptable organoleptic and textural attributes3. In a related study, Olagunju et al. (2020) employed acetylated pigeon pea starch to overcome the functional limitations associated with native starch in yogurt formulation19. Furthermore, Pang et al. (2018) evaluated ADP in comparison with native starch in yogurt and found that the modified starch exhibited greater resistance to acidic environments and shear force, albeit with reduced solubility and viscosity relative to native starch13. Overall, these studies indicated that while native starch, modified starch, and gelatin each offer functional benefits in yogurt system, they also present certain drawbacks. Given the lack of research focused on optimizing combinations of native starch, ADP, and gelatin to offset these limitations, we used the statistical mixture design approach to identify the stabilizer mixture for stirred yogurt with the highest organoleptic and technological properties. Moreover, we assessed the interactions among components and their influence on the physicochemical and sensory properties of stirred yogurt. D-optimal is a specialized variant of mixture design which is extensively employed to optimize formulations. It enhances the precision of model coefficients estimates by minimizing their variance and achieves this by selecting an optimal subset of experimental runs that maximize the determinant of the information matrix20. This method optimizes complex blending processes, identifies significant interactions between components, and minimizes experimental trials which makes it appropriate to optimize food formulations21. Materials and methods Materials Commercial UHT milk (3% fat, Mihan), UHT cream (25% fat, Sabah), commercial stirred yogurt (4% fat, Seven, Kaleh), and NaCl were obtained from a local supermarket in Iran. Non-fat dry milk was purchased from Pegah, Azarbayejan Sharghi, Iran. Native corn starch and modified starch (E1414, Acetylated Distarch Phosphate) were obtained from Farhikhtegan Zarnam Research and Industrial Group, Iran. Gelatin (230 bloom, halal) was purchased from Gerede Jelatin, Turkey. Freeze-dried lactic culture (YF-L811) was purchased from Christian Hansen, Horsholm, Denmark. Stirred yogurt production Different combinations of stabilizers were designed using Design Expert (7.1.5) software, the D-optimal mixture design method. The stabilizer included gelatin (0–0.4%), modified starch (0–2%), and native starch (0–2%), resulting in 14 formulations containing 3 replicated samples as outlined in Table 1. Yogurt samples (fat content of 4%) were produced following the method of Nagaoka (2018) with slight modification22. Initially full-fat milk (3% fat) was standardized with cream (25% fat) using a Thermomix® (Thermomix TM6-1, Vorwerk, Wuppertal, Germany). For each sample, a complete dispersion of stabilizer mixture (gelatin, modified starch and native starch) was prepared in a part of milk. The milk was then heated to 95 °C. Meanwhile, the stabilizer-milk mixture was added to the Thermomix and maintained at 95 °C for 3 min to ensure uniform dispersion and sterilization. Following this, the sterilized milk was immediately transferred to sterile containers and the temperature decreased to 45 °C. After inoculation, the containers were sealed and kept at 45 °C until the desired pH of 4.6 was achieved. Subsequently, NaCl at a concentration of 0.5% was added to the yogurts and gently stirred to ensure complete dispersion, resulting a uniform texture. The samples were then kept at 4 °C for further analysis. Table 1Different stabilizer mixtures for stirred yogurt formulation obtained by D-optimal mixture designSamplesNative starchModified starchGelatin101.80.2201.60.430.80.80.441.250.450.350.451.250.361.600.471.450.450.1802091.800.210200112001202013110141.600.4 Table 1Different stabilizer mixtures for stirred yogurt formulation obtained by D-optimal mixture designSamplesNative starchModified starchGelatin101.80.2201.60.430.80.80.441.250.450.350.451.250.361.600.471.450.450.1802091.800.210200112001202013110141.600.4 Different stabilizer mixtures for stirred yogurt formulation obtained by D-optimal mixture design Table 2Analysis of variance (ANOVA) and estimated coefficients of models for syneresis and viscosity during storage.ANOVA resultsSynersis Day 1Syneresis Day 7Syneresis Day 28Viscosity D1Viscosity D28modellinearquadraticcubiccubiccubicF value37.907.3253.6951.60132.02P-value< 0.0001< 0.00010.00370.00390.001R-Squared0.87330.98920.99380.99360.9975Adj R-Squared0.85020.98140.97530.97430.9899estimated coefficientsNative starch (A)6.58*31.45*32.42*16,100*14,400*Modified starch (B)5.37*23.53*25.70*17,100*17,700*Gelatin (C)−24.05*−324.38*−7473.46*−1.114E + 006*2.094E + 006*AB-−4.181.26*−2400*4600.00*AC-−548.07*12158.781.919E + 006*−3.485E + 006*BC-−525.80*12404.501.983E + 006*−3.381E + 006*ABC--−10184.84−1.761E + 006*2.826E + 006*AB (A-B)--−115.689240.74*3848.15AC (A-C)--−4898.92−8.413E + 005*1.512E + 006*BC (B-C)--−5291.83−9.367E + 005*1.319E + 006**Significant at level of 0.05. Table 2Analysis of variance (ANOVA) and estimated coefficients of models for syneresis and viscosity during storage.ANOVA resultsSynersis Day 1Syneresis Day 7Syneresis Day 28Viscosity D1Viscosity D28modellinearquadraticcubiccubiccubicF value37.907.3253.6951.60132.02P-value< 0.0001< 0.00010.00370.00390.001R-Squared0.87330.98920.99380.99360.9975Adj R-Squared0.85020.98140.97530.97430.9899estimated coefficientsNative starch (A)6.58*31.45*32.42*16,100*14,400*Modified starch (B)5.37*23.53*25.70*17,100*17,700*Gelatin (C)−24.05*−324.38*−7473.46*−1.114E + 006*2.094E + 006*AB-−4.181.26*−2400*4600.00*AC-−548.07*12158.781.919E + 006*−3.485E + 006*BC-−525.80*12404.501.983E + 006*−3.381E + 006*ABC--−10184.84−1.761E + 006*2.826E + 006*AB (A-B)--−115.689240.74*3848.15AC (A-C)--−4898.92−8.413E + 005*1.512E + 006*BC (B-C)--−5291.83−9.367E + 005*1.319E + 006**Significant at level of 0.05. Analysis of variance (ANOVA) and estimated coefficients of models for syneresis and viscosity during storage. *Significant at level of 0.05. Syneresis The extent of whey separation (syneresis) was evaluated using the centrifugation method described by Wang et al. (2020)3. Specifically, twenty g of each sample was centrifuged at 222 ×g (10 min) (ROTOFIX 32 A, Hettich, Germany). The separated whey was weighted and the percentage of syneresis was reported as the weight of separated whey (g)/initial sample weight (g). This measurement was carried out on days 1, 7, 14, 21 and 28. pH and titratable acidity The pH values of samples were evaluated by a pH meter (Metrohm 912 digital, Switzerland) following the method of Ge et al. (2024)23. Titratable acidity measurement was according to the AOAC method (method 942.15, 2000). Briefly, yogurt samples were diluted with distilled water at a ratio of 1:4 and then titrated with 0.1 N NaOH to a final pH of 8.1. All measurements were taken on days 1, 7, 14, 21, and 28 post-productions. The results were calculated using the equation provided in the AOAC method and reported as the lactic acid percentage24. Texture profile analysis (TPA) The textural characteristics of samples were assessed according to the method of Zhang et al. (2024) with slight modification25. TPA was conducted using a Stable Micro Systems (TA. XT Plus 100 C, Surrey, UK) equipped with a cylindrical flat-ended probe of 25 mm diameter. Briefly, 80 g of each sample was placed into cylindrical plastic cup (volume of 100 ml) and the following test parameters were applied: pre-test speed and test speed at 1 mm/s, post-test speed at 10 mm/s, distance of 30 mm, and trigger force of 10 g. The following textural properties were then determined using texture exponent software: Hardness (N, maximum force developed with penetration), Cohesiveness (area 1/area 2), Gumminess (hardness × cohesiveness), Adhesiveness (N×s, total negative area). The measurements were at 1, 7, 14, 21, and 28 days post-production. Viscosity The yogurt viscosity was evaluated by a high-range rotational Brookfield viscometer (AMETEK HBDVE230, USA) at room temperature according to the method outlined by Saleh et al. (2020) with some changes26. The spindle No. 63 was used and the measurements were taken at a constant rotational speed of 50 rpm. Viscosity measurements were recorded on days 1, 7, 14, 21 and 28. The results were reported in centipoise (cP). Sensory evaluation Organoleptic evaluation of the yogurt samples was conducted after 7 days of storage using a five-point hedonic scale, where 5 indicated “excellent” and 1 indicated “very poor”. The sensory attributes included appearance, color, odor, taste, and overall acceptability27. Samples were evaluated by 21 trained panelists (aged 20–40 years, 10 men, and 11 women) who were post graduate student of food science and technology. Two training sessions were conducted to provide instruction about terminology and product characteristics. Separate booths under natural light conditions were arranged for each panelist. An initial session was conducted before experiment to provide a detailed discussion and explanation of sample characteristics. Deionized water was provided for panelists to cleanse their palates between samples28. Optimization and validation of model Optimization of stirred yogurt formulations were carried out by design expert (7.1.5) using the four most important parameters which included as: hardness targeted as 65–75 N at day 1 and 28, lowest syneresis after 28 days of storage, and highest overall acceptability. According to these factors, two optimized formulations were selected according to the highest desirability. These formulations were then produced and the actual values of hardness at days 1 and 28, syneresis, and overall acceptability were measured and compared with predicted values. Additionally, the rheological properties of the two optimized formulations were evaluated. Following the selection of the optimized formulations, a two-tailed confirmation test was conducted to assess the predictive accuracy and validate the model. A 95% confidence level was applied to determine whether the experimentally observed values fell within the corresponding predicted intervals29. Rheological measurements The rheological properties of the fresh yogurt formulations were evaluated using a modular compact rheometer (Anton Paar MCR-102, Graz, Austria), equipped with a concentric cylinder geometry following the modified method of Wang et al. (2020)30. The spindle CC24-77410 was specifically employed for the tests. An external circulation bath connected to the Peltier system was used to maintain the temperature at 10 °C. The apparent viscosity of the yogurt samples was measured under steady state conditions as a function of shear rates. The shear rate was varied from 0.1 to 100 s− 1 and 21 data points were recorded. Additionally, viscoelastic attributes (Gʹ, Gʹʹ, and tan δ) of the samples was also determined through dynamic rheological measurements. The linear viscoelastic region (LVR) of the samples was first identified by employing an amplitude sweep test at a constant frequency of 1 Hz and strain range from 0.01 to 100%. A strain range of 0.1–10% was found to represent the LVR. Following this, a frequency sweep test was performed within the LVR at a constant strain of 0.1%. The frequency range was 0.01–100 Hz. Storage modulus (Gʹ) and loss modulus (Gʹʹ) were recorded as a function of frequency, and loss tangent (tan δ) was calculated as Gʹʹ/Gʹ. Statistical and data analysis Response values were analyzed using Design Expert (7.1.5) software and were fitted using the linear, quadratic and cubic models (Eqs. (1–3)).1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm Y= AX_{1}+BX_{2}+CX_{3} (linear)}$$\end{document} 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm Y= A_{X}1+BX_{2}+CX_{3}+ ABX_{1}X_{2}+ ACX_{1}X_{3}+BCX_{2}X_{3} (quadratic)}$$\end{document} 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm Y= AX_{1}+BX_{2}+CX_{3}+ ABX_{1}X_{2}+ ACX_{1}X_{3}+BCX_{2}X_{3}+ ABCX_{1}X_{2}X_{3} (cubic)}$$\end{document} The dependent response variables (Y) included syneresis, viscosity, textural and sensory characteristics. A, B and C are the coefficient estimates for linear and non-linear models and X denotes the proportion of real ingredients. The statistically significant effects of the independent variables on the measured responses were evaluated using the Fisher test at a 95% confidence level. For each response parameter, the most appropriate model was selected based on its statistical significance (p < 0.05), and the non-significance of lack of fit. Model adequacy was further assessed using the regression coefficient (R2) and adjusted regression coefficient (Adj-R2)31. All analysis were performed in triplicate and results are reported as mean value ± standard deviation. Differences among samples and storage periods were evaluated using one-way analysis of variance (ANOVA) in JMP statistical software (SAS) (version 10, SAS Institute Inc., Cary, NC, USA) following Tukey’s post hoc test to examine significant differences at p < 0.05. Results and discussion Physicochemical properties pH and acidity Ingredients in yogurt can change its physicochemical properties, which are influenced by their concentration and interactions with milk proteins, particularly casein32. The pH and acidity of different samples during storage are presented in Table 1, supplementary. According to the ANOVA results, no significant model was found for pH value and acidity, indicating that the addition of ingredients and their interactions did not significantly affect these parameters. These findings are consistent with those reported by Olagunju et al. (2020) where the addition of native and acetylated pigeon pea starch, had no significant effect on pH value19. However, Storage time had a significant effect on both pH and acidity with notable changes observed between Days 1 and 28 (p < 0.05). For most samples containing a combination of native starch, modified starch and gelatin, the greatest pH reduction occurred after 7 days, while for samples containing only native or modified starch, the most significant changes in pH and acidity were observed after 14 days. The commercial yogurt, used as a control in the present study, showed no significant changes during storage. Acidity results primarily from the fermentation of lactose to lactic acid by lactobacillus bacteria. In this study, starches and gelatin did not change the activity of these bacteria. The continuous activity of lactic acid bacteria was primary factor responsible for acid production during storage33. Syneresis Syneresis refers to the accumulation of water on the surface of yogurt, which negatively affects its sensory properties and consumer acceptance34. While whey separation and syneresis are considered defects in stirred yogurt, attributes such as thickness and firmness are indicative of a high-quality product35. According to Fig. 1 (supplementary) gelatin had a significant influence on preventing syneresis. Modified starch when combined with gelatin positively influenced water retention within the yogurt matrix. This effect is particularly evident when comparing samples 1 and 9, which are only different in their starch composition. Samples with the highest gelatin concentration (0.4%) demonstrated the lowest syneresis. Even at low concentration (0.1–0.2%), the reduction in syneresis emphasized the critical role of gelatin in maintaining water within the protein matrix. Following the results shown in Table 2 the best models for syneresis at days 1, 7 and 28 were linear, quadratic and cubic, respectively. The linear model indicated that this parameter was influenced by the contribution of individual component. This model demonstrated high statistical significance (p < 0.0001) along with high R2 and adj-R2 (0.8733 and 0.8502, respectively). Quadratic and cubic models represent non-linear relationships; the quadratic model accounts for both individual effects and binary interactions (AB, AC, BC) whereas the cubic model additionally incorporates ternary interactions (ABC). As presented in Table 2, the statistical significance of the models for day 7 and 28 (p < 0.01), together with high R2 and adj-R2 values confirms their adequacy in describing the experimental data. Notably, a linear model best described syneresis on day 1, whereas a quadratic model was more appropriate on day 7, and a cubic model was most appropriate on day 28, suggesting that interaction effects among ingredients became progressively more complex over the storage period. No significant model was found for days 14 and 21. Each ingredient influenced syneresis similarly during storage. Different starches increased syneresis, while gelatin effectively prevented it. Native starch had a higher coefficient than modified starch indicating a greater tendency for syneresis. Interactions between ingredients did not significantly affect syneresis immediately after production, but they had a notable impact after 7 days of storage. At day 7, both native and modified starches increased syneresis, while interactions between gelatin and starches resulted in a syneresis reduction. After 28 days of storage, the ingredients had significant effects on syneresis. While interactions decreased syneresis, it was not statistically significant. It suggests that increasing syneresis during storage is more likely related to the retrogradation of starches than to the interaction between different ingredients. It has been reported that the relatively large granule size of native starch limits its ability to interact effectively with milk proteins, thereby hindering the formation of a cohesive and strong network and reducing its capacity to prevent syneresis. In particular, the spaces between casein micelles are not adequately filled by these large starch granules. Upon heating in milk, starch exhibits enhanced gelling capacity and water absorption; however, during storage, retrogradation may occur, leading to water release and deterioration of textural properties36. Furthermore, the incorporation of starch at elevated concentrations can interfere with the formation of the casein network and promote phase separation, likely due to restricted accessibility of casein to water and impaired network development37. Addition of protein and increasing the protein: starch ratio has been shown to favor the formation of a continuous protein matrix in which starch granules are dispersed, resulting in a more stable structure. This process consequences in decreasing whey separation associated with starch retrogradation38. Additionally, released water might be retained by gelatin macromolecules, contributing to the enhanced structured water39. Moreover, gelatin strengthens the three-dimensional network by interacting with the casein matrix and connecting aggregates of casein micelles, thereby improving water holding capacity16. The rearrangement of the protein network further immobilizes whey proteins, ultimately reducing syneresis. These findings align with previous studies such as Ares et al. (2007) who reported that gelatin was the most effective component in preventing syneresis and a concentration of 0.6% gelatin completely prevented it 40. Mudgil et al. (2018) reported similar effect for gelatin in camel milk yogurt due to the immobilization of water within the yogurt’s gel network and the acceleration of gel network formation34. Gonçalvez et al. (2005) compared the addition of gelatin (0.3–0.6%) and oxidized starch (0.1–10%) on the syneresis of stirred yogurt and found that gelatin was more effective in preventing water release than oxidized starch40. In the present study, samples stabilized with modified starch showed lower syneresis than those with native starch. This is likely related to the acid-resistant properties of modified starch which are essential for yogurt stability during storage. During the acetylation process, the substitution of hydrophilic groups enhances water retention in the gel network via hydrogen bindings41. Similarly, Olagunju et al. (2020) demonstrated that native starch addition increased syneresis during yogurt storage, while acetylated pigeon pea starch reduced syneresis because of bulk acetyl groups which decreased starch-starch interactions19. Textural properties The hardness results for different yogurt samples at days 1 and 28 are shown in Fig. 1a, (data for days 7, 14, and 21 are not shown). Hardness increased significantly over a 28-day storage period; however, the trend was not consistent across all samples. Samples 14 and 6, which contained the highest level of gelatin, showed the highest hardness at day 28, whereas sample 7 showed the lowest hardness throughout the storage period. Most gelatin-contained samples demonstrated an increase in hardness during storage. Experimental data in Table 3 indicated that hardness follows cubic models at days 1 and 28. Both models showed p < 0.05 and high R2 (0.9981 and 0.809 for day 1 and 28, respectively) and adj-R2 (0.9878 and 0.9235 for day 1 and 28, respectively). Native and modified starches had significant positive effects on yogurt hardness after production, while gelatin showed a softening effect (p < 0.05). The interaction between ingredients also significantly influenced hardness of yogurt at this stage. Regression analysis between hardness and viscosity at day 1 showed a strong correlation with R2 value of 0.78. After 28 days, all ingredients had significant hardening effects, with gelatin showed higher coefficients. However, ingredient interactions were not significantly effective at day 28 (p > 0.05). As mentioned above, similar trend was observed for syneresis. Table 3Analysis of variance (ANOVA) and estimated coefficients of models for textural parameters.ANOVA resultsHardness D1Hardness D28Adhesiveness D1Adhesiveness D28Cohesiveness D1Cohesiveness D28GumminessD1Gumminess D28modelcubiccubiccubiccubiclinearquadraticcubiccubicF value62.2917.0957.4215.3326.517.6352.7162.66P-value0.00300.01970.00070.0090.00010.00940.00080.003R-Squared0.99470.98090.99230.97180.84130.84490.99160.9947Adj R-Squared0.97870.92350.97500.90840.80960.73420.97280.9788estimated coefficientsNative starch (A)63.53**69.77**493.55**391.93**0.75**0.77**48.77*52.65*Modified starch (B)73.12**71.25**550.48**517.64**0.78**0.76**56.24*54.59*Gelatin (C)−6051.94**1407.20**−1.764E + 005**−2.57E + 005**1.23**−1.33**−11547.13*5091.14*AB10.34*6.7084.06494.24-−0.16*6.11−5.72AC10169.81**−1406.893.006E + 005**3.542E + 005-2.91*19431.32*−7791.02BC10458.29**−2143.013.019E + 005**3.44E + 005-3.02*19549.40*−8366.70ABC−8726.32**465.25−2.658E + 005**−3.057E + 005--−16445.19*6178.06AB (A-B)−16.55−62.02−1278.83**−2587.84*--−3.32−56.74AC (A-C)−4211.38**−200.69−1.298E + 005**−1.560E + 005--−8180.50*2733.11BC (B-C)−4647.39**987.36−1.337E + 005**−1.421E + 005--−8310.30*3659.51*Significant at level of 0.05.**Significant at level of 0.001. Table 3Analysis of variance (ANOVA) and estimated coefficients of models for textural parameters.ANOVA resultsHardness D1Hardness D28Adhesiveness D1Adhesiveness D28Cohesiveness D1Cohesiveness D28GumminessD1Gumminess D28modelcubiccubiccubiccubiclinearquadraticcubiccubicF value62.2917.0957.4215.3326.517.6352.7162.66P-value0.00300.01970.00070.0090.00010.00940.00080.003R-Squared0.99470.98090.99230.97180.84130.84490.99160.9947Adj R-Squared0.97870.92350.97500.90840.80960.73420.97280.9788estimated coefficientsNative starch (A)63.53**69.77**493.55**391.93**0.75**0.77**48.77*52.65*Modified starch (B)73.12**71.25**550.48**517.64**0.78**0.76**56.24*54.59*Gelatin (C)−6051.94**1407.20**−1.764E + 005**−2.57E + 005**1.23**−1.33**−11547.13*5091.14*AB10.34*6.7084.06494.24-−0.16*6.11−5.72AC10169.81**−1406.893.006E + 005**3.542E + 005-2.91*19431.32*−7791.02BC10458.29**−2143.013.019E + 005**3.44E + 005-3.02*19549.40*−8366.70ABC−8726.32**465.25−2.658E + 005**−3.057E + 005--−16445.19*6178.06AB (A-B)−16.55−62.02−1278.83**−2587.84*--−3.32−56.74AC (A-C)−4211.38**−200.69−1.298E + 005**−1.560E + 005--−8180.50*2733.11BC (B-C)−4647.39**987.36−1.337E + 005**−1.421E + 005--−8310.30*3659.51*Significant at level of 0.05.**Significant at level of 0.001. Analysis of variance (ANOVA) and estimated coefficients of models for textural parameters. *Significant at level of 0.05. **Significant at level of 0.001. The softening effect of gelatin observed after production may be attributed to a reduction in the strength of the acid induced caseinate gel during acidification. At pH values above 5.0, negatively charged gelatin interacts with positively charged regions of κ-casein, thereby limiting the rearrangement of the casein network. However, an increase in hardness during cold storage has been associated with swelling of milk protein particles and intermolecular interactions42. Crosslinking between gelatin and milk protein, along with enhanced protein -protein interactions, resulted in the formation of a denser network with smaller pore size, contributing to increased hardness34. Consequently, the final hardness reflects a balance between the weakening effects during acidification and the strengthening mechanisms that occur through cold storage. The increased hardness observed in samples with higher gelatin concentrations can be attributed to the formation of a denser gelatin network in the continuous phase, which resulted in a strengthened gel network43. Additionally, gelatin triple helix interaction with milk protein promotes in protein polymerization, significantly enhancing the hardness of yogurt matrix44. In a similar study it was shown that addition of pectin to yogurt increased hardness during 21 days of storage which was related to the network formation by interaction between pectin and whey proteins45. It has been proposed that a starch concentration of 1.5% is optimal for achieving desirable hardness in stirred yogurt attributed to electrostatic interactions and hydrogen binding that enhance gel stability16,41. However, improving the gel strength of set yogurt with starch was reported up to 1.5% and above this concentration destroyed the gel network and decreased hardness18. Our results employed the maximum combined starch level of 2% with both native and modified starch which improved hardness significantly. The adhesiveness of samples, as depicted in Fig. 1b, varied over the 28 days storage period. The adhesiveness of most samples containing starch-gelatin mixture increased significantly during storage. As indicated in Table 3, the adhesiveness values followed the cubic model at both days 1 and day 28 with p < 0.05. The linear mixture components were statistically significant, which starches showed a positive effect and gelatin had a negative effect on adhesiveness. Interaction terms were significant at day 1, but by day 28, only the independent variables (individual component effects) remained significant. The effect of component interactions on adhesiveness was similar to their influence on hardness: significant after production but not after 28 days. Cohesiveness followed different statistical models over time: linear model on day 1 and a quadratic model on day 28. Both models had significant p-value and high R2 and adj-R2. Moving from a linear model to a quadratic model during storage indicates the interaction progress during storage which affected cohesiveness of yogurt. On day 1, a positive significant coefficient was observed for all ingredients (p < 0.05), indicating their individual contributions to enhancing cohesiveness. However, by day 28, the coefficient of gelatin became negative, while native starch, modified starch and their interaction with gelatin maintained positive contributions. These findings align with previous work by Ares et al. (2007) who reported that both starch and gelatin improve yogurt consistency46. Gumminess followed a cubic model at both day 1 and day 28 which were highly significant as shown in Table 3. This means that gumminess was influenced by a high order and interactive effect of components. The individual addition of each ingredient significantly increased gumminess. On day 1, the interaction of native starch-gelatin and modified starch-gelatin increased gumminess (p < 0.05). The effect of native starch-modified starch was not significant. The interaction of three ingredients decreased gumminess significantly. Viscosity According to the results (Fig. 1c) the lowest viscosity after production was observed in samples 1, 7 and the commercial yogurt (control). Viscosity of all samples (except 7, 10, and 11) increased during storage. Notably, the commercial yogurt showed the lowest viscosity at the end of storage. Table 2 shows that the viscosity at days 1 and 28 followed the cubic model. These models were identified by their high significancy (p-value of 0.004 and 0.001 for day 1 and 28, respectively) R2 (about 0.99) and adj-R2 (0.97 and 0.99, respectively). No statistically significant model was found for intermediate time points (day 7, 14 and 21; data not shown). On day 1, the addition of both native and modified starch significantly increased viscosity, while gelatin exerted a significant negative effect (p < 0.05). By day 28, the effect of starches remained significantly positive and consistent with day 1; however, gelatin showed a positive effect on day 28. The interaction among all three components (ABC) and starch–starch (AB) demonstrated a negative influence at day 1 but a positive one by day 28. These results indicate that storage time modulates the nature of interactions between starch and protein components. While the interaction between native and ADP decreased viscosity one day after production, it increased viscosity after 28 days of storage. It is probable that the interaction of these starches may not be enough to increase viscosity after production, but during storage particle-particle interactions likely contributed to increased viscosity. The viscosity-enhancing role of two types of starches during the storage time is attributed to water absorption and swelling of starch granules. The phase volume of starch expands during pasteurization, significantly affecting both viscosity and texture3. Comparing of starch and gelatin in stirred yogurt by Lie and Nikoo (2015) showed that both starch and gelatin enhanced viscosity and 0.6% gelatin yielded the highest viscosity. This effect is attributed to the high content of hydrophilic amino acids in gelatin, which improves water absorption and increases viscosity47. Samples containing ADP showed higher viscosity compared to those formulated with native starch. This difference may be attributed to the higher degree of retrogradation in native starch samples. While cross-linking of starch typically reduces solubility and swelling power of granules, acetylation increases these properties by weakening hydrogen bonding48. 13. Furthermore, addition of gelatin may mitigate the effect of retrogradation-induced water loss, as its hydrophilic nature allows for reabsorption of released water. These findings are supported by Gheshlaghi et al. (2021) who identified 0.3% fish gelatin hydrolysate as the optimal concentration for improving the textural properties of set yogurt through enhancement of continuous phase viscosity44. Sensory attributes The sensory evaluation results were compared against a commercial stirred yogurt, which served as the control. Color, taste and odor were not affected by the incorporation of the tested ingredients (data not shown). This observation may be associated with the white color and neutral taste of added ingredients. Moreover, acidity is a key determinant of yogurt flavor acceptability, and no significant changes in acidity were observed following formulation with the new ingredients. This is particularly important for the dairy market, as one of the main requirements for novel yogurt formulations is to modify composition and nutritional value while maintaining the original sensory attributes, including color, taste, texture and appearance. According to Fig. 1d, among the experimental formulations, samples, 6 and 14 received the highest score for overall acceptability. Samples containing gelatin were more acceptable than those without it, indicating the positive influence of gelatin on panelist preference. The samples prepared using combination of two types of starches and gelatin had higher overall acceptability than the commercial control. One of the primary hypotheses contributing to the improved overall acceptability of the new yogurt formulations might be increasing hardness compared to the control sample. All formulated yogurts showed higher hardness than the commercial yogurt. Regression analysis between hardness (day 7) and overall acceptability showed R2 value of 0.61 indicating a moderate positive correlation between these parameters. Consistent with these finding, Bruzzone et al. (2013) reported that acceptance of thicker yogurt-based systems has been increased in recent years, which has encouraged the broader use of texture enhancing ingredients such as starches to improve yogurt consistency49. As shown in Table 4, the sensory attributes of appearance and overall acceptability followed special cubic and cubic models, respectively. Special cubic model for appearance was significant with p < 0.5, R2 of 0.83 and adj-R2 of about 0.7. Cubic model for overall acceptability showed p < 0.05 and R2 and adj-R2 of about 0.94 and 0.83. These results mean that individual component, binary and ternary interactions influenced appearance and overall acceptance of yogurt. The addition of starches and gelatin significantly enhanced the appearance score (p < 0.05). However, their interactions did not yield significant effects on this attribute. All three component positively attributed to overall acceptability and gelatin showed the most substantial effect. Binary interactions between either type of starch and gelatin negatively influenced the overall acceptability. In contrast, the ternary interaction among native starch, modified starch and gelatin improved this attribute, suggesting a synergistic effect. Table 4Analysis of variance (ANOVA) and estimated coefficients of models for sensory characteristic.ANOVA resultsappearanceOverall acceptability model Special cubiccubicF value5.938.35P-value0.0170.028R-Squared0.83550.9495Adj R-Squared0.69460.8358estimated coefficientsNative starch (A)3.58*3.47*Modified starch (B)3.74*3.47*Gelatin (C)6.68*543.26*AB−0.0350.67AC−1.24−888.43*BC−4.04−918.34*ABC7.29772.93*AB (A-B)-2.07AC (A-C)-362.46*BC (B-C)-402.72**Significant at level of 0.05. Table 4Analysis of variance (ANOVA) and estimated coefficients of models for sensory characteristic.ANOVA resultsappearanceOverall acceptability model Special cubiccubicF value5.938.35P-value0.0170.028R-Squared0.83550.9495Adj R-Squared0.69460.8358estimated coefficientsNative starch (A)3.58*3.47*Modified starch (B)3.74*3.47*Gelatin (C)6.68*543.26*AB−0.0350.67AC−1.24−888.43*BC−4.04−918.34*ABC7.29772.93*AB (A-B)-2.07AC (A-C)-362.46*BC (B-C)-402.72**Significant at level of 0.05. Analysis of variance (ANOVA) and estimated coefficients of models for sensory characteristic. *Significant at level of 0.05. The high overall acceptability of samples containing 0.4% gelatin is consistent with findings by Ares et al. (2007) who reported that the inclusion of 0.6% gelatin improved the overall acceptability of low-fat yogurt by improving mouth feel, creaminess and viscosity46. Similarly, gelatin addition to camel milk improved textural and overall acceptability without adversely affecting flavor and taste34. Comparative studies have shown that gelatin improves the viscosity more effectively than oxidized starch. Also, gelatin increased the ropyness of samples. Stabilizers increased the solid behavior of samples as detected by panelists40. The results of sensory evaluation indicated that the combination of the selected ingredients enhanced overall sensory acceptability without adversely affecting color, taste and odor, in line with current market preferences. Moreover, this combination also offers a cost-effective alternative to animal-derived gelatin. Optimization and model validation The optimization process identified two formulations with the highest desirability scores (Table 5). Also, the observed values for these optimized formulations are summarized in Table 5. Comparison between observed and predicted values revealed that the hardness of optimized formulation 1 comprising 0.998% native starch, 0.872% modified starch, and 0.129% gelatin was higher than predicted values (71.5 vs. 66.72 on day 1 and 74.1 vs. 65.49 on day 28). Nonetheless, these values remained within the acceptable range defined for the optimization process. Importantly, the hardness of this sample changed insignificantly throughout storage, indicating desirable textural stability for stirred yogurt. Syneresis was slightly higher (20.78 vs. 17.6) than predicted values while overall acceptability closely matched expected values. For optimized formulation 2, hardness was higher than the predicted value on day 1 (79.8 vs. 73.34), with a more pronounced increase on day 28 (94.64 vs. 75). These findings suggest that the higher gelatin concentration exerts a marked hardening effect on yogurt texture over time. The hardness of this formulation is comparable to that of samples containing 0.4% gelatin (samples 2, 14, 6), which exhibited the highest hardness after 28 days. Syneresis and overall acceptability of formulation 2 were almost consistent with expected values. Table 5Optimum formulations, predicted values and observed values for characteristics.Optimum formulations and predicted values for characteristicsNative starchModified starchGelatinHardness-D1Hardness-D 28Syneresis-D 28Overall acceptabilityDesirability10.9980.8720.12966.7265.4917.604.100.78821.1160.5260.35873.3475.006.813.750.704Observed values for optimum formulations and commercial yogurt (control)Native starchModified starchGelatinHardness-D1Hardness-D 28Syneresis-D 28Overall acceptability10.9980.8720.12971.5074.1020.784.1621.1160.5260.3579.8094.646.683.50Control---44.4062.8926.332.90 Table 5Optimum formulations, predicted values and observed values for characteristics.Optimum formulations and predicted values for characteristicsNative starchModified starchGelatinHardness-D1Hardness-D 28Syneresis-D 28Overall acceptabilityDesirability10.9980.8720.12966.7265.4917.604.100.78821.1160.5260.35873.3475.006.813.750.704Observed values for optimum formulations and commercial yogurt (control)Native starchModified starchGelatinHardness-D1Hardness-D 28Syneresis-D 28Overall acceptability10.9980.8720.12971.5074.1020.784.1621.1160.5260.3579.8094.646.683.50Control---44.4062.8926.332.90 Optimum formulations, predicted values and observed values for characteristics. To validate the model, the results of optimized formulation 1 were evaluated and are presented in Table 6. Following analysis, the experimental values were compared with the corresponding lower and upper predicted intervals. All results observed for optimized formation 1 fell within the corresponding predicted ranges. Therefore, we can confirm the model’s precision and predictive reliability. In the present study, a lower concentration of gelatin in combination with starches was associated with greater overall acceptability. In contrast, Mudgil et al. (2018) reported that gelatin as the sole stabilizer in camel milk set yogurt resulted in the optimal textural and sensory properties at the concentration of 0.75 −1%. The need for higher concentration of gelatin in camel milk has been attributed to its lower k-casein content and consequently weaker gel network, which requires greater gelatin concentration to compensate for gel structural deficiencies34. Table 6Model’s validation using a two-sided confirmation run with 95% confidence (p < 0.05) between high and low predicted intervals (PI).ResponsePredictionSE MeanSE Pred95% PI low95% PI highHardness-D-166.721.391.5961.8471.95Hardness-D-2865.495.175.9146.9384.53Syneesis-D2817.603.613.616.1229.08Overall acceptability4.100.120.163.654.55 Table 6Model’s validation using a two-sided confirmation run with 95% confidence (p < 0.05) between high and low predicted intervals (PI).ResponsePredictionSE MeanSE Pred95% PI low95% PI highHardness-D-166.721.391.5961.8471.95Hardness-D-2865.495.175.9146.9384.53Syneesis-D2817.603.613.616.1229.08Overall acceptability4.100.120.163.654.55 Model’s validation using a two-sided confirmation run with 95% confidence (p < 0.05) between high and low predicted intervals (PI). The hardness, gumminess, and adhesiveness of optimized formulations at days 1 and 28 are illustrated in Fig. 2a and b, respectively. Both gumminess and hardness showed similar increasing trends; however, hardness change was not statistically significant for optimized formulation 1. In contrast, all parameters increased significantly for optimized formulation 2. Adhesiveness was significantly lower in formulation 1 compared to formulation 2 and it decreased significantly during storage for formulation 1, whereas it increased significantly over time for formulation 2. Fig. 1(a) Hardness changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (b). Adhesiveness changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (c). Viscosity changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (d). Overall acceptability of different samples. Different letters indicates significant differences among samples (p < 0.05). (a) Hardness changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (b). Adhesiveness changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (c). Viscosity changes of different samples during 28 days of storage. Different letters indicates significant differences among samples (p < 0.05). (d). Overall acceptability of different samples. Different letters indicates significant differences among samples (p < 0.05). Fig. 2Hardness and gumminess (a) and adhesiveness (b) of optimum yogurt formulations at day 1 and day 28. Different letters indicates significant differences between samples and storage time (p < 0.05). Hardness and gumminess (a) and adhesiveness (b) of optimum yogurt formulations at day 1 and day 28. Different letters indicates significant differences between samples and storage time (p < 0.05). Fig. 3Rheological characteristics of optimum 1 (a), optimum 2 (b), and commercial yogurt (c). Storage modulus (A), loss modulus (B), apparent viscosity (C) and loss tangent (D). Rheological characteristics of optimum 1 (a), optimum 2 (b), and commercial yogurt (c). Storage modulus (A), loss modulus (B), apparent viscosity (C) and loss tangent (D). Rheological properties The rheological properties of two optimized formulations were evaluated and compared with a commercial stirred yogurt. A frequency sweep test was applied to measure the storage modulus (Gʹ) and loss modulus (G″) presented in Fig. 3A and B, respectively. All samples showed frequency dependence and viscoelastic characteristics with higher Gʹ than G″, indicating predominantly elastic behavior. The highest Gʹ and G″ were observed in optimized formulation 2, which contained a higher gelatin concentration, while the commercial control exhibited the lowest Gʹ value. The difference between Gʹ and G″ increased with frequency, becoming pronounced at frequencies up to 100 rad/s, whereas at lower frequencies, this difference was minimal. Superior gel strength of optimized formulation 2 at higher frequencies likely results from enhanced water absorption by starch in the continuous phase, coupled with increased protein interactions that reinforce the yogurt gel network32. Soft starch granules form a network within the continuous phase, thereby enhancing the elasticity of yogurt texture3. Moreover, increasing the gelatin concentration to 0.0.35% in optimum 2 formulation increased both Gʹ than G″, might be related to the higher cross-linking and stronger gelation of yogurt50. In a similar study, gelatin increased the elastic modulus of camel milk yogurt at the concentrations exceeding 0.75%, whereas lower concentrations (0.5%) were less effective34. The interaction between casein and tapioca acetylated distarch phosphate was investigated by Yu et al. (2020), who reported a shear thinning behavior18. An increase in starch concentration led to enhanced Gʹ and G″ with Gʹ consistently exceeding G″ at all frequencies. The apparent viscosity curve (Fig. 3C) showed similar pattern for all samples. A decrease in viscosity was observed with increasing shear rate, consistent with pseudoplastic flow behavior of all samples. Substantial differences were observed among formulations, particularly within the shear rate range of 0.09–0.9 (1/s). The formulation designated as optimum 2, incorporated with higher concentration of gelatin and greater proportion of native starch exhibited the highest viscosity, which shows that these ingredients diminished viscosity reduction induced by increasing shear rate. In previous studies it has indicated that increasing the gelatin concentration results in higher viscosity due to water holding capacity of gelatin which reduces water mobility39. These results are consistent with textural properties of yogurt sample. The shear thinning behavior of tapioca ADP/casein gel were also demonstrated by Yu et al. (2020)18. The loss tan (tan δ = G″/Gʹ) is shown in Fig. 3D. Across the entire frequency spectrum, tan δ values were 0.18–0.27 indicating solid like structure of all samples. However, Optimum 2 showed the highest tan δ values, while optimum 1 showed the lowest tan δ below the frequency of 10 Hz, indicating more solid-like gel structure. The commercial yogurt sample displayed intermediate viscoelasticity, with tan δ values falling between those of optimum 1 and optimum 2. Conclusion In the present study, we have used native starch (0–2%), modified starch (0–2%) and gelatin (0.1–0.4%) to optimize a stabilizer mixture formulation for stirred yogurt applications. Gelatin resulted in the most pronounced reduction in syneresis under storage condition of 4 °C, 28 days. The ingredients and their interactions significantly influenced hardness immediately after production with gelatin showing a softening effect. After 28 days, gelatin had the most substantial effect on increasing hardness. The results indicated that a combination of three ingredients (ABC) produced the stirred yogurt with highest sensory acceptability. Finally, a formulation containing 0.998% native starch, 0.872% acetylated distarch phosphate, and 0.129% gelatin was recommended as optimum formulation. A limitation of current study was that texture of almost all samples changed more than control sample during storage time. This might be related to the high level of interactions between ingredients. Further researches are needed to understand how storage time effects the interactions between starch-starch and starch-gelatin in both model and yogurt matrix systems. We recommend to evaluate milk protein interactions with added gelatin, native and modified starch in model system. Also, it is likely that at high concentration of gelatin, physicochemical and textural properties deviate from expected patterns, making it an important consideration for future studies. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1