Based on UPLC-Q-TOF / MS E A method for fuzzy identification of bioactive components of traditional Chinese medicine with different effects based on BP neural network association analysis with integrated effect

Abstract

The present invention provides a method for fuzzy identification of bioactive components with different effects of traditional Chinese medicine by BP neural network correlation analysis based on UPLC-Q-TOF / MS<supgt;E< / supgt> and integration effect, comprising the following steps: classifying most components in traditional Chinese medicine into different types of component groups by fuzzy chemical identification method; and obtaining the integration effect values of different effects of traditional Chinese medicine through animal experiment evaluation and / or multi-attribute comprehensive indexes; then using BP neural network to correlate the component contents determined by UPLC-Q-TOF / MSE analysis with the integration effect values to realize the correlation analysis between traditional Chinese medicine components and different effects. The method established by the present invention can provide a scientific basis for expanding the clinical application scope of rhubarb and further developing related products. This method can also be used to study the bioactive components of other traditional Chinese medicines or natural functional products, and has strong generality.

Description

Technical Field

[0001] The present invention relates to a research method for active ingredients of traditional Chinese medicine, and in particular to a method based on UPLC-Q-TOF / MS E A method for fuzzy identification of bioactive ingredients of traditional Chinese medicine with different effects was proposed based on BP neural network association analysis with integrated effect. Background Art

[0002] As one of the oldest continuously practiced traditional medical systems in the world, Chinese medicine (TCM) has been used for thousands of years, and its use has increased in recent years in both developing and developed countries around the world. In China, nearly half of commercial drugs are botanicals, which are derived from TCM and widely used to treat various chronic diseases, including cardiovascular disease and cancer. Isolation and purification of active ingredients from natural products, especially TCM, is also considered a successful approach for new drug identification and development. However, the pharmacologically active ingredients of plant-derived TCM are not always the original natural molecules, but may be their host-specific metabolites or molecular complexes formed by some different types of natural compounds. This complexity has brought significant scientific challenges to the study of active ingredients in TCM. The multi-component nature of TCM generates multiple potential molecular interactions, multiple targets, and numerous metabolic byproducts, which further lead to multiple effects, indicating that the traditional reductionist approach has limitations in identifying bioactive ingredients, and a more comprehensive, network-oriented approach is more desirable.

[0003] Rhubarb belongs to the genus Rheum in the Polygonaceae family. It is one of the oldest and most important medicinal plants in many countries. Its use can be traced back to the ancient book "Shennong's Herbal Classic" in 270 BC. Rhubarb is one of the most effective laxatives and is widely used to treat intestinal constipation around the world. It is not only officially included in the "Chinese Pharmacopoeia", but also appears in the "British Pharmacopoeia" and "European Pharmacopoeia". Many classic works of traditional Chinese medicine, including the "Chinese Pharmacopoeia", record that its main effects are "purgative and attacking accumulation" (E1), "clearing heat and purging fire" (E2), "cooling blood and detoxifying" (E3), "removing blood stasis and promoting menstruation" (E4) and "removing dampness and jaundice" (E5). Rhubarb has also shown a variety of pharmacological effects in modern pharmacological research, such as laxative, gallbladder, liver and kidney protection, anti-tumor, lipid-lowering, and blood circulation and stasis removal. Although there have been many reports on the biological activities of several types of rhubarb compounds, including anthraquinones, flavonoids and their glycosides, stilbene glycosides, chromones, gallic acid and gallotannins, phenylbutyric acid and its glycosides, and naphthoglycosides, free and bound anthraquinones are considered to be the main biologically active components of rhubarb, but the contribution of these components to their different effects is still unknown. Therefore, the approach and methodology developed by the present invention is to conduct a comprehensive investigation of the biologically active components of rhubarb and quantify their contributions to the five main effects. These works will promote our knowledge and understanding of the biologically active components of rhubarb and provide a scientific basis for expanding its clinical application range and further developing rhubarb-related products. Summary of the invention

[0004] Purpose of the invention: Although many methods have been developed in recent years to study the bioactive components of the main efficacy or pharmacological effects of traditional Chinese medicine, no one has tried to touch on the study of multiple efficacy, because each efficacy of traditional Chinese medicine corresponds to many different bioactive components. The present invention aims to develop a generally applicable approach and method, that is, to use UPLC-Q-TOF / MS E Fuzzy identification of bioactive components of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effects. Most components of traditional Chinese medicine can usually be classified into one of multiple chemical structure types. They will be classified into different chemical component groups by fuzzy chemical identification method without the need to identify the exact chemical structure, and their relative contents in different extracts of traditional Chinese medicine can be detected by UPLC-Q-TOF / MS E Determination. The present invention can also obtain the integrated effect value of different effects of traditional Chinese medicine through animal experiment evaluation and / or multi-attribute comprehensive index analysis. Subsequently, the BP neural network is used to associate the component content with the integrated effect, thereby realizing the correlation analysis between the components of traditional Chinese medicine and different effects. In this way, the contribution of each component to the different effects of traditional Chinese medicine is quantified, and then the contribution of the effects of a certain type of components is superimposed to illustrate their influence on the common and different characteristics of the different effects of traditional Chinese medicine.

[0005] Technical solution: In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A UPLC-Q-TOF / MS-based E A method for fuzzy identification of biologically active ingredients of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effect, characterized in that it comprises the following steps:

[0007] Most of the ingredients in traditional Chinese medicine are classified into different types of ingredient groups using fuzzy chemical recognition methods; and the integrated effect values ​​of different effects of traditional Chinese medicine are obtained through animal experimental evaluation and / or multi-attribute comprehensive indicators; then the BP neural network is used to associate the ingredient content determined by UPLC-Q-TOF / MSE analysis with the integrated effect value to achieve the correlation analysis between traditional Chinese medicine ingredients and different effects.

[0008] As a preferred solution, the above-mentioned method based on UPLC-Q-TOF / MS E A method for fuzzy identification of biologically active ingredients of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effect is characterized in that the traditional Chinese medicine contains anthraquinone effective ingredients.

[0009] As a preferred solution, the above-mentioned method based on UPLC-Q-TOF / MS E A method for fuzzy identification of biologically active ingredients of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effect, characterized in that it comprises the following steps:

[0010] (1) Preparation of rhubarb sample solution and chemical reference substance:

[0011] Rhubarb sample solution: Preparation process: accurately weigh multiple portions of rhubarb powder, first fully soak them with ethanol aqueous solutions of different concentrations, then add ethanol aqueous solutions of different concentrations to reflux and boil, filter while hot, combine the filtrates, and concentrate under reduced pressure to obtain ethanol extracts of different concentrations for later use;

[0012] Reference solution: accurately weigh 28 chemical reference substances: Chrysophanol, Chrysophanol 1-ObD-glucoside, Chrysophanol 8-ObD-glucoside, Emodin, Emodin 1-ObD-glucoside, Emodin 8-ObD-glucoside, Physcion, Physcion 1-ObD-glucoside, Physcion 8-ObD-glucoside, Aloe-emodin, Aloe-emodin 8-ObD-glucoside, Rhein, Rhein 8-ObD-glucoside, SennosideA, Sennoside B, Sennoside C, Cianidanol, (-)-Epicatechin, (-)-Epicatechin gallate, Procyanidin B2, Gallic acid, Resveratroloside, Rhapontigenin, Torachrysone 8-O-glucoside, 4-(4-Hydroxyphenyl)-2-butanone, Raspberryketone glucoside, Lindleyin, and 5-Acetyl-7-hydroxy-2-methyl-chromone were respectively dissolved with DMSO, and then methanol was added and completely dissolved into a transparent solution by ultrasonication to obtain 28 chemical reference substance mother solutions; then, the 28 chemical reference substance mother solutions were mixed to obtain a mixed reference substance solution;

[0013] (2) UPLC-Q-TOF / MS E analyze:

[0014] The rhubarb sample solution and the reference substance solution of step (1) are respectively injected into an ultra-high performance liquid chromatograph for analysis, the mass spectrometry scanning acquisition time is 0.1s, and the ultraviolet detection wavelength range is set to 190-420nm; the mobile phase A is 0.1% formic acid water, and B is acetonitrile, and the gradient elution program is: 0-2min, 3-10% B; 2-5min, 10-15% B; 5-8min, 15% B; 8-11min, 15-23% B; 11-16min, 23-30% B; 16-20min, 30-60% B; 20-24min, 60-80% B; 24-25min, 80-3% B;

[0015] (4) Fuzzy chemical identification:

[0016] Through the fuzzy chemical recognition method, according to the peak order, fragmentation information and cleavage pathway of different categories of components, the rhubarb full spectrum information database in Table 1 was established (Table S1), and the corresponding compound group network was established through the rhubarb full spectrum information database to identify unknown components;

[0017] (4) BP neural network association method:

[0018] The BP neural network was used to associate the relative content of each component in the rhubarb sample solution with the five integrated effects. It mainly includes three steps: (1) data normalization: y = (x-MinValue) / (MaxValue-MinValue), where x and y are the values ​​before and after conversion, respectively, and MaxValue and MinValue are the maximum and minimum values ​​in the sample, respectively; (2) data set division: all component contents are taken as input, and the corresponding integrated effect values ​​are taken as output; and all groups 1 to 18 data are used as training samples, among which 90% EW-L, ethanol-S and ethanol-L are used as test samples; (3) dose-effect association: based on the BP neural network algorithm, different parameters are adjusted to optimize the model to study the correlation between components and efficacy, and reveal the influence of each component on the overall efficacy.

[0019] As a preferred solution, the above-mentioned method based on UPLC-Q-TOF / MS E The method for fuzzy identification of bioactive ingredients of different efficacy of traditional Chinese medicine by BP neural network association analysis with integrated effect also includes establishing a mouse model and determining the efficacy index to obtain the integrated effect value of different efficacy of traditional Chinese medicine between step (3) and step (4); the efficacy index includes constipation model, blood stasis syndrome evaluation and cholestasis evaluation.

[0020] Beneficial effect: The present invention uses UPLC-Q-TOF / MS E The BP-ANN correlation analysis between the effects of rhubarb and integration was performed, and a generally applicable fuzzy identification strategy for bioactive ingredients with different effects in traditional Chinese medicines such as rhubarb was proposed. Based on the fuzzy chemical identification method, the present invention classified more than 100 ingredients into ten categories, and then established three mouse models to explore the five effects recorded in the Chinese Pharmacopoeia. The results showed that bound anthraquinones, flavanols and their polymers may be the common characteristics of rhubarb with multiple effects, while diphenylethylene glycosides, anthraquinones and their dimers, free anthraquinones, chromones, gallic acid and gallic tannins, phenylbutyric acid and its glycosides are the different characteristics of rhubarb. These results will provide a scientific basis for expanding the clinical application scope of rhubarb and further developing related products. This method can also be used to study the bioactive ingredients of other traditional Chinese medicines or natural functional products, and has strong versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1The total ion maps and PCA maps of 28 mixed standards and rhubarb samples are shown.

[0022] Figure 2 The network diagram of the compound group with high content of anthraquinones (A), flavanols and their polymers, and gallic acid and gallic tannins (B) as examples (the circles are the molecular weights determined by the molecular ion peaks in the positive and negative ion modes; the boxes are the numbers of the components in the rhubarb sample, and the red ones are chemically

[0023] Figure 3 The E1 index chart is used as a control, and the first black stool time (a1), black stool number (a2), stool weight within 12 hours (a3) ​​(b1), colon and stomach organ coefficient after dissection (b2), mouse serum mtl (c1), ss (c2), vip (c3), and duodenal tissue pain (d). (n=8, mean ± sd) confirmed ingredients).

[0024] Figure 4 Figure 2 is the E2 index: mouse serum TG (a), duodenal tissue Na+-K+-ATPase (b1), TNF-(b2), IL-1 (b3), expressed protein bands (c) and the corresponding grayscale ratios of p-NFBP65 (c1), NF-B p65 (c2), p-p38 (c3), p38 (c4), p-ERK (c5), (c6), p-JNK (c7), (c8), TL4 (c9) in colon tissue of each group (n=8, mean±SD).

[0025] Figure 5 E3 index diagram: the second measurement of rectal temperature in mouse serum (a); hsp-70 (b1), sod (b2), no (b3), tlf-(b4), il-1 (b5), il-6 (b6); representative exposed protein bands (c) and corresponding grayscale ratios of colon p-nfb-p65 (c1), -BP65 (c2), p-p38 (c3), p38 (c4), p-erk (c5), jnk (c6), p-(c7), 4 (c8) (n=6, mean±SD).

[0026] Figure 6Figure 4 is the E4 index: TT (A1), PT (A2), APTT (A3), FIB (A4), TXB2 (B1), 6-keto-PGF1 (B2), the ratio of TXB2 to 6-keto-PGF1 (B3), PGE2 (B4), ET-1 (B5), MG2+ (B6), and CA2+ (B7) in mouse plasma (n=6, mean±SD).

[0027] Figure 7 Figure 3. Cholestasis before and after morphological dissection (a), weight change, representative serum color (b), liver and gallbladder (c) after morphological dissection (n = 8, mean ± sd).

[0028] Figure 8 This is the E5 index chart: liver organ coefficient (A1), gallbladder coefficient (A2); T-SOD (B1), MDA (B2), GSH (B3), Fe3+ (B4); serum ALT (C1), AST (C2), ALP (C3), Fe3+ (C4), GST (C5), GGBIL (C6), TBILI (C7), DBIL (C8), TBA (C9). (n=8, mean±SD). DETAILED DESCRIPTION

[0029] The present invention is further described below by means of specific examples, but those skilled in the art should be aware that the examples described herein do not limit the present invention in any way.

[0030] Example 1

[0031] 1. Experimental materials and reagents. Anhydrous ethanol (analytical grade ≥99.7%), methanol (chromatographic grade), and dimethyl sulfoxide (DMSO, ≥99.9%) were purchased from Merck. Lipopolysaccharide (LPS, L2880) was from Sigma. -Naphthalene isothiocyanate (ANIT, E1909091) was from Aladdin. White sugar (20180404), Yili full-fat sweet milk powder (20180409), Jinluo edible lard (20180317), Yujiangyuan extra virgin olive oil (20191008), and epinephrine hydrochloride (Adr, veterinary 1 mg / mL, 20190307) were purchased from the market. Drugs include cisapride tablets (20180518, Anlikang), compound danshen dripping pills (170912, Tasly), ursodeoxycholic acid capsules (L19001A, Kangzhe), and 28 rhubarb chemical reference substances (purity ≥ 98%, Nanjing Jinyibai Biotechnology Co., Ltd.). Mouse enzyme-linked immunosorbent assay (ELISA) kit indicators include MTL, SS, VIP, AchE, TG, Na + -K +-ATPase, TNF-, IL-1, IL-6, HSP-70, (T-)SOD, NO, TXB2, 6-keto-PGF1, PGE2, ET-1, Mg 2+ , Ca 2+ , MDA, GSH, ALT, AST, ALP, GST, GGT, TBIL, DBIL and TBA were purchased from Shanghai ELISA Biotechnology Co., Ltd. The coagulation test kits of TT, PT, APTT and FIB were from Taizhou Zhongqin Shidi Biotechnology Co., Ltd. 3+ The assay kit (colorimetric method, A039-1-1, A039-2-1) was purchased from Nanjing Jiancheng Bioengineering Institute.

[0032] 2. Main instruments: EnSpire (Molecular Devices), protein electrophoresis instrument (PowerPac, BioRad), gel imager (#1708195, BioRad), platelet aggregation and coagulation factor analyzer (LG-PABER-1, Zhongqin Shidi), UHPLC Acquity TM System coupling Synapt TM Q-TOF mass spectrometer (Waters Corp.), etc.

[0033] 3. Preparation of rhubarb solution

[0034] A rhubarb sample library was designed from two aspects: 9 different concentrations of ethanol-water (EW) solvents and 2 extraction times. Short time (S) means the first decoction was 10 minutes and the second was 5 minutes; long time (L) means the first was 120 minutes and the second was 80 minutes. A total of 18 rhubarb groups were set up: water-S, water-L, 10% EW-S, 10% EW-L, 20% EW-S, 20% EW-L, 35% EW-S, 35% EW-L, 50% EW-S, 50% EW-L, 65% EW-S, 65% EW-L, 80% EW-S, 80% EW-L, 90% EW-S, 90% EW-L, ethanol-S and ethanol-L. The highest dose of rhubarb specified in the Chinese Pharmacopoeia was selected to be converted into a mouse dose of 1.95g / kg. Preparation process: accurately weigh the rhubarb powder that has passed the No. 4 sieve of the pharmacopoeia, fully soak it with 2 times the amount of extraction solvent, then add 8 times the amount of the corresponding solvent and reflux and boil twice. Filter while hot, combine the filtrate, and concentrate it at 50℃ by rotary evaporation until there is no alcohol taste, and finally dilute it with water to a suspension of 0.195g / mL for use.

[0035] Before rotary evaporation, 1 mL of each drug solution was accurately pipetted and the corresponding solvent was added to a concentration of 0.03 g / mL. 100 μL of each sample was mixed as the quality control (QC) sample of rhubarb.

[0036] 4. Preparation of reference solution:

[0037] Weigh 1 mg of each of the 28 chemical reference substances accurately, add a small amount of DMSO to dissolve, then add methanol to 1 mL, and completely dissolve into a transparent solution by ultrasonication. Take 100 μL of each mother solution and dilute it with 1 mL of methanol, and then take 100 μL of the dilution solution and mix it as a mixed standard. All the above samples are filtered through a 0.22 μm microporous membrane and placed in a HPLC glass vial, and stored at 4°C for testing.

[0038] 5. UPLC-Q-TOF / MS E analyze:

[0039] The rhubarb sample solution and the reference solution in step (1) were respectively injected into an ultra-high performance liquid chromatograph for analysis. The mass spectrometer scan range was m / z 100-1000 in positive and negative ESI modes, the acquisition time was 0.1 s, the ion source temperature was 120°C, the desolvation nitrogen temperature was 400°C, the flow rate was 600 L / h, the cone gas flow rate was 50 L / h, the collision gas flow rate was 0.15 mL / min, the capillary voltage was 3 kV in positive ion mode and 2.5 kV in negative ion mode, the cone voltage was 30 V, the ion energy was 1 V, and the nebulizer pressure was 40 psi. E In this mode, energy collisions at 6eV and 20-60eV simultaneously generate fragments; Leucine-enkephalin (ESI + :m / z 556.2771,ESI - : m / z 554.2615) was used as mass calibration solution for simultaneous accurate mass calibration. Mobile phase A was 0.1% formic acid in water, B was acetonitrile, and the gradient elution program was: 0-2 min, 3-10% B; 2-5 min, 10-15% B; 5-8 min, 15% B; 8-11 min, 15-23% B; 11-16 min, 23-30% B; 16-20 min, 30-60% B; 20-24 min, 60-80% B; 24-25 min, 80-3% B. The UV detection wavelength range was set to 190-420 nm. At the beginning of the analytical batch, the QC rhubarb sample was injected five times continuously to balance the system. Then 18 rhubarb samples were analyzed in sequence, each in parallel three times. The corresponding blank solvents and QC samples were regularly inserted to evaluate the reliability and repeatability of the instrument status in the entire workflow. Under the same conditions, 28 rhubarb reference solutions and mixed standard samples were tested in sequence.

[0040] 6. Fuzzy chemical identification

[0041] Through the fuzzy chemical recognition method, according to the peak order, fragmentation information and cleavage pathway of different categories of components, the rhubarb full spectrum information database in Table 1 was established (Table S1), and the corresponding compound group network was established through the rhubarb full spectrum information database to identify unknown components;

[0042] Through the fuzzy chemical identification method, the ingredients of traditional Chinese medicine can be quickly identified. The core of this method is to classify uncertain ingredients, because compounds with the same parent nucleus have similar action properties and interaction laws. This method not only avoids some defects such as the complex identification process of a large number of ingredients and the lack of chemical reference substances, but also reflects the fuzzy and holistic view of traditional Chinese medicine.

[0043] 7. Mouse model and efficacy index determination

[0044] BALB / c mice, weighing 22±2g, were purchased from Shanghai Sippr-BK Laboratory Animal Co., Ltd. (License No.: SCXK(Hu)2018-0006). Mice were free to eat regular feed and drink water. Before modeling, they were stratified and marked according to body weight and randomly divided into 21 groups, including a control group, a model group, a positive drug group, and 18 rhubarb administration groups.

[0045] 7.1 Evaluation of Constipation Models E1 and E2

[0046] This food accumulation and gastrointestinal heat accumulation constipation model refers to the existing method (Chen, JQ; Li, DW; Chen, YY; Tao, HJ; Pu, ZJ; Zhang, J.; Tan, YJ; Shi, XQ; Yue, SJ; Zhou, GS; Tang, YP; Duan, JAJ Ethnopharmacol. 2019, 238, 111868.). The positive drug is cisapride tablets for the treatment of gastrointestinal motility diseases, and the maximum dose converted to mice is 3.9 mg / kg. We divided 168 male mice (No. 20180006004662) into 8 groups on average. From the first day of modeling, each mouse was given 0.8 mL of homemade high-calorie feed (mixed with white sugar, milk powder and lard) by gavage, 3 times a day, 6 hours apart each time; the modeling was continued for 7 consecutive days, but on the sixth and seventh days, it was changed to twice a day. The positive drug and rhubarb administration groups were given the corresponding drug solution (0.01mL / g) in the morning of the sixth to eighth days, while the control and model groups were gavaged with an equal dose of water at the same time. 30 minutes after administration on the eighth day, all mice were gavaged with India ink (0.01mL / g) and immediately placed in metabolic cages to observe defecation characteristics, including the time of the first black stool, the number of black stools, and the total weight of feces in 12 hours. On the ninth day, the mouse serum was collected after weighing, and then the duodenum 0.5cm below the pylorus, stomach, colon and its contents were dissected. MTL, SS, VIP, TG in mouse serum and AchE and Na in duodenal tissue + -K + -ATPase, TNF-, and IL-1 biochemical index levels were detected by ELISA, and the expressions of common inflammatory proteins such as p-NF-B p65, NF-B p65, p-p38, p38, p-ERK, ERK, p-JNK, JNK, and TLR4 in the colon were detected by WB.

[0047] 7.2 Evaluation of Blood Stasis Syndrome E3 and E4

[0048] LPS was used to simulate the process of exogenous virus infection. The mouse modeling dose was 5 mg / kg, which was prepared by 0.9% normal saline (NS). Veterinary Adr injection can help achieve the effect of blood stasis in a short time. This model uses compound Danshen dripping pills, which are commonly used in clinical practice to regulate qi and promote blood circulation, as a positive drug (the maximum dose for mice is 105.3 mg / kg). 252 female mice (No.20180006007816) were divided into 12 mice per group. On the first day of modeling, except for the control group injected with an equal dose of 0.9% NS, all mice were intraperitoneally injected with LPS 0.01 mL / g for the first time and the rectal temperature was measured on time 4 hours later. If the body temperature is abnormal, the modeling is successful; the corresponding liquid is given once every night (0.01 mL / g, the control and model groups are given equal doses of water) for 6 consecutive days. On the fifth day, the mice were intraperitoneally injected with LPS 0.01 mL / g again, and the rectal temperature was measured 4 hours later. On the sixth day, Adr 0.005mL / g was injected subcutaneously twice with an interval of 4h; the control group was injected with 0.9% NS accordingly. After fasting for at least 12h, the mice were weighed and killed on the seventh day. Six mice were randomly selected from each group for orbital blood sampling, and 3.2% sodium citrate anticoagulant (1:9, v / v) was immediately added to the blood. The plasma was centrifuged and the four items of TT, PT, APTT, and FIB coagulation were detected. Secondly, the serum of the remaining mice in each group was collected for ELISA detection of HSP-70, SOD, NO, TNF-, IL-1, IL-6, TXB2, 6-keto-PGF1, PGE2, ET-1, and Mg 2+ and Ca 2+ Then the spleen and thymus of all mice were precisely weighed. In addition, WB was also used to detect the expression levels of the above 9 common inflammatory proteins in colon tissue.

[0049] 7.3 Evaluation of Cholestasis E5

[0050] ANIT was used to establish a cholestatic jaundice model. Through preliminary tests, the mouse modeling dose was determined to be 75 mg / kg. The positive drug in this experiment is ursodeoxycholic acid capsules, which can effectively promote bile secretion when the maximum equivalent dose in mice is 91 mg / kg. Male mice (No.20180006009949) were divided into 8 groups. The positive drug and rhubarb groups were given preventive medication (0.01 mL / g) once a day for 3 consecutive days; the control and model groups were given equal doses of water. On the fourth day, except for the control group (equal doses of olive oil), all mice were gavaged with ANIT 0.01 mL / g dissolved in olive oil, and the corresponding drug solution was given once after 6 hours. The drug was given again on the evening of the fifth day, and fasting was performed at night. On the sixth day (ie 48 hours after ANIT modeling), the mice were weighed, blood was taken to store serum, and the entire liver and gallbladder were accurately weighed respectively. The kit was used to detect T-SOD, MDA, GSH, Fe 3+and serum ALT, AST, ALP, GST, GGT, TBIL, DBIL, TBA, Fe 3+ The liver homogenate was prepared as follows: about 50 mg of liver tissue was added with 0.9% NS (20 μL / mg) and two steel balls, and placed in an automatic grinder for 3 min. The homogenate was placed on ice for 30 min, centrifuged at 4°C, and the supernatant was taken for testing.

[0051] 7.4 BP Neural Network Association Method

[0052] The BP neural network was used to associate the relative content of each component in 18 rhubarb samples with the five integrated effects. It mainly includes three steps. (1) Data normalization: y = (x-MinValue) / (MaxValue-MinValue), where x and y are the values ​​before and after conversion, respectively, and MaxValue and MinValue are the maximum and minimum values ​​in the sample, respectively. (2) Data set division: all component contents are used as input, and the corresponding integrated effect values ​​are used as output; and all groups 1 to 18 data are used as training samples, of which 90% EW-L, ethanol-S and ethanol-L are used as test samples. (3) Quantity-effect association: Based on the BP neural network algorithm, different parameters are adjusted to optimize the model to study the correlation between components and efficacy, revealing the impact of each component on the overall efficacy.

[0053] 8 Statistical analysis. The experimental data were processed by GraphPad Prism 7.0 and SPSS 22.0 and expressed as “mean ± standard deviation (SD)”. One-way analysis of variance or independent sample t-test was used to compare data between groups. A two-sided P value less than 0.05 was considered statistically significant.

[0054] 9 Results and Discussion

[0055] Based on UPLC-Q-TOF / MS E The semi-quantitative and qualitative analysis of rhubarb components by fuzzy chemical identification. The mixed standard samples were basically well separated within 25 minutes, which can provide a reference for the qualitative analysis of various compounds. In the PCA graph, the QC samples were all close to the center of the Hotelling T2 ellipse, while the 18 rhubarb groups were clearly separated, but the same group of repeated sampling was clustered together. This shows that UPLC-Q-TOF / MS E The stability of the method ( Figure 1 ). Combining the retention time (RT) and total ion current of three batches of rhubarb liquor, we screened out 108 unknown components with high responsiveness and different change trends from 18 rhubarb samples. The peak area, i.e. the relative content of all components, was calculated by EZinfo 2.0 software.

[0056] In addition, according to the MS of the reference E The characteristic fragment pattern was matched with the above 108 components and 25 components were confirmed. Referring to the database of the full spectrum of chemical components of rhubarb in Table S1, we constructed a network of compound groups of different categories ( Figure 2 ) to classify the remaining components (Table 1).

[0057] Table 1. Classification of 108 rhubarb components based on fuzzy chemical recognition

[0058]

[0059]

[0060] ▲ Rh-01 to 10 represent free anthraquinone, bound anthraquinone, anthrone and its dimer, flavanol and its polymer, gallic acid and gallotannin, stilbene glycosides, naphthyl glycosides, butyrophenone and its glycosides, chromones, and flavonoid (flavonol) glycosides.

[0061] Evaluation of five integration effects based on factor analysis.

[0062] TCM syndromes often cannot be completely cured by a single therapeutic effect, so we explore two or more therapeutic effects in one animal model to improve research efficiency and fit with actual symptoms. In the first constipation model, the time of the first black stool, the amount of black stool, the total weight of stool in 12 hours, the weight of colon contents, the organ coefficient of colon and stomach, MTL, SS, VIP, AchE can reflect E1( Figure 3 ).TG、Na + -K + The expression levels of -ATPase, TNF-, IL-1 and the grayscale ratio of the 9 colon inflammatory proteins mentioned above can reflect the expression of E2( Figure 4 ).

[0063] The treatment of the second heat toxicity and blood stasis syndrome also includes cooling blood and activating blood circulation. The model mice showed obvious blood stasis symptoms of spleen enlargement and thymus reduction (P<0.001), while the spleen of the drug-treated group recovered to a certain extent. The levels of rectal temperature, HSP-70, SOD, NO, TNF-, IL-1, IL-6 and the same 9 colon inflammatory proteins were measured as E3 indicators ( Figure 5), because from the perspective of Western medicine, E3 and E2 are similar and both play an anti-inflammatory role. The fever response of mice to LPS maintained an opposite low body temperature state (P<0.001), and some drug-treated groups could significantly antagonize this toxic reaction. Compared with the model group, the content of inflammatory factors in most groups was downregulated (P<0.05), especially the phosphorylated protein and TLR4 levels in the rhubarb administration group with high proportion of ethanol extraction (P<0.001). LPS can exert its effect by activating the NF-B signaling pathway mediated by the TLR family associated with proinflammatory cytokines. On the other hand, E4 indicators include TT, PT, APTT, FIB, TXB2, 6-keto-PGF1, the ratio of TXB2 to 6-keto-PGF1, PGE2, ET-1, Mg 2+ and Ca 2+ ( Figure 6 ). The positive drug group showed better improvement in the four coagulation parameters, but most rhubarb groups only had significant corrections in PT and FIB (P<0.05). Alcohol-extracted rhubarb corrected the imbalance of the ratio of TXB2 to 6-keto-PGF1 (P<0.001) and promoted blood circulation. Prostaglandins can dilate blood vessels, while endothelins can constrict blood vessels. They all showed different degrees of correction trends in the drug groups.

[0064] The third cholestasis model is Figure 7 As shown in the results, the model mice had a significant decrease in body weight (P<0.01), the color of serum became darker, the gallbladder was enlarged, the bile was black, and the liver was obviously damaged with dense bleeding spots on the surface. The overall condition of the positive drug group recovered well, and these pathological conditions of some rhubarb-administered groups also improved to a certain extent. Then we studied the measurement indicators related to jaundice, including the liver and gallbladder organ coefficient, T-SOD, MDA, GSH, tissue Fe 3+ , ALT, AST, ALP, serum Fe 3+ , GST, GGT, TBIL, DBIL and TBA to evaluate E5( Figure 8 ). Long-term decocted rhubarb can regulate the level of oxidative stress in liver tissue (P<0.05), which may be one of the main mechanisms of its anti-jaundice effect. In addition to GGT and DBIL, serum liver function levels were generally increased in model mice, and most groups were significantly reduced after administration (P<0.05).

[0065] As shown in Table 2, there were significant differences between the model group and the control group. The therapeutic effects of the positive drugs were very obvious in the five effects, while the efficacy of rhubarb prepared by different methods was different. E1 was significant in the 10% EW-S~35% EW-L and 80% EW-L~ethanol-L groups. The high-concentration alcohol extraction group had a better E2 effect overall. The E3 effect of rhubarb extracted with more than 20% EW was very obvious, especially in the group with long-term decoction, which is in line with the traditional Chinese medicine theory of "making wine, steaming and drying, and clearing the heat and toxins in the upper part of the blood." In contrast, the E4 effect of rhubarb was weaker and was only effective when extracted with water or more than 80% alcohol. E5 was significant in the 10% EW-L~ethanol-L group, among which 35% EW-L, 50% EW-L and 65% EW-L were the best, suggesting that the "water and wine co-decoction method" may be more helpful in exerting the therapeutic effect.

[0066] Table 2. Rhubarb integration effect based on factor analysis (n=6-8, mean±SD).

[0067]

[0068]

[0069] # P<0.05, ## P<0.01, ### P < 0.001 compared with the control group;

[0070] * P<0.05, ** P<0.01, *** P<0.001 compared with the model group.

[0071] The dose-effect correlation method was established based on BP neural network.

[0072] Advanced BP-ANN simulates neurons based on the structure and function of biological brains, and has high prediction accuracy and optimization capabilities. It can be intelligently applied to fitting association analysis to help determine the contribution of each chemical component to the corresponding efficacy. Therefore, the present invention attempts to objectively associate as many detected components as possible with the efficacy. Through repeated debugging of the following parameters, a reasonable dose-effect association method is established.

[0073] (1) Number of hidden layers: In theory, the more hidden layers there are, the stronger the ability to fit the function will be. But in fact, too many layers can easily lead to overfitting and increase the difficulty of training, making it difficult for the model to converge. Under the premise of setting the input layer to 108 nodes and the output to 1 node, we explored the fitting results of 1, 2, and 3 hidden layers. The number of nodes in different hidden layers was randomly set to 10, and the run was repeated 20 times to compare the relative error between the predicted value and the true value.

[0074] (2) The number of neurons in each hidden layer: According to the empirical formula and the setting principle of hidden neurons, within the range of 108 nodes in the input layer to 1 node in the output layer, we first set a value every 20 as the number of neurons (i.e., 20, 40, 60, 80, and 100); then we added 5, 10, and 15 between 0 and 20, 70 and 75 between 60 and 80, and 130, 160, and 190 after 100. The second hidden layer was consistent with the first hidden layer. We ran 5 times with each number of neurons and obtained 5 sets of data for three test samples (Table S2). Considering the predicted value, the true value, and their relative error values, we set the number of neurons to 20, at which time the relative error was relatively stable overall.

[0075] (3) Learning rate: In general, we tend to choose a small adaptive learning rate to ensure the stability of the system, but a learning rate that is too small will result in a long training time. In order to obtain better results, the test samples were run 5 times at learning rates of 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7 and 0.8 (Table S3). The results show that there is no obvious trend change. Relatively speaking, a learning rate of 0.02 can make the gradient descent BP algorithm have better performance.

[0076] (4) Operation effect: We adjusted the display frequency to 100, the number of training times to 1000, and the minimum error of the training target to 0.0000001 to further improve the prediction accuracy of the method. All training samples and the corresponding integrated effect values ​​were substituted into the optimized neural networks. When the mean square error is less than its set value or the gradient reaches the set value, the training is stopped; when the generalization ability verification check error increases instead of decreasing for 6 consecutive times, the training is also forcibly terminated. The results show that their fitting correlation effects are very good.

[0077] (5) Calculation of decision weights: The above neural network training results only reflect the relationship between neurons. It is necessary to clarify the relative importance of each input factor to the output information. Therefore, the decision weights of the input components are calculated using the following neural network learning algorithm.

[0078] (Input layer: n=108, first hidden layer: m=20, second hidden layer: p=20, output layer: l=1)

[0079] The influence of the o-th unit in the second hidden layer on the k-th output relative to all units:

[0080] F ok =|w ok | / |w ok|o=1,2…p,k=1,2…l

[0081] The influence of the jth unit in the first hidden layer on the oth unit in the second hidden layer relative to all units is:

[0082] F jo =|w jo | / |w jo |j=1,2…m,o=1,2…p

[0083] The influence of the i-th unit in the input layer on the j-th unit in the first hidden layer relative to all units:

[0084] F ij =|w ij | / |w ij |i=1,2…n,j=1,2…m

[0085] Therefore, the impact of the i-th input on the k-th output is:

[0086] F ik =F ij *F jo *F ok

[0087] Then the decision weight of the input component can be expressed as: S i =F ik / F ik

[0088] The main bioactive ingredients of the five effects based on the association analysis between ingredients and efficacy. We analyzed the ingredients ranked in the final decision weight table S4 with a contribution greater than 0.01, of which E1 had 30 (Nos. 42>21>35>37>34>46>11>98>36>27>59>88>89>18>108>78>105>72>6>55>106>45>61>51>90>32> 41>95>38>107); E2 has 28 (Nos.42>7>107>83>29>28>61>98>108>8>80>35>58>81>40>15>92>33>106>36>50>79>11>100>57>19>101>39); E3 has 35 (Nos.24>18>12>86>72>37>51>80>26 >55>34>21>16>46>97>42>22>33>88>28>92>27>2>67>70>59>36,44>90>1>82>6>73>4,30); E4 has 32 (Nos.40>81>56>75>69>53>9>64>100>58>49>29>74>83>90>52>63>42>66>106>67 >48>31>62>71>89>87>92>20>60>46>36); E5 has 33 (Nos.19>55>62>12>33>52>22>35>87>64>69>95>59>42>88>46>43>25>107>72>29>104>71>94>89>99>92>81>36>45>58>108>102). Table 3 summarizes the total contribution of each category of rhubarb ingredients.

[0089] Table 3. Total contribution of rhubarb components with weight greater than 0.01 in each category.

[0090]

[0091] ▲ Rh-01 to 10 represent free anthraquinone, bound anthraquinone, anthrone and its dimer, flavanol and its polymer, gallic acid and gallotannin, stilbene glycosides, naphthyl glycosides, butyrophenone and its glycosides, chromones, and flavonoid (flavonol) glycosides.

[0092] The serial number following the numerical value indicates the ranking of each category's contribution in each efficacy.

[0093] Table S1 Database of full spectrum information of chemical components of rhubarb

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133]

[0134] Table S3 Prediction value (p), true value (t) and relative error (r) of different numbers of neurons in each hidden layer

[0135]

[0136]

[0137] Table S4 Prediction value (p), true value (t) and relative error (r) of test samples at different learning rates

[0138]

[0139]

[0140] Table S5 Total weights of rhubarb ingredients (the ordinal numbers after the values ​​represent the 10 largest orders of each effect, and weights < 0.01 are ignored by “-”) The decision weights for rhubarb ingredients are greater than 0.01.

[0141]

[0142]

[0143]

Claims

1. A UPLC-Q-TOF / MS-based E The method for fuzzy identification of biologically active components of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effect is characterized in that: The following steps are involved: (1) Preparation of rhubarb sample solution and chemical reference substance: Rhubarb sample solution: Preparation process: accurately weigh multiple portions of rhubarb powder, first fully soak them with ethanol aqueous solutions of different concentrations, then add ethanol aqueous solutions of different concentrations to reflux and boil, filter while hot, combine the filtrates, and concentrate under reduced pressure to obtain ethanol extracts of different concentrations for later use; Reference solution: accurately weigh 28 chemical reference substances: Chrysophanol, Chrysophanol1-ObD-glucoside, Chrysophanol 8-ObD-glucoside, Emodin, Emodin 1-ObD-glucoside, Emodin 8-ObD-glucoside, Physcion, Physcion 1-ObD-glucoside, Physcion 8-ObD-glucoside, Aloe-emodin, Aloe-emodin 8-ObD-glucoside, Rhein, Rhein 8-ObD-glucoside, Sennoside A, Sennoside B, Sennoside C, Cianidanol, (-)-Epicatechin, (-)-Epicatechin gallate, Procyanidin B2, Gallic acid, Resveratroloside, Rhapontigenin, Torachrysone 8-O-glucoside, 4-(4-Hydroxyphenyl)-2-butanone, Raspberryketone glucoside, Lindleyin, and 5-Acetyl-7-hydroxy-2-methyl-chromone were respectively dissolved with DMSO, and then methanol was added and completely dissolved into a transparent solution by ultrasonication to obtain 28 chemical reference substance mother solutions; then, the 28 chemical reference substance mother solutions were mixed to obtain a mixed reference substance solution; (2) UPLC-Q-TOF / MS E analysis: The rhubarb sample solution and the reference solution in step (1) were respectively injected into an ultra-high performance liquid chromatograph for analysis, and the mass spectrometer scan range was m / z 100-1000 in positive and negative ESI modes, the acquisition time was 0.1 s, the ion source temperature was 120°C, the desolvation nitrogen temperature was 400°C, the flow rate was 600 L / h, the cone gas flow rate was 50 L / h, the collision gas flow rate was 0.15 mL / min, the capillary voltage was 3 kV in positive ion mode and 2.5 kV in negative ion mode, the cone voltage was 30 V, the ion energy was 1 V, the nebulizer pressure was 40 psi, and the MS E In this mode, energy collisions at 6eV and 20-60eV simultaneously generate fragments; ESI of leucine-enkephalin + :m / z556.2771,ESI - :m / z 554.2615, used as mass calibration liquid for simultaneous accurate mass calibration; mobile phase A is 0.1% formic acid water, B is acetonitrile, gradient elution program: 0-2min, 3-10% B; 2-5min, 10-15% B; 5-8min, 15% B; 8-11min, 15-23% B; 11-16min, 23-30% B; 16-20min, 30-60% B; 20-24min, 60-80% B; 24-25min, 80-3% B; UV detection wavelength range is set to 190-420nm; (3) Fuzzy chemical identification: Through fuzzy chemical recognition method, a database of rhubarb spectrum information was established according to the peak order, fragment information and cleavage pathway of different categories of components. The corresponding compound group network was established through the database of rhubarb spectrum information to identify unknown components. (4) BP neural network association method: The BP neural network was used to associate the relative content of each component in the rhubarb sample solution with the five integrated effects, which mainly included three steps: (1) data normalization: y = (x-MinValue) / (MaxValue-MinValue), where x and y are the values ​​before and after conversion, respectively, and MaxValue and MinValue are the maximum and minimum values ​​in the sample, respectively; (2) data set division: all component contents were taken as input, and the corresponding integrated effect values ​​were taken as output; and all groups 1 to 18 data were used as training samples, of which 90% EW-L, ethanol-S and ethanol-L were used as test samples; (3) dose-effect association: based on the BP neural network algorithm, different parameters were adjusted to optimize the model to study the correlation between components and efficacy, and reveal the impact of each component on the overall efficacy.

2. A method based on UPLC-Q-TOF / MS according to claim 1. E The method for fuzzy identification of biologically active components of traditional Chinese medicine with different effects by BP neural network association analysis with integrated effect is characterized in that: Between step (3) and step (4), a mouse model is established and efficacy indexes are determined to obtain the integrated effect values ​​of different efficacy of traditional Chinese medicine; the efficacy indexes include constipation model, blood stasis evaluation and cholestasis evaluation.

Images