Novel integrated analysis platform for evaluating quality of complex natural products

Abstract

The present invention provides a novel integrated platform capable of verifying the production process, standardizing and evaluating complex natural products, thereby enabling "as an integral system" characterization of complex mixtures contained in products (based on or consisting of natural substances) with therapeutic and / or health care effects, the characterization of the metabonomics is not limited to a reductive chemical method aiming at a single marker, and qualitative and quantitative data and fingerprint characterization metabonomics of the mixture are integrated with physical and biophysical activity data, so that the characterization of the product is realized, which is the basis of the verification process of the production process; and the reproducibility of different batches of products in the aspects of efficacy and safety can be ensured. The analysis platform described herein can ensure that a complex natural product is a whole, and all quality and efficacy characteristics of the product are considered, thereby ensuring reproducibility of effect and safety through a multi-dimensional analysis method.

Description

Technical Field

[0001] This invention relates to the validation, standardization, and evaluation of the manufacturing processes of complex natural products. Specifically, it provides a novel integrated platform for validating, standardizing, and evaluating complex natural products, enabling the characterization of complex mixtures based on or composed of natural substances "as a whole system" for the production of products with therapeutic effects (e.g., medical devices) or products that contribute to the physiological functions of organisms (e.g., dietary supplements), without limiting characterization to chemical methods targeting single markers. The platform integrates qualitative and quantitative fingerprinting data of the mixtures with metabolomics and physical and biophysical activity data, providing characterization of the product that forms the basis of the manufacturing process validation process, ensuring reproducibility in efficacy and safety between different batches of such products.

[0002] The analytical platform described in this paper can guarantee the quality of complex natural products as a whole, taking into account their overall characteristics, and thus ensure the reproducibility of effects and safety through multidimensional analysis methods. Background Technology

[0003] In recent years, complex families of natural products have become increasingly important in treating so-called "self-limiting diseases" (such as cough, sore throat, constipation, etc.) for specific patient populations (e.g., children and the elderly) and for significant "unmet medical needs," particularly syndromes and functional disorders for which classical pharmacological approaches fail to provide a satisfactory benefit-risk ratio. This last parameter, playing a crucial role in limited therapy, has evolved in recent years from solely assessing the patient to simultaneously evaluating the environmental impact of therapeutic treatments. This is because the introduction of non-biodegradable substances into the environment has direct and measurable negative impacts on human health, creating an unsustainable vicious cycle.

[0004] Clearly, in this new context, which can be defined as “extending to the benefit-risk relationship between humans and the environment,” the use of natural substances defined as biodegradable represents an opportunity to explore all their potential uses.

[0005] Regulation 2017 / 745 concerning medical devices provides for the possibility of treating and preventing conditions using complex natural products and establishes a clear regulatory framework that complements the existing drug regulatory framework. The substantial difference between the two regulatory contexts relates to the mechanism of action that achieves therapeutic activity: for drugs, this must be through pharmacological (or immunological or metabolic) action, that is, the interaction between a specific substance and a cellular target (often defined as a receptor); while for medical devices, the therapeutic target must be achieved through any mechanism of action other than those reported above. Therefore, under this legislation, therapeutic compositions / preparations are limited to drugs or medical devices based on their mechanism of action for exerting a therapeutic effect. As for natural substances with therapeutic effects, these can be classified as either drugs or medical devices. However, for drugs, it is necessary to demonstrate that the therapeutic activity is achieved through a specific substance (often defined as a marker or active ingredient) contained in the natural substance that acts on a specific cellular target; while for medical devices, this must be linked to the overall complexity of the matrix and cannot be traced back to a specific mechanism of action.

[0006] From a technical and regulatory perspective, this means that, given that therapeutic effects are achieved by a single molecule or combination of molecules through pharmacological, immunological, or metabolic mechanisms (Section 2001 / 83, Paragraph 2), a drug must be able to restore, correct, or alter physiological function, while a medical device must act through non-pharmacological, immunological, or metabolic mechanisms, and therefore must alter physiological or pathological processes or states (Section 2017 / 745, Paragraph 2). [Capone, L., Gerachi, A., Giovagnoni, E., Marcoaldi, R., and Palazzino, G. (2012) "Elementi di valutazione edi discernimento tra dispositivo medico e medicinale. Roma: Istituto Superioredi Sanità. Rapporti ISTISAN 12 / 30; Racchi, M., Govoni, S., Lucchelli, A., Capone, L., and Giovagnoni, E. (2016). Insights into] the definition of terms in European medical device regulation.Expert Rev.Med.Devices 13(10),907–917.doi:10.1080 / 17434440.2016.1224644].

[0007] In the context of medical devices made of natural substances, the therapeutic effects of natural products are due to a set of components that characterize them. Therefore, it seems necessary to establish standards for characterizing and validating these innovative products to ensure the reproducibility of the benefit-risk ratio behind their therapeutic uses.

[0008] Given the variability of natural substances, it is necessary to provide production processes and quality control methods that can ensure that individual batches of products fall within a defined range of variation determined experimentally and are verified to maintain the proven risk-benefit ratio.

[0009] As required by Annex I of Regulation 2017 / 745, paragraph 13.3, this characterization must be placed on the basis of demonstrating the effectiveness and safety of the medical device and must be reported on the product label as a guarantee of its characterization.

[0010] This result can be achieved through methods that are not limited to determining the product composition from a quantitative chemical perspective. In fact, while such methods may be sufficient to characterize products composed of purified substances, they cannot limit the reproducibility of complex substances, because for complex substances, it is necessary to be able to describe the product from both physicochemical and biological perspectives.

[0011] Based on the above, and also because complex substances obtained from natural biological matrices (UVCB substances, i.e., unknown or variable composition, biological materials) cannot be defined in the same way as pure chemical substances, but can be defined and verified within a specific range through analytical techniques, such as fingerprinting and physical and biological characterization.

[0012] Therefore, integrated and advanced analytical methods are needed to characterize its structural and effectiveness performance and to define the correlations between various parameters.

[0013] Studies on the correlation between fingerprint data and activity of complex natural products have begun to appear in the literature. For example, in the publication Jing Wang, Hongwei Kong, Zimin Yuan, Peng Gao, Weidong Dai, Chunxiu Hu, Xin Lu, Guowang Xu, A novel strategy to evaluate the quality of traditional Chinese medicine based on the correlation analysis of chemical fingerprint and biological effect. Journal of Pharmaceutical and Biomedical Analysis 83(2013)57-64, an analysis of the correlation between the fingerprints of a certain number of compounds contained in complex natural products and their biological effects is described, thereby identifying a certain number of quality marker components related to the biological effects of such products.

[0014] However, even in these studies, standardization was based on a single marker.

[0015] Currently, there are no available procedures to characterize the complex mixtures contained in natural products as a complete "system," and therefore, to avoid limiting characterization to a few selected markers. This would ensure validation of the manufacturing process and / or reproducibility between different batches of complex natural products. Furthermore, the lack of such procedures prevents us from ensuring the reproducibility of the risk-benefit ratio in clinical studies. Summary of the Invention

[0016] This invention provides an innovative analytical platform that integrates advanced analytical techniques to characterize complex mixtures contained in natural products as a "complete system," rather than limiting characterization to reductionistic chemistry methods targeting individual markers. This is achieved by integrating qualitative and quantitative fingerprinting data of the mixtures with metabolomics and physical and biophysical activity data. The inventors of this invention have successfully characterized products (e.g., products on which manufacturing process validation procedures are based), ensuring reproducibility between different batches of the product.

[0017] The analytical platform described here enables us to ensure the reproducibility of effects and safety of complex natural products as a whole, taking into account all their quality and efficacy characteristics, through multidimensional analytical methods.

[0018] In fact, complex natural systems express characteristics that cannot be described by a single technology. This invention provides a method that can correlate qualitative and quantitative parameters with efficacy parameters to evaluate complex products.

[0019] This invention provides a method for characterizing a product having therapeutic and / or health-promoting properties, the product comprising a complex natural system (e.g., a composition obtained by mixing natural substances, their extracts, and fractions, preferably obtained through a physical production process with low environmental impact, the composition being capable of presenting different pharmaceutical forms, such as syrups, tablets, granule sachets, capsules, sprays, suspensions, aerosols, and powders for aerosols), the method comprising the following steps:

[0020] a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining fingerprint data X data blocks;

[0021] b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining a data block Y of biophysical properties;

[0022] c. The data are correlated through multivariate analysis, which involves the correlation between data block X and data block Y. This is done by correlating the correlations between each data point obtained in a and each data point obtained in b, thereby obtaining a regression line for each pair of values ​​and calculating its regression coefficient R. 2 And mean squared estimation error (RMSEE), choose R from them. 2 Data pairs with a ratio of ≥0.8 provide a model capable of predicting each biophysical characteristic of the standard formulation based on the fingerprint used.

[0023] The present invention also provides a method for producing and validating one or more batches of a product having therapeutic and / or health-promoting properties, the product comprising a complex natural system, the method comprising the following steps:

[0024] a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining fingerprint data X data blocks;

[0025] b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining biophysical property data Y data block;

[0026] c. Use multivariate analysis to integrate the data for correlation, which involves the correlation between data block X and data block Y. Integrate the correlation between each data point obtained in a. and each data point obtained in b. to obtain a regression line for each pair of values, and calculate its regression coefficient R. 2 And mean squared estimation error (RMSEE), choose R from them. 2 Data pairs ≥0.8 provide a model capable of predicting each biophysical characteristic of a standard formulation based on the fingerprint used;

[0027] d. Using the regression model developed in c., perform qualitative and quantitative fingerprint analysis on one or more batches of the product to be validated using the same method used in a. to predict the biophysical characteristics described in point b.

[0028] Glossary

[0029] According to the present invention, the term "product" comprising a complex natural system and having health and / or therapeutic properties refers to a product made from natural substances, such as a medical device (a product with therapeutic properties) as required by Annex I of EC Regulation 745 / 2017, or a food supplement (a product with health-promoting activity) composed of complex natural substances as required by Article 2 of Directive 2002 / 46 / EC and Article 2 of DLgs169 / 2004.

[0030] In every part of the specification and claims, the product referred to is not a product naturally occurring in nature, but rather a product formulated from complex natural substances. Therefore, the term "product comprising a complex natural system, having health-promoting and / or therapeutic activity" may be replaced in any part of this document and claims with the following: "a formulation comprising (or consisting of) a complex natural system, having health-promoting and / or therapeutic properties" or "a composition comprising (or consisting of) a natural system complex, having health-promoting and / or therapeutic properties" or "a mixture comprising (or consisting of) a complex natural system, having health-promoting and / or therapeutic properties".

[0031] Furthermore, in any part of the specification and claims, the phrase "comprising complex natural systems" can be replaced with "consisting of complex natural systems".

[0032] As for products with health-promoting activity, these refer to products that, as described above, are beneficial to or have a positive effect on health when ingested through any suitable route, especially those beneficial to healthy individuals. These fall within the definition of products with health-promoting activity (e.g., supplements, foods for special medical purposes, mixtures of nutrients such as proteins, vitamins, enzymes, oils, etc.).

[0033] The terms “preparation” or “quality sample” are also defined herein as a “standard preparation” or “target product” of product x, indicating a product preparation (including complex natural systems as defined above) having therapeutic and / or health-promoting properties, wherein said product meets the characteristics required for a medical device as defined in Annex I, EC Regulation 745 / 2017, or as a food supplement made from complex natural substances as defined in Article 2 of Directive 2002 / 46 / EC and Article 2 of DLgs 169 / 2004. According to the foregoing, a medical device is defined as a composition having a therapeutic effect consisting of substances or combinations of substances (including materials of biological origin), and therefore can be in various pharmaceutical forms, such as syrups, tablets, granule sachets, capsules, sprays, suspensions, aerosols, powders for aerosols, etc. Preparations of the same type can also have health-promoting compositions, such as the food supplements defined herein.

[0034] In this invention, the term "poor quality formulation (or sample or ingredient)" refers to a formulation that differs from a standard formulation (or sample) because it contains the same ingredient in a different percentage compared to the sample, or because it uses the same ingredient present in a standard formulation but has quality characteristics different from those used in the product definition.

[0035] Data integration refers to building a correlation model between qualitative and quantitative fingerprint data and biophysical activity. With the help of this model, by measuring a segment of data from a new batch (e.g., NIR fingerprint), corresponding data (e.g., biophysical activity) can be predicted.

[0036] Collecting this information from multiple batches allows for the definition of very stringent quality ranges by considering multiple aspects based on correlation models—qualitative and quantitative fingerprints and biophysical activity. Therefore, the quality range can be specifically defined for the product of interest.

[0037] Characterization in this specification: The term “characterization” refers to a qualitative and quantitative description (limited) of the complex natural product of interest and its biophysical properties.

[0038] Qualitative and quantitative description of complex natural products: This refers to the results of product analysis using fingerprinting techniques such as metabolomics (e.g., NIR spectroscopy and / or mass spectrometry).

[0039] NIR spectroscopy: studies the infrared spectrum of matter (4000-12500 cm⁻¹). -1 The absorption and emission of light within the body.

[0040] Mass spectrometry is a method for studying and measuring specific spectra. In the case of mass spectrometry, the mass of a sample is measured by its mass-charge ratio after passing through an electromagnetic field.

[0041] Biophysical property description of complex systems: refers to performing tests that define known biophysical properties of the product that are related to its therapeutic efficacy, such as tests that measure barrier effects, adhesion and antioxidant activities, neutralizing activities, moisturizing activities, and lubricating activities.

[0042] In this invention, the term "poliresin" is used to refer to a plant complex based on polysaccharides, resins, and flavonoids, which is obtained by subjecting three starting plants (Planttain, Grindelia, and Helichrysum) to a low-environment physical extraction process (LPME process, i.e., liquid-phase microextraction, a process characterized by the liquid-phase extraction of plant parts into microparticles), which produces extract fractions and standardizes their content in the finished product.

[0043] In this specification, OPLS (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) refers to a well-known multivariate data analysis method that effectively utilizes all relevant data with minimal loss of information. This is the most recent known extension of PLS ​​in the literature. OPLS separates systematic or parallel variance from unpredictable or orthogonal variance.

[0044] PLS (Projections to latent structures by means of partial least squares) is a well-known statistical method used to link two matrices X and Y through a multivariate linear model. PLS discovers a linear or polynomial relationship between a data block Y (dependent variable) and a variable block X (predictor variable). The objective function of PLS ​​is to maximize the covariance between X and Y.

[0045] PLS2 (Partial Least Squares 2) is the first known algorithm in the current literature for PLS regression and is one of the most commonly used implementations of PLS. PLS2 refers to a model with multiple dependent variables. Given a response matrix Y and a prediction matrix X, the goal of PLS2 is to calculate the regression coefficient matrix B that produces the linear regression model.

[0046] According to this specification, complex natural ingredients refer to those components present in products based on natural substances, and are therefore not synthetic, including multiple components such as chopped or crushed plant parts, plant extracts or plant parts extracted by water and alcohol, fractions of said extracts, such as fractions obtained by semi-permeable membrane filtration (microfiltration, ultrafiltration, nanofiltration), or fractions obtained by selective precipitation by controlled evaporation of water and alcohol solutions, or fractions obtained by treatment on adsorption resins. Attached Figure Description

[0047] Figure 1 R-squared values ​​of OPLS models for NIR spectroscopy (near-infrared fingerprint, i.e., fingerprints or spectra obtained through near-infrared spectroscopy) and adhesion effects (PINCL) 2 and Q 2 .

[0048] The OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections tolatent structures) better separates the systematic variation of information because it separates orthogonal information from parallel information and only considers the predicted or parallel information.

[0049] Information for block X is given by the NIR fingerprint, while information for block Y is given by the adhesion effect (PINCL).

[0050] This figure shows the cumulative diagnostic parameters (fit summary) R. 2 and Q 2 The bar charts (represented by light and dark gray respectively) show the parameters of the OPLS model in terms of the correlation between the NIR fingerprint of the data and the adhesion effect. 2 Q is a measure of the model's goodness of fit, while Q... 2 It is a measure of the goodness of prediction (prediction quality). Diagnostic parameter R 2 and Q 2The closer these two parameters are to each other and tend towards 1 (y-axis), the more likely the diagnostic parameter R is to be considered... 2 and Q 2 The test met the requirements. The number of components used is reported on the horizontal axis (x-axis) and indicated by square brackets. The two bars on the left represent the R values ​​of the parallel component and the zero-orthogonal component (comp[1+0]). 2 and Q 2 The value of . The two bars in the middle represent the R values ​​of the parallel component and the first orthogonal component (comp[1+1]). 2 and Q 2 The value of . The two bars on the right represent the R values ​​of the parallel component and the second orthogonal component (comp[1+2]). 2 and Q 2 The value of .

[0051] Therefore, in this case, the optimal OPLS model is one using one parallel component and two orthogonal components (comp[1+2]), because R 2 and Q 2 The values ​​are close to each other and tend to 1.

[0052] The image in the bottom right corner shows the software used and the reprocessing date.

[0053] Figure 2a A scatter plot integrating NIR fingerprint data (near-infrared) obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with the adhesion values ​​(PINCL) of the slope.

[0054] The information for each block is presented as predicted (or calculated) data on the x-axis. Each circle shown in the figure represents an experimental measurement of the adhesion of each analyzed sample to the slope (see Table 7), characterized by different compositions as described herein (see Table 6). The samples are identified by the acronym “DOE_EXP 1 to 8” and correspond to samples referred to as “Experiment 1 to 8” in Table 7.

[0055] The dashed lines in the middle of the graph represent the relationships between the data. The key feature is the equation y = 0.9966x + 0.1152 shown in the upper left corner of the graph, with a correlation coefficient R0. 2 It equals 0.993.

[0056] This straight line represents a low error value for the estimate, with RMSEE equal to 2.03219 (Root Mean Square Error of Estimation); and it also represents a low cross-validation error, RMSE. CV This equals 2.6701 (Root Mean Square Error of CrossValidation), which will enable very accurate prediction of biophysical activity values ​​(Y data block) based on the NIR fingerprint (X data block).

[0057] RMSEE value and RMSE CV The values ​​are reported on the x-axis under the heading YPRED[1](PINCL). The vertical bars on the right represent experimental values ​​of the Y-block used to create the model, in this case, measurements of adhesion to the slope. The graph in the lower right corner shows the software used and the date of reprocessing.

[0058] Figure 2b A scatter plot integrating NIR fingerprint data (near-infrared) obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with the adhesion values ​​(PINCL) of the slope was used to predict the TARGET adhesion effect based on the NIR fingerprint.

[0059] The diagram and Figure 2a The graphs are basically the same, and there are also predicted TARGET samples (DoE_Target).

[0060] Therefore, all the details described in the previous figure remain valid. Furthermore, in this case, the NIR fingerprint of the TARGET sample has been added to the information of block X, which consists of the NIR fingerprint.

[0061] The sample data used to build the model is represented by light gray circles, while the TARGET samples in the prediction model are represented by dark gray circles.

[0062] By utilizing a regression model, the fingerprint information of the TARGET sample is used to predict the information of the Y block, in this case, to predict the adhesion information to the slope.

[0063] The dashed line in the middle of the figure represents the interpolation between data including the TARGET sample, characterized by the equation y = 0.9969x - 0.2445 shown in the upper left corner of the figure, with a correlation coefficient R. 2 It equals 0.9893.

[0064] This line represents a low prediction error value, with RMSEP equal to 3.13957 (Root Mean Square Error of Prediction, root of the mean square error of prediction), indicating that the prediction of biophysical activity values ​​is very accurate.

[0065] The RMSEP value is reported on the x-axis as YPredPS[1](PINCL).

[0066] Figure 3 R-squared NIR spectroscopy (near-infrared fingerprint, a fingerprint or spectrum obtained through near-infrared spectroscopy) and the OPLS model of the barrier effect. 2 and Q 2 .

[0067] The OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections tolatent structures) better separates the systematic variation of information because it separates orthogonal information from parallel information and only considers the predicted information or parallel information.

[0068] Information about block X is given by the NIR fingerprint, while information about block Y is given by the barrier effect (BARR).

[0069] The graph shows the cumulative diagnostic parameters (fit summary) R. 2 and Q 2 The bar charts (represented by light and dark gray respectively) show the parameters reflecting the model performance of the correlation between NIR fingerprint data and the barrier effect. 2 Q is a measure of the model's goodness of fit, while Q... 2 It is a measure of the goodness of prediction (prediction quality). Diagnostic parameter R 2 and Q 2The closer these two parameters are to each other and tend towards 1 (y-axis), the more likely the diagnostic parameter R is to be considered... 2 and Q 2 The test met the requirements. The number of components used is reported on the horizontal axis (x-axis) and indicated by square brackets. The two bars on the left represent the R values ​​of the parallel component and the zero-orthogonal component (comp[1+0]). 2 and Q 2 The values ​​of R are shown in the middle bars, which represent the parallel components and the first, second, and third orthogonal components (comp[1+1], comp[1+2], comp[1+3]), respectively. 2 and Q 2 The values ​​of . The last two bars on the right represent the R values ​​of the parallel component and the fourth orthogonal component (comp[1+4]). 2 and Q 2 The value of .

[0070] Therefore, in this case, the optimal OPLS model is the one using 1 parallel component and 4 orthogonal components (comp[1+4]), because R 2 and Q 2 The value is close to and tends towards 1.

[0071] The image in the bottom right corner shows the software used and the reprocessing date.

[0072] Figure 4a A scatter plot integrating NIR fingerprint data (near-infrared) obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with barrier effect measurements (BARR).

[0073] The information of the blocks is presented as predicted (or calculated) data on the horizontal axis (x-axis). Each circle shown in the figure represents the experimental measurement of the barrier effect for each sample analyzed (see Table 7), characterized by the different compositions as described herein (see Table 6). The samples are identified by the acronym “DOE_EXP 1 to 8” and correspond to the samples referred to as “Experiment 1 to 8” in Table 7.

[0074] The dashed lines in the middle of the figure represent the relationships between the data, characterized by the equation y = 1 / x + 9.003e10 shown in the upper left corner of the figure. -5 Correlation coefficient R 2 It equals 0.9991.

[0075] This straight line represents the low error value of the estimate, RMSEE equals 1.07321 (Root Mean Square Error of Estimation, Root of the root mean square error of the estimate); and it also represents the cross-validation error, RMSE. CV This equals 2.14445 (root mean square error of cross-validation), which will enable very accurate prediction of biophysical activity values ​​(Y data block) based on NIR fingerprints (X data block).

[0076] RMSEE value and RMSE CV The values ​​are reported as YPRED[1] (BARR EFFECT) on the x-axis. The vertical bars on the right are color-coded to represent the experimental values ​​of the Y-block used to create the model, in this case, the measured values ​​of the barrier effect. The graph in the lower right corner shows the software used and the date of reprocessing.

[0077] Figure 4b A scatter plot integrating NIR fingerprint data (near-infrared) obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with barrier effect values ​​(BARR) is used to predict the TARGET barrier effect based on the NIR fingerprint.

[0078] The diagram and Figure 4a The diagram is essentially the same, except that a predicted TARGET sample (DoE_Target) is also present. Therefore, all the details described in the previous diagram remain valid. Furthermore, in this case, the NIR fingerprint of the TARGET sample has been added to the information of block X, which consists of the NIR fingerprint.

[0079] The sample data used to build the model is represented by light gray circles, while the TARGET samples in the prediction model are represented by dark gray circles.

[0080] By utilizing a regression model, the fingerprint information of the TARGET sample is used to predict the information of the Y block, in this case, to predict the barrier effect information.

[0081] The dashed line in the middle of the figure represents the interpolation between data including the TARGET sample, characterized by the equation y = 1.003x + 0.1146 shown in the upper left corner of the figure, with a correlation coefficient R. 2 It equals 0.9892.

[0082] This line represents a low prediction error value, with RMSEP equal to 3.76766 (Root Mean Square Error of prediction), indicating that the prediction of biophysical activity values ​​is very accurate.

[0083] The value of RMSEP is reported on the x-axis as YPredPS[1](BARR EFFECT).

[0084] Figure 5 UHPLC-qToF spectroscopy (fingerprint of ultra-high performance liquid chromatography-quadrupole time of flight, fingerprint or spectrum obtained by ultra-high performance liquid chromatography-quadrupole time of flight) and R-squared of the OPLS model of antioxidant effect 2 and Q 2 .

[0085] The OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections tolatent structures) better separates the systematic variation of information because it separates orthogonal information from parallel information and only considers the predicted information or parallel information.

[0086] Information for block X was obtained from UHPLC-qToF fingerprinting (UNT), while information for block Y was obtained from antioxidant effects (ORAC).

[0087] This figure shows the cumulative diagnostic parameters (fit summary) R. 2 and Q 2 The bar charts (represented by light and dark gray respectively) show the parameters reflecting the performance of the OPLS model in the correlation between UHPLC-qToF fingerprints and antioxidant effects (model performance). 2 Q is a measure reflecting the goodness of the model (goodness of fit), while Q... 2 It is a measure reflecting the goodness of prediction (prediction quality). Diagnostic parameter R 2 and Q2 The closer these two parameters are to each other and tend towards 1 (y-axis), the more likely the diagnostic parameter R is to be considered... 2 and Q 2 The test met the requirements. The number of components used is reported on the horizontal axis (x-axis) and indicated by square brackets. The two bars represent the R values ​​of the parallel components and the non-orthogonal components (comp[1+0]). 2 and Q 2 The value of .

[0088] Therefore, in this case, due to only the parallel components of R 2 and Q 2 The value of is very high and close to 1, so no orthogonal components are used. Therefore, the optimal OPLS model is the one that uses one parallel component and no orthogonal components (comp[1+0]).

[0089] The image in the bottom right corner shows the software used and the reprocessing date.

[0090] Figure 6a A scatter plot integrating fingerprint data obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with UHPLC-qToF (Ultra-High Performance Liquid Chromatography-Quadrupole Time-of-Flight Fingerprint) and antioxidant effect (ORAC) measurements.

[0091] The information for the blocks is presented as predicted (or calculated) data on the horizontal axis (x-axis). Each circle shown in the figure represents the experimental measurement of the antioxidant effect of each sample analyzed (see Table 7), characterized by different compositions as described herein (see Table 6). The samples are identified by the acronym “DOE_EXP 1 to 8” and correspond to samples referred to as “Experiment 1 to 8” in Table 7.

[0092] The dashed lines in the middle of the figure represent the relationships between the data, characterized by the equation y = 1 / x + 8.809e10 shown in the upper left corner of the figure. -5 Correlation coefficient R 2 It equals 0.9591.

[0093] This straight line represents the low error value of the estimate, RMSEE equals 296.847 (Root Mean Square Error of Estimation, Root of the mean squared error of the estimate); and it also represents the cross-validation error, RMSE. CV This equals 339.527 (root mean square error of cross-validation), which will enable very accurate prediction of biophysical activity values ​​(Y data block) based on the UHPLC-qToF fingerprint (X data block).

[0094] RMSEE value and RMSE CV The values ​​are reported on the x-axis (x-axis) under the heading YPRED[1](ORAC). The vertical bars on the right are color-gradients representing the experimental values ​​of the Y-block used to create the model, in this case representing the antioxidant activity values ​​(ORAC values). The graph in the lower right corner shows the software used and the date of reprocessing.

[0095] Figure 6b A scatter plot integrating UHPLC-qToF fingerprint data (ultra-high performance liquid chromatography fingerprint-quadrupole time of flight) obtained using the OPLS algorithm (Orthogonal Partial Least Squares / Orthogonal projections to latent structures) with measurements of antioxidant effect (ORAC) was used to predict the antioxidant effect of TARGET based on the UHPLC-qToF fingerprint.

[0096] The diagram and Figure 6a The figures are essentially the same, except that a predicted TARGET sample (DoE_Target) is also present. Therefore, all the details described in the previous figures remain valid. Furthermore, in this case, the UHPLC-qToF fingerprint of the TARGET sample is added to the information in block X, which consists of the UHPLC-qToF fingerprint.

[0097] The sample data used to build the model is represented by light gray circles, while the TARGET samples in the prediction model are represented by dark gray circles.

[0098] By utilizing a regression model, fingerprint information from the TARGET sample is used to predict information about the Y block, which in this case is used to predict information about antioxidant effects.

[0099] The dashed line in the middle of the figure represents the interpolation between data including the TARGET sample, characterized by the equation y = 1 / x - 38.7 shown in the upper left corner of the figure, with a correlation coefficient R. 2 It equals 0.9513.

[0100] This line represents a low prediction error value, with RMSEP equal to 346.658 (Root Mean Square Error of prediction, root mean square error of prediction), indicating that the prediction of biophysical activity values ​​is very accurate.

[0101] RMSEP values ​​are reported on the x-axis as YPredPS[1](ORAC). Detailed Implementation

[0102] This invention provides a method for characterizing a product having therapeutic and / or health-promoting properties, the product comprising (or composed of) a complex natural system, the method comprising the following steps:

[0103] a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining fingerprint data X data blocks;

[0104] b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining a data block of biophysical properties;

[0105] c. Perform correlation integration on the data through multivariate analysis, which involves the correlation between data block X and data block Y. Integrate the correlation between each data point obtained in a. and each data point obtained in b. to obtain a regression line for each pair of values, and calculate its regression coefficient R. 2 And mean squared estimation error (RMSEE), choose R from them. 2 Data pairs with a ratio of ≥0.8 provide a model capable of predicting each biophysical characteristic of the standard formulation based on the fingerprint used.

[0106] In this way, the regression model can monitor product quality not only from the perspective of qualitative and quantitative fingerprints, but also from the perspective of desired biophysical activities, which represents an unprecedented advance to date in the quality assessment of complex natural products.

[0107] The data obtained in section c. can be used to classify products as "acceptable quality" or "non-acceptable quality" according to specific product standards. Acceptable standards are variable and depend on the specific product.

[0108] The present invention also provides a method for producing and validating one or more batches of a product having therapeutic and / or health-promoting properties, the product comprising a complex natural system, the method comprising the following steps:

[0109] a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining a block;

[0110] b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining biophysical property data block Y;

[0111] c. Perform correlation integration on the data through multivariate analysis, which involves the correlation between data block X and data block Y. Integrate the correlation between each data point obtained in a. and each data point obtained in b. to obtain a regression line for each pair of values, and calculate its regression coefficient R. 2 And mean squared estimation error (RMSEE), choose R from them. 2 Data pairs ≥0.8 provide a model capable of predicting each biophysical characteristic of a standard formulation based on the fingerprint used;

[0112] d. Using the model developed in c., perform qualitative and quantitative fingerprint analysis on one or more batches of the product to be validated using the same method used in a. to predict the biophysical properties described in point b.

[0113] Therefore, the data obtained in d. can be used to classify products as "acceptable quality" or "unacceptable quality" according to specific product standards. Acceptable standards are variable and depend on the specific product, aiming to maintain qualitative, quantitative, and biophysical activity parameters.

[0114] For each specific product, acceptance criteria are established based on fingerprint and biophysical measurements performed on products that differ from the standard formulation. These values ​​represent compliance limits for each characteristic, and exceeding these limits necessitates an investigation into the nonconforming batch.

[0115] The availability of this analytical platform enables us to monitor qualitative, quantitative, and biophysical activity parameters of various production batches and to eliminate batches that deviate from the parameters and cannot guarantee that the biophysical activity will remain within a predetermined range based on an ad hoc control chart.

[0116] Collecting this information from numerous batches allows us to define very stringent quality ranges by considering multiple aspects, based on correlation models—qualitative and quantitative fingerprints and biophysical activity. Therefore, quality ranges can be specifically defined according to the product of interest.

[0117] In this way, the regression model can monitor product quality not only from the perspective of qualitative and quantitative fingerprints, but also from the perspective of biophysical activity, which represents an unprecedented advancement to date in the quality assessment of complex natural products.

[0118] According to the present invention, the product comprising a complex natural system and having therapeutic properties is preferably a product composed of natural substances, such as medical devices as required by paragraph 13.3 of Annex I to EC Regulation 745 / 2017 and / or food supplements with health-promoting activity composed of complex natural substances as defined in Article 2 of Directive 2002 / 46 / EC and Article 2 of Legislative Decree 169 / 2004.

[0119] Therefore, examples of such products can be medical devices or food supplements as defined above, obtained by mixing natural substances and / or their extracts and / or fractions of said extracts. Preferably, the components used to prepare said medical device or said food supplement are obtained through a physical production process with low environmental impact.

[0120] Non-limiting examples of natural substances according to the invention may include plant parts, powders of such parts, extracts of plants or their parts (leaves, flowers, roots, bark, stems, seeds, fruits, petioles, flower parts, etc.), fractions of said extracts (e.g., hydro-alcohol fractions or other fractions), essential oils, natural gums, natural resins, minerals, honey, propolis, animal substances, minerals, etc.

[0121] This product may be available in various forms, such as syrups, tablets, capsules, granules, powders, solutions, suspensions, aerosols, powders for aerosols, sachets of powder, sachets of granules, operculums, sprays, creams, emulsions, soft or hard gelatin capsules, or other suitable formulations commonly used in industry.

[0122] According to the present invention, formulations that differ from the standard product in terms of the presence of the same component in different percentages or in different percentages of the same component compared to the sample are formulations representing low-quality products. These formulations can be prepared by altering the basic components that provide therapeutic or health-promoting activity.

[0123] To reasonably study the contribution of each component to the properties of the compounded product, an appropriate number of formulations must be prepared to assess their impact on the product's properties. Therefore, it is necessary to employ a suitable design of experiments (DOE) to identify the component conditions that affect the response.

[0124] For classic medicines containing one or two specific active ingredients (compounds), a "full-factorial" DOE is used, which involves preparing a 2K formulation, where K = the number of active ingredients. For complex natural products, the components contributing to therapeutic effects are diverse, and the number of low-quality formulations at point b is countless.

[0125] For example, for a product composed of 11 complex natural ingredients, it would be necessary to create more than 1,000 formulations to describe all the non-mass space surrounding the standard formulation.

[0126] Preliminary studies can thus identify the component conditions necessary to maintain product characteristics. This can be achieved by using a mass (or standard) sample representing the target (or the center point of the experimental design) and a low-mass sample, which differ from the mass sample in that the low-mass sample contains the same component at a different percentage compared to the target, or contains the same component at a different percentage.

[0127] To select low-quality formulations, the product ingredients of interest can be grouped together.

[0128] In embodiments of the invention, low-quality formulations can be selected by grouping the components into groups, with three basic groups. The formulation properties are altered by changing the components (in Table 4 of the Examples section, these groups are defined as "factors," namely A, B, and C). Another group can also be identified, where all other components of the product can be defined as independent factorial groups in Table 4, belonging to the group defined as factor D. Using a computer-aided experimental design with a non-orthogonal factorial design, different percentages of the grouped components (factors A, B, and C) can be varied in random combinations, and factor D can be combined in a complementary manner up to 100. Through this calculation process, formulations "suitable for generating regression lines" can be selected, i.e., formulations most likely to construct a line that can be used as a model to predict new data. For the products analyzed in the Examples section, eight formulations were selected, shown in Table 6.

[0129] Therefore, for "total factorial" DOE studies, the number of low-quality formulations to be studied is significantly reduced.

[0130] Qualitative and quantitative fingerprint analysis at point a can be performed using known techniques capable of describing the complexity of the product of interest: for example, infrared (IR) and near-infrared (NIR) spectroscopy, mass spectrometry (MS and various possible couplings), nuclear magnetic resonance (NMR), and ultraviolet-visible spectroscopy (UV-Vis). According to one embodiment, the fingerprint analysis is performed by acquiring chromatograms via UHPLC-qToF, NIR spectral recording, IR spectral recording, or by coupling other mass spectrometers (ion traps, triple quadrupoles, single quadrupoles) with different chromatographic systems (e.g., gas chromatography, liquid chromatography), or by directly injecting mass spectrometry, NMR, or UV-Vis spectroscopy, which is also coupled with a chromatographic system (e.g., liquid chromatography).

[0131] According to the present invention, at point b, one or more biophysical properties related to the therapeutic efficacy of the product of interest are analyzed. Non-limiting examples of such properties include mucosal adhesion properties, barrier effects, antioxidant properties, neutralizing properties, moisturizing properties, lubricating properties, etc.

[0132] Therefore, in one embodiment of the present invention, at point b, the analysis of biophysical properties can be performed by one or more of the following methods: measuring mucosal adhesion, measuring barrier effect, measuring antioxidant effect, measuring neutralizing activity, measuring thermal insulation activity, and measuring lubricating activity.

[0133] The analysis of these properties can be performed using any method known to experts in the field. Non-limiting examples of such techniques include, for example, measuring (mucus) adhesion effects by inclined plane tests, measuring barrier effects by in vitro tests on human fibroblasts or by tests of reduced throughput of dextran-luciferin molecules (70,000 MW) on a semipermeable membrane (transport chamber), and measuring antioxidant effects by ORAC tests.

[0134] In a preferred embodiment, the OPLS algorithm is used to integrate the data between points c.

[0135] In OPLS, data is divided into predictive (or parallel) information and irrelevant (or orthogonal) information, thereby removing variances in X (descriptive variable block) that are irrelevant to Y (feature variable block) and generating a highly predictive correlation model. Therefore, the OPLS model is better because it separates orthogonal and parallel information, considering only predictive or parallel information.

[0136] Algorithms or functions suitable for scaling NIR fingerprint data include, for example, the SNV (Standard Normal Variable) algorithm, first derivative, and second derivative; for fingerprints analyzed by mass spectrometry, the PARETO and MEAN-CENTERING algorithms are used; and for biophysical data, the UV (Unit Variance) or CENTERING algorithms, which are well-known to those skilled in the art, are used.

[0137] Algorithms or functions applicable to data related to the analysis of each biophysical property through transformation processing include, for example, the EXPONENTIAL function (e.g., squared exponent, quarter exponent) or the LOGARITHM function.

[0138] The variables being processed can be used to calculate the model diagnostic parameters R. 2 and Q 2 Conduct an evaluation. Select R. 2 Sufficient data pairs, such as R 2 ≥0.8, preferably R 2 ≥0.9, preferably ≥0.95, or even more preferably greater than or equal to 0.99.

[0139] In any part of this specification, the term "comprising" may be replaced by the term "consisting of".

[0140] The following examples are created on a standard formulation of syrup A (Grintuss, manufactured by Aboca) as defined below, and on formulations that differ from the standard formulation due to the presence of the same ingredient in different percentages or different percentages of the same ingredient. The examples provided are intended to illustrate possible ways of carrying out some steps of the invention or the claimed process, and should therefore not be construed as limiting it in any way.

[0141] Example

[0142] Method used

[0143] 1. Methods for determining NIR fingerprints - reflection patterns

[0144] 1.1. Instruments

[0145] Bruker NIR spectrometer, model MPA (Multi-function Analyzer)

[0146] 1.2. Parameters of the reflection method

[0147] Resolution: 16cm -1 Wavenumber reproducibility: better than 0.04 cm⁻¹ -1 Wavenumber accuracy: better than 0.1cm -1 Photometric accuracy: 0.1%T; Wavenumber range: 4000 to 12000 cm⁻¹ -1 Background scan: 64; Sample acquisition scan count: 64

[0148] 1.3. Sample Preparation

[0149] For NIR analysis, use a disposable plastic pipette to place 3 ml of each syrup sample in the center of the culture dish. Place a metal spacer above the droplet, taking care to eliminate any air bubbles that may form between the spacer and the plate.

[0150] 1.4. Sample Collection

[0151] For each sample, three technical repeat tests were performed, with three separate data acquisitions to monitor data reproducibility. Background subtraction was performed before each acquisition.

[0152] 1.5. Data Preprocessing

[0153] To develop reprocessing methods, it is also necessary to evaluate data preprocessing, which utilizes... The software can thus choose to use SNV-standardized methods as preprocessing methods.

[0154] 1.6. Use Conduct data evaluation

[0155] New samples introduced into the NIR model are evaluated using statistical tests to demonstrate whether they fall within the variance range of the validated model. The aforementioned diagnostic test is the Hotellings T... 2 And DModX.

[0156] 2. Methods for determining NIR fingerprints - transmission mode

[0157] The difference between this method and the previous one is the way the sample is presented for instrument reading, which in this case involves using a test tube with a 2mm optical path.

[0158] 3. Method for determining UHPLC-qToF fingerprint chromatograms

[0159] 3.1. Reagents

[0160] Methanol / ultrapure water 50:50 (v / v), HPLC grade methanol (MeOH), ultrapure water

[0161] 3.2. Materials

[0162] Chromatographic column: Waters C18 1.6μm 2.1x100mm; Precolumn: Waters Cortecs C18 1.6μm 2.1x5mm VanGuard TM Pre-Column 3 / Pk; Syringe filter, 0.2μm RC membrane, Phenomenex

[0163] 3.3. Instruments

[0164] UHPLC-qToF system: Agilent 1290 Infinity coupled with a 6545 quadrupole time-of-flight mass spectrometer (qToF), with a resolution of 2 GHz, using dual AJS ESI electrospray ionization sources (Agilent Technologies). Ultrasonic bath. Analytical balance (resolution 0.00001 g).

[0165] 3.4. Sample Preparation

[0166] Weigh 0.500 g ± 0.005 g of sample into a 50 mL volumetric flask using a glass pasteur pipette. Dissolve the sample in approximately 45 mL of 50% MeOH for 15 minutes using sonication at 35 °C ± 3 °C, followed by LC-MS. At room temperature, bring the volume to 50 mL using the same extraction solvent. Filter 3 mL of the extract through a 0.2 μm cellulose acetate filter into a 4 mL amber vial, discarding the first 0.5 mL of filtrate. Transfer 200 μL of the filtered sample using a calibrated 200 μL micropipette to an amber vial with a glass insert and screw cap. Inject into a UHPLC-qToF system in negative polarity mode [ESI(-)] to determine the metabolomics profile.

[0167] 3.5. UHPLC Chromatographic Conditions

[0168] Injection volume: 3 μl; Column temperature: 40℃. Elution buffer: A. H₂O, 0.1% HCOOH; B. MeOH, 0.1% HCOOH

[0169] Time period: 0 waste liquid; 2 min MS; Flow rate: 0.3 mL / min. Elution protocol:

[0170] Time (min) A(％) B(％) 0 99 1 1 99 1 2 75 25 10 50 50 11 50 50 15 25 75 17 15 85 19 1 99 19.5 1 99 21 99 1

[0171] Post-time: 3 min; Iso pump: 0.05 mL / min

[0172] 3.6. Reprocessing of fingerprint data obtained by UHPLC-qToF

[0173] As a first step, feature extraction (data extraction, data.extraction) is performed using the XCMS software package (version 3.2.0), which employs the following algorithm:

[0174] ·centWave

[0175] The algorithm performs operations such as peak picking and alignment. The algorithm is executed using the parameters shown in Table 2.

[0176] Table 2. Definition of the parameter set used for data extraction function execution

[0177]

[0178]

[0179] As a second step, a filtering function for the obtained variables is used, which is called pre-processing.

[0180] The parameters shown in Table 3 are used to execute the function.

[0181] Table 3. Parameter definitions set for executing the preprocessing function

[0182]

[0183] The final step in variable extraction is to subtract the analytical blank. This is done using the blanksubact.centWave function (Table 4).

[0184] Table 4. Parameter definitions used to execute the blanksubtract.centWave function

[0185]

[0186]

[0187] *Optimize via IPO, using the following range: t1 = 500, 5 <min> 20，30 <max>60

[0188] The final data matrix thus obtained is converted into a.csv file, which is loaded into software in order to perform a statistical reprocessing using OPLS.

[0189] 4. Method for determining the adhesion activity on a slope

[0190] A reference standard solution (RSS) of a physiological saline solution of 0.9% w / w sodium alginate is prepared. The sample to be tested (SAMPLE) (syrup, SYRUP A) is analyzed without any manipulation.

[0191] Before starting the analysis, ensure that the temperature of the RSS and SAMPLE sample is 20°C.

[0192] Using a micropipette, place an equal number of drops of SAMPLE and RSS at the start of a plate with a defined slope. Place the drops on the same line, equidistant from each other. Due to the action of gravity, SAMPLE and RSS move along the plate. If the adhesion capacity of SAMPLE is higher than that of RSS, the SAMPLE test is positive.

[0193] 5. Method for determining the barrier effect

[0194] The assay is based on the principle that a reduction in the passage of dextran-fluorescein molecules (70000 MW) compared to the untreated control (no application of the formulation, representing the positive control) after application of the product is considered an indicator of the protective efficacy of the formulation.

[0195] The test requires the setting of two chambers physically separated by a semi-permeable membrane (transport chamber). The product is applied to the surface of the transport chamber and 10 minutes later the dextran-fluorescein is applied. Taking into account that the barrier effect can be determined by measuring the fluorescence of the hole solution in the receiving plate, the solution in the receiving plate is collected one hour after the application of the dextran and read on a plate reader by setting the excitation wavelength at 494 nm and the emission wavelength at 518 nm. The white (application of the product without dextran) is also taken into account. This condition is used to evaluate the contribution of the potential diffusion of the product inside the receiving chamber.

[0196] After calculating the amount of fluorescein dextran present in the sample and the positive control, the percentage of barrier effect is calculated using the following formula:

[0197] % Barrier effect = 100 - ((sample dextran concentration / average positive control dextran concentration) * 100

[0198] Reagents and Instruments

[0199] • Dextran fluorescein D1823 70000MW (Invitrogen)

[0200] ·PBS (Gibco)

[0201] HTS 96 HTS

[0202] Corning 3605-96 Round-bottomed white plate + lid

[0203] • Plate fluorescence reader (Varioskan-Thermo Fisher)

[0204] ·Eppendorf 1.5ml

[0205] Thermo Fisher p10, p20, p100, p1000 pipettes and tips

[0206] Multipette R Xstream Eppendorf + 5ml suction head

[0207] Falcon 15 and 50ml

[0208] · Vortex oscillator

[0209] • Timer

[0210] Perform measurement

[0211] The reagents required for the test are:

[0212] · Dextran fluorescein As shown in the datasheet, the powder needs to be resuspended in PBS at a concentration of 25 mg / ml. This is a stock solution, and it is necessary to prepare aliquots and store them at -20°C (vortex the stock solution before aliquoting). The concentration used in the test is 500 μg / ml, therefore it must be diluted with PBS at a ratio of 1:50.

[0213] · sample Before starting the test, the sample must be mixed and approximately 5 ml of the sample should be transferred to a 15 ml centrifuge tube (Falcon).

[0214] · PBS Divide the required amount of PBS into 50ml bottles. Store PBS at room temperature.

[0215] The following are the steps for setting up the experiment and conducting the test. Each sample must be made in triplicate.

[0216] Positive control (CTRL+): 15 μl of PBS was placed on top, 90 μl of PBS was placed at the bottom, and 2 μl of 500 μg / ml dextran fluorescein was placed on top.

[0217] Sample: Place 15 μl of sample on top, 90 μl of PBS on the bottom, and 2 μl of 500 μg / ml dextran fluorescein on top.

[0218] Blank: 15 μl of sample was placed on top, 90 μl of PBS on the bottom, and 2 μl of PBS on top. This condition was used to assess the contribution of the product to the potentially widely distributed portion within the receiving chamber.

[0219] 1. Mix the samples to be tested.

[0220] 2. Prepare a 500 μg / ml dextran solution of fluorescein and store it away from light.

[0221] 3. Add 15 μl of syrup (for samples and blanks) or 15 μl of PBS (for Ctrl+) to the surface of the HTS 96 transfer chamber and incubate at RT for 10 minutes. Replace the pipette tip for each transfer chamber.

[0222] 4. At the end of 10', add 90 μl of PBS to the bottom, vortex the 500 μg / ml dextran solution with fluorescein for the first time, and add 2 μl to the top for the sample and Ctrl+. For the blank, add 2 μl of PBS to the top.

[0223] 5. Incubate in the dark at RT for 1 hour.

[0224] 6. Prepare the calibration curve in Eppendorf and store it in the dark at RT for 1 hour.

[0225] 7. Transfer 75 μl (two copies) of each curve point and sample to a 96-well U-shaped white plate and read the fluorescence using varioskan (Ex 494 nm, Em 521 nm).

[0226] Quantitative analysis of dextran fluorescein

[0227] Once fluorescence readings are obtained by setting the excitation and emission wavelengths to 494 nm and 521 nm, respectively, a fluorophore calibration curve for dextran can be generated using the GRAPHPAD PRISM software. A function is set to subtract the average fluorescence value of the blank wells from the sample fluorescence value. The concentration of fluorophore in the samples and positive controls is then calculated by interpolating the fluorescence values ​​of the corresponding wells using a standard curve.

[0228] Results Analysis

[0229] After calculating the levels of dextran fluorescein present in the samples and positive controls, the percentage of barrier effect was calculated using the following formula:

[0230] % Barrier effect = 100 - ((sample dextran concentration / average dextran concentration ctrl + )) * 100

[0231] Compared with the untreated control group (no formulation applied), the reduction in the amount of dextran fluorescein passing through the product was considered an indicator of the protective efficacy of the formulation.

[0232] 6. Method for determining antioxidant properties through ORAC measurement (oxygen radical absorption / antioxidant capacity)

[0233] The method was developed to evaluate the protective effects of antioxidants against reactive hydroxides and peroxides in organisms. It is designed to test human and animal serum samples, plant products, foods, food ingredients, pharmaceuticals, and pet food. Currently, it is considered the only method capable of measuring the ability of antioxidants to inhibit free radicals. The unit of measurement for antioxidant capacity is named the ORAC unit.

[0234] This is a highly sensitive method that uses fluorescein as a fluorescent label and utilizes 2,2'-azobis(2-amidinepropane)dihydrochloride (AAPH)—a water-soluble azo compound—which thermally decomposes to form aqueous peroxide radicals at a constant rate. fluorescein fluorescence is extremely sensitive to the conformation and chemical integrity of the molecule itself. Under appropriate conditions, fluorescence loss in the presence of peroxide radicals is an indicator of oxidative damage to the reactant species. In a given sample, the inhibition of fluorescein degradation (reflected in the retention of its fluorescence) benefits from the protective effect of the present antioxidants and is a measure of the sample's antioxidant capacity against the reactant species. As long as antioxidants can capture free radicals, they protect the fluorescent label from decay; once the antioxidant effect ends, free radicals react with fluorescein, leading to loss of fluorescence. The fluorescence decay time is directly proportional to the amount and activity of the antioxidants present in the sample.

[0235] Therefore, the ORAC assay measures the decrease in fluorescence of the labeled molecule fluorescein over time, which is due to damage caused by oxygen-focused free radicals.

[0236] The results are quantified by allowing the reaction to proceed to completion and then integrating the area under the kinetic curve relative to a blank reaction without added antioxidants. The area under the curve (AUC) is proportional to the concentration of all antioxidants present in the sample. This method combines two parameters—the time and percentage of inhibition of the reactants due to the total antioxidants—into a single quantity for calculation; therefore, it is superior to similar methods that use either the percentage of inhibition at a fixed time or the time of inhibition at a fixed percentage of inhibition.

[0237] The final result was calculated using the difference in fluorescein decay AUC (relative fluorescence versus time) between the test sample and the blank.

[0238] The standard (6-hydroxy-2,5,7,8-tetramethyltryptane-2-carboxylic acid) is a water-soluble analog of vitamin E.

[0239] ORAC unit = (AUC sample - AUC blank) / (AUC standard - AUC blank) x [standard] / [sample] = μmol / g

[0240] The results are reported based on the following equivalence: 1 ORAC unit = 1 μM per 100g sample. equivalent.

[0241] Example 1: Characterization of Complex Products

[0242] Grintuss, also referred to herein as "Syrup A," is used to treat coughs and is characterized and manufactured through a highly standardized production process. It consists of a mixture of the following ingredients:

[0243] • Polysaccharides, resins and flavonoids-based polysaccharides (LPME extract fractions of plantain, guarana and helichrysum obtained by a low-environment-impact physical extraction process (LPME process, i.e., liquid-phase microextraction, characterized by liquid-phase extraction of plant parts into microparticles) of three starting plants (plantain, guarana, and helichrysum, the physical extraction process being able to produce extract fractions and standardize their content in the finished product), honey, sugarcane, water, essential oils of eucalyptus, star anise and lemon, natural lemon flavoring, gum arabic and xanthan gum.

[0244] All experiments described below were performed in triplicate.

[0245] 1. Design of Experiments (DOE) and Characterization of Experimental Samples

[0246] To obtain a correlation model that integrates the qualitative and quantitative characteristics of a product with its corresponding biophysical features, different quality formulations of the original complex natural product were defined. This model, used to characterize a large number of quality samples, will allow for the strict definition of the quality levels of the same product.

[0247] - For syrup formulation A, Table 5 shows all the components identified as important for studying its effects on the properties of the complex natural product. These components are categorized as Factor A (represented by the components polysaccharides, gum arabic, and xanthan gum), Factor B (represented by the component honey), and Factor C (represented by the components sugar and water). Factor D is reported because it represents other components of the formulation (components: water, aroma, lemon, and eucalyptus oil), but for the purposes of this study, it is considered an independent factor.

[0248] Table 5. Composition of DOE Research Factors

[0249]

[0250] Note 1. Mixture D is considered an independent factor.

[0251] Based on the composition of the TARGET product (syrup A), computer-aided design (CAD) was applied to increase or decrease factors A, B, and C to cover the composition space (box) centered on the TARGET, thereby generating a grid of possible combinations. Only grid elements that met the formulation constraints were included in the candidate experimental set, which was sampled to select eight formulations for consideration in the response study. Therefore, Table 6 shows the accepted factor combinations and the percentage increase or decrease compared to the target. Specifically, the variation range of factor A was found to be ±80%, while the variation ranges of factors B and C were ±20%.

[0252] The computer-executed combinatorial program aims to obtain the DOE matrix with the minimum "condition number". Therefore, due to the use of the formulation experimental design procedure, the distance between the candidate formulation and the fully factorial design experiment has been minimized to a minimum of 2. 3 (2 K Because three factors (A, B, C) were considered, it is equivalent to 8 experiments.

[0253] Therefore, Table 6 reports the formulations under investigation. The DOE matrix above supports linear and interaction models with a condition number less than 3.6.

[0254] Table 6. Percentage Changes in Experimental Samples Used by the DOE Institute Compared to the AC Factor Target

[0255]

[0256] The formulations, referred to as Experiments 1 through 8, were used to record fingerprints and biophysical properties.

[0257] 1.1 - Fingerprints using NIR spectroscopy.

[0258] Choosing between 4000 and 12500cm -1 The INR field was calculated, and background noise was subtracted. Spectral recording can be performed in either transmission or reflection mode, where in spectroscopy, transmittance T is the intensity of the transmitted light (I0). t The reflectivity is the ratio of the intensity of the incident ray (I0) to the intensity of the reflected ray (I2). r The ratio of the intensity of the incident light to the intensity of the incident light (I0).

[0259] NIR spectra were converted into a data matrix, where variables were characterized by wavelengths and corresponding transmittance or reflectance values. The matrix containing data from all studied samples was then aligned and preprocessed, specifically transformed using the SNV (Standard Normal Variable) algorithm and scaled using the MEAN-CENTERING function. Further preprocessing functions could be first and second derivatives, or combinations thereof with the SNV algorithm. The resulting final data matrix was used in the next stage for integration with biophysical data using OPLS.

[0260] The software packages available for NIR spectral analysis are proprietary to NIR instruments, such as OPUS software for Bruker.

[0261] Another software used for NIR spectral analysis is (Sartorius)

[0262] 1.2- Fingerprints obtained by UHPLC-qToF mass spectrometry.

[0263] Feature extraction (these are variables) is performed from the fingerprint using peak extraction and peak operation alignment. The parameters shown in Table 2 are used to execute the algorithm.

[0264] As a second step, a filtering function is applied to the variables obtained using the parameters reported in Table 3. The final step in extracting the variables involves subtracting the analysis blank (Table 4). The resulting final data matrix is ​​converted to a .csv file and used in subsequent stages for integration with the activity data for statistical reprocessing using OPLS correlation.

[0265] As an example of a software package that can be used for data extraction, data processing, and data analysis, XCMS and... (Sartorius).

[0266] 1.3 Measurement of Adhesion Properties

[0267] This biophysical property is measured by placing a complex natural product (e.g., syrup A) on an inclined plane, where the product moves under gravity. The tested product must be more viscous than a reference product, which is verified by measuring the distance (in mm) that the product and the reference product travel on the inclined plane.

[0268] The measurements are performed under constant temperature conditions, and the inclined plane has a defined and fixed angle to ensure that the measurements are reliable and reproducible.

[0269] The detailed procedures used in this case are given as examples in the Materials and Methods section.

[0270] 1.4 Measurement of Barrier Properties

[0271] This biophysical property was evaluated by placing the product on a semipermeable membrane (transfer chamber) and measuring its ability to prevent the passage of the compound dextran-fluorescein (molecular weight 70,000).

[0272] The detailed procedures used in this case are given as examples in the Materials and Methods section.

[0273] 1.5 Measurement of Antioxidant Properties

[0274] This property is measured using the ORAC test, according to standard protocols known to those skilled in the art. Detailed procedures employed in this case are given as examples in the Materials and Methods section.

[0275] 2. Calculation of Multiple Linear Regression

[0276] Table 7 reports the experimental measurements of the biophysical properties used to calculate the multiple linear regression.

[0277] Table 7. Measurements of antioxidant activity, adhesion activity, and barrier effect of the experimental samples used in the DOE study.

[0278]

[0279] Note 2. (*) This data is a prediction calculation result and will be used to build the model.

[0280] Before assessing the relationship between fingerprints and biophysical responses (to be performed using OPLS), biophysical responses were analyzed based on multiple linear regression (classical least squares regression) to verify the existence of a linear relationship between the components of factors A, B, and C in eight different formulations (Experiments 1 to 8). Therefore, different correlation coefficients were evaluated based on Equation 1, and the results are reported in Table 8.

[0281] Equation (1) f(Y)=aA+bB+cC+dAB+eAC+fBC+g

[0282] Where: f(Y) is the transformation function of the evaluated response.

[0283] a, b, and c = regression coefficients, representing the influence of variations in parameters A, B, and C on antioxidant biophysical activity (a), adhesion effect (b), and barrier effect (c).

[0284] d, e, and f are regression coefficients, representing the effects of variations in the combined parameters AB, AC, and BC on antioxidant biophysical activity (a), adhesion effect (b), and barrier effect (c).

[0285] A, B, and C = the variable component mixtures from Experiments 1-8

[0286] g = intercept

[0287] Data related to biophysical parameters were transformed using a base-10 logarithmic function, a function with an exponent of 1 / 4, and a squared exponent function (also known as the Box-Cox transformation; Box, GEP and Cox, DR (1964). An analysis of transformations, Journal of the Royal Statistical Society, Series B, 26, 211-252.) to analyze antioxidant activity, adhesion activity, and barrier effect.

[0288] Table 8 shows the results of the multiple linear regression assessment. For the three activities assessed, Q... 2 and R 2 The values ​​are all very good.

[0289] Table 8. Evaluation of Interaction Models Using Multiple Linear Regression Based on Biophysical Activity Data

[0290]

[0291] Note 3: Coeff.SC represents the scaling factor, which represents the logarithmic function with base 10, the function with an exponent of 1 / 4 or 0.25, and the exponential square function, respectively.

[0292] The coefficient 'a' represents the correlation of factor A in mixtures 1-8, and therefore shows a good correlation (positively proportional) with antioxidant activity and a good correlation (inversely proportional) with adhesion effect. Factor A is related to the presence of poliresin and gums (xanthan gum and gum arabic).

[0293] The coefficient b represents the correlation of factor B in mixtures 1-8, and therefore shows a good correlation (inversely proportional) with the adhesion effect and a good correlation (inversely proportional) with the barrier effect. Factor B is related to the presence of honey.

[0294] The coefficient c represents the correlation of the C factor in mixtures 1-8, and therefore has a good correlation (directly proportional) with the adhesion effect and a relatively good correlation (inversely proportional) with the barrier effect. Factor C is related to the presence of sugar and water.

[0295] The coefficient f represents the correlation of the combined BC factors of mixtures 1-8, and therefore has a good correlation (proportional) with the adhesion effect. The combined BC factors are related to the presence of honey, sugar, and water.

[0296] By inserting the data from Table 8 into Equation 1, the corresponding regression equation was obtained, enabling the study of the relationship between the responses of factors A, B, and C in experimental formulations 1-8 (Table 7) while considering the interaction model. The Box-Cox transformation was applied to the responses to obtain normally distributed residuals.

[0297] A base-10 logarithmic transformation of the ORAC response yields a linear model with the following equation:

[0298] log 10 (ORAC)=3.259(p<0.001)+0.317(p<0.001)A-0.085(p=0.014)C

[0299] The condition number is 1.81, and the p-value for the ANOVA is less than 0.001. The goodness of fit of the multiple linear regression model is R0. 2 = 0.97. The result of leave-one-out (loo) cross-validation is Q. 2 =0.93.

[0300] The slope adhesion response is transformed by a power of 0.25, resulting in an interaction model with the following equations:

[0301] (Adhesive on inclined plane) 0.25 =2.296(p<0.001)-0.288(p<0.001)A-0.262(p<0.001)B-0.555(p<0.001)C+0.068(p=0.022)BC

[0302] The condition number is 2.96, and the p-value obtained from ANOVA is less than 0.001. The model shows R... 2 =0.99 and Q 2 =0.98.

[0303] The "barrier effect" response is subjected to a square power transformation, resulting in a linear model with the following equation:

[0304] (Barrier effect) 2 =8820(p<0.001)-1130(p=0.047)A-1300(p=0.129)B-3300(p=0.011)C.

[0305] The response value of Experiment 5 was excluded from the calculation because it was found to be saturated. The condition number was 3.46, and the p-value obtained from the ANOVA test was 0.022. The model showed R... 2 =0.78, Q 2 =0.21. Therefore, the response estimate in Experiment 5 is 110.

[0306] 3. Integrating the correlation between qualitative and quantitative data and biophysical characteristics using OPLS

[0307] Based on the positive results obtained from the multiple linear regression study of the biophysical properties measured in mixtures 1-8, we constructed a correlation model between qualitative and quantitative fingerprint data and biophysical data using the Orthogonal Projection Latent Structure (OPLS) algorithm. In OPLS, experimental data are divided into predictive (or parallel) information and uncorrelated (or orthogonal) information, thereby eliminating the variation of X (descriptor variable) that is uncorrelated with Y (characteristic variable). Therefore, OPLS models are superior because they only consider predictive or parallel information, separating it from unpredictable or orthogonal information.

[0308] Similar correlations can also be obtained using PLS and PLS2 algorithms (see pages 14-15-16 for definitions, Mattoli L, Gianni M, Burico M. Mass spectrometry-based metabolomics analysis as a tool for quality control of natural complex products. Mass Spectrom Rev. 2022 Mar 3:e21773.doi:10.1002 / mas.21773.Epub ahead of print.PMID:35238411; see pages 430-431 User Guide to SIMCA 15 By Sartorius (https: / / www.sartorius.com / download / 544940 / simca-15-user-guide-en-b-00076-sartorius-data.pdf; see: Stocchero M., Riccadonna S., Franceschi P., Projection to latent structures with orthogonal constraints for metabolomics). data,Journal of Chemometrics.2018;e2987.https: / / doi.org / 10.1002 / cem.2987)

[0309] 3.1. OPLS Correlation Integration of NIR Fingerprint Data and Adhesion Data

[0310] Fingerprint data was preprocessed using the SNV (Standard Normal Variance) algorithm and scaled using the MEAN-CENTERING function. Data related to adhesion activity measurements were transformed using an exponential function of 1 / 4 (or 0.25) and scaled using the UV (Unit Variance) function. Generally, transformation algorithms can be exponential or logarithmic, such as Log, NegLog, Logit, or exponential. Other transformations can be evaluated to obtain normally distributed variables.

[0311] By calculating the model diagnostic parameter R 2 and Q 2 The parameters were evaluated to assess the variables, and the results were excellent, equaling 0.996 and 0.984 respectively. Therefore, the model consists of one parallel component and two orthogonal components, also written as [1+2], as follows: Figure 1 (R of one parallel component and two orthogonal components) 2 and Q 2 The OPLS_NIR & adhesion effect of the model diagnostic parameters is shown in the figure.

[0312] Figure 2a The graph shows the experimental values ​​of the slope adhesion on the vertical axis and the predicted values ​​on the horizontal axis. The graph also shows the regression line as an expression of the experimental data. In the case of an ideal straight line, if the regression coefficient equals 1, the predicted value will equal the experimental value. Since there is always a deviation between theoretical and experimental values, the regression coefficient R is calculated... 2 By calculating the root mean square error (RMSEE) of the estimated values, we can estimate the goodness of the model and the deviation of the predicted values ​​from the experimental values.

[0313] therefore, Figure 2a The model reported by the figure represents a high correlation between samples, and therefore has high predictive power, R. 2 The value is very high, >0.99, and the RMSEE (root mean square error of the estimate) is equal to 2.0 (based on a scale of 70). Figure 2b In this study, the model was used to predict the slope adhesion effect of TARGET samples based on NIR fingerprints. It can be seen that the prediction is very good because R... 2 The value is very high, >0.99, and the mean square prediction error (RMSEP) is also very good, equal to 3.1 (based on a scale of 70).

[0314] In summary, Figure 2 shows an OPLS scatter plot integrating NIR fingerprint data and slope adhesion values, where a correlation was observed between the experimental adhesion effect (vertical axis) and the predicted adhesion effect (horizontal axis). Figure 2a ) and the adhesion effect of TARGET predicted based on NIR fingerprints ( Figure 2b ).

[0315] The graph in Figure 2 shows that when new samples are projected (e.g., samples in production batches or product reformulation studies), the effect size estimates will be very accurate due to the very low error.

[0316] 3.2. OPLS Correlation Integration of NIR Fingerprint Data and Barrier Effect Data

[0317] Fingerprint data was preprocessed using the SNV (Standard Normal Variable) algorithm and scaled using the MEAN-CENTERING function. Data related to barrier effect measurements were transformed using the squared exponential function and scaled using the UV (Unit Variance) function. Generally, transformation algorithms can be exponential or logarithmic, such as log, NegLog, Logit, and exponential. Other transformations can be evaluated to obtain normally distributed variables.

[0318] By calculating the model diagnostic parameter R 2 and Q 2 The parameters were evaluated to assess the variables treated, and the results were excellent, at 0.994 and 0.954 respectively. Therefore, the model consists of one parallel component and four orthogonal components, and is written as [1+4], as follows: Figure 3 (R of one parallel component and four orthogonal components) 2 and Q 2 The OPLS NIR & barrier effect of the model diagnostic parameters is shown in the figure.

[0319] exist Figure 4a In the graph, the experimental values ​​of the barrier effect are shown on the vertical axis, while the predicted values ​​are shown on the horizontal axis. The graph shows the regression line as an expression of the experimental data. In the case of an ideal straight line, if the regression coefficient equals 1, the predicted value will equal the experimental value. Since there is always a deviation between theoretical data and experimental values, the regression coefficient R is calculated... 2 The root mean square error (RMSEE) of the estimated values ​​can be used to estimate the goodness of the model and the deviation of the predicted values ​​from the experimental values.

[0320] therefore, Figure 4a The model shown in the figure represents a high correlation between samples, and therefore has high predictive power, R. 2 The value is very high, >0.99, and the RMSEE (root mean square error of the estimate) is equal to 1.0 (on a scale of 100). Figure 4b In this study, the model was used to predict the barrier effect of TARGET samples based on NIR fingerprints. It can be seen that the prediction is very good because R... 2 The value is very high, >0.98, and the mean square prediction error (RMSEP) is equal to 3.7 (based on a scale of 100).

[0321] In summary, Figure 4 shows an OPLS scatter plot integrating NIR fingerprint data and barrier effect values, where the experimental barrier effect (vertical axis) and the predicted barrier effect (horizontal axis) are observed. Figure 4a ), and predicting the TARGET barrier effect based on NIR fingerprints ( Figure 4b ).

[0322] The graph in Figure 4 shows that when new samples are projected (e.g., samples in production batches or product reformulation studies), the effect size estimates will be very accurate due to the very low error.

[0323] 3.3. Correlation between UHPLC-qToF fingerprint data and antioxidant effect data integrated by OPLS

[0324] UHPLC-qToF fingerprint data were scaled using the PARETO and MEAN-CENTERING functions to ensure a normal distribution of variables. Data related to antioxidant effect measurements were not transformed; only the UV (unit variance) function was used for scaling. Generally, transformation algorithms can be exponential or logarithmic, such as log, NegLog, Logit, or exponential. Other transformations can be evaluated to obtain normally distributed variables.

[0325] By calculating the model diagnostic parameter R 2 and Q 2 The parameters were evaluated to assess the variables treated, and the results were excellent, at 0.959 and 0.929 respectively. Therefore, the model consists of one parallel component and zero orthogonal components, and can be written as [1+0], as follows. Figure 5 (R of a parallel component and a zero-orthogonal component) 2 and Q 2 The OPLS UHPLC-qToF and antioxidant effects of the model diagnostic parameters are shown in the figure.

[0326] Figure 6a The graph shows the experimental values ​​of the antioxidant effect on the vertical axis and the predicted values ​​on the horizontal axis. The graph also shows the regression line representing the experimental data. In the case of an ideal straight line, if the regression coefficient equals 1, the predicted value will equal the experimental value. Since there is always a deviation between theoretical and experimental data, the regression coefficient R is calculated... 2 The root mean square error (RMSEE) can be used to estimate the goodness of the model and the deviation of the predicted values ​​from the experimental values.

[0327] therefore, Figure 6a The model reported in the figure represents a high correlation between samples, and therefore has high predictive power, R0. 2 The value is very high, >0.95, and the RMSEE (root mean square error of the estimate) is equal to 297 (based on a scale of 4000). Figure 6b In this study, the model was used to predict the antioxidant effect of the TARGET sample based on UHPLC-qToF fingerprinting. It can be seen that the prediction is very good because R... 2 The value is very high, >0.95, and the mean square prediction error (RMSEP) is equal to 347 (based on a scale of 4000).

[0328] In summary, Figure 6 shows an OPLS scatter plot of UHPLC-qToF fingerprint data and surface adhesion values, where a correlation was observed between the experimental antioxidant effect (ordinate) and the predicted antioxidant effect (bb). Figure 6a ), and predicting the antioxidant effect of TARGET based on UHPLC-qToF fingerprint ( Figure 6b ).

[0329] The graph in Figure 6 shows that when new samples are projected (e.g., samples in production batches or product reformulation studies), the effect size estimates will be very accurate due to the very low error.< / max> < / min>

Claims

1. A method for characterizing a product having therapeutic and / or health-promoting properties, said product comprising a complex natural system, said method comprising the following steps: a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining a block; b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining biophysical property data block Y; c. The data are correlated through multivariate analysis, involving the correlation between data block X and data block Y. Correlation is integrated between each data point obtained in a. and each data point obtained in b. to obtain a regression line for each pair of values, and the regression coefficient R is calculated. 2 And mean squared estimation error (RMSEE), choose R where 2 Those data pairs with a value ≥0.8 provide a model capable of predicting each biophysical property of the standard formulation based on the fingerprint used.

2. A method for producing and validating one or more batches of a product having therapeutic and / or health-promoting properties, said product comprising a complex natural system, said method comprising the following steps: a. Perform qualitative and quantitative fingerprint analysis on the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining fingerprint data X data blocks; b. Analyze one or more biophysical properties of the standard formulation of the product and the formulation that differs from the standard formulation due to the presence of the same component of the same mass in different percentages or due to the presence of the same component of different masses in the same percentages, thereby obtaining biophysical property data block Y; c. The data are correlated through multivariate analysis, involving the correlation between data block X and data block Y. Correlation is integrated between each data point obtained in a. and each data point obtained in b. to obtain a regression line for each pair of values, and the regression coefficient R is calculated. 2 And mean squared estimation error (RMSEE), choose R where 2 Those data pairs ≥0.8 provide a model capable of predicting each biophysical property of the standard formulation based on the fingerprints used; d. Using the regression model developed in c., qualitative and quantitative fingerprint analysis is performed on one or more batches of the products to be verified using the same method used in a., thereby predicting the biophysical characteristics described in point b.

3. The method according to any one of claims 1 or 2, wherein the qualitative and quantitative fingerprint analysis is performed by: recording NIR spectra, recording IR spectra; or acquiring chromatograms using a chromatographic system coupled to a mass spectrometer, such as: gas chromatography, liquid chromatography and qToF, single quadrupole mass spectrometry, triple quadrupole mass spectrometry, ion trap mass spectrometry; or mass spectrometry, NMR, UV-Vis spectroscopy performed by direct injection into the aforementioned mass spectrometer, which is also coupled to a chromatographic system—a liquid chromatography system.

4. The method according to any one of claims 1 to 3, wherein the one or more biophysical properties are selected from: mucosal adhesion ability, barrier effect, antioxidant effect, neutralizing activity, moisturizing activity, and lubricating activity.

5. The method according to any one of claims 1 to 4, wherein the correlation integration in c is performed using PLS, OPLS, or PLS2.

6. The method according to any one of claims 1 to 5, comprising a complex natural system, wherein the product having therapeutic properties is in the form of a medical device composed of natural substances, and wherein the product having health-promoting properties is in the form of a food supplement composed of natural substances.

7. The method according to any one of claims 1 to 6, wherein the natural substance is selected from parts of plants, powders of such parts, extracts of plants or parts thereof such as leaves, flowers, roots, bark, stems, seeds, fruits, petioles, flower fractions, essential oils, natural gums, natural resins, minerals, honey, propolis, animal matter, and minerals.

8. The method according to any one of claims 1 to 7, wherein the product is in the form of syrup, tablet, capsule, lozenge, granule, powder, solution, suspension, aerosol, powder for aerosol, sachet powder, sachet granule, cap, spray, cream, emulsion, soft or hard gelatin capsule.

9. The method according to any one of claims 1 to 8, wherein the R 2 ≥0.

9.

10. The method according to any one of claims 1 to 9, wherein the correlation integration of the data in c is performed using OPLS.

Images