PLS-SEM-based method for mining critical path of drug effect substance group and drug effect index group of traditional Chinese medicine

Abstract

The invention discloses a method for mining a key path of a traditional Chinese medicine efficacy substance group and a efficacy index group based on PLS-SEM, and the method comprises the steps: building a PLS-SEM model based on the experimental data of the efficacy substances and efficacy indexes of the traditional Chinese medicine, selecting a plurality of efficacy substances in the traditional Chinese medicine as independent variables, selecting a plurality of efficacy indexes as dependent variables, optimizing the variable grouping, and obtaining the key path of the efficacy substance group and the efficacy index group of the traditional Chinese medicine; and establishing a causal path model of reflection type reflection type forming or forming type reflection type forming. By adopting the method to discuss the pharmacodynamic data of different traditional Chinese medicine formulas, pharmacodynamic substance groups with similar efficacy, pharmacodynamic index groups with similar characterization and optimal paths of pharmacodynamic substances acting on pharmacodynamic indexes can be mined, and experiments show that the method is feasible and effective, supports are provided for clarification of traditional Chinese medicine principles, and references are provided for development of new drugs.

Description

Technical Field

[0001] The present invention belongs to the technical field of Chinese medicine formula efficacy mining, and in particular relates to a method for mining key paths of Chinese medicine efficacy substance groups and efficacy index groups based on PLS-SEM. Background Art

[0002] The chemical composition of traditional Chinese medicine is complex. Its single herbs have different composition combinations, its compound medicines have different compatible drugs, and some drug components are present in the body's serum. Therefore, traditional Chinese medicine has multiple components, multiple targets, multiple pathways, and multiple pharmacodynamic effects, which makes it unclear what the material components are that exert their efficacy, and makes it unclear what the relationship between different traditional Chinese medicine components and different pharmacodynamic indicators is, which brings difficulties to the rational use of traditional Chinese medicine in clinical practice.

[0003] Existing research usually uses spectrum-effect kurtosis correlation, network pharmacology or system pharmacology, correlation analysis, association analysis, partial least squares method, etc. to study the components related to efficacy as component groups. In the prior art, high performance liquid chromatography (HPLC) is used to analyze the relationship between the efficacy components and the efficacy index, which can reflect the characteristics of the comprehensive effect of multiple efficacy components on the efficacy index, and is suitable for the screening of efficacy-related components in complex systems of traditional Chinese medicine. However, this method requires the use of a more complex HPLC, and the results reflected by the peak area of ​​the chromatogram are not intuitive enough. Although the relevant traditional Chinese medicine efficacy research methods have made extraordinary achievements in recent years, they still fail to well reflect the overall efficacy, overall regulation and effective intervention of traditional Chinese medicine on diseases. The methodology and methodology of traditional Chinese medicine efficacy research are still bottleneck problems in traditional Chinese medicine pharmacodynamics.

[0004] The research and development team of inventor Nie Bin used the orthogonal partial least squares method (PLS) to study the relationship between individual active ingredients and efficacy indicators of traditional Chinese medicine, and found that there is a complex relationship between them. However, they only established PLS regression equations for individual active ingredients, and did not group all active ingredients for research, resulting in a scattered relationship between individual active ingredients and efficacy indicators in the research results, which is not systematic and targeted.

[0005] Partial Least Squares Structural Equation Modeling (PLS-SEM) is a powerful second-generation statistical analysis method that uses partial least squares (PLS) technology to define a variable explanatory structure with a linear combination of variables, and then uses the regression principle to test the prediction and explanatory relationship between principal components. This method has advantages that other methods do not have: ① It can overcome the influence of small samples, non-normal distribution, multivariate collinearity, and measurement errors, and can ensure convergence, making statistical estimates more accurate; ② It is suitable for the calculation and interpretation of complex relationship structure models such as multiple independent variables, multiple dependent variables, interactions between variables, direct and indirect path patterns, etc. It can be seen that the advantages of PLS-SEM are more consistent with the analysis of Chinese medicine efficacy data. Summary of the invention

[0006] The purpose of the present invention is to provide a method for mining the key paths of the effective substance group and the effective index group of traditional Chinese medicine based on PLS-SEM, and constructing the PLS-SEM model based on the experimental data of the effective substance and the effective index of the traditional Chinese medicine formula. The method of the present invention is used to explore the effective data of different traditional Chinese medicine formulas, and can mine the effective substance group with similar efficacy, the effective index group with similar characteristics, and the path of the effective substance acting on the effective index. If the P value of each parameter is significant, it is considered to be a key path. Experimental verification shows that the method is feasible and effective, which provides support for the clarification of the principle of traditional Chinese medicine and provides a reference for the development of new drugs.

[0007] The technical solution of the present invention is as follows:

[0008] A method for mining the key paths of effective substance groups and effective index groups of traditional Chinese medicine based on PLS-SEM, constructing a PLS-SEM model based on the experimental data of effective substances and effective indexes of traditional Chinese medicine, selecting several effective substances in traditional Chinese medicine as independent variables, and several effective indexes as dependent variables, optimizing the variable grouping, and establishing a causal path model of reflection-reflection-formation or formative-reflection-formation.

[0009] The causal path model includes an initial causal path model, a grouped causal path model (according to actual needs, it can also include an optimized grouped causal path model, a re-optimized grouped causal path model, etc.), which are all composed of corresponding external models and internal models. The external model includes an independent variable external model and a dependent variable external model. The independent variable external model uses the corresponding pharmacological substance as the measured independent variable and the corresponding pharmacological substance group as the exogenous latent variable. The dependent variable external model uses the corresponding pharmacological index as the measured dependent variable and the corresponding pharmacological index group as the endogenous latent variable. The internal model is composed of the corresponding pharmacological substance group and the corresponding pharmacological index group to form a causal path.

[0010] Preferably, the pharmacological substance is derived from a serum drug component of a traditional Chinese medicine compound, a compatible drug of a traditional Chinese medicine compound, or a compatible drug of a traditional Chinese medicine component; the external mode is a reactive type or a formative type, and the internal mode is a reactive type.

[0011] A method for mining the key paths of traditional Chinese medicine efficacy substance groups and efficacy index groups based on PLS-SEM specifically comprises the following steps:

[0012] S1. Establish an initial causal path model, wherein the initial causal path model includes an external model 1 and an internal model 1. The external model 1 includes an independent variable external model 1 and a dependent variable external model 1. The independent variable external model 1 uses all pharmacodynamic substances as measured independent variables and an initial pharmacodynamic substance group as an exogenous latent variable. The dependent variable external model 1 uses all pharmacodynamic indicators as measured dependent variables and an initial pharmacodynamic indicator group as an endogenous latent variable. The internal model 1 is composed of the initial pharmacodynamic substance group and the initial pharmacodynamic indicator group to form a one-cause-one-effect path. The PLS-SEM model is used for calculation to select a better path and optimize the variable grouping.

[0013] S2. On the basis of step S1, a grouped causal path model is established, wherein the causal path model includes an external model 2 and an internal model 2. The external model 2 includes an independent variable external model 2 and a dependent variable external model 2. The independent variable external model 2 uses each group of pharmacological substances optimized in step S1 as a measured independent variable and the grouped pharmacological substance group as an exogenous latent variable. The dependent variable external model 2 uses each group of pharmacodynamic indicators optimized in step S1 as a measured dependent variable and the grouped pharmacological indicator group as an endogenous latent variable. The internal model 2 is composed of the grouped pharmacological substance group and the grouped pharmacological indicator group to form a causal path. The PLS-SEM model is used to calculate and select a better path optimization variable grouping.

[0014] S3, based on step S2, optimization is performed to establish a causal path model for optimized grouping. The method is the same as the causal path model for grouping in step S2. Through PLS-SEM model calculation, a better path is selected to optimize the variable grouping until it cannot be optimized, or it is stopped according to actual needs;

[0015] S4. Arrange the results to obtain groups of pharmacological substances with similar efficacy, groups of pharmacological efficacy indicators and their key paths.

[0016] Preferably, the variable optimization criteria of the external model include but are not limited to any one or more of the following:

[0017] (1) Check the reliability of each measurement variable: the absolute value of the factor loading (indicating the size of the correlation) should be above moderate or strong (factor loading ≥ 0.70 is strong, 0.40 ≤ factor loading < 0.70 is moderate, and 0 ≤ factor loading < 0.40 is weak), that is, the absolute value of the factor loading should be large, and the critical value of the large factor loading should be 0.7 or 0.4. Different signs should be separated, and the same sign should be combined into a group, and the significance of the P value and T value of the factor loading should be checked;

[0018] (2) Check the combined reliability of each latent variable: When CR (the standard of exploratory research, representing internal consistency) is positive (indicating good consistency between variables), select variables with CR ≥ 0.60 (CR ≥ 0.70 is strong, 0.60 ≤ CR < 0.70 is acceptable) and merge them into a group; for variables with negative CR (indicating possible negative correlation between variables), recheck their factor loadings and adjust the correlation;

[0019] (3) Check convergent validity AVE, AVE ≥ 0.5;

[0020] (4) Check the correlation, Cronbach's alpha coefficient is positive;

[0021] (5) Check the discriminant validity and use the Fomell-Larcker criterion or cross-loading to distinguish and group variables. The Fomell-Larcker criterion is: the square root of the AVE value of each dimension must be greater than its correlation coefficient with other dimensions (latent variables); the cross-loading value is: the highest factor loading value of each variable should be in the dimension to be measured, and this value will become smaller in other dimensions.

[0022] When the external mode is formative, the following additional criteria should be considered:

[0023] (6) Multicollinearity: VIF < 5;

[0024] (7) External model indicator weight;

[0025] (8) The significance of the P-value and T-value of the external model indicator weight.

[0026] Preferably, in step S1, the variable optimization criteria of the external model 1 include: checking the reliability of each variable: the absolute value of the factor loading, the larger the value, the larger the critical value is 0.7 or 0.4, different signs are separated, and the same sign is combined into a group.

[0027] Preferably, the variable optimization criteria of the internal model include but are not limited to any one or more of the following:

[0028] ① Check the significance of the P value and T value of the path coefficient (whose size indicates the strength of the relationship between two latent variables);

[0029] ② Check the explanatory power of each variable R 2 , the value is selected as moderate or above (R 2 ≥0.670 is strong, 0.333≤R 2 <0.670 is moderate, 0.190≤R 2 <0.333 is weak);

[0030] ③ Prediction ability and cross-validation verification: Q 2 >0(Q 2 That is, Q2predict, Q 2 ≥0.35 is strong, 0.15≤Q 2 <0.35 is moderate, 0.02≤Q 2 <0.15 is weak).

[0031] Preferably, the medicinal substance is derived from serum drug components of a traditional Chinese medicine compound, the traditional Chinese medicine compound is Dachengqi Decoction, the Dachengqi Decoction includes the original Dachengqi Decoction formula and a uniformly designed Dachengqi Decoction combination formula, and the experimental data is experimental data of Dachengqi Decoction in treating acute pancreatitis.

[0032] The serum drug components include rhein, rhein, chrysophanol, aloe-rhein, rhein methyl ether, magnolol, honokiol, hesperidin, and hesperetin. The efficacy indicators include calcium ions, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate.

[0033] The pharmacodynamic substance group obtained in step S4 is pharmacodynamic substance group 1 or pharmacodynamic substance group 2, the pharmacodynamic substance group 1 includes magnolol and hesperidin, and the pharmacodynamic substance group 2 includes magnolol, honokiol and hesperidin; the obtained pharmacodynamic index group includes pharmacodynamic index group 1 and pharmacodynamic index group 2, the pharmacodynamic index group 1 includes amylase, interleukin-6, lymphocyte function-associated antigen 1a, and pancreatic lipase, and the pharmacodynamic index group 2 includes interleukin-10 and calcium ions.

[0034] The obtained key path includes: the pharmacodynamic substance group 1 or the pharmacodynamic substance group 2 reduces the index value of the pharmacodynamic index group 1 and increases the index value of the pharmacodynamic index group 2.

[0035] Preferably, the medicinal substance is derived from the serum drug component of a traditional Chinese medicine compound, the traditional Chinese medicine compound is Maxing Shigan Decoction, the Maxing Shigan Decoction includes the original Maxing Shigan Decoction formula and a uniformly designed Maxing Shigan Decoction combination formula, and the experimental data is the experimental data of Maxing Shigan Decoction for reducing fever.

[0036] The serum drug components include ephedrine, pseudoephedrine, methylephedrine, amygdalin, prunin, liquiritin, liquiritigenin, and glycyrrhetinic acid, and the efficacy indicators include prostaglandin E2, body temperature response index, and 6h fever inhibition rate.

[0037] The pharmacodynamic substance group I obtained in step S4 includes ephedrine, methylephedrine, and pseudoephedrine, and the obtained pharmacodynamic index group includes pharmacodynamic index group I and pharmacodynamic index group II, wherein the pharmacodynamic index group I includes the body temperature response index, and the pharmacodynamic index group II includes the 6h fever inhibition rate.

[0038] The obtained key path includes: the pharmacological substance group I increases the index value of the pharmacological index group I, and reduces the index value of the pharmacological index group II.

[0039] Preferably, the medicinal substance is derived from a compound drug of a traditional Chinese medicine, the traditional Chinese medicine formula is Dachengqi Decoction, the Dachengqi Decoction includes the original Dachengqi Decoction formula and a uniformly designed Dachengqi Decoction compound formula, and the experimental data is experimental data of Dachengqi Decoction in treating acute pancreatitis.

[0040] The compatible drugs include rhubarb, magnolia bark, bitter orange, and sodium sulfate, and the efficacy indicators include calcium ions, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate.

[0041] The obtained effective substance group A in step S4 includes Magnolia officinalis, and the obtained effective index group includes effective index group A and effective index group B, wherein the effective index group A includes amylase and pancreatic lipase, and the effective index group B includes interleukin-10, calcium ions, and survival rate.

[0042] The obtained key path includes: the pharmacological substance group A reduces the index value of the pharmacological index group A and increases the index value of the pharmacological index group B.

[0043] Preferably, the medicinal substance is derived from a Chinese medicine component compatibility drug, the Chinese medicine component compatibility drug is a Polygonum multiflorum component compatibility drug, and the experimental data is experimental data of the Polygonum multiflorum component compatibility drug against hypoxia / reoxygenation.

[0044] The drugs combined with the polygonum multiflorum component include orientin, quercetin and vitexin, and the drug efficacy indicators include cell viability, lactate dehydrogenase leakage rate and nitric oxide level.

[0045] The pharmacodynamic substance group a obtained in step S4 includes quercetin, and the pharmacodynamic index group a obtained includes the leakage rate of lactate dehydrogenase and the nitric oxide level.

[0046] The obtained key path includes: the pharmacological substance group a increases the index value of the pharmacological index group a.

[0047] Beneficial effects of the present invention:

[0048] 1. The method of the present invention is to establish a relationship model between pharmacological substances and pharmacodynamic indicators, and to obtain pharmacological substance groups and pharmacodynamic indicator groups with similar efficacy (increasing or decreasing) and the action pathways therebetween from numerous component pharmacodynamic indicators and numerous complex effects through PLS-SEM quantitative calculation and selective iteration, so as to better reflect the characteristics of component action pharmacodynamic indicators and their action pathways.

[0049] 2. The present invention proposes a group of pharmacodynamic substances with similar efficacy (increased or decreased), a group of pharmacodynamic indexes, and related concepts, and the concept of a critical path from a pharmacodynamic substance group to a pharmacodynamic index group. These concepts are realized by screening parameters such as factor loading and Cronbach's coefficient. Experimental verification shows that this method is feasible and effective, which provides support for the clarification of the principles of traditional Chinese medicine and a reference for the development of new drugs.

[0050] 3. Through the verification results of the embodiments, it can be seen that the present invention establishes the key paths of the pharmacodynamic substance groups and pharmacodynamic index groups of the Chinese medicine compound serum drug components, the Chinese medicine compound compatible drugs and the Chinese medicine component compatible drugs, respectively, which has strong feasibility, especially the verification effect of the serum drug components of Dachengqi Decoction is better. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the PLS-SEM model of the reactive type, reactive type and formative type of the present invention;

[0052] Figure 2 It is a schematic diagram of the PLS-SEM model of the formation type, reflection type and formation type of the present invention;

[0053] Figure 3 This is a schematic diagram of factor loading, path coefficient and explanatory power in the first optimization grouping path model of Example 2 of the present invention;

[0054] Figure 4 This is the significant result of Bootstrapping in the first optimization grouping model of Example 2 of the present invention;

[0055] Figure 5 It is a schematic diagram of factor loading, path coefficient and explanatory power in the initial path model of Example 3 of the present invention;

[0056] Figure 6 This is the significance result of Bootstrapping in the initial path model of Example 3 of the present invention;

[0057] Figure 7 It is a schematic diagram of factor loading, path coefficient and explanatory power in the initial path model of Example 4 of the present invention;

[0058] Figure 8It is a schematic diagram of factor loading, path coefficient and explanatory power in the initial path model of Example 5 of the present invention;

[0059] Fig. 9 This is the significance result of Bootstrapping in the initial path model of Example 5 of the present invention.

[0060] in, Figure 1 Middle: The rectangular box represents the measured variable, the circular box represents the latent variable, χ represents the independent variable, y represents the dependent variable, X represents the exogenous latent variable explained by the independent variable χ, Y represents the endogenous latent variable explained by the dependent variable y, λ represents the factor loading, and γ represents the path coefficient);

[0061] Figure 2 Chinese: Each symbol refers to the meaning and Figure 1 same;

[0062] Figure 3 , 5 , 7, and 8: Except for the numbers in the circular boxes of the efficacy index group, which represent the explanatory power of the efficacy index group, the meanings of the remaining numbers, rectangular boxes, and circular boxes are the same as those in Figure 1 same;

[0063] Figure 4 , 6 , 9: All numbers represent significant P values, and the meanings of the other rectangular and circular boxes are the same as Figure 1 same. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Unless otherwise specified, the reagents involved in the embodiments of the present invention are all commercially available products and can be purchased through commercial channels.

[0066] 1. Explanation of some proper nouns in this invention:

[0067] 1. Pharmacodynamic substance group: refers to a group of pharmacological ingredients (pharmacological substances) with the same or similar effects on pharmacodynamic indicators. In the present invention, it mainly refers to a group of pharmacological substances with similar functions and indications.

[0068] 2. Pharmacological efficacy index group: refers to a collection of a group of medicinal efficacy indicators that reflect the same or similar medicinal efficacy (functions and indications).

[0069] 3. Critical path: The path by which a group of pharmacological substances acts on a group of pharmacological indicators, and the path is significant in statistical tests.

[0070] 4. Reflective model: the latent variable reflects the measured variable, the latent variable is the cause, and the measured variable is the effect. The latent variable is believed to produce the observed variable (for example, the personality trait of the latent variable and the attitude of the observed variable).

[0071] 5. Formative model: explores the latent variables formed by causes, with measured variables as causes and latent variables as effects. Formative indicators produce obvious variables of unobserved theoretical structures (e.g., social status structure generated by occupation, income, and place of residence).

[0072] 2. Software and Algorithm Settings

[0073] The PLS-SEM calculation of the present invention uses SmartPLS software, and some of its parameter settings are shown in Table 1 (reference "Ringle, CM, Wende, S., and Becker, J.-M. 2022.", software type "SmartPLS 4." Oststeinbek: SmartPLS GmbH, http: / / www.smartpls.com.).

[0074] Table 1 PLS-SEM algorithm and Bootstrapping parameter settings

[0075]

[0076]

[0077] 3. Mining the key paths of serum pharmacodynamic substance groups and pharmacodynamic index groups of traditional Chinese medicine compound based on PLS-SEM Example 1 Dachengqi decoction for the treatment of acute pancreatitis (the optimization criterion of factor loading in grouping model and optimization grouping model is strong ≥ 0.70)

[0078] The PLS-SEM model was constructed based on the experimental data of serum drug components in Dachengqi Decoction for the treatment of acute pancreatitis. Several serum drug components in Dachengqi Decoction were selected as independent variables, and several efficacy indicators were selected as dependent variables. The variable grouping was optimized to establish a reflective-type, reflective-type, and formative-type path model. The model diagram is shown in Figure 1 .

[0079] Experimental data on the treatment of acute pancreatitis with serum drug components of Dachengqi Decoction:

[0080] The experimental data acquisition method is as follows: Based on the experimental data of Dachengqi decoction compatibility drugs in the existing literature 1 (Qiandan, Yu Riyue, Yu Xiaojuan, Wan Panting, Liu Xinhui, Nie Bin, Chen Yinfang, Peng Hong, Xu Guoliang. Experimental study on the treatment of acute pancreatitis with Dachengqi decoction compatibility [J]. Liaoning Journal of Traditional Chinese Medicine, 2017, 44(1): 190-192.), the animals were given medication (Dachengqi decoction decoction) in the animal experiment and then measured according to the measurement method in literature 2 (Shen Fengyun, Wei Huizhen, Sun Yongbing, Wang Yuesheng, Lv Shang, Gao Meng, Zeng Lianqing, Rao Yi. Uplc-Ms / Ms simultaneous determination of the contents of 9 active ingredients in rat plasma of Dachengqi decoction [J]. China Journal of Traditional Chinese Medicine, 2014, 39(12): 2345-2350.) The Dachengqi decoction serum drug components of the animals were matched with the efficacy index data in literature 1 to obtain Tables 2 and 3.

[0081] Serum drug ingredients: Emodin, Rhein, Chrysophanol, Aloe Emodin (AE), Emodin methyl ether (EME), Magnolol, Honokiol, hesperidin, hesperetin.

[0082] Efficacy index: calcium ion (Ca 2+ ), pancreatic lipase (PL), interleukin-6 (IL6), interleukin-10 (IL10), lymphocyte function associated antigen 1a (LFA1a), amylase (AMY), organ indicators (OI), survival rate (SR).

[0083] Table 2 Ratio of serum drug components of Dachengqi Decoction (dose, g / kg)

[0084]

[0085]

[0086] Table 3 Results of Dachengqi Decoction serum drug components in the treatment of acute pancreatitis

[0087]

[0088] 1. Establish an initial causal path model

[0089] (1) Model construction:

[0090] All independent variables and dependent variables are involved in modeling.

[0091] External mode: Reflective

[0092] Independent variable out-of-control model: Rhein, rhein, chrysophanol, aloe-rhein, physophanol methyl ether, magnolol, honokiol, hesperidin, and hesperetin were used as measured independent variables, and the initial pharmacodynamic substance group was used as a latent variable.

[0093] Dependent variable extra-model: calcium ion, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate were used as measured dependent variables, and the initial efficacy index group was used as a latent variable.

[0094] Internal mode: formative

[0095] The initial group of pharmacodynamic substances is the cause, and the initial group of pharmacodynamic indicators is the effect, forming a cause-effect path.

[0096] (2) Calculation results

[0097] The model and PLS-SEM calculation results are shown in Table 4-7.

[0098] Table 4 Reliability and convergent validity of the initial path model of Example 1

[0099]

[0100] Table 5 Fomell-Larcke criterion of the initial path model of Example 1

[0101] Initial efficacy index group Initial active substance group Initial efficacy index group 0.735 Initial active substance group -0.805 0.498

[0102] Table 6 Path coefficients / total effects of the initial path model of Example 1

[0103]

[0104] Table 7 Explanatory power R of the initial path model of Example 1 2

[0105] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Initial efficacy index group 0.648 0.604

[0106] (3) The results are described and analyzed as follows:

[0107] ① Reliability of measurement variables: The value of factor loading can be positive or negative, large or small. It indicates positive or negative correlation, and the size of the correlation. The T value is less than 1.96, which is not significant. The P value is greater than 0.05, which is not significant.

[0108] ②Combined reliability of latent variables: CR value is greater than 0.8, indicating that the facet has good internal consistency.

[0109] ③ Convergent validity: If the AVE value corresponding to the efficacy index is greater than 0.5, there is convergent validity; if the AVE value corresponding to the efficacy substance is less than 0.5, there is no convergent validity.

[0110] ④Correlation: Cronbach's coefficient item, the efficacy index is negative, indicating that the indicators may be negatively correlated.

[0111] ⑤ Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE. It is not greater than the correlation coefficient value below, which also indicates that there is no discriminant validity.

[0112] ⑥Path coefficient: The path coefficient is not significant.

[0113] ⑦Explanatory ability: R 2 better.

[0114] ⑧ Prediction ability and cross-validation verification: In the structural equation, the sample size is smaller than the number of variables, so 10-fold cross-validation is not possible, and there is no Q 2 value.

[0115] (4) Conclusion

[0116] Select the independent variable and the dependent variable, and group them. Examine the independent variable and the dependent variable, select the largest factor loading, separate the ones with different signs, and combine the ones with the same sign; separate the Cronbach's coefficients of the dependent variable with different signs, and combine the ones with the same sign. Get:

[0117] Pharmacodynamic substance group 0′: aloe-emodin, rhein, rhein, chrysophanol, physophanol methyl ether, hesperidin;

[0118] Pharmacodynamic substance group 1′: magnolol, honokiol, hesperidin;

[0119] Efficacy index group 1′: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6;

[0120] Efficacy index group 2′: interleukin-10, calcium ions, and survival rate.

[0121] 2. Establish a grouped causal path model

[0122] (1) Model construction:

[0123] External mode: Reflective

[0124] Independent variable out-of-control mode (reflective type): aloe-emodin, emodin, rhein, chrysophanol, physophanol methyl ether, and hesperidin are used as measured independent variables, and the active substance group 0″ is used as the latent variable. Magnolol, honokiol, and hesperidin are used as measured independent variables, and the active substance group 1″ is used as the latent variable.

[0125] External dependent variable model (reflective type): Amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, and interleukin-6 are used as measured dependent variables, and the efficacy index group 1″ is used as a latent variable. Interleukin-10, calcium ions, and survival rate are used as measured dependent variables, and the efficacy index group 2″ is used as a latent variable.

[0126] Internal mode: formative

[0127] With the medicinal effect substance group 0″ and the medicinal effect substance group 1″ as causes, and the medicinal effect index group 1″ and the medicinal effect index group 2″ as effects, a two-cause and two-effect path is formed crosswise.

[0128] (2) Calculation results

[0129] The model and PLS-SEM calculation results are shown in Tables 8-11.

[0130] Table 8 Reliability and convergent validity of the group path model of Example 1

[0131]

[0132] Table 9 Fomell-Larcke criterion of the group path model of Example 1

[0133] Efficacy index group 1" Efficacy index group 2" Pharmacological substance group 0" Medicinal substance group 1" Efficacy index group 1" 0.802 Efficacy index group 2" -0.734 0.861 Pharmacological substance group 0" 0.066 -0.399 0.884 Medicinal substance group 1" -0.746 0.744 -0.224 0.824

[0134] Table 10 Path coefficients / total effects of the grouped path model of Example 1

[0135] path Original sample Sample mean Standard Deviation T-value P-value Pharmacological effect substance group 0" → Pharmacological effect index group 1" -0.107 -0.014 0.358 0.298 0.766 Pharmacological effect substance group 0″→Pharmacological effect index group 2″ -0.245 -0.193 0.330 0.743 0.458 Pharmacological effect substance group 1″ → Pharmacological effect index group 1″ -0.770 -0.727 0.301 2.558 0.011 Pharmacological effect substance group 1" → Pharmacological effect index group 2" 0.689 0.690 0.330 2.087 0.038

[0136] Table 11 Explanatory power R of the group path model of Example 1 2

[0137] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy index group 1" 0.567 0.443 Efficacy index group 2" 0.611 0.499

[0138] (3) Variable optimization criteria

[0139] The variable optimization criteria for the external model include:

[0140] ① Check the reliability of each measurement variable: the absolute value of the factor loading (indicating the size of the correlation), the value is selected as strong (factor loading ≥ 0.70 is strong, 0.40 ≤ factor loading < 0.70 is moderate, 0 ≤ factor loading < 0.40 is weak), that is, the absolute value of the factor loading is selected as large, the critical value of the large is 0.7, different signs are separated, the same sign is combined into a group, and the significance of the P value and T value of the factor loading is checked;

[0141] ② Check the combined reliability of each latent variable: When CR (the standard of exploratory research, representing internal consistency) is positive (indicating good consistency between variables), select variables with CR ≥ 0.60 (CR ≥ 0.70 is strong, 0.60 ≤ CR < 0.70 is acceptable) and combine them into a group. For variables with negative CR (indicating possible negative correlation between variables), recheck their factor loadings and adjust the correlation.

[0142] ③ Check convergent validity AVE, AVE ≥ 0.5;

[0143] ④ Check the correlation, Cronbach's alpha coefficient is positive;

[0144] ⑤ Check the discriminant validity, use the Fomell-Larcke criterion or cross-loading amount for discrimination and variable grouping. The Fomell-Larcker criterion is: the square root of the AVE value of each dimension must be greater than its correlation coefficient with other dimensions (latent variables); the cross-loading amount is: the highest factor loading amount of each variable should be in the dimension to be measured, and this value will become smaller in other dimensions.

[0145] The variable optimization criteria of the internal model include:

[0146] ① Check the significance of the P value and T value of the path coefficient (whose size indicates the strength of the relationship between two latent variables);

[0147] ② Check the explanatory power of each variable R 2 , the value is selected as moderate or above (R 2 ≥0.670 is strong, 0.333≤R 2 <0.670 is moderate, 0.190≤R 2 <0.333 is weak), merged into a group;

[0148] ③ Prediction ability and cross-validation verification: Q 2 >0(Q 2 That is, Q2predict, Q 2 ≥0.35 is strong, 0.15≤Q 2 <0.35 is moderate, 0.02≤Q 2 <0.15 is weak).

[0149] (4) The main results are described and analyzed as follows:

[0150] ① Reliability of measurement variables: The value of factor loading, positive and negative, has been separated. Honokiol is 0.543, less than 0.7; T value is greater than 1.96, which is significant.

[0151] ②Combined reliability of latent variables: CR value is greater than 0.8, indicating that the facet has good internal consistency.

[0152] ③ Convergent validity: The AVE values ​​corresponding to efficacy index group 1″ and efficacy index group 2″ are greater than 0.5, which shows convergent validity; the AVE values ​​corresponding to efficacy substance group 0″ and efficacy substance group 1″ are greater than 0.5, which shows convergent validity.

[0153] ④Correlation: All Cronbach's coefficients are positive.

[0154] ⑤Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE, which is greater than the correlation coefficient value below, also indicating that the discriminant validity is very good.

[0155] ⑥Path coefficient: There are two paths whose path coefficients are not significant.

[0156] ⑦Explanatory ability R 2 : Moderate.

[0157] ⑧Prediction ability and cross-validation verification: Since the number of samples is smaller than the number of variables, there is no such result.

[0158] The above results show that the test path coefficient is significant and it is found that:

[0159] The components of Magnolia officinalis and Citrus aurantium showed a decreasing effect of active substance group 1″→active index group 1″, and an increasing effect of active substance group 1″→active index group 2″, which was significant.

[0160] The P values ​​of effective substance group 0″→effectiveness index group 1″ and effective substance group 0″→effectiveness index group 2″ were 0.766 and 0.458, respectively, which were not significant. However, other parameters showed that the rhubarb component of effective substance group 0″ had a reducing effect on the efficacy index values ​​of the two groups.

[0161] (5) Conclusion

[0162] Select the independent variables and dependent variables, group them. Separate the efficacy indicators of positive correlation and negative correlation. Keep the paths with high reliability and validity and significant path coefficients. Get:

[0163] Pharmacodynamic substance group 1″: magnolol, hesperidin;

[0164] Efficacy index group 1″: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6;

[0165] Efficacy index group 2″: interleukin-10, calcium ions, and survival rate.

[0166] 3. Optimizing the Causal Path Model for Grouping

[0167] (1) Model construction:

[0168] External mode: Reflective

[0169] Independent variable-exclusive model (reflective type): magnolol and hesperidin are used as measured independent variables, and active substance group 1 is used as a latent variable.

[0170] Dependent variable external mode (reflective type): amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, and interleukin-6 are used as measured dependent variables, and drug efficacy index group 1 is used as a latent variable. Interleukin-10, calcium ions, and survival rate are used as measured dependent variables, and drug efficacy index group 2 is used as a latent variable.

[0171] Internal mode: formative

[0172] Taking the drug efficacy substance group 1 as the cause, and the drug efficacy index group 1 and the drug efficacy index group 2 as the effects, the intersection constitutes a causal path.

[0173] (2) Calculation results

[0174] The model and PLS-SEM calculation results are shown in Tables 12-16.

[0175] Table 12 Reliability and convergent validity of the optimized grouping path model in Example 1

[0176]

[0177]

[0178] Table 13 Example 1 Optimizing the Fomell-Larcke Criterion of the Grouping Path Model

[0179] Efficacy index group 1 Efficacy index group 2 Pharmacological substance group 1 Efficacy index group 1 0.804 Efficacy index group 2 -0.742 0.862 Pharmacological substance group 1 -0.700 0.756 0.961

[0180] Table 14 Path coefficients / total effects of the optimized grouping path model in Example 1

[0181] path Original sample Sample mean Standard Deviation T-value P-value Pharmacological effect substance group 1 → Pharmacological effect index group 1 -0.700 -0.726 0.311 2.255 0.024 Pharmacological effect substance group 1 → Pharmacological effect index group 2 0.756 0.706 0.392 1.927 0.054

[0182] Table 15 Explanatory power R of the optimized grouping path model in Example 1 2

[0183] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 <!-- 12 -->]]> Efficacy index group 1 0.490 0.426 Efficacy index group 2 0.571 0.517

[0184] Table 16 Prediction ability and cross-validation verification of optimized grouping path model in Example 1

[0185]

[0186] Note: RMSE stands for root mean square error, MAE stands for mean absolute error, PLS-SEM_RMSE stands for the root mean square error of the PLS-SEM test, PLS-SEM_MAE stands for the mean absolute error of the PLS-SEM test, LM_RMSE stands for the root mean square error of the linear regression test, and LM_MAE stands for the mean absolute error of the linear regression test.

[0187] (3) Variable optimization criteria

[0188] Refer to the causal path model construction process of the aforementioned grouping.

[0189] (4) The main results are described and analyzed as follows:

[0190] ① Reliability of measurement variables: factor loadings are all greater than 0.70, T values ​​are all greater than 1.96, and P values ​​are all significant.

[0191] ②Composition reliability of latent variables: CR is greater than 0.70

[0192] ③ Convergent validity: AVE is greater than 0.50. The effect is very good.

[0193] ④Correlation: Cronbach's coefficients are all positive.

[0194] ⑤Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE, which is greater than the correlation coefficient value below, indicating good discriminant validity.

[0195] ⑥Path coefficient: One path coefficient is close to 0.05, and the other is significant.

[0196] ⑦Interpretation ability: Moderate.

[0197] ⑧ Prediction ability and cross-validation verification: For Q2, the measurement mode PL prediction is less than 0, and other indicators are greater than 0, which is acceptable. For Q2, all indicators are greater than 0, which is acceptable. PLS-SEM performs better than linear (LM) regression.

[0198] The above results show that the overall efficacy index is very good, indicating that Magnolol and Hesperidin are the most useful pharmacological substances.

[0199] (5) Conclusion

[0200] Pharmacodynamic substance group 1: magnolol, hesperidin;

[0201] Efficacy index group 1: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6;

[0202] Efficacy index group 2: interleukin-10, calcium ions.

[0203] Critical path: Pharmacodynamic substance group 1 reduces the index value of pharmacological index group 1 and increases the index value of pharmacological index group 2.

[0204] Example 2 Dachengqi Decoction for the treatment of acute pancreatitis (grouping model, optimization criteria for factor loading when optimizing grouping model is moderate or above, factor loading ≥ 0.4)

[0205] The difference between this embodiment and Embodiment 1 is that the optimization criterion of the factor loading in the grouping model is moderate or above, that is, the factor loading ≥ 0.4, and the rest remains unchanged.

[0206] 1. Establish an initial causal path model

[0207] The process and results are the same as those in Example 1.

[0208] 2. Establish a grouped causal path model

[0209] (1) Model construction: Same as Example 1.

[0210] (2) Calculation results: Same as in Example 1.

[0211] (3) Variable optimization criteria

[0212] In the variable optimization criteria of the external model 2, the absolute value of the factor loading is selected to have a maximum critical value of 0.4, and the rest is the same as in Example 1.

[0213] (4) Conclusion

[0214] Select the independent variables and dependent variables, group them. Separate the efficacy indicators of positive correlation and negative correlation. Keep the paths with high reliability and validity and significant path coefficients. Get:

[0215] Pharmacodynamic substance group 2″: magnolol, honokiol, hesperidin;

[0216] Efficacy index group 1″: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6;

[0217] Efficacy index group 2″: interleukin-10, calcium ions, and survival rate.

[0218] 3. Causal Path Model for Optimized Grouping

[0219] (I) First Modeling

[0220] (1) Model construction:

[0221] External mode: Reflective

[0222] Independent variable-exclusive model (reflective type): magnolol, honokiol, and hesperidin are used as measured independent variables, and the active substance group 2″′ is used as a latent variable.

[0223] External dependent variable mode (reflective type): amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, and interleukin-6 are used as measured dependent variables, and the efficacy index group 1″′ is used as a latent variable. Interleukin-10, calcium ions, and survival rate are used as measured dependent variables, and the efficacy index group 2″′ is used as a latent variable.

[0224] Internal mode: formative

[0225] The medicinal substance group 2″′ is the cause, and the medicinal index group 1″′ and the medicinal index group 2″′ are the effects, which cross-constitute a causal path.

[0226] (2) Calculation results

[0227] The model and PLS-SEM calculation results are shown in Figure 3 , 4 ,Depend on Figure 4 It can be seen that the T value and P value of the efficacy indicator "survival rate" were not obtained, so the "survival rate" was removed and the model was rebuilt.

[0228] (II) Second modeling

[0229] (1) Model construction:

[0230] External mode: Reflective

[0231] Independent variable-exclusive model (reflective type): magnolol, honokiol, and hesperidin were used as measured independent variables, and the active substance group 2 was used as a latent variable.

[0232] Dependent variable external mode (reflective type): amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, and interleukin-6 are used as measured dependent variables, and drug efficacy index group 1 is used as a latent variable. Interleukin-10 and calcium ion are used as measured dependent variables, and drug efficacy index group 2 is used as a latent variable.

[0233] Internal mode: formative

[0234] With the pharmacodynamic substance group 2 as the cause, and the pharmacodynamic index group 1 and the pharmacodynamic index group 2 as the effects, the intersection constitutes a causal path.

[0235] (2) Calculation results

[0236] The model and PLS-SEM calculation results are shown in Tables 17-21.

[0237] Table 17 Reliability and convergent validity of the optimized grouping path model in Example 2

[0238]

[0239] Table 18 Example 2 Optimizing the Fomell-Larcke Criterion of the Grouping Path Model

[0240] Efficacy index group 1 Efficacy index group 2 Pharmacological Substance Group 2 Efficacy index group 2 0.802 Efficacy index group 1 -0.610 0.910 Pharmacological Substance Group 2 -0.745 0.663 0.824

[0241] Table 19 Path coefficients / total effects of the optimized grouping path model in Example 2

[0242] path Original sample Sample mean Standard Deviation T-value P-value Pharmacological effect substance group 2 → Pharmacological effect index group 1 -0.745 -0.769 0.293 2.543 0.011 Pharmacological effect substance group 2 → Pharmacological effect index group 2 0.663 0.633 0.386 1.718 0.086

[0243] Table 20 Example 2 Optimization of the explanatory power of the grouping path model R 2

[0244] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy index group 1 0.555 0.500 Efficacy index group 2 0.440 0.370

[0245] Table 21 Prediction ability and cross-validation verification of optimized grouping path model in Example 2

[0246]

[0247] (3) Variable optimization criteria

[0248] In the variable optimization criteria of the external model 2, the absolute value of the factor loading is selected to have a maximum critical value of 0.4, and the rest is the same as in Example 1.

[0249] (4) The main results are described and analyzed as follows:

[0250] ① Reliability of measurement variables: The factor loading of magnolia officinalis was 0.545, the other indicators were greater than 0.70, the T values ​​were greater than 1.96, and the P values ​​were significant.

[0251] ② The reliability of the latent variables: CR is greater than 0.70.

[0252] ③ Convergent validity: AVE is greater than 0.50. The effect is very good.

[0253] ④Correlation: Cronbach's coefficient has the same sign.

[0254] ⑤Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE, which is greater than the correlation coefficient value below, indicating good discriminant validity.

[0255] ⑥Path coefficient: one is significant, and one is relatively significant.

[0256] ⑦Interpretation ability: Moderate.

[0257] ⑧ Prediction ability and cross-validation verification: For Q 2, only the measurement mode PL prediction is less than 0, and the other indicators are greater than 0, which is acceptable. PLS-SEM performs better than LM (linear) regression.

[0258] (5) Conclusion

[0259] Pharmacodynamic substance group 2: magnolol, honokiol, hesperidin;

[0260] Efficacy index group 1: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6;

[0261] Efficacy index group 2: interleukin-10, calcium ions.

[0262] Critical path: Pharmacodynamic substance group 2 reduces the index value of pharmacodynamic index group 1 and increases the index value of pharmacodynamic index group 2.

[0263] Embodiment 3 Ma Xing Shi Gan Tang antipyretic

[0264] The PLS-SEM model was constructed based on the experimental data of the antipyretic effect of serum drug components in Maxing Shigan Decoction. Several serum drug components in Maxing Shigan Decoction were selected as independent variables, and several efficacy indicators were selected as dependent variables. The variable grouping was optimized to establish a reflective-type, reflective-type, and formative-type path model. The model diagram is shown in Figure 1 .

[0265] Experimental data on the antipyretic effect of serum drug components of Maxing Shigan Decoction:

[0266] Based on the experimental data of antipyretic effect of Ma Xing Shi Gan Tang combined with other drugs in the existing literature 3 (Cui Yanru, Qu Fei, Xu Jing, et al. Effect of changes in combined dosage on the antipyretic effect of Ma Xing Shi Gan Tang [J]. Chinese Journal of Experimental Traditional Chinese Medicine, 2014, 20(06):122-126.), in the animal experiment, the drug components of Ma Xing Shi Gan Tang serum were measured after the animals were given the drug (Ma Xing Shi Gan Tang decoction), and the drug components of Ma Xing Shi Gan Tang serum were matched with the efficacy index data in literature 3 to obtain Tables 22 and 23.

[0267] Serum drug ingredients: Ephedrine, Pseudoephedrine, Methylephedrine, Amygdalin, Prunasin, Liquiritin, Liquiritigenin, Glycyrrhetinic Acid.

[0268] Efficacy indicators: prostaglandin E2 (PGE2), temperature response index (TRI), 6-hour fever suppression rate (ShFSR).

[0269] Table 22 Ratio of serum drug components in Maxingshigan Decoction

[0270]

[0271] Table 23 Results of serum drug antipyretic test of Ma Xing Shi Gan Tang

[0272]

[0273]

[0274] 1. Establish an initial causal path model

[0275] (1) Model construction:

[0276] External mode: Reflective

[0277] Independent variable out-of-control mode: ephedrine, pseudoephedrine, methylephedrine, amygdalin, prunin, liquiritin, liquiritigenin, and glycyrrhetinic acid were used as measured independent variables, and the initial pharmacodynamic substance group was used as a latent variable.

[0278] External dependent variable model: Prostaglandin E2, temperature response index, and 6-hour fever suppression rate were used as measured dependent variables, and the initial efficacy index group was used as a latent variable.

[0279] Internal mode: formative

[0280] The initial group of pharmacodynamic substances is the cause, and the initial group of pharmacodynamic indicators is the effect, forming a cause-effect path.

[0281] (2) Calculation results

[0282] The model and PLS-SEM calculation results are shown in Figure 5 , 6 .

[0283] (3) Conclusion

[0284] Select independent variables and dependent variables, and group them. When examining independent variables and dependent variables, select large factor loadings, separate factors with different signs, and combine factors with the same sign; separate Cronbach's coefficients of dependent variables with different signs, and combine factors with the same sign.

[0285] from Figure 5It can be seen that the absolute values ​​of the negative part of the factor loading of the efficacy substance on the left are very small, so they are discarded; the factor loading of the efficacy index part "prostaglandin E2" on the right is very small, so it is discarded. For the remaining measurement variables, only the positive factor loading values ​​on the left are one group, and the positive and negative values ​​on the right are separated, so they can be divided into two groups and remodeled. We get:

[0286] Pharmacological substance group I′: ephedrine, methylephedrine, pseudoephedrine;

[0287] Efficacy index group I′: temperature response index;

[0288] Efficacy index group II′: 6h fever inhibition rate.

[0289] 2. Establish a grouped causal path model

[0290] (1) Model construction:

[0291] Independent variable-external model (reflective type): Ephedrine, methylephedrine, and pseudoephedrine are used as independent variables, and pharmacodynamic substance group I is used as a latent variable.

[0292] Dependent variable external mode (reflective type): body temperature reaction index is used as the measured dependent variable, and efficacy index group I is used as the latent variable. 6h fever suppression rate is used as the measured dependent variable, and efficacy index group II is used as the latent variable.

[0293] Internal mode: formative

[0294] Taking the pharmacodynamic substance group I as the cause, and the pharmacodynamic index group I and pharmacodynamic index group II as the effects, the intersection constitutes a causal path.

[0295] (2) Calculation results

[0296] The model and PLS-SEM calculation results are shown in Tables 24-28.

[0297] Table 24 Reliability and convergent validity of the grouping path model of Example 3

[0298]

[0299] Table 25 Fomell-Larcke criterion of the group path model of Example 3

[0300] Pharmacodynamic substance group I Efficacy Index Group I Efficacy Index Group II Pharmacodynamic substance group I 0.994 Efficacy Index Group I 0.796 1.000 Efficacy Index Group II -0.716 -0.939 1.000

[0301] Table 26 Path coefficients / total effects of the grouped path model of Example 3

[0302]

[0303] Table 27 Explanatory power R of the grouping path model of Example 3 2

[0304] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy Index Group I 0.634 0.600 Efficacy Index Group II 0.512 0.468

[0305] Table 28 Predictive ability and cross-validation verification of the grouping path model in Example 3

[0306]

[0307]

[0308] (3) Variable optimization criteria

[0309] Refer to Example 1.

[0310] (4) The main results are described and analyzed as follows:

[0311] ① Reliability of measurement variables: factor loadings are all greater than 0.70, T values ​​are all greater than 1.96, and P values ​​are all significant.

[0312] ② The reliability of the latent variables: CR is greater than 0.70.

[0313] ③ Convergent validity: AVE is greater than 0.50. The effect is very good.

[0314] ④Correlation: Cronbach's coefficient is positive.

[0315] ⑤Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE, which is greater than the correlation coefficient value below, indicating good discriminant validity.

[0316] ⑥Path coefficient: significant.

[0317] ⑦Explanatory ability: R 2 good.

[0318] ⑧ Prediction ability and cross-validation verification: For Q2, the measurement mode indicators are all greater than 0, and are good and acceptable. For Q2, the latent variable mode is greater than 0 and greater than 0.35, indicating strong prediction ability.

[0319] The above results show that ephedrine, methylephedrine and pseudoephedrine are pharmacodynamic substance groups; temperature response index and 6h fever inhibition rate are pharmacodynamic indicator groups (single); the pharmacodynamic substance group and the pharmacodynamic indicator group have good reliability and validity, which may be the key path.

[0320] (5) Conclusion

[0321] Pharmacological substance group I: ephedrine, methylephedrine, pseudoephedrine.

[0322] Efficacy index group I: body temperature response index.

[0323] Efficacy index group II: 6h fever inhibition rate;

[0324] Critical path: Pharmacodynamic substance group I increases the index value of pharmacological index group I and decreases the index value of pharmacological index group II.

[0325] 4. Critical Paths of Exploring the Effective Substance Group and Effectiveness Index Group of Traditional Chinese Medicine Compounds Based on PLS-SEM Example 4: Dachengqi Decoction for the Treatment of Acute Pancreatitis

[0326] The PLS-SEM model was constructed based on the experimental data of Dachengqi decoction combined with other drugs for the treatment of acute pancreatitis. Four combined drugs in Dachengqi decoction were selected as independent variables, and eight efficacy indicators were selected as dependent variables. The variable grouping was optimized to establish a formative-reflective-formational path model. The model diagram is shown in Figure 2 .

[0327] Experimental data on the treatment of acute pancreatitis with Dachengqi decoction combined with other drugs:

[0328] The experimental data are derived from the existing literature 1 (Qiandan, Yu Riyue, Yu Xiaojuan, Wan Panting, Liu Xinhui, Nie Bin, Chen Yinfang, Peng Hong, Xu Guoliang. Experimental study on the treatment of acute pancreatitis with Dachengqi decoction [J]. Liaoning Journal of Traditional Chinese Medicine, 2017, 44(1): 190-192.), see Table 29 and Table 30 for details.

[0329] Serum drug ingredients: Radix et Rhizoma Rhei (RRR), Magnolia officinalis (MOC), Fructus Aurantii Immaturus (FAI), Mirabilite.

[0330] Efficacy index: calcium ion (Ca 2+ ), pancreatic lipase (PL), interleukin-6 (IL-6), interleukin-10 (IL10), lymphocyte function associated antigen 1a (LFA1a), amylase (AMY), organ indicators (OI), survival rate (SR).

[0331] Table 29 Drug ratios of Dachengqi Decoction

[0332] Group Rhubarb (g / kg) Magnolia officinalis (g / kg) Citrus aurantium (g / kg) Sodium Sulfate(g / kg) original party 54.00 40.50 40.50 20.25 Ratio 1 21.33 12.64 162.00 9.00 Ratio 2 0.00 28.44 14.22 45.00 Ratio 3 32.00 144.00 0.00 63.00 Ratio 4 48.00 0.00 48.00 54.00 Ratio 5 9.48 8.43 6.32 27.00 Ratio 6 162.00 96.00 72.00 36.00 Ratio 7 14.22 216.00 32.00 18.00 Ratio 9 108.00 18.96 21.33 81.00 Ratio 10 72.00 42.67 9.48 0.00

[0333] Table 30 Results of Dachengqi decoction combined with other drugs in the treatment of acute pancreatitis

[0334]

[0335]

[0336] 1. Establish an initial causal path model

[0337] (1) Model construction:

[0338] All independent variables and dependent variables are involved in modeling.

[0339] Independent variable external model (formative type): Rhubarb, Magnolia officinalis, Citrus aurantium, and Sodium Sulfate are used as measured independent variables, and the initial group of effective substances is used as latent variables.

[0340] External dependent variable model (reflective type): calcium ion, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate are used as measured dependent variables, and the initial efficacy index group is used as a latent variable.

[0341] Internal mode: formative

[0342] The initial group of pharmacodynamic substances is the cause, and the initial group of pharmacodynamic indicators is the effect, forming a cause-effect path.

[0343] (2) Calculation results

[0344] The model and PLS-SEM calculation results are shown in Figure 7 (factor loading, path coefficient and explanatory power), and its Bootstrapping modeling failed.

[0345] (3) Conclusion

[0346] Select independent variables and dependent variables, and group them. Examine the independent variables and dependent variables, select the largest factor loading, separate those with different signs, and combine those with the same sign, and you will get:

[0347] Pharmacodynamic substance group A′: Magnolia officinalis;

[0348] Pharmacological substance group B′: rhubarb, immature bitter orange, and sodium sulfate;

[0349] Efficacy index group A′: amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6, and lung index;

[0350] Efficacy index group B′: interleukin-10, calcium ions, and survival rate.

[0351] 2. Establish a grouped causal path model

[0352] (1) Model construction:

[0353] Independent variable external model (formational type): Magnolia officinalis is used as the measured independent variable, and the active substance group A″ is used as the latent variable. Rhubarb, Citrus aurantium, and Sodium Sulfate are used as the measured independent variables, and the active substance group B″ is used as the latent variable.

[0354] External dependent variable model (reflective type): Amylase, pancreatic lipase, lymphocyte function-associated antigen 1a, interleukin-6, and lung index are used as measured dependent variables, and the efficacy index group A″ is used as a latent variable. Interleukin-10, calcium ions, and survival rate are used as measured dependent variables, and the efficacy index group B″ is used as a latent variable.

[0355] Internal pattern: Cross-forming

[0356] With the medicinal substance group A″ and the medicinal substance group B″ as causes, and the medicinal index group A″ and the medicinal index group B″ as effects, a two-cause and two-effect path is formed crosswise.

[0357] (2) Calculation results

[0358] The model and PLS-SEM calculation results are shown in Tables 31-37.

[0359] Table 31 Reliability and convergent validity of the group path model of Example 4

[0360]

[0361] Table 32 Fomell-Larcke criterion of group path model in Example 4

[0362] Pharmacodynamic substance group A″ Efficacy index group A″ Efficacy index group A″ Pharmacodynamic substance group A″ 1.000 Efficacy index group A″ -0.812 0.710 Efficacy index group B″ 0.739 -0.819 0.862

[0363] Table 33 External mode weights of grouping path model in Example 4

[0364]

[0365]

[0366] Table 34 Collinearity analysis of grouping path model in Example 4

[0367]

[0368] Table 35 Path coefficients / total effects of the grouped path model of Example 4

[0369] path Original sample Sample mean Standard Deviation T-value P-value Pharmacological effect substance group A″→Pharmacological effect index group A″ -0.797 -0.735 0.269 2.967 0.004 Pharmacological effect substance group A″→Pharmacological effect index group B″ 0.634 0.566 0.326 1.943 0.055 Pharmacological effect substance group B″→Pharmacological effect index group A″ 0.071 0.190 0.389 0.182 0.856 Pharmacodynamic substance group B″→Pharmacodynamic index group B″ -0.49 -0.507 0.382 1.306 0.194

[0370] Table 36 Explanatory power R of the grouping path model of Example 4 2

[0371] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy index group A″ 0.784 0.723 Efficacy index group B″ 0.664 0.568

[0372] Table 37 Predictive ability and cross-validation verification of the grouping path model in Example 4

[0373]

[0374] (3) Variable optimization criteria

[0375] On the basis of Example 1, three optimization criteria for forming the external model are added: multicollinearity VIF < 5; external model indicator weight; significance of P value and T value of external model indicator weight; the rest are the same as Example 1.

[0376] (4) Conclusion

[0377] Select the independent variables and dependent variables, and group them. Separate the positively correlated and negatively correlated efficacy indicators. Keep the paths with high reliability and validity and significant path coefficients (P<0.05). As shown in Table 31, the factors with absolute values ​​of factor loadings of efficacy substances greater than 0.7 are kept, and the positive and negative are separated; the efficacy indicators are separated into positive and negative groups, and can be divided into two groups. We get:

[0378] Pharmacodynamic substance group A″: Magnolia officinalis;

[0379] Efficacy index group A″: amylase, pancreatic lipase;

[0380] Efficacy index group B″: interleukin-10, calcium ions, and survival rate.

[0381] 3. Causal Path Model for Optimized Grouping

[0382] (1) Model construction:

[0383] Independent variable-external model (formative type): Magnolia officinalis is used as the measured independent variable, and the active substance group A is used as the latent variable.

[0384] Dependent variable external mode (reflective type): amylase and pancreatic lipase are used as measured dependent variables, and drug efficacy index group A is used as a latent variable. Interleukin-10, calcium ions, and survival rate are used as measured dependent variables, and drug efficacy index group B is used as a latent variable.

[0385] Internal pattern: Cross-forming

[0386] Taking the pharmacodynamic substance group A as the cause, and the pharmacodynamic index group A and the pharmacodynamic index group B as the effects, the intersection constitutes a causal path.

[0387] (2) Calculation results

[0388] The model and PLS-SEM calculation results are shown in Tables 38-43.

[0389] Table 38 Reliability and convergent validity of the optimized grouping path model in Example 4

[0390]

[0391] Table 39 Example 4 Optimizing the Fomell-Larcke Criterion of the Grouping Path Model

[0392] Pharmacodynamic substance group A Efficacy Index Group A Efficacy Index Group B Pharmacodynamic substance group A 1.000 Efficacy Index Group A -0.714 0.852 Efficacy Index Group B 0.796 -0.840 0.856

[0393] Table 40 Example 4 Optimizing the external mode weights of the grouping path model

[0394]

[0395] Table 41 Collinearity analysis of optimized grouping path model in Example 4

[0396]

[0397] Table 42 Path coefficients / total effects of the optimized grouping path model in Example 4

[0398] path Original sample Sample mean Standard Deviation T-value P-value Pharmacological effect substance group A → Pharmacological effect index group A -0.714 -0.715 0.211 3.375 0.001 Pharmacological effect substance group A → Pharmacological effect index group B 0.796 0.796 0.223 3.570 0.001

[0399] Table 42 Example 4 Optimization of the explanatory power of the grouping path model R2

[0400] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy Index Group A 0.509 0.448 Efficacy Index Group B 0.634 0.588

[0401] Table 43 Predictive ability and cross-validation verification of the optimized grouping path model in Example 4

[0402]

[0403] (3) Variable optimization criteria

[0404] On the basis of Example 1, three optimization criteria for forming the external model are added: multicollinearity VIF < 5; external model indicator weight; significance of P value and T value of external model indicator weight; the rest are the same as Example 1.

[0405] (4) Results description and analysis:

[0406] The above results show that the factor loading values ​​are all greater than 0.7, the combined reliability is greater than 0.800, the convergent validity AVE is greater than 0.7, and the path coefficients are -0.714 and 0.796, respectively, which are relatively high; the explanatory power R 2 The values ​​were 0.509 and 0.634 respectively, indicating that the explanatory power was acceptable; P<0.05, passing the test. Pharmacodynamic substance group A contained magnolia bark, efficacy index group A contained amylase and pancreatic lipase, and efficacy index group B contained interleukin-10, calcium ions, and survival rate.

[0407] There are two key paths, namely, the path of pharmacological substance group A acting on pharmacological index group A, and the path of pharmacological substance group B acting on pharmacological index group B.

[0408] (5) Conclusion

[0409] Pharmacodynamic substance group A: Magnolia officinalis;

[0410] Efficacy index group A: amylase, pancreatic lipase;

[0411] Efficacy index group B: interleukin-10, calcium ions, and survival rate.

[0412] Critical path: Pharmacodynamic substance group A reduces the index value of pharmacological index group A and increases the index value of pharmacological index group B.

[0413] V. Mining the key paths of the pharmacodynamic substance group and pharmacodynamic index group of Chinese medicine composition based on PLS-SEM Example 5: Polygonum multiflorum against hypoxia / reoxygenation

[0414] The PLS-SEM model was constructed based on the experimental data of anti-hypoxia / reoxygenation effects of the components and compatibility drugs of Polygonum multiflorum. The three components and compatibility drugs in Dachengqi Decoction were selected as independent variables, and the three efficacy indicators were selected as dependent variables. The variable grouping was optimized to establish a formative-reflective-formational pathway model. The schematic diagram of the model is shown in Figure 2 .

[0415] Experimental data of Polygonum multiflorum against hypoxia / reoxygenation:

[0416] The experimental data comes from existing literature (Siqi Wan; Jinfang Zhang; Rong Hou; Minsi Zheng; Liya Liu; Mingshuo Zhang; Zhiyong Li; Xiulan Huang; A strategy for component-based Chinese medicines design approach of Polygonum orientale L. against hypoxia / reoxygenation based on uniform design-stepwise regression-simulated annealing[J]; Biomedicine & Pharmacotherapy, 2021, 135), see Table 44 for details.

[0417] The components of Polygonum multiflorum are: orientin, quercitrin and vitexin.

[0418] Efficacy indicators: Cell viability, LDH leakage rate, and NO level.

[0419] Table 44: Ratio of drug combinations and anti-hypoxia / reoxygenation results of Polygonum multiflorum

[0420]

[0421] 1. Establish an initial causal path model

[0422] (1) Model construction:

[0423] All independent variables and dependent variables are involved in modeling.

[0424] Independent variable extra-model (formative type): orientin, quercetin, and vitexin are used as measured independent variables, and the initial pharmacodynamic substance group is used as a latent variable.

[0425] External dependent variable model (reflective type): Cell viability, lactate dehydrogenase, and nitric oxide levels are used as measured dependent variables, and the initial efficacy index group is used as a latent variable.

[0426] Internal mode: formative

[0427] The initial group of pharmacodynamic substances is the cause, and the initial group of pharmacodynamic indicators is the effect, forming a cause-effect path.

[0428] (2) Calculation results

[0429] The model and PLS-SEM calculation results are shown in Figure 8 , 9 .

[0430] (3) Conclusion

[0431] Select independent variables and dependent variables, and group them. Examine the independent variables and dependent variables, select the largest factor loading, separate factors with different signs, and combine factors with the same sign. Figure 8 It can be seen that the cell viability value of the efficacy indicator is negative. According to the selection of greater than 0.7, and the positive and negative signs are separated, we can get:

[0432] Pharmacodynamic substance group a′: quercetin;

[0433] Efficacy index group a′: lactate dehydrogenase and nitric oxide levels.

[0434] 2. Establish a grouped causal path model

[0435] (1) Model construction:

[0436] Independent variable extra-model (formative type): Quercetin is used as the measured independent variable, and the active substance group a is used as the latent variable.

[0437] External dependent variable model (reflective type): lactate dehydrogenase and nitric oxide levels are used as measured dependent variables, and efficacy index group a is used as a latent variable.

[0438] Internal pattern: Cross-forming

[0439] Taking the pharmacological substance group a as the cause and the pharmacological index group a as the effect, without crossing, a causal path is formed.

[0440] (2) Calculation results

[0441] The model and PLS-SEM calculation results are shown in Tables 45-51.

[0442] Table 45 Reliability and convergent validity of the grouping path model of Example 5

[0443]

[0444] Table 46 Fomell-Larcke criterion of the group path model of Example 5

[0445] Efficacy index group a Efficacy index group a 0.912

[0446] Table 47 External mode weights of grouping path model in Example 5

[0447] Latent variables Measured variables External mode weight T-value P-value Pharmacodynamic substance group a Quercetin 1.000 n / a n / a Efficacy index group a Lactate dehydrogenase 0.496 1.337 0.169 Efficacy index group a Nitric oxide levels 0.598 1.689 0.092

[0448] Table 48 Collinearity analysis of grouping path model in Example 5

[0449] VIF Lactate dehydrogenase 1.800 Nitric oxide levels 1.800 Quercetin 1.000

[0450] Table 49 Path coefficients / total effects of the grouped path model of Example 5

[0451] path Original sample Sample mean Standard Deviation T-value P-value Pharmacodynamic substance group a → Pharmacodynamic index group a 0.564 0.507 0.336 1.678 0.094

[0452] Table 50 Explanatory power R of the grouping path model of Example 5 2

[0453] <![CDATA[R 2 ]]> <![CDATA[Adjusted R 2 > Efficacy index group a 0.318 0.220

[0454] Table 51 Predictive ability and cross-validation verification of the grouping path model in Example 5

[0455]

[0456]

[0457] (3) Variable optimization criteria

[0458] On the basis of Example 1, three optimization criteria for forming the external model are added: multicollinearity VIF < 5; external model indicator weight; significance of P value and T value of external model indicator weight; the rest are the same as Example 1.

[0459] (4) The main results are described and analyzed as follows:

[0460] ① Reliability of measurement variables: factor loading is greater than 0.70, T value is greater than 1.96, and P value is significant.

[0461] ② The reliability of the latent variables: CR is greater than 0.70.

[0462] ③Convergent validity: AVE is greater than 0.50.

[0463] ④Correlation: Cronbach's coefficients have the same sign and the same function.

[0464] ⑤Discriminant validity: The diagonal line of the Fomell-Larcker criterion is the square root of AVE, which is greater than the correlation coefficient value below.

[0465] ⑥Factor loading weight: P value is not significant.

[0466] ⑦Collinearity analysis: no correlation exists.

[0467] ⑧Path coefficient: 0.094, not significant, but less than 0.1.

[0468] ⑨Explanatory power R 2 : The value is 0.318, which is acceptable.

[0469] ⑩ Prediction ability and cross-validation verification: The predicted value of the indicator is greater than 0. The cross-validation prediction ability test is not significant enough, which may be affected by the small number of experimental samples or other reasons, and it is worth further study.

[0470] The above results show that the external model variable factor loading is good, the path coefficient is good, the explanatory power is acceptable, there is no correlation between variables, and the cross-validation prediction ability is insufficient. It has a certain suggestive effect on the combination of components, and quercetin may act on lactate dehydrogenase and nitric oxide levels. The key path.

[0471] (5) Conclusion

[0472] Select the independent variables and dependent variables, and group them. Examine the independent variables and dependent variables, select the largest factor loading, separate the ones with different signs, and merge the ones with the same sign. Separate the efficacy indicators of positive correlation and negative correlation. Keep the paths with high reliability and validity and more significant path coefficients. Get:

[0473] Pharmacodynamic substance group a: quercetin;

[0474] Efficacy index group a: lactate dehydrogenase and nitric oxide levels.

[0475] Critical path: Pharmacodynamic substance group a increases the index value of pharmacodynamic index group a.

Claims

1. A method for mining the key paths of Chinese medicine efficacy substance groups and efficacy index groups based on PLS-SEM, characterized by: The PLS-SEM model is constructed based on the experimental data of the effective substances and efficacy indicators of traditional Chinese medicine. Several effective substances in traditional Chinese medicine are selected as independent variables, and several efficacy indicators are selected as dependent variables. The variable grouping is optimized to establish a causal path model of reflective-reflective-formative or formative-reflective-formative. The causal path model includes an initial causal path model and a grouped causal path model in accordance with the modeling steps, both of which are composed of corresponding external models and internal models. The external model includes an independent variable external model and a dependent variable external model. The independent variable external model uses the corresponding pharmacodynamic substance as the measured independent variable and the corresponding pharmacodynamic substance group as the exogenous latent variable. The dependent variable external model uses the corresponding pharmacodynamic index as the measured dependent variable and the corresponding pharmacodynamic index group as the endogenous latent variable. The internal model is composed of the corresponding pharmacodynamic substance group and the corresponding pharmacodynamic index group to form a causal path.

2. The method according to claim 1, characterized in that: The following steps are involved: S1. Establish an initial causal path model, wherein the initial causal path model includes an external model 1 and an internal model 1. The external model 1 includes an independent variable external model 1 and a dependent variable external model 1. The independent variable external model 1 uses all pharmacodynamic substances as measured independent variables and an initial pharmacodynamic substance group as an exogenous latent variable. The dependent variable external model 1 uses all pharmacodynamic indicators as measured dependent variables and an initial pharmacodynamic indicator group as an endogenous latent variable. The internal model 1 is composed of the initial pharmacodynamic substance group and the initial pharmacodynamic indicator group to form a one-cause-one-effect path. The PLS-SEM model is used for calculation to select a better path and optimize the variable grouping. S2. On the basis of step S1, a grouped causal path model is established, wherein the causal path model includes an external model 2 and an internal model 2. The external model 2 includes an independent variable external model 2 and a dependent variable external model 2. The independent variable external model 2 uses each group of pharmacological substances optimized in step S1 as a measured independent variable and the grouped pharmacological substance group as an exogenous latent variable. The dependent variable external model 2 uses each group of pharmacodynamic indicators optimized in step S1 as a measured dependent variable and the grouped pharmacological indicator group as an endogenous latent variable. The internal model 2 is composed of the grouped pharmacological substance group and the grouped pharmacological indicator group to form a causal path. The PLS-SEM model is used to calculate and select a better path optimization variable grouping. S3, based on step S2, optimization is performed to establish a causal path model for optimized grouping. The method is the same as the causal path model for grouping in step S2. Through PLS-SEM model calculation, a better path is selected to optimize the variable grouping until it cannot be optimized, or it is stopped according to actual needs; S4. Arrange the results to obtain groups of pharmacological substances with similar efficacy, groups of pharmacological efficacy indicators and their key paths.

3. The method according to claim 1, characterized in that: The pharmacological substance is derived from the serum drug component of the traditional Chinese medicine compound, the compatible drug of the traditional Chinese medicine compound or the compatible drug of the traditional Chinese medicine component; the external mode is the reflection type or the formation type, and the internal mode is the reflection type.

4. The method according to claim 1, characterized in that: The variable optimization criteria of the external model include any one or more of the following: (1) Check the reliability of each variable: factor loading (Outer loadings), select the largest absolute value, the critical value of which is 0.7 or 0.4, separate the factors with different signs, and combine the factors with the same sign into groups, and check the significance of the P value and T value of the factor loading; (2) Check the combined reliability of each variable: When the CR is positive, select the variables with CR ≥ 0.60 to merge into a group. For the variables with negative CR, recheck their factor loading and adjust the correlation. (3) Check convergent validity AVE, AVE ≥ 0.5; (4) Check the correlation, the Cronbach's alpha is positive; (5) Check discriminant validity and use the Fomell-Larcker criterion or cross-loading to discriminate and group variables. The Fomell-Larcker criterion is that the square root of the AVE value of each dimension must be greater than its correlation coefficient with other dimensions; the cross-loading value is that the highest factor loading of each variable should be on the dimension to be measured; (6) Multicollinearity: VIF < 5; (7) External model indicator weights; (8) The significance of the P-value and T-value of the external model indicator weight.

5. The method according to claim 1, characterized in that: The variable optimization criteria of the internal model include any one or more of the following: ① Check the significance of the P value and T value of the path coefficient; ② Check the explanatory power of each variable R 2 , whose values ​​are selected as moderate and above, are merged into groups; ③ Prediction ability and cross-validation verification: Q 2 >0.

6. The method according to claim 4, characterized in that: In step S1, the variable optimization criteria of external model 1 include: checking the reliability of each variable: the absolute value of the factor loading, the value is selected to be larger, the critical value of the larger value is 0.7 or 0.4, different signs are separated, and the same sign is combined into a group.

7. The method according to any one of claims 1 to 6, characterized in that: The medicinal substance is derived from the serum drug component of a traditional Chinese medicine compound, the traditional Chinese medicine compound is Dachengqi Decoction, the Dachengqi Decoction includes the original Dachengqi Decoction and a uniformly designed Dachengqi Decoction combination, and the experimental data is the experimental data of Dachengqi Decoction in treating acute pancreatitis; The serum drug components include rhein, rhein, chrysophanol, aloe-rhein, rhein methyl ether, magnolol, honokiol, hesperidin, and hesperetin, and the efficacy indicators include calcium ions, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate; The obtained pharmacodynamic substance group is pharmacodynamic substance group 1 or pharmacodynamic substance group 2, the pharmacodynamic substance group 1 includes magnolol and hesperidin, and the pharmacodynamic substance group 2 includes magnolol, honokiol and hesperidin; the obtained pharmacodynamic index group includes pharmacodynamic index group 1 and pharmacodynamic index group 2, the pharmacodynamic index group 1 includes amylase, interleukin-6, lymphocyte function-associated antigen 1a, and pancreatic lipase, and the pharmacodynamic index group 2 includes interleukin-10 and calcium ions; The obtained key path includes: the pharmacodynamic substance group 1 or the pharmacodynamic substance group 2 reduces the index value of the pharmacodynamic index group 1 and increases the index value of the pharmacodynamic index group 2.

8. The method according to any one of claims 1 to 6, characterized in that: The pharmacological substance is derived from the serum drug component of a traditional Chinese medicine compound, the traditional Chinese medicine compound is Maxing Shigan Decoction, the Maxing Shigan Decoction includes the original Maxing Shigan Decoction and a uniformly designed Maxing Shigan Decoction combination, and the experimental data is the experimental data of Maxing Shigan Decoction for reducing fever; The serum drug components include ephedrine, pseudoephedrine, methylephedrine, amygdalin, prunin, liquiritin, liquiritigenin, and glycyrrhetinic acid, and the efficacy indicators include prostaglandin E2, temperature response index, and 6h fever inhibition rate; The obtained pharmacodynamic substance group I includes ephedrine, methylephedrine, and pseudoephedrine, and the obtained pharmacodynamic index group includes pharmacodynamic index group I and pharmacodynamic index group II, wherein the pharmacodynamic index group I includes the body temperature response index, and the pharmacodynamic index group II includes the 6h fever inhibition rate; The obtained key path includes: the pharmacological substance group I increases the index value of the pharmacological index group I, and reduces the index value of the pharmacological index group II.

9. The method according to any one of claims 1 to 6, characterized in that: The medicinal substance is derived from a compound medicine of a traditional Chinese medicine, the traditional Chinese medicine formula is Dachengqi Decoction, the Dachengqi Decoction includes the original Dachengqi Decoction formula and a uniformly designed Dachengqi Decoction compound formula, and the experimental data is experimental data of Dachengqi Decoction in treating acute pancreatitis; The compatible drugs include rhubarb, magnolia bark, immature bitter orange, and sodium sulfate, and the efficacy indicators include calcium ion, pancreatic lipase, interleukin-6, interleukin-10, lymphocyte function-associated antigen 1a, amylase, lung index, and survival rate; The obtained pharmacodynamic substance group A includes magnolia bark, and the obtained pharmacodynamic index group includes pharmacodynamic index group A and pharmacodynamic index group B, wherein the pharmacodynamic index group A includes amylase and pancreatic lipase, and the pharmacodynamic index group B includes interleukin-10, calcium ions, and survival rate; The obtained key path includes: the pharmacological substance group A reduces the index value of the pharmacological index group A and increases the index value of the pharmacological index group B.

10. The method according to any one of claims 1 to 6, characterized in that: The pharmacological substance is derived from a Chinese medicine component and a drug combination, the Chinese medicine component and a drug combination is a Polygonum multiflorum component and a drug combination, and the experimental data is experimental data of the Polygonum multiflorum component and a drug combination for resisting hypoxia / reoxygenation; The drugs combined with the Polygonum erythrorhizon component include orientin, quercetin, and vitexin, and the efficacy indicators include cell viability, lactate dehydrogenase leakage rate, and nitric oxide level; The obtained pharmacodynamic substance group a includes quercetin, and the obtained pharmacodynamic index group a includes the leakage rate of lactate dehydrogenase and the nitric oxide level; The obtained key path includes: the pharmacological substance group a increases the index value of the pharmacological index group a.

Images