A method for in vitro prediction of the glycemic index of a surface product

Abstract

This invention discloses a method for predicting the glycemic index (GI) of flour products in vitro, belonging to the field of food testing technology. The invention systematically acquires multi-dimensional characteristic data of flour products, including nutritional components, structural features, gelatinization properties, and in vitro digestion kinetics. Using machine learning algorithms such as gradient boosting trees, four key parameters are selected: total starch, damaged starch, gelatinization temperature, and 90-minute hydrolysis index. Combined with in vivo GI values ​​obtained from the ISO 26642:2010 gold standard, a high-precision prediction model is constructed through nonlinear regression fitting. The model constructed in this invention relies solely on in vitro detection data, eliminating the need for human trials, and achieves rapid, low-cost, and high-precision prediction of the GI value of flour products. It can be widely applied to raw material selection, formulation optimization, and quality control of low-GI foods.

Description

Technical Field

[0001] This invention relates to a method for predicting the glycemic index of flour products in vitro, belonging to the field of food testing technology. Background Technology

[0002] As a traditional staple food, wheat products are a core source of dietary energy and carbohydrates, and their glycemic index (GI) directly affects postprandial blood glucose levels. Currently, in pursuit of taste and shelf life, wheat product processing commonly employs high-refined flour and processes that may excessively damage starch structure, easily leading to high GI values ​​in the final products. This makes it difficult to meet the urgent needs of diabetic patients and healthy individuals for low-GI staple foods. Therefore, accurate assessment and targeted control of the GI value of wheat products are of great significance for the development of healthy products and the dietary management of chronic diseases.

[0003] Currently, the international gold standard for GI value determination (ISO 26642:2010) relies on human clinical trials. Although this method can reflect the true glycemic response, it requires the recruitment of at least 10 subjects, involves complex ethical reviews, is greatly affected by individual differences, is costly, and has a testing cycle of about two weeks. It cannot meet the needs of enterprises for high-throughput and rapid evaluation in the raw material screening and product development stages.

[0004] Existing in vitro prediction methods have advantages such as speed, good repeatability, and low cost, but they still have shortcomings. Related patent technologies, such as CN115032350A and CN113238010A, have proposed GI prediction methods for noodles and carbohydrate foods, respectively, but the models only consider simple linear relationships under two-dimensional conditions. Although Bellmann et al. used the TIMcarb dynamic in vitro system to simulate the glycemic response of various foods, including pasta, the equipment is complex to operate and has high maintenance and operating costs, making it difficult to widely apply to the accurate prediction of GI values ​​for various noodle products.

[0005] Building upon this, cooking methods significantly impact the glycemic index (GI) of food. Even if the raw material itself is low-GI, improper cooking methods can significantly increase its GI value, leading to a heightened glycemic response. Starchy foods (such as brown rice, whole wheat, or potatoes), if cooked for too long or with too much water, will have their starch granules fully gelatinized, becoming looser and more easily broken down into glucose by digestive enzymes. This accelerates glucose absorption in the intestines, pushing up postprandial blood glucose spikes. For diabetic patients, choosing food solely based on the raw material's GI value while ignoring the changes brought about by cooking methods can lead to misjudging the actual glycemic effect of food. Long-term consumption of these "hidden high-GI" foods can easily cause blood sugar fluctuations, increasing the difficulty of blood sugar control and thus raising the risk of complications.

[0006] Therefore, developing a method for constructing and applying a GI evaluation model for the entire process of flour products from processing to consumption has extremely high practical and economic value. Summary of the Invention

[0007] To address the aforementioned problems, this invention aims to establish a machine learning-based GI prediction model for flour products. This method systematically integrates multi-dimensional data such as raw material composition, gelatinization characteristics, structural features, and in vitro digestion parameters. It then uses feature selection and nonlinear fitting by correlating these data with measured in vivo GI values, thereby efficiently constructing a high-precision prediction model. This invention establishes a machine learning-based GI prediction model for flour products. It achieves this by systematically integrating multi-dimensional data such as raw material composition, gelatinization characteristics, structural features, and in vitro digestion parameters, and then using feature selection and nonlinear fitting by correlating these data with measured in vivo GI values.

[0008] The first objective of this invention is to provide a method for predicting the glycemic index of flour products in vitro. This method involves substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index (0-90 min) of the flour product into an equation to calculate the glycemic index. The equation is as follows:

[0009] Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

[0010] In one embodiment, the dough product to be tested may be either uncooked or cooked.

[0011] In one embodiment, steaming / boiling includes boiling in water (100°C) for 4 to 20 minutes; further, boiling in water for 4 to 10 minutes.

[0012] In one embodiment, the total starch content was determined according to the enzymatic hydrolysis method in GB 5009.9-2016; the damaged starch content was determined according to AACC Method 76-31.01, the American Association of Cereal Chemists' standard, "Determination of Damaged Starch (Maltose Method)"; the gelatinization temperature was determined using a rapid viscosity analyzer; and the starch digestibility and hydrolysis index from 0 to 90 min was determined by an in vitro simulated digestion experiment.

[0013] In one embodiment, the starch digestion and hydrolysis index for 0-90 min is determined by the Englyst in vitro simulated digestion method.

[0014] In one embodiment, the noodle product sample to be tested includes one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, and mixed noodles made from raw materials such as wheat, oats, and buckwheat.

[0015] A second objective of this invention is to provide an apparatus for predicting the glycemic index of in vitro products, the apparatus comprising a nutrient composition determination module, a gelatinization property detection module, and an in vitro simulated digestion module; The nutritional component determination module measures the total starch, protein, lipids, amylose, dietary fiber, damaged starch, and moisture content of the noodle product being tested. The gelatinization property detection module measures the gelatinization temperature, peak viscosity, valley viscosity, final viscosity, disintegration value, and retrogradation value of the dough product under test. The in vitro simulated digestion module was used to determine the glucose release, area under the starch hydrolysis curve, hydrolysis index, reaction rate constant, theoretical maximum hydrolysis rate, and hemihydrolysis time of the tested noodle products. The device calculates the glycemic index by substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index (0-90 min) of the tested noodle product into an equation; the equation is:

[0016] Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

[0017] In one embodiment, the noodle product sample to be tested includes one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, and mixed noodles made from raw materials such as wheat, oats, and buckwheat.

[0018] A third objective of this invention is to provide the application of any of the described apparatus in determining the GI value of noodle products.

[0019] The fourth objective of this invention is to provide a method for improving the accuracy of in vitro prediction of the glycemic index (GI) of flour products. This method involves substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index (0-90 min) of the flour product under test into an equation to calculate the glycemic index. The equation is as follows:

[0020] Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

[0021] In one embodiment, the total starch content was determined according to the enzymatic hydrolysis method in GB 5009.9-2016; the damaged starch content was determined according to AACC Method 76-31.01, the American Association of Cereal Chemists standard, "Determination of Damaged Starch (Maltose Method)"; the gelatinization temperature (GT) was determined using a rapid viscosity analyzer; and the starch digestibility and hydrolysis index from 0 to 90 min was determined by an in vitro simulated digestion experiment. The method for detecting the starch digestion and hydrolysis index from 0 to 90 min was the Englyst in vitro simulated digestion method.

[0022] In one embodiment, the noodle product sample to be tested includes one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, and mixed noodles made from raw materials such as wheat, oats, and buckwheat.

[0023] Beneficial effects of the present invention (1) This invention uses machine learning to deeply analyze the data of the entire process of "raw material-processing-digestion" to identify the core regulatory factors. The model established is highly consistent with the results of the gold standard in vivo, and achieves accurate prediction of the GI value of flour products (the coefficient of determination R² can reach 0.941).

[0024] (2) This invention is based entirely on in vitro experiments, which avoids the ethical, cycle and cost limitations of human trials. It can complete the evaluation within hours and provides an efficient solution for high-throughput sample screening and rapid process optimization.

[0025] (3) This invention is applicable to various flour product systems with different raw materials and processes, has good versatility and stability, can be directly used for the research and development and production quality control of low-GI products, and has clear industrial transformation value. Attached Figure Description

[0026] Figure 1 The images show the state of different noodle products before and after steaming or boiling.

[0027] Figure 2 This is a density diagram of protein crosslinking points in buckwheat flour.

[0028] Figure 3 The gelatinization characteristic curves are for different flour products.

[0029] Figure 4 The in vitro digestion curves of different flour products are shown.

[0030] Figure 5 Comparison of GI values ​​of flour products predicted by mathematical models with clinical GI values. Detailed Implementation

[0031] The preferred embodiments of the present invention are described below. It should be understood that the embodiments are for better explanation of the present invention and are not intended to limit the present invention.

[0032] Test method: Total starch content was determined according to the enzymatic hydrolysis method in GB 5009.9-2016. Damaged starch content was determined according to AACC Method 76-31.01, the American Association of Cereal Chemists' standard, "Determination of Damaged Starch (Maltose Method)".

[0033] Structural characterization: Cooked or dried flour products were fixed with paraformaldehyde and then labeled with fluorescein isothiocyanate (for protein) and rhodamine B (for starch). The distribution of proteins and starch and the gluten protein network structure were observed under a laser confocal microscope.

[0034] Gelatinization temperature (GT) was determined using a rapid viscosity analyzer (RVA) (RVA 4500, Boton Australia). The specific method was as follows: 3.5 g of sample (moisture content equilibrated to 14%) was accurately weighed into an RVA-specific aluminum box, 25 mL of deionized water was added, and the peak viscosity, valley viscosity, final viscosity, disintegration value, reversion value, and gelatinization temperature of the sample were determined using a 13-minute standard program.

[0035] HI90 was determined using the Englyst in vitro simulated digestion method. The specific method was as follows: noodles were cut into 1 cm segments and cooked according to standard methods. Approximately 100 mg (dry basis) of the cooked noodle product was accurately weighed and placed in a centrifuge tube. Glass beads, deionized water, and sodium acetate buffer (0.2 M, pH 5.4, containing 4 mM calcium chloride) were added, and the mixture was vortexed to simulate physical disruption. Subsequently, a mixture of α-amylase (150 U / ml) and amyloglucosidase (3 U / ml) was added, and the mixture was placed in a shaker at 37 ℃ for 180 minutes to simulate the small intestinal digestive environment. Samples were taken at different time points (0, 5, 10, 15, 20, 30, 45, 60, 90, 120, 180 min), and the enzyme reaction was immediately inactivated with ethanol. After centrifugation, the supernatant was collected. The reducing sugar content at 540 nm wavelength was determined using the DNS method, digestion kinetic curves were plotted, the area under the curve was calculated, and the starch hydrolysis index was further calculated.

[0036] Clinical GI values ​​were obtained according to ISO 26642:2010.

[0037] Other nutritional components: Protein content was determined by the Kjeldahl method according to GB 5009.5-2016; Fat content was determined by the Soxhlet extraction method according to GB 5009.6-2016; Dietary fiber content was determined by the enzymatic gravimetric method according to GB 5009.88-2023 (soluble / insoluble dietary fiber); Amylose content was determined by the method according to GB / T 15683-2008; Moisture content was determined by the direct drying method according to GB5009.3-2016.

[0038] Example 1: A method for predicting the glycemic index of flour products in vitro 1. Fitting equation Based on the MATLAB R2021b platform, an in vitro prediction formula for the glycemic index of noodle products was constructed.

[0039] A dataset was constructed based on in vitro detection data of flour product samples and corresponding in vivo glycation end product (GI) measurements. Missing and outlier values ​​were handled, and the samples were divided into training and validation sets. In the training set, a CatBoost regression model was used to rank the feature importance of candidate in vitro indicators (using a regression loss function and an early stopping strategy). Based on this, TS, DS, GT, and HI-90 were selected as key variables.

[0040] Subsequently, a simplified GI function expression was constructed based on key variables, and nonlinear least squares regression was used to minimize the error between the predicted GI in the training set and the measured GI in vivo to determine the model coefficients and constant terms, forming the final fixed prediction equation. R², RMSE, and MAE were calculated using an independent validation set, and the Bland-Altman method was used to evaluate the consistency between the predicted GI and the in vivo GI. The equation is as follows (constructed using the CatBoost algorithm based on gradient boosting decision trees):

[0041] Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

[0042] 2. Detect the GI value of flour products prepared by different methods. (1) Oat noodles Commercially available oat noodle samples were selected (Oat Noodle 1 and Oat Noodle 2 are Maijia Maiximai 0 Fat Oat Noodles and Wuyue Mingfang Vegetarian Life Pure Oat Noodles, respectively). The noodles were boiled in boiling water for 5 minutes, and 100 mg of the cooked sample (dry basis) was taken to determine its total starch content, damaged starch content, gelatinization temperature and starch hydrolysis index.

[0043] The total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index are substituted into the equation to calculate the GI value.

[0044] (2) Pasta sample Commercially available pasta samples (Carboff low-sugar pasta) were selected. The pasta was boiled in boiling water for 6 minutes and drained. 100 mg of the cooked sample (dry basis) was taken and its total starch content, damaged starch content, gelatinization temperature and starch hydrolysis index were determined.

[0045] The total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index are substituted into the equation to calculate the GI value.

[0046] (3) Buckwheat semi-dried noodles Commercially available semi-dried buckwheat noodle samples were selected (Buckwheat noodle 1 and buckwheat noodle 2 were Baixiang bitter buckwheat noodles and Jinshahe low-fat pure buckwheat noodles, respectively). The semi-dried noodles were boiled in boiling water for 4 minutes, and 100 mg of cooked sample (dry basis) was taken. The total starch content, damaged starch content, gelatinization temperature and starch hydrolysis index were determined respectively.

[0047] The total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index are substituted into the equation to calculate the GI value.

[0048] 3. Results Statistics The different flour products selected in section 2 before and after steaming are as follows: Figure 1 As shown; the density diagram of protein cross-linking points in buckwheat flour is as follows. Figure 2 As shown.

[0049] The nutritional composition of different flour products is shown in Table 1; the gelatinization characteristic parameters of different flour products are shown in Table 2, and the gelatinization characteristic curves are shown in... Figure 3 As shown in Table 3, the in vitro digestion parameters of different flour products are shown in Table 4, and the in vitro digestion curves are shown in Table 5. Figure 4 As shown in Table 4; the area under the starch digestion curve (AUC) of different flour products is shown in Table 5; the starch hydrolysis index (HI) of different flour products is shown in Table 6; the predicted GI and clinical GI values ​​of different flour products are also shown in Table 5. Figure 5 And as shown in Table 6 Table 1 Nutritional components

[0050] Table 2 Gelatinization characteristics

[0051] Table 3 In vitro digestion parameters

[0052] Table 4. Area Under the Digestion Curve (AUC)

[0053] Table 5 Starch Hydrolysis Index (HI)

[0054] Table 6 GI Predicted Values ​​and Clinical Values

[0055] The prediction model constructed in this patent exhibits excellent performance and robustness (Table 6). Figure 5 The model's predicted GI values ​​for five types of pasta products showed a high degree of agreement with clinical values, demonstrating its reliable predictive ability across product types. The differences in GI values ​​among the samples stemmed from their varying in vitro digestibility. As shown in Tables 3 and 5, the high-GI oat flour 2 had a significantly higher HI180 value (74.53%), while the low-GI pasta had the lowest HI180 value (40.87%). Digestion curves ( Figure 4 The GI (Glycemic Index) directly reflects the rate of digestion, which is completely consistent with the GI ranking. This difference in digestibility is determined by the material composition of the samples. As shown in Tables 1 and 2, high-GI oat noodles 2 have a higher damaged starch content (8.37%) and a lower gelatinization temperature (88.7℃), meaning their starch is more easily hydrolyzed by enzymes; while low-GI pasta exhibits a higher gelatinization temperature (92.7℃) and an extremely low disintegration value (27 cP), indicating a more stable starch gel structure and stronger resistance to digestion.

[0056] In summary, the results show that Example 1, through multidimensional parameter screening, identified total starch, damaged starch, gelatinization temperature, and 90-minute hydrolysis index as the core influencing factors, and constructed a high-precision nonlinear GI prediction model. Verification showed that this model is applicable to noodles of different types and processes, with a prediction error rate of less than 5%. This invention provides efficient and reliable technical support for raw material selection, process optimization, and quality control of low-GI noodles.

[0057] Comparative Example 1: Constructing equations using different factors Substituting the gelatinization temperature of the flour product to be tested into the equation, the glycemic index is calculated; the equation is: GI = -2.97 × GT + 328.5 Wherein, GI value is the glycemic index of the tested noodle product, and GT is the gelatinization temperature (°C) of the tested noodle product. 2 =0.72, MAE=10.1. The model prediction results are shown in Table 7.

[0058] Table 7. GI values ​​of flour products predicted based on a univariate regression model.

[0059] Comparative Example 2: Constructing Equations Using Different Factors The glycemic index is calculated by substituting the 90-minute hydrolysis index and gelatinization temperature of the flour product under test into the equation; the equation is: GI = 0.65 × HI90 - 0.3 × GT + 18, Wherein, GI value is the glycemic index of the tested noodle product, HI90 is the 90-minute hydrolysis index (%) of the tested noodle product, and GT is the gelatinization temperature (°C) of the tested noodle product. 2 =0.71, MAE= 5.8. The model prediction results are shown in Table 8.

[0060] Table 8. GI values ​​of flour products predicted based on a binary regression model.

[0061] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various modifications and alterations without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the claims.

Claims

1. A method for predicting the glycemic index of flour products in vitro, characterized in that, The glycemic index was calculated by substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestibility / hydrolysis index (0-90 min) of the tested flour product into the equation. The equation is as follows: Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

2. The method according to claim 1, characterized in that, Total starch content was determined by enzymatic hydrolysis according to GB 5009.9-2016; damaged starch content was determined according to AACC Method 76-31.01, American Association of Cereal Chemists standard, "Determination of Damaged Starch (Maltose Method)"; gelatinization temperature was determined using a rapid viscosity analyzer; starch digestibility and hydrolysis index from 0 to 90 min was determined by in vitro simulated digestion experiments.

3. The method according to claim 2, characterized in that, The method for detecting the starch digestion and hydrolysis index from 0 to 90 min was the Englyst in vitro simulated digestion method.

4. The method according to claim 1, characterized in that, The noodle product samples to be tested include one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, or mixed noodles made from wheat, oats, or buckwheat.

5. A device for predicting the glycemic index of in vitro products, characterized in that, The device includes a nutrient component determination module, a gelatinization property detection module, and an in vitro simulated digestion module. The nutritional component determination module measures the total starch, protein, lipids, amylose, dietary fiber, damaged starch, and moisture content of the noodle product being tested. The gelatinization property detection module measures the peak viscosity, valley viscosity, final viscosity, disintegration value, retrogradation value, and gelatinization temperature of the dough product under test. The in vitro simulated digestion module was used to determine the glucose release, hydrolysis index, area under the starch hydrolysis curve, reaction rate constant, theoretical maximum hydrolysis rate, and hemihydrolysis time of the tested noodle products. The device calculates the glycemic index by substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestion and hydrolysis index (0-90 min) of the tested noodle product into an equation; the equation is: Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

6. The apparatus according to claim 5, characterized in that, The noodle product samples to be tested include one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, or mixed noodles made from wheat, oats, or buckwheat.

7. The application of the apparatus according to any one of claims 5 to 7 in determining the GI value of noodle products.

8. A method for improving the accuracy of in vitro prediction of GI values ​​for noodle products, characterized in that, The glycemic index was calculated by substituting the total starch content, damaged starch content, gelatinization temperature, and starch digestibility / hydrolysis index (0-90 min) of the tested flour product into the equation. The equation is as follows: Wherein, GI value is the glycemic index of the noodle product being tested; TS is the total starch content (%) of the noodle product being tested; DS is the damaged starch content (%) of the noodle product being tested; GT is the gelatinization temperature (°C) of the noodle product being tested; and HI90 is the starch digestion and hydrolysis index (%) from 0 to 90 min.

9. The method according to claim 8, characterized in that, Total starch content was determined according to the enzymatic hydrolysis method in GB 5009.9-2016; damaged starch content was determined according to AACC Method 76-31.01, American Association of Cereal Chemists standard, "Determination of Damaged Starch (Maltose Method)"; gelatinization temperature was determined using a rapid viscosity analyzer; starch digestibility and hydrolysis index from 0 to 90 min was determined through in vitro simulated digestion experiments. The method for detecting the starch digestion and hydrolysis index from 0 to 90 min was the Englyst in vitro simulated digestion method.

10. The method according to claim 8, characterized in that, The noodle product samples to be tested include one or more types of dried noodles, semi-dried noodles, pre-cooked noodles, or mixed noodles made from wheat, oats, or buckwheat.

Images