A method for evaluating the equivalence of traditional decoction and decoction of traditional Chinese medicine formula granules based on multi-index dimension reduction discriminant analysis

Abstract

The application relates to a traditional Chinese medicine formula granule decoction and traditional decoction equivalence evaluation method based on multi-index dimension reduction discriminant analysis, which can effectively solve the problem that the prior art method is relatively single, only a single category component index is compared, and multi-dimensional comprehensive index evaluation such as components and efficacy is not performed, and the technical solution is that the information of multiple batches of samples meeting various quality standards is preliminarily calculated, the relative stability of the quality of the multiple batches of samples is judged according to TD, after necessary dimension reduction of the multi-index information, a sample principal component analysis score graph that can explain a certain variation under a certain confidence is formed, so that the equivalence relationship of the two is directly judged, a function is constructed on the basis of the most integrated, most important and most abstract integration of sample information, and the function is passed through the PCA score graph, the relationship between the samples can be quickly and directly and comprehensively presented, and qualitative and quantitative description is realized, which is an innovation of the traditional Chinese medicine formula granule decoction and traditional decoction equivalence evaluation method.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine analysis technology, and in particular to a method for evaluating the equivalence between traditional Chinese medicine formula granules and decoctions based on multi-index dimensionality reduction discriminant analysis. Background Technology

[0002] Traditional decoctions (TD) are the earliest and most widely used dosage form in my country. They allow for the addition or subtraction of herbs according to the patient's condition and serve as a primary vehicle for showcasing the advantages and characteristics of traditional Chinese medicine formulas. However, in today's fast-paced lifestyle and the modernization of traditional Chinese medicine, their inconvenience and other drawbacks severely restrict their application and development. At this juncture, decoction granules (DGD), a relatively new form of traditional Chinese medicine, have emerged due to their ease of use and other advantages, gaining popularity. However, the question remains whether DGD is equivalent to TD. Therefore, a suitable and intuitive method is needed to analyze and assess their equivalence.

[0003] Previous literature on the quality evaluation of DGD and TD has largely relied on HPLC fingerprint chromatograms to assess similarity. While this provides numerical similarity, many studies only evaluated single batches of the same sample without parallel evaluations, making it impossible to determine sample distribution. Furthermore, the methods are relatively simplistic, comparing only single-category component indicators without comprehensive evaluation of components, efficacy, and other multi-dimensional indicators. There are no standards for determining whether TD and DGD are interchangeable, or what level of similarity qualifies them, nor are confidence levels provided, and no graphical, intuitive distinction or judgment is possible. Therefore, it is imperative to develop a scientific method more suitable for evaluating the equivalence of DGD and TD. Summary of the Invention

[0004] In response to the above situation and to overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method for evaluating the equivalence of traditional Chinese medicine formula granules and decoctions based on multi-index dimensionality reduction discriminant analysis. This method can effectively solve the problems of existing technologies being relatively simple, comparing only single-category component indicators without evaluating them in terms of components, efficacy, and other comprehensive indicators; whether the two (TD and DGD) can be substituted, what similarity level is required for substitution, lack of standards, inability to provide confidence levels, and inability to achieve intuitive graphical differentiation and judgment.

[0005] The technical solution provided by this invention is that the equivalence evaluation method includes the following steps:

[0006] 1) Construct the initial discriminant D:

[0007]

[0008] Where c T c represents the coefficient of variation (RSD) of each variable in the traditional decoction group.D Let be the coefficient of variation (RSD) of each variable value in a certain formula granule decoction group, p be the number of variables in each sample group, and i can be 1, 2, 3...n (n≤p). D≥0 indicates that the variation within the DGD sample group is less than that within the TD group, and the sample quality is relatively stable. D<0 indicates that the variation within the DGD sample group is greater than that within the TD group, indicating that the quality of the DGD sample in this group is unstable and will be judged as unqualified drug. If DGD is unqualified drug, no further intuitive discrimination analysis is required.

[0009] The above formula is based on the principle that the intra-group variation of all DGDs should not be greater than the intra-group variation of TDs (although the preparation of formulation granules is relatively complex, its purpose is to standardize the decoction, so its standard should at least be equivalent to the decoction). Otherwise, it indicates that the quality of the sample group is unstable and will be judged as unqualified drugs.

[0010] 2) Construct the discriminant function F:

[0011]

[0012] E1 is an ellipse representing the quality of TD, and E2 is an ellipse representing the quality of DGD. x and y are the coordinates of each sample point. TD is used as a reference standard. For samples that meet the discriminant function, they can be considered equivalent to TD. If they do not meet the discriminant function, they are not equivalent. They can be corrected and re-discriminated. After verification and adjustment, it can be determined whether they are equivalent.

[0013] The principle upon which the above discriminant is based is as follows: taking the quality data of qualified DGD samples and the quality data of TD samples, and using the relevant knowledge of two-dimensional random variable probability theory, an elliptical region (confidence interval) of the two-dimensional random sample set can be obtained under a certain confidence level. When the confidence levels α are 95.0%, 90.0%, and 68.3%, the abscissa or ordinate of the sample point appears in the regions of μ±1.96σ, μ±1.65σ, and μ±σ, respectively. The joint probability density function of the two-dimensional continuous random variable is calculated, and the confidence interval is constructed accordingly. The reason it is an ellipse is that the probability density function of the two-dimensional normal distribution (not necessarily the standard normal distribution) is a three-dimensional figure similar to a "hat". The major and minor axes of an ellipse represent the ranges of the components (projections) of a two-dimensional variable on the two coordinate axes. The confidence interval of the ellipse is centered at the ellipse, with its major axis rotated by an angle β = arctg(b / a) along the longest median of the triangle, where b is the vertical axis and a is the horizontal axis. The angle can also be expressed as arctg(s). o / s t ) indicates that s o Let s be the standard deviation of the projection of all samples onto the ordinate. tis the standard deviation of the projection of all samples onto the x-axis.

[0014] 3) Constructing the discrimination criteria: In the two-dimensional principal component plot obtained by PCA, each sample point is a two-dimensional random variable. If the explained variance reaches 80% or more, an ellipse can be constructed to visually represent the relationship between the two groups of samples. However, if the explained variance is less than 80%, a three-dimensional or higher dimension needs to be constructed to represent the relationship between the samples. In the two-dimensional plane, it is assumed that the points are represented by (x, y). At this time, x and y are orthogonal. The randomly distributed regions can be constructed as any possible ellipse equations E1 and E2 in the two two-dimensional plane plots, where x and y are the coordinates of each sample point, A, B, and C are coefficients, and u, v, and w are constants.

[0015] E 1： A1(x-u1) 2 +2B1(x-u1)(y-v1)+2C1(y-v1) 2 +w1=0 Formula 3

[0016] E 2： A2(x-u2) 2 +2B²(x-u²)(y-v²)+2C²(y-v²) 2 +w2=0 Formula 4

[0017] The above steps are then visualized using MATLAB to intuitively evaluate the equivalence between DGD and TD.

[0018] In step 2), if the discriminant function is not met, that is, if the two ellipses in the graph are separate and cannot be substituted, the equivalent ratio of the granules in the formula can be corrected and adjusted as necessary, and then the discriminant function can be tested again. If the discriminant function has a solution at this time, that is, if the two ellipses intersect or contain each other, it is considered that the two can be substituted to different degrees after a certain correction. If the discriminant function still cannot be passed, that is, if the discriminant function has no solution, then the two are considered to be substituted.

[0019] This invention proposes a novel approach to the study of Disseminated Glycoside Digestion (DGD) and presents a scientific and feasible method for discriminating the similarity between DGD granules and traditional decoctions (TD) with a certain confidence level. This method is not only simple, fast, and intuitive to operate, providing comprehensive feedback on the relationship between samples, but also offers qualitative and quantitative descriptions with accurate results. It represents an innovation in the evaluation method for the equivalence of DGD granules and traditional decoctions. Attached Figure Description

[0020] Figure 1 This is a flowchart of the evaluation method of the present invention.

[0021] Figure 2The two-dimensional PCA score diagrams (a) of multiple batches of roasted licorice from Factory A before DGD and TDTD correction; and (b) of roasted licorice from Factory A after DGD and TD correction.

[0022] Figure 3 The two-dimensional PCA score diagrams (a) of multiple batches of roasted licorice from Factory B before DGD and TDTD correction; and (b) of roasted licorice from Factory B after DGD and TD correction.

[0023] Figure 4 The two-dimensional PCA score diagrams (a) of multiple batches of roasted licorice from Factory C before DGD and TDTD correction; and (b) of roasted licorice from Factory C after DGD and TD correction.

[0024] Figure 5 The images show the two-dimensional PCA score diagrams (a) and (b) of different batches of Amomum villosum from manufacturer E before and after correction of DGD and TD. Detailed Implementation

[0025] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0026] Example 1: Equivalence evaluation of multiple batches of prepared licorice root granule decoction from Manufacturer A compared with traditional decoction.

[0027] Data on the quality of multiple batches of processed licorice root granules (DGD) and traditional processed licorice root decoction (TD) from Factory A were collected using HPLC and UPLC-MS / MS. Differences between DGD and TD were analyzed using dimensionality reduction and discriminant analysis of various indicators. First, the quality stability of DGD was determined for each batch of TD and DGD using the discriminant expression D. If D≥0, the next step was to perform PCA dimensionality reduction on multiple indicators, obtaining two-dimensional principal components that explained up to 80% of the variance. The differences between the two were then visualized using MATLAB.

[0028] Calculations showed that the difference between the DGD and TD groups of processed licorice root from Factory A was smaller than that from Factory TD, satisfying discriminant D, indicating stable quality between batches. Dimensionality reduction and visualization analysis of the two groups of sample data were performed. The figure below shows the two-dimensional PCA score plots of the TD and DGD samples of processed licorice root at a 95% confidence level. The principal components explained over 80% of the variance. Figure 2 It can be seen intuitively that at this confidence level, the two ellipses of the TD sample and the DGD sample from Factory A intersect, which means that TD and DGD are substitutable, but the degree of substitutability is not high. After correction, the degree of intersection of the two ellipses increases, which means the degree of substitutability is improved.

[0029] PCA processing and discriminant function Matlab model method:

[0030]

[0031]

[0032] Example 2: Equivalence evaluation of multiple batches of roasted licorice root formula granules from Manufacturer B compared with traditional decoctions

[0033] Data on the quality of multiple batches of granulated decoction of prepared licorice root (DGD) and traditional decoction of prepared licorice root (TD) from Factory B were collected using HPLC and UPLC-MS / MS. Differences between DGD and TD were analyzed by dimensionality reduction and discriminant comparison of various indicators. First, the quality stability of DGD was determined for each batch of TD and DGD using the discriminant expression D. If D≥0, the next step was to perform PCA dimensionality reduction on multiple indicators to obtain two-dimensional principal components that explained up to 80% of the variance. The differences between the two were then visualized using MATLAB.

[0034] Calculations showed that the difference between batch-specific and batch-specific (DGD) values ​​for processed licorice from factory B was less than that from group TD, satisfying discriminant D. This indicates stable quality across batches. Data from factory B was then analyzed using dimensionality reduction and visualization. Figure 3 The two-dimensional PCA score plots of the TD and DGD samples of roasted licorice at a 95% confidence level show that, at this confidence level, the TD sample from factory B and the DGD sample are contained within each other in the two ellipses, indicating that they are equivalent. However, the principal component explanation of the variance is less than 80%, so they cannot be directly determined to be equivalent in the two-dimensional plane. After correction, the principal component explanation of the variance increases, and the two ellipses are contained within each other, so they can be considered equivalent.

[0035] PCA processing and discriminant function Matlab model method:

[0036]

[0037]

[0038] Example 3: Equivalence evaluation of multiple batches of prepared licorice root granule decoction from manufacturer C compared with traditional decoction.

[0039] Data on the quality of multiple batches of processed licorice root decoction (DGD) and traditional processed licorice root decoction (TD) from Factory C were collected using HPLC and UPLC-MS / MS. Differences between DGD and TD were analyzed using dimensionality reduction and discriminant analysis of various indicators. First, the quality stability of DGD was determined for each batch of TD and DGD using the discriminant expression D. If D≥0, the next step was to perform PCA dimensionality reduction on multiple indicators, obtaining two-dimensional principal components that explained up to 80% of the variance. The differences between the two were then visualized using MATLAB.

[0040] Calculations showed that the difference between the C group and the DGD group of roasted licorice was smaller than that of the TD group, satisfying discriminant D, indicating stable quality between batches. Dimensionality reduction and visualization analysis of the two groups of sample data were performed. The figure below shows the two-dimensional PCA score plots of the TD and DGD samples of roasted licorice at a 95% confidence level. The principal components explained over 80% of the variance. Figure 4 It can be seen intuitively that at this confidence level, the two ellipses of the TD sample and the DGD sample from Factory C are separated, indicating that TD and DGD are not interchangeable. After correction, the two ellipses intersect, meaning that the two can be partially interchangeable.

[0041] PCA processing and discriminant function Matlab model method:

[0042]

[0043]

[0044] Example 4: Equivalence evaluation of the formula granules of Amomum villosum from Manufacturer D compared with traditional decoctions

[0045] Quality data for multiple batches of Amomum villosum granules (DGD) and traditional Amomum villosum decoction (TD) from manufacturer D were collected using HPLC and UPLC-MS / MS methods. Differences between DGD and TD were analyzed using dimensionality reduction and discriminant analysis of various indicators. First, the quality stability of each batch of TD and multiple batches of DGD was determined using the discriminant expression D. If D ≥ 0, the next step was to perform PCA dimensionality reduction on multiple indicators to obtain two-dimensional principal components explaining up to 80% of the variance. The differences between the two were then visualized using MATLAB.

[0046] Calculations show that the DGD difference between different batches of Amomum villosum from Factory D is ( (n=3, average RSD=0.13) is greater than TD( (n=7), average RSD=0.07), does not satisfy discriminant D, multiple batches of DGD from factory D are identified as having large intra-group differences and poor quality stability, therefore, multi-index dimensionality reduction discriminant analysis will no longer be performed on them.

[0047] Example 5: Equivalence evaluation of Amomum villosum granule decoction from Manufacturer E compared with traditional decoction

[0048] Data on the quality of multiple batches of Decoction Granules (DGD) and Traditional Decoction (TD) of Amomum villosum from Factory E were collected using HPLC and UPLC-MS / MS. Differences between DGD and TD were analyzed by dimensionality reduction and discriminant comparison of various indicators. First, the quality stability of DGD was determined for each batch of TD and DGD using the discriminant expression D. If D≥0, the next step was to perform PCA dimensionality reduction on multiple indicators to obtain two-dimensional principal components explaining up to 80% of the variance. The differences between the two were then visualized using MATLAB.

[0049] Calculations showed that the difference between the DGD and TD groups of Amomum villosum from Factory E was smaller than that from the TD group, satisfying discriminant D, indicating stable quality between batches. Dimensionality reduction and visualization analysis of the two groups of sample data were performed. The figure below shows the two-dimensional PCA score plots of the TD and DGD Amomum villosum samples at a 95% confidence level. The principal components explained over 80% of the variance. Figure 5 It can be seen intuitively that at this confidence level, the data characterizing the quality of Factory E are formed into two separate ellipses through MATLAB. The TD sample and the DGD sample cannot be substituted. After correction by the correction coefficient (Figure b), the two groups of samples still cannot be cross-referenced. This may be because they contain more volatile components. There are many influencing factors between DGD and TD, and simple calibration cannot make them equivalent.

[0050] PCA processing and discriminant function Matlab model method:

[0051] clc;

[0052] clear;

[0053] X = [Original data matrix]

[0054] [Y,lbd,cos2,CC]=pca(X,2)

[0055] clc;

[0056] clear;

[0057] load 'pca dimensionality reduction results';

[0058] X1 = Y;

[0059] x1 = X1(TD group sample variables, 1)

[0060] y1 = X1(TD group sample variable, 2)

[0061] figure(4)

[0062] plot(x1,y1,'.'); grid off;

[0063] xlabel('PC1'); ylabel('PC2'); title(");

[0064] hold on

[0065] meanx1 = mean(X1(TD group sample variables, 1))

[0066] meany1 = mean(X1(TD group sample variables, 2))

[0067] stdx1 = std(X1(TD group sample variables, 1))

[0068] stdy1 = std(X1(TD group sample variables, 2))

[0069] k1 = atan(stdy1 / stdx1)

[0070] a1 = 1.96 * stdx1; 90% = 1.65; 68.3% = 1; 95% = 1.96 (This can form ellipses at different confidence levels)

[0071] b1 = 1.96 * stdy1;

[0072] r1=min(min(corrcoef(x1,y1)));

[0073] if r1>0

[0074] k1=k1

[0075] else

[0076] k1=-k1

[0077] end

[0078] tuoyuan(meanx1,meany1,a1,b1,k1)

[0079] x2 = X1 (DGD group sample variables, 1)

[0080] y2 = X1(DGD group sample variable, 2)

[0081] hold on

[0082] plot(x2,y2,'o'); grid off;

[0083] hold on

[0084] meanx2 = mean(X1(DGD group sample variables, 1))

[0085] meany2 = mean(X1(DGD group sample variables, 2))

[0086] stdx2 = std(X1(DGD group sample variables, 1))

[0087] stdy2 = std(X1(DGD group sample variables, 2))

[0088] k2 = atan(stdy2 / stdx2)

[0089] a2=1.96*stdx2;%90%=1.65 68.3%=1;95%=1.96

[0090] b2 = 1.96 * stdy2;

[0091] r2=min(min(corrcoef(x2,y2)));

[0092] if r2>0

[0093] k2=k2

[0094] else

[0095] k2=-k2

[0096] end

[0097] tuoyuan(meanx2,meany2,a2,b2,k2)

[0098] It should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any modifications or alterations made by those skilled in the art without departing from the scope of the present invention to create equivalent embodiments based on the disclosed technical content shall fall within the protection scope of the present invention.

[0099] In summary, this invention performs preliminary calculations on information from multiple batches of samples that meet various quality standards. Based on the TD (Total Determination) method, it determines the relative stability of the quality among multiple batches of DGD (Digital Grade Digestive) samples. After necessary dimensionality reduction of multiple index information, it generates a sample principal component analysis (PCA) score map that can explain a certain degree of variation at a certain confidence level. This allows for a direct determination of the equivalence relationship between the two. Based on the most holistic, important, and abstract integration of sample information, a function is constructed and presented through the PCA score map. This allows for a rapid, intuitive, and comprehensive presentation of the relationship between samples, providing both qualitative and quantitative descriptions. This is an innovation in the equivalence evaluation method between traditional Chinese medicine formula granules and decoctions, and it has good application and promotion value.

Claims

1. A method for evaluating the equivalence between traditional Chinese medicine formula granules and decoctions based on multi-index dimensionality reduction discriminant analysis, characterized in that, Includes the following steps: 1) Construct the initial discriminant D: Official 1 Where c T The coefficient of variation for each variable in the traditional decoction group. Let D be the coefficient of variation of each variable value in a certain formula granule decoction group, p be the number of variables in each type of sample, i take the values ​​1, 2, 3...n, n≤p. If D≥0, it means that the variation within the DGD sample group is less than that within the TD group, and the sample quality is relatively stable; if D<0, it means that the variation within the DGD sample group is greater than that within the TD group, indicating that the quality of the DGD sample group is unstable and will be judged as unqualified drug. If DGD is an unqualified drug, no further intuitive discrimination analysis is required. DGD refers to traditional Chinese medicine formula granule decoction, and TD refers to traditional decoction. 2) Construct confidence ellipses E1 and E2 for the distribution of traditional decoction sample group and formula granule decoction sample group in the dimension-reduced space respectively; determine equivalence based on the positional relationship between the two confidence ellipses: if E2 falls within E1 or the two ellipses intersect, it is determined that the two can be substituted; if the two ellipses are separate, it is determined that they cannot be substituted. 3) Construction of discrimination criteria: In the two-dimensional principal component plot obtained by PCA, each sample point is a two-dimensional random variable. If the explained variance reaches 80% or more, the constructed ellipse can intuitively represent the relationship between the two groups of samples. However, if the explained variance is less than 80%, a three-dimensional or higher dimension needs to be constructed to represent the relationship between the samples. In the two-dimensional plane, it is assumed that the points are represented by (x, y). At this time, x and y are orthogonal. The randomly distributed regions are constructed as any possible ellipse equations E1 and E2 in the two two-dimensional plane plots, where x and y are the coordinates of each sample point, A, B, and C are coefficients, and u, v, and w are constants. Official 2 Official 3 The above steps were visualized using MATLAB to intuitively evaluate the equivalence between DGD and TD.

2. The method for evaluating the equivalence between traditional Chinese medicine formula granules and decoctions based on multi-index dimensionality reduction discriminant analysis according to claim 1, characterized in that, In step 2), data on the quality of qualified DGD samples and TD samples are used. Based on the relevant knowledge of probability theory of two-dimensional random variables, an elliptical region for the occurrence of a two-dimensional random sample set is obtained under a certain confidence level. When the confidence levels α are 95.0%, 90.0%, and 68.3%, the abscissa or ordinate of the sample point appears in the regions μ±1.96σ, μ±1.65σ, and μ±σ, respectively. The joint probability density function of the two-dimensional continuous random variable is calculated, and a confidence interval is constructed accordingly. The reason it is an ellipse is that the probability density function of the two-dimensional normal distribution is a hat-shaped three-dimensional figure. The major axis and minor axis of the ellipse are the value ranges of the components of the two-dimensional variable on the two coordinate axes, respectively. The elliptical confidence interval is centered on the center of the ellipse, and its major axis is rotated by an angle β=arctg(b / a) in the direction of the longest median of the triangle, where b is the ordinate and a is the abscissa. Alternatively, the angle can be expressed as arctg(s). o / s t ) indicates that s o Let s be the standard deviation of the projection of all samples onto the ordinate. t is the standard deviation of the projection of all samples onto the x-axis.

3. The method for evaluating the equivalence between traditional Chinese medicine formula granules and decoctions based on multi-index dimensionality reduction discriminant analysis according to claim 1, characterized in that, In step 2), the two ellipses in the graph are separated and cannot be substituted. After necessary correction and adjustment of the granule equivalent ratio of the formula, the test and judgment are carried out. If the two ellipses satisfy the relationship of inclusion or intersection after correction, it is considered that the two can be substituted to different degrees after a certain correction. If they are still separated after correction, it is considered that the two cannot be substituted.

Images