Systematic method for predicting and verifying cold-heat property of traditional Chinese medicine
Through the combination of random forest model and micro-calorimeter, the SMILES structure and biothermal parameters of traditional Chinese medicine compounds were analyzed, and the modern scientific support problem of insufficient cold and heat attribute determination of traditional Chinese medicine was solved, and efficient and accurate identification of the cold and heat properties of traditional Chinese medicine was achieved.
Patent Information
- Application Number
- CN202510691381.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-27
- Publication Date
- 2025-09-02
AI Technical Summary
The determination of the cold and heat properties of traditional Chinese medicines in the prior art mainly relies on ancient books or medical experience. Without the support of modern scientific mechanisms, it is difficult to achieve standardized and efficient identification of cold and heat properties of traditional Chinese medicines.
The random forest model was used to combine with the micro-calorimeter to analyze the SMILES structure and biothermal parameters of traditional Chinese medicine compounds, and a prediction model for the cold and heat properties of traditional Chinese medicine was established. The compounds were screened using OB, Caco-2 permeability and DL, and the biothermal changes in bacteria were detected in combination with the micro-calorimeter, and the biothermal parameters that distinguish the cold and heat properties were summarized.
The standardization and accuracy of the identification of cold and heat properties of traditional Chinese medicine has been achieved, subjective judgments of manual experience have been avoided, and the efficiency and accuracy of the identification of cold and heat properties of traditional Chinese medicine has been improved.
Smart Images
Figure CN120581099A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the application of artificial intelligence and machine learning to complex network prediction of traditional Chinese medicine properties, and in particular to a method for predicting the cold and hot properties of traditional Chinese medicine based on a random forest model and the use of a microcalorimeter to summarize the effects of cold and hot traditional Chinese medicines on bacterial biothermodynamic parameters. The method can be used to classify the cold and hot properties of traditional Chinese medicines and study the changes in the properties of traditional Chinese medicines before and after preparation. Background Art
[0002] The theory of the Four Qi and Five Flavors of Chinese Medicine (TCM), including the cold and heat properties, is a fundamental theory of Traditional Chinese Medicine (TCM). It was developed based on extensive clinical practice and combined with its efficacy, and has served as the theoretical basis for the clinical application of TCM for thousands of years. Clinically, the cold and heat properties of TCM are the primary properties of TCM, and cold and heat syndrome differentiation is the primary basis for TCM differentiation. "Treating cold with hot herbs, treating heat with cold herbs" is a fundamental principle of TCM treatment. However, current methods for determining the cold and heat properties of TCM are largely based solely on ancient texts or the personal clinical experience of physicians, lacking modern scientific support. There is an urgent need to develop a prediction method for TCM cold and heat properties that integrates traditional TCM theory with modern bioinformatics to address the challenges of determining and distinguishing these properties. Modern research on TCM property theory has focused on the relationship between the cold and heat properties of TCM and its chemical composition. The primary basis for TCM properties is efficacy, which in turn is based on its chemical composition. Cold-property TCMs regulate heat syndromes, while hot-property TCMs regulate cold syndromes, and these herbs should possess corresponding chemical compositions. Therefore, the study of the cold and hot properties of traditional Chinese medicine needs to start with the analysis and characterization of the material components of traditional Chinese medicine, and characterize the material components of traditional Chinese medicine through traditional Chinese medicine inorganic substances, primary substances of traditional Chinese medicine, traditional Chinese medicine image properties and efficacy indications, genomics, proteomics, metabolomics, three-dimensional fluorescence spectroscopy, molecular descriptors of traditional Chinese medicine compounds and chemical fingerprint maps of traditional Chinese medicine. Summary of the Invention
[0003] Purpose of the invention: The purpose of the present invention is to provide a method for predicting the cold and hot properties of traditional Chinese medicine based on a random forest model. On the basis of model prediction and verification, a microcalorimeter is used to detect the effects of traditional Chinese medicine with cold and hot properties on the biothermodynamic changes of bacteria. The biothermodynamic parameters for distinguishing the cold and hot properties of traditional Chinese medicine are summarized to solve the problems existing in the above-mentioned prior art.
[0004] Technical solution: The traditional Chinese medicine cold and hot property prediction model based on the random forest model of the present invention comprises the following steps:
[0005] Step S1: Searching for Chinese medicines with "cold and hot" properties in the Pharmacopoeia of the People's Republic of China to screen the Chinese medicine dataset with cold and hot properties;
[0006] Step S2: Search the database for compounds corresponding to cold Chinese medicines and hot Chinese medicines to establish cold Chinese medicine and hot Chinese medicine compound datasets. Then, delete duplicate compounds from the two datasets. Then, search the PubChem database for the SMILES structures of key compounds of cold Chinese medicines and hot Chinese medicines to finally obtain the cold and hot Chinese medicine datasets.
[0007] Step S3: performing vector expression processing and weighting on the collected SMILES structural formulas of the traditional Chinese medicine compounds; using MACCSkeys to convert the SMILES structures of the compounds in the two data sets into a binary language that can be recognized by a computer; preferably, the final calculation is performed into 169×206-dimensional vector data;
[0008] Step S4: Classify and train the compound structure data of traditional Chinese medicine using the random forest algorithm, and then establish a model for distinguishing the cold and hot properties of traditional Chinese medicine.
[0009] The step S1 includes the following specific steps: selecting a Chinese medicine whose properties are clearly defined as "cold" or "hot" in the pharmacopoeia.
[0010] The step S2 includes the following specific steps: querying and screening compounds of Chinese medicines with cold and hot properties through the TCMSP, TCMID and TCM Bank databases, establishing a cold Chinese medicine compound data set and a hot Chinese medicine compound data set respectively, and then arranging the compound information of the two data sets according to the format in the TCMSP database; deleting duplicate compound data that exists in both data sets (because duplicate Chinese medicine compounds will interfere with model construction), and finally obtaining cold Chinese medicine and hot Chinese medicine compound data sets; using three methods for screening Chinese medicine compounds, namely oral bioavailability (OB), Caco-2 permeability and drug likeness (DL), to screen the SMILES structures of the included cold Chinese medicine and hot Chinese medicine compounds, and finally obtaining the cold and hot Chinese medicine compound data sets. Since the importance of each Chinese herbal compound component to each Chinese herbal medicine varies and their clinical application frequency is also different, in order to improve the accuracy of the cold and hot properties of Chinese herbal medicine discrimination model established later, oral bioavailability (OB), Caco-2 permeability and drug likeness (DL) were selected.
[0011] The step S3 includes the following specific steps: using RDkit to generate a Morgan fingerprint for the SMILES representation of each drug, and expressing the Morgan fingerprint as a binary vector to obtain a molecular fingerprint expression vector of the molecule, and calculating the traditional Chinese medicine fingerprint feature expression with oral availability, Caco-2 permeability and drug-like properties, and then using a normalization method to summarize the statistical distribution of the unified sample.
[0012] Preferably, the step S3 comprises the following specific steps: using RDkit to generate a Morgan fingerprint for the SMILES representation of each drug, and representing the Morgan fingerprint as a binary vector to obtain the N-dimensional molecular fingerprint expression vector of the molecule: X′=(x 1 , x 2 , x 3 ,Λ,x n ); The three molecular attributes OB, Caco-2 permeability and DL are represented by a, b, and c respectively, and multiplied with each element in X' to obtain the new expression vector of molecule X: X'abc = (x 1 abc,x 2 abc,x 3 abc,Λ,x n abc); for the expression and normalization of Chinese medicine fingerprint features, assume that each Chinese medicine W contains m Chinese medicine active molecules, then the expression vector of each active molecule in the Chinese medicine W is Where j = (1, 2, 3, Λ, m), the expression vector of traditional Chinese medicine W is: W = (w1, w2, w3, Λ, w n ),in After completing the expression of the fingerprint characteristics of traditional Chinese medicine compounds, since the number of traditional Chinese medicine compounds contained in each traditional Chinese medicine is different, it is necessary to use the normalization method to summarize the statistical distribution of the unified sample. The fingerprint characteristics of the traditional Chinese medicine after normalization are: in
[0013] The step S4 includes the following specific steps: using 15-25% of the data as a training set and 75-85% of the data as a test set for prediction, and creating a random forest regression model classifier; using 15-25% of the data as a training set and 75-85% of the data as a test set for prediction, and creating a support vector machine model classifier.
[0014] A prediction method using the prediction model is provided, wherein the SMILES structure weighted data of the compound in the Chinese medicine to be tested is input, and the calculated output is 1 for a cold Chinese medicine, and the calculated output is 0 for a hot Chinese medicine.
[0015] The method for verifying the prediction results of the prediction model uses a microcalorimeter to detect the effect of traditional Chinese medicine on the biothermodynamic changes of bacteria, thereby summarizing and distinguishing the biothermodynamic parameters of hot and cold traditional Chinese medicines, including the following steps:
[0016] Step 1: Inoculate the strain in the exponential growth phase into the culture medium;
[0017] Step 2: Add the culture medium solution containing the strain and water into the bottle, seal it, and obtain a blank solution; preferably, use an ampoule bottle.
[0018] Step 3: Add the culture medium solution containing the strain and the Chinese herbal medicine extract culture medium solution into the bottle, seal it, and obtain a sample solution; the volume of the sample solution is the same as that of the blank solution in step 2; the drug extract culture medium solution is prepared by dissolving the Chinese herbal medicine freeze-dried powder in a sterile culture medium solution; preferably, an ampoule bottle is selected.
[0019] Step 4: Place the blank solution and the sample solution in a microcalorimeter at 34-39° C., respectively, and record the thermal power-time (Pt) curve of the strain growth process. The test is terminated when the thermal spectrum curve returns to the baseline again, thus obtaining the bacterial growth metabolic thermal spectrum curve after the intervention of traditional Chinese medicine;
[0020] Step 5: After recording, extract the experimental time and thermal power data. Use Origin software to plot the time as the horizontal axis and the power as the vertical axis to obtain the Pt curve and calculate the total calorific value Q.
[0021] Step 6: Calculate the biothermodynamic parameters based on the Pt curve. Preferably, the biothermodynamic parameters are calculated using the following formula (1):
[0022] ln P=ln P0+kt (1)
[0023] Where: P is the heat production power corresponding to time t, P0 is the heat production power corresponding to time t=0, and k is the exponential growth rate.
[0024] Preferably, k1 (growth rate in the first exponential growth phase) and P1 (heat production power at the first exponential growth phase) are used as control indicators to distinguish between cold Chinese medicine and hot Chinese medicine. Escherichia coli selects either k1 or P1 as an indicator, where k1 is 2×10 -4 As defined, P1 is 1×10 -3 The value below the limit is cold, and the value above the limit is hot. For Staphylococcus aureus, either k1 or P1 is selected as the indicator. K1 is 9×10 -5 As defined, P1 is 4×10 -3 The value below the limit is cold, and the value above the limit is hot.
[0025] The strains in step 1 are Escherichia coli and Staphylococcus aureus.
[0026] Preferably, the culture medium in step 1 and step 2 is LB culture medium; more preferably, LB culture medium: 10 g peptone, 5 g yeast extract, 10 g NaCl, dissolved in 1000 mL distilled water, pH = 7.2.
[0027] The relative OD value of the culture medium solution containing the strain in step 3 is 0.2-0.4; the drug concentration of the traditional Chinese medicine extract culture medium solution is 5-15 mg / mL.
[0028] The Chinese medicine extract culture medium solution in step 3 is prepared from Chinese medicine freeze-dried powder, which is obtained by extracting the medicinal materials with water, filtering the obtained filtrate, and concentrating and drying.
[0029] The preparation process of the Chinese medicine freeze-dried powder is as follows: 20-30g of medicinal materials are weighed and placed in a container, and water is used as the extraction solvent with a solid-liquid ratio of 1:8-12 for heating and reflux extraction. The extraction is performed 1-5 times, each time for 0.5-1.5 hours. After filtering, the filtrate is concentrated under reduced pressure and vacuum dried at low temperature to prepare the freeze-dried powder.
[0030] Beneficial effects: The present invention uses OB, Caco-2 permeability and DL three indicators to screen Chinese medicine compounds and combines the SMLIES structural formula of Chinese medicine compounds to perform vector weighting as the overall characterization of Chinese medicine at the molecular level. The correlation between different cold and hot medicinal properties of Chinese medicine and its compounds is studied through feature screening. Based on this data, a cold and hot medicinal property identification prediction model is constructed in combination with informatics machine learning methods. The cold and hot medicinal property identification prediction model is evaluated based on the model evaluation parameter index to obtain a better cold and hot medicinal property identification prediction model as a Chinese medicine cold and hot medicinal property identification model to predict and analyze the medicinal properties of the Chinese medicine samples to be identified. On the basis of the model prediction results, the present invention uses a microcalorimeter to detect the biothermodynamic parameters of Chinese medicines with cold and hot medicinal properties, and summarizes the biological indicators that can distinguish between Chinese medicines with cold and hot medicinal properties. The idea of identifying cold and hot medicinal properties is consistent with the holistic view of traditional Chinese medicine. The method and indicators are objective and specific. They can replace the traditional subjective judgment method of cold and hot medicinal properties of Chinese medicines that relies on manual experience, realize standardized identification and judgment of cold and hot medicinal properties of Chinese medicines, improve the efficiency and accuracy of identifying cold and hot medicinal properties of Chinese medicines, and avoid generalizing. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 Schematic diagram of the prediction of the cold and hot medicinal properties of traditional Chinese medicine by random forest model and verification by microcalorimetry;
[0032] Figure 2ROC curves for the random forest prediction model and the support vector machine model for the test set data (A: ROC curve for the random forest prediction model; B: ROC curve for the support vector machine prediction model);
[0033] Figure 3 The accuracy of the random forest model and support vector machine model on the training set and test set (A: confusion matrix of the random forest prediction model; B: confusion matrix of the support vector machine prediction model);
[0034] Figure 4 Prediction of the cold and hot properties of traditional Chinese medicine by random forest model and support vector machine model (A: verification results of random forest prediction model; B: verification results of support vector machine prediction model). DETAILED DESCRIPTION
[0035] The technical solution of the present invention is further described below with reference to the accompanying drawings. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system, such as a set of computer-executable instructions, and that, although a prediction model is shown in the flowcharts, in some cases, the steps shown or described can be performed using a machine learning model different from the one shown here.
[0036] Example 1
[0037] With the development of bioinformatics, the research methods of the composition and efficacy of traditional Chinese medicine are becoming increasingly rich, providing a new way to reveal the material basis of the efficacy of traditional Chinese medicine. In classification research, traditional single-classification models are often not very accurate and are prone to overfitting problems. Effectively integrating classifiers and concentrating their advantages to improve prediction accuracy solves many problems that single classifiers cannot solve or are difficult to obtain effective results. The random forest model uses random resampling technology and node random splitting technology to construct multiple decision trees, and obtains the final classification results through voting. A large number of theoretical and empirical studies have proved that it has a high prediction accuracy, good tolerance to outliers and noise, and is not prone to overfitting. Therefore, the present invention uses a random forest model to predict the cold and hot medicinal properties of traditional Chinese medicine compounds using the SMILES structural formula, as shown in the schematic diagram. Figure 1 .
[0038] 1 Experimental methods
[0039] 1.1 Screening of Chinese medicines with cold and hot properties
[0040] Select Chinese herbal medicines with "cold or hot" properties through the Pharmacopoeia of the People's Republic of China. The selection principles include: the pharmacopoeia clearly stipulates that the properties of Chinese herbal medicines are "cold" or "hot".
[0041] 1.2 Data screening of traditional Chinese medicine compounds with cold and hot properties
[0042] Compounds of TCMs with cold and hot properties were screened using the TCMSP, TCMID, and TCM Bank databases. Compound information from both datasets was then arranged according to the format used in the TCMSP database. Duplicate TCM compounds could interfere with model building, so duplicate compound data was removed from the database to obtain the final dataset of TCM compound components for both cold and hot properties.
[0043] 1.3 Screening of SMILES structural data of traditional Chinese medicine compounds with cold and hot properties
[0044] SMILES structures were retrieved through the PubChem database. Since the importance of each Chinese medicine compound component to each Chinese medicine is different and the frequency of their clinical application is also different, in order to improve the accuracy of the cold and hot property discrimination model of Chinese medicine established later, three methods of screening drug compound components, OB, Caco-2 permeability and DL, were used to screen the SMILES structures of the included cold and hot Chinese medicine compounds, and finally a SMILES structure dataset of Chinese medicine compounds with cold and hot properties was obtained.
[0045] 1.4 Data Preprocessing
[0046] The collected SMILES structural formulas of traditional Chinese medicine compounds are processed and weighted. The molecular fingerprint of the SMILES structure is analyzed using the MCCSkeys key on the RDKit chemical information platform for each drug compound to obtain the N-dimensional molecular fingerprint expression vector of the molecule: X′=(x 1 , x 2 , x 3 ,Λ,x n ), then the three molecular attributes OB, Caco-2 permeability and DL are represented by a, b, and c respectively, and multiplied with each element in X′ to obtain the new expression vector of molecule X: X′abc=(x 1 abc,x 2 abc,x 3 abc,Λ,x n Secondly, it is necessary to perform fingerprint feature expression and normalization of traditional Chinese medicine. Assuming that each traditional Chinese medicine W contains m active molecules, the expression vector of each active molecule in the traditional Chinese medicine W is Where j = (1, 2, 3, Λ, m). Then the expression vector of Chinese medicine W is: W = (w1, w2, w3, Λ, w n ),in After completing the expression of the fingerprint characteristics of traditional Chinese medicine compounds, since the number of traditional Chinese medicine compounds contained in each traditional Chinese medicine is different, it is necessary to use the normalization method to summarize the statistical distribution of the unified sample. The fingerprint characteristics of the traditional Chinese medicine after normalization are: in
[0047] 1.5 Modeling
[0048] Because SMILES structural formulas represent drug compounds, MACCSkeys was used to convert the compound structures associated with cold and hot properties into computer-recognizable small molecule structures. Based on the encoding properties of the SMILES structure, the structural formulas of drug compounds can be converted to binary form. Pandas is a tool based on NumPy, created for data analysis tasks. Therefore, Pandas was used for data preprocessing, converting compound expressions into binary characters of 0 and 1, expressed as MACCSkeys. The released version contains 166 keys, each corresponding to a specific molecular feature. Therefore, RF was used to classify the compound dataset for each Chinese herbal medicine with cold and hot properties, and the voting results were finally obtained, which can distinguish the corresponding cold and hot properties of the Chinese herbal medicine.
[0049] 2 Experimental results
[0050] The results of screening Chinese medicines with cold and hot properties through the Pharmacopoeia of the People's Republic of China are shown in Table 1:
[0051] Table 1 Data of Chinese medicines with cold and hot properties
[0052]
[0053] The TCMSP, TCMID, and TCM Bank databases were used to search for compounds corresponding to 108 hot Chinese medicines and compounds corresponding to 98 cold Chinese medicines. The SMILES structural formulas corresponding to the compounds were retrieved in the Pubchem database. A total of 1,324 compounds were finally obtained and divided into cold Chinese medicine datasets and hot Chinese medicine datasets based on the source Chinese medicine. The results are shown in Tables 2 and 3 below.
[0054] Table 2 Dataset of cold-natured Chinese medicinal compounds (partial)
[0055]
[0056] Table 3 Dataset of heat-property Chinese herbal medicine compounds (partial)
[0057]
[0058]
[0059] Based on the SMILES structures corresponding to the cold and hot properties of traditional Chinese medicine compounds, the MACCSKeys algorithm was used to extract the 167-dimensional molecular fingerprints of the compounds in the dataset. Then, the 167-dimensional molecular fingerprints were normalized and expressed as fingerprint vectors, resulting in vector representations of 206 cold and hot properties of traditional Chinese medicine. Based on the above dataset, 20% of the data was used as a training set and 80% of the data was used as a test set for prediction. A random forest regression model classifier was created: n_estimators = 100, random_state = 10, and the final ROC curve AUC value was 0.90. The results are shown in the figure. Figure 2 . At the same time, a support vector machine model is created to assist in the verification of the random forest regression model. The support vector machine model uses 20% of the data as the training set and 80% of the data as the test set. The support vector machine model classifier is: kernel = 'linear', C = 1.0, random_state = 42, probability = True. The final area under the ROC curve and the coordinate axis, the AUC value, is 0.83. The AUC value is the area under the ROC curve and the coordinate axis. The closer the AUC is to 1.0, the higher the accuracy of the model. When the AUC is higher than 0.7, it means that the accuracy of the model training is high. At the same time, the model is supervised by the confusion matrix. The results show that there is a certain deviation between the training set and the test set in the model. The specific accuracy deviation results are shown in Figure 3 Its rows represent the true categories, and its columns represent the predicted categories. The value of each cell represents the number of samples whose true category is the category corresponding to the row and is predicted to be the category corresponding to the column. Finally, the data of Curculigo, Clematis, Phellodendron, and Rhizoma Anemarrhenae are brought into the model for prediction verification, and the output result is set to 0 for hot and 1 for cold. The results are shown in Figure 4 , which is displayed as 0, 0, 1, 1, indicating that Curculigo and Clematidis are both hot Chinese medicines in the model, while Phellodendron and Anemarrhena are both cold Chinese medicines in the model. The prediction results are accurate, indicating that the model is reliable and can be used for discriminant analysis of the hot and cold properties of Chinese medicines.
[0060] Example 2
[0061] Effects of Chinese herbal medicine extracts with different cold and hot properties on biothermodynamic changes of Escherichia coli
[0062] In the following embodiments, the microcalorimeter used is a TAM Air microcalorimeter (TA Instruments, USA). In addition, in the following embodiments, all tests performed in the microcalorimeter are performed at 37°C.
[0063] This study investigates the correlation between hot and cold Chinese herbal medicines and biothermodynamic parameters, and whether the processing of hot and cold Chinese herbal medicines alters their properties. It is generally believed that the changes in the thermal effect produced by a Chinese herbal medicine after it acts on an organism can, to a certain extent, measure the hot and cold properties of the Chinese herbal medicine. Herbal medicines that produce a greater thermal effect in an organism are generally hot herbs, which can promote metabolism, while those that produce a lower thermal effect are cold herbs, which can inhibit metabolism. Using a microcalorimeter, the minute changes in thermal effect exhibited by Chinese herbal medicines acting on microorganisms can be monitored in real time, helping to examine the relationship between biothermodynamic parameters and the hot and cold properties of a drug, as well as the impact of hot and cold combinations on changes in its properties.
[0064] 1 Experimental Methods Traditional Chinese Medicine
[0065] 1.1 Preparation of TCM aqueous extracts
[0066] Weigh 25g of each herb and extract with water at a 1:10 solid-liquid ratio by heating and refluxing. Extract three times for 1 hour each. Filter and combine the filtrates, concentrate under reduced pressure, and dry in a vacuum oven to produce a lyophilized powder. For microcalorimetry, prepare a suspension equivalent to 1g of the original herb per milliliter in LB medium, achieving an initial concentration of 1g / mL. Filter through a 0.22μm filter to remove bacteria and impurities, obtaining a clarified sterile aqueous extract medium. Aliquot into 15mL tubes and store at 4°C until ready for use. Cinnamon bark (SRG), dried ginger (SGJ), Notopterygium wilfordii (SQH), Curculigo orchioides (SXM), Clematis chinensis (SWLX), yellow rice wine (HJ), Notopterygium wilfordii (JQH), Curculigo orchioides (JXM), Clematis chinensis (JWLX), salt solution (Y), Anemarrhena rhizome (SZM), Phellodendron chinense (SHB), salt Anemarrhena rhizome (YZM), salt Phellodendron chinense (YHB)
[0067] 1.2 Culture medium
[0068] LB liquid medium: Take 2 g of peptone, 1 g of sodium chloride, 1 g of yeast extract, and 200 mL of distilled water, adjust the pH to 7.0, sterilize with high-pressure steam at 121°C for 20 min, cool, and place in a refrigerator at 4°C until use.
[0069] 1.3 Experimental steps
[0070] The ampoule method was used for determination under sterile conditions at normal pressure and a constant temperature of 37°C. Sterile forceps were used to pick up the prepared ampoules. 5 mL of culture medium containing the Chinese herbal medicine extract (crude drug concentration was 10 mg / mL) and 5 mL of Escherichia coli culture medium were added to each ampoule, bringing the total volume of the system to 10 mL. Sterile liquid culture medium and deionized water, equal in volume to the reaction cell, were added to each ampoule in the reference cell as a reference. The ampoule was capped, sealed, and placed in the measurement channel of the calorimeter. The growth metabolic thermal spectrum curve was recorded and stopped when the curve returned to the baseline, which took approximately 20-30 hours. After recording, the experimental time and thermal power data were extracted. Using Origin software, a plot was constructed with time as the horizontal axis and power as the vertical axis to obtain the Pt curve, and the total calorific value, Q, was calculated. Biothermodynamic parameters were calculated based on the Pt curve.
[0071] The biothermodynamic parameters are calculated using the following formula (1):
[0072] ln P=ln P0+kt (1)
[0073] Where: P is the heat production power corresponding to the exponential growth period t, P0 is the heat production power corresponding to t=0, k is the exponential growth rate, and t is the time taken to reach the exponential growth period.
[0074] 2 Experimental results
[0075] The microcalorimeter response data for Chinese herbal medicine samples with different medicinal properties were analyzed, and the results are shown in Tables 4 and 5. In the table, P1 is the heat production power required to reach the first exponential growth phase, P2 is the heat production power required to reach the second exponential growth phase, Δt1 is the time required to reach the first exponential growth phase, Δt2 is the time required to reach the second exponential growth phase, and Q is the integrated area under the Pt curve. k1 is the growth rate during the first exponential growth phase, and k2 is the growth rate during the second exponential growth phase. From the results in Tables 4 and 5, it can be seen that under the action of a crude drug concentration of 10 mg / mL, it was found that the thermal power P1 of the first exponential growth period and the growth rate k1 of the first exponential period of hot Chinese medicines were generally much higher than those of cold Chinese medicines. In comparing the biothermodynamic parameters of raw and processed products on Escherichia coli, it was found that after wine-fried Rhizoma Notopterygii, Curculigo and Clematis increased the P1, k1 and Q of Escherichia coli, while after salt-fried Rhizoma Phellodendri and Rhizoma Anemarrhenae reduced the P1, k1 and Q of Escherichia coli. This indicates that the thermal power value P1 of the first exponential growth period and the growth rate k1 of the first exponential period can be used as discriminant indicators for hot and cold Chinese medicines, and P1, k1 and Q can be used as discriminant indicators for distinguishing wine-fried and salt-fried Chinese medicines.
[0076] Table 4 Thermodynamic data of different Chinese medicines on Escherichia coli
[0077]
[0078] As shown in Table 4, the k1 values of the first exponential growth rate of E. coli by different Chinese medicines are all higher than 2×10 -4 The k1 values of cold Chinese medicines are all lower than 2×10 -4 , and the same rules are found on P1, so k1 and P1 are used as control indicators of Escherichia coli in the cold and hot properties of traditional Chinese medicine.
[0079] Table 5 Thermodynamic data of different Chinese medicines on Staphylococcus aureus
[0080]
[0081]
[0082] As shown in Table 5, in terms of the first exponential growth rate k1 of different Chinese medicines against Staphylococcus aureus, the k1 values of hot Chinese medicines except rice wine were all higher than 9×10 -5 The k1 values of cold Chinese medicines are all lower than 9×10 -5 , and there is the same rule on P1, so k1 and P1 are used as control indicators of Staphylococcus aureus in the cold and hot properties of traditional Chinese medicine.
[0083] The results of the above two bacterial strains showed that using K1 and P1 as control indicators can distinguish between cold Chinese medicine and hot Chinese medicine. -4 As defined, P1 is 1×10 -3 As defined; Staphylococcus aureus K1 is 9×10 -5 As defined, P1 is 4×10 -3 This result is consistent with the predicted result.
[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements of strains and models, and improvements made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.
Claims
1. A model for predicting the cold and hot properties of traditional Chinese medicine based on a random forest model, characterized in that: The following steps are involved: Step S1: Searching for Chinese medicines with "cold and hot" properties in the Pharmacopoeia of the People's Republic of China to screen the dataset of Chinese medicines with cold and hot properties; Step S2: Search the database for compounds corresponding to cold Chinese medicines and hot Chinese medicines to establish cold Chinese medicine and hot Chinese medicine compound datasets. Then, delete duplicate compounds from the two datasets. Then, search the PubChem database for the SMILES structures of key compounds of cold Chinese medicines and hot Chinese medicines to finally obtain the cold and hot Chinese medicine datasets. Step S3: performing vector expression processing and weighting on the collected SMILES structural formulas of traditional Chinese medicine compounds; The MACCSkeys key was used to convert the SMILES structures of the compounds in the two datasets into a binary language that can be recognized by computers; Step S4: Classify and train the compound structure data of traditional Chinese medicine using the random forest algorithm, and then establish a model for distinguishing the cold and hot properties of traditional Chinese medicine.
2. The prediction model according to claim 1, characterized in that The step S1 includes the following specific steps: selecting a Chinese medicine whose properties are clearly defined as "cold" or "hot" in the pharmacopoeia.
3. The prediction model according to claim 1, characterized in that Step S2 includes the following specific steps: querying and screening compounds of traditional Chinese medicines with cold and hot properties through the TCMSP, TCMID, and TCM Bank databases, establishing a cold traditional Chinese medicine compound data set and a hot traditional Chinese medicine compound data set respectively, and then arranging the compound information of the two data sets according to the format in the TCMSP database; The duplicate compound data that existed in both datasets were deleted, and finally the compound datasets of cold Chinese medicine and hot Chinese medicine were obtained; The SMILES structures of the included cold and hot Chinese medicine compounds were screened using three methods of screening Chinese medicine compounds: oral availability, Caco-2 permeability, and drug-like properties, and finally a dataset of Chinese medicine compounds with cold and hot properties was obtained.
4. The prediction model according to claim 1, characterized in that The step S3 includes the following specific steps: using RDkit to generate a Morgan fingerprint for the SMILES representation of each drug, and expressing the Morgan fingerprint as a binary vector to obtain a molecular fingerprint expression vector of the molecule, and calculating the traditional Chinese medicine fingerprint feature expression with oral availability, Caco-2 permeability and drug-like properties, and then using a normalization method to summarize the statistical distribution of the unified sample.
5. The prediction model according to claim 1, characterized in that The step S4 includes the following specific steps: using 15-25% of the data as a training set and 75-85% of the data as a test set for prediction, and creating a random forest regression model classifier; using 15-25% of the data as a training set and 75-85% of the data as a test set for prediction, and creating a support vector machine model classifier.
6. A method for prediction using the prediction model according to claim 1, characterized in that: Input the SMILES structure weighted data of the compound in the Chinese medicine to be tested. The calculated output is 1 for a cold Chinese medicine, and the calculated output is 0 for a hot Chinese medicine.
7. A method for verifying the prediction results of the prediction model according to claim 1, characterized in that: The effects of Chinese medicine on bacterial biothermodynamics were detected using a microcalorimeter to summarize the biothermodynamic parameters for distinguishing between hot and cold Chinese medicines, including the following steps: Step 1: Inoculate the strain in the exponential growth phase into the culture medium; Step 2: Add the culture medium solution containing the strain into the bottle and seal it to obtain a blank solution; Step 3: Add the culture medium solution containing the strain and the Chinese herbal medicine extract culture medium solution into the bottle and seal it to obtain a sample solution; The volume of the sample solution is consistent with that of the blank solution in step 2; the drug extract culture medium solution is prepared by dissolving the lyophilized powder of traditional Chinese medicine in a sterile culture medium solution; Step 4: Place the blank solution and the sample solution in a microcalorimeter at 34-39° C., respectively, and record the thermal power-time (Pt) curve of the strain growth process. The test is terminated when the thermal spectrum curve returns to the baseline again, thus obtaining the bacterial growth metabolic thermal spectrum curve after the intervention of traditional Chinese medicine; Step 5: After recording, extract the experimental time and thermal power data. Use Origin software to plot the time as the horizontal axis and the power as the vertical axis to obtain the Pt curve and calculate the total calorific value Q. Step 6: Calculate the biothermodynamic parameters based on the Pt curve.
8. The method according to claim 7, characterized in that The strains in step 1 are Escherichia coli and Staphylococcus aureus.
9. The method according to claim 7, characterized in that The relative OD value of the culture medium solution containing the strain in step 3 is 0.2-0.4; the drug concentration of the traditional Chinese medicine extract culture medium solution is 5-15 mg / mL.
10. The method according to claim 7, characterized in that The Chinese medicine extract culture medium solution in step 3 is prepared from Chinese medicine freeze-dried powder, which is obtained by extracting the medicinal materials with water, filtering the obtained filtrate, and concentrating and drying.