Method and device for predicting pharmacodynamic quality markers of radix paeoniae rubra based on neural network

Abstract

The invention discloses a method and a device for predicting a pharmacodynamic quality marker of red paeony root based on a neural network, and the method comprises the steps: importing a plurality of traditional Chinese medicine marker peak areas and a plurality of pharmacodynamic index data into a data processing system, and calculating and determining a neural network model based on a prediction set and a training set in the index data, and training a neural network model based on experimental data, determining the prediction precision of the neural network model, and predicting the pharmacodynamic quality marker of the red peony root by using the verified neural network model. According to the invention, the BP neural network model is used for the first time to combine the chemical fingerprint spectrum information of the radix paeoniae rubra with the biological effect information so as to predict the effect of the coagulant drug. According to the method, the chemical fingerprint spectrum information and the biological effect information of the medicine can be fully utilized, and the accuracy of a prediction result is improved.

Description

Technical Field

[0001] The present invention relates to a method and device for predicting red peony root efficacy quality markers based on neural network, and in particular to the field of drug efficacy prediction. Background Art

[0002] In the field of biomedicine, drug efficacy prediction is an important research direction. Traditional drug efficacy prediction methods are mainly carried out through biological experiments. Although this method has high accuracy, it takes a lot of time and resources. At the same time, although existing technologies can use neural network models to predict drug efficacy, there are still some problems. First, most of the existing models are based on the chemical fingerprint information of the drug for prediction, while ignoring the biological effect information of the drug. This may lead to insufficient accuracy of the prediction results. Secondly, most of the existing models are trained based on a large amount of simulated data, and there is currently no technology for training and prediction based on the biological characteristics data of red peony root. Summary of the invention

[0003] The purpose of the present invention is to provide a method to solve the technical problem of constructing a model that can fully utilize the chemical fingerprint information and biological effect information of the drug, adapt to the situation of insufficient data in practical applications, and select a suitable neural network structure, as proposed in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions: In one embodiment, a neural network-based prediction method for red peony root efficacy quality markers includes the following steps: importing data containing multiple Chinese medicine marker peak areas and multiple pharmacodynamic indexes into a data processing system; constructing a neural network model based on the prediction set and training set in the index data to calculate and determine the neural network model; training the neural network model based on experimental data to determine the prediction accuracy of the neural network model; and using the verified neural network model to predict red peony root efficacy quality markers.

[0005] Furthermore, the peak areas of the multiple Chinese medicine markers are the peak areas of the seven most important Chinese medicine markers that affect the coagulation efficacy of red peony root; and the efficacy index is the coagulation effect index of red peony root measured by four different measurement methods.

[0006] Furthermore, the efficacy indices are indices determined by APTT test method, PT test method, TT test method, and FIB test method, respectively.

[0007] Furthermore, the calculation and determination of the neural network model based on the prediction set and training set in the indicator data further includes: standardizing and normalizing the imported data to form processed data with a standard unified range and controllable range; constructing four neural network models with the peak area of ​​traditional Chinese medicine markers as input and pharmaceutical indicators as output layers; determining the structure of the neural network algorithm based on the correlation coefficient and root mean square interpolation.

[0008] Furthermore, the training of the neural network model based on the experimental data and determining the prediction accuracy of the neural network model further include: training the model based on the training samples in the experimental data, adjusting the parameters in the model, and generating a trained neural network model; predicting all sample data through the trained neural network model, and performing error evaluation between the predicted values ​​and the measured values.

[0009] Furthermore, the sample data were randomly screened to determine 51 samples as training sets, 11 samples as prediction sets, and 11 samples as validation sets.

[0010] Furthermore, the minimum value of the root mean square interpolation (RMSE) of the training set and the prediction set was used as an indicator to determine the number of nodes in the hidden layer, and a BP network with one hidden layer was established. The entire network structure adopted the 10-5-1 type, where 10 is the number of input layer nodes, 5 is the number of hidden layer nodes, and 1 is the number of output layer nodes.

[0011] Furthermore, the input features loaded during the training process are the peak areas of TCM markers, and the target outputs are pharmacodynamic indicators.

[0012] Furthermore, the BP neural network model obtained after training was used to predict the coagulation quality markers of Radix Paeoniae Rubrae in 73 groups of sample data, and the error was evaluated with the actual measured values.

[0013] In another embodiment, a neural network-based prediction device for red peony root efficacy quality markers includes: a data import module for importing data containing multiple Chinese medicine marker peak areas and multiple pharmacodynamic indexes into a data processing system; a model determination module for constructing and determining the neural network model based on the prediction set and training set in the index data; a training verification module for training the neural network model based on experimental data to determine the prediction accuracy of the neural network model; and a red peony root quality marker prediction module for predicting red peony root efficacy quality markers using the verified neural network model.

[0014] Compared with the existing technology, The beneficial effects of the technical solution are as follows: 1. The present invention proposes for the first time to use the BP neural network model to combine the chemical fingerprint information of red peony root with the biological effect information to predict the coagulation efficacy (4 coagulation efficacy-related target values). This method can not only make full use of the chemical fingerprint information and biological effect information of the drug to improve the accuracy of the prediction results, but also can directly predict the biological effect from the chemical fingerprint without the need to detect the biological effect, greatly shortening the time and resource consumption of drug efficacy prediction. 2. The present invention establishes a BP network containing one hidden layer, and the entire network structure adopts a 10-5-1 type, where 10 is the number of input layer nodes, 5 is the number of hidden layer nodes, and 1 is the number of output layer nodes. This network structure can adapt to the situation of insufficient data in practical applications, and can select a suitable neural network structure to improve the promotion and application of the model. 3. The BP neural network model of the present invention can be used for coagulation efficacy prediction after training and learning the data. The measured value and the predicted value data are regressed and analyzed. It can be seen from the regression line of the measured values ​​of the four coagulation indicators and the BP neural network predicted values ​​that the goodness of fit between the two groups of data is more than 95%. This shows that the BP neural network model of the present invention has high accuracy and reliability. 4. The construction process of the BP neural network model of the present invention is simple and fast. It only needs to guide the chromatographic peak information of the red peony root fingerprint to the input layer, and the prediction result of its anticoagulant efficacy information can be obtained through the established BP-ANN model. This provides a simple and fast method for the quality evaluation of red peony root, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 The flowchart of the method for predicting the medicinal effect quality marker of red peony root based on neural network of the present invention; Figure 2 It is a flowchart of the present invention for calculating and determining the neural network model based on the prediction set and the training set in the indicator data; Figure 3 It is a flow chart of the present invention for training a neural network model based on experimental data and determining the prediction accuracy of the neural network model; Figure 4 It is the training and prediction results of the origin-main component-APTT BP neural network model; Figure 5 It is the training and prediction result of the origin-main component-PT BP neural network model; Figure 6 It is the training and prediction result of the origin-main component-TT BP neural network model; Figure 7 It is the training and prediction results of the origin-main component-FIB BP neural network model; Figure 8 It is the regression curve of measured value X and predicted value Y of APTT BP neural network model; Fig. 9 It is the regression curve of measured value X and predicted value Y of PT BP neural network model; Fig.10 It is the regression curve of measured value X and predicted value Y of TT BP neural network model; Fig.11 It is the regression curve of measured value X and predicted value Y of FIB BP neural network model; Fig.12 It is a schematic diagram of a prediction device for red peony root efficacy quality marker based on a neural network provided in an embodiment of the present disclosure; Fig.13 A schematic diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0017] This embodiment mainly describes a method for predicting the efficacy quality marker of red peony root based on a neural network, and the specific implementation steps mainly include the following contents: S101 imports data including peak areas of multiple Chinese medicine markers and multiple pharmacodynamic indexes into a data processing system; S102 calculating and determining the neural network model based on the prediction set and the training set in the indicator data; S103 training a neural network model based on the experimental data to determine the prediction accuracy of the neural network model; S104 Use the validated neural network model to predict the efficacy quality markers of red peony root.

[0018] This embodiment provides a simple and rapid method for predicting drug efficacy. According to this solution, a neural network structure is selected and determined, and the model is trained in a targeted manner to determine a special model that can be used to predict the quality markers of red peony root efficacy. Based on the special model, the user only needs to import the Chinese medicine marker front of the red peony root fingerprint into the input layer of the neural network model, and the prediction results of its anticoagulant efficacy information can be obtained through the established BP-ANN model. This not only solves the problem of disconnection between ingredients and efficacy in quality evaluation and control, but also provides a simpler and faster quality evaluation solution.

[0019] Step S101 imports data including multiple peak areas of Chinese medicine markers and multiple pharmacodynamic indexes into a data processing system. Specifically, 73 sets of data are imported into MATLAB R2023b software, each set of data including peak areas of 7 Chinese medicine markers screened by spectrum-effect relationship and 4 pharmacodynamic indexes (APTT, PT, TT, FIB).

[0020] In this embodiment, 73 batches of red peony root from different collection areas were used as samples, and spectral analysis was performed on the samples. The peak areas of multiple common peaks in the 73 batches of red peony root after spectral analysis were used as variables for principal component analysis, and 7 traditional Chinese medicine markers that affect the coagulation efficacy of red peony root were determined, and the frontal area of ​​each traditional Chinese medicine marker in each batch was determined.

[0021] In order to comprehensively determine the coagulation effect of red peony root, different reagents were used to measure from multiple dimensions, and the coagulation four-item kit determination method was used. The specific determination method is as follows: APTT test method: Pipette 50 μL of the plasma to be tested into a blood coagulation cup, add 50 μL of sample solution, add 50 μL of APTT reagent, incubate at 37°C for 5 minutes, and finally add 50 μL of 37°C pre-warmed calcium chloride solution (0.025 mol·L-1) to measure the coagulation time. Replace the sample with a blank solvent and measure the coagulation time in the same way as above, which is the control.

[0022] PT test method: Pipette 50 μL of plasma to be tested into a blood coagulation cup, add 50 μL of sample solution, add 100 μL of PT reagent, incubate at 37°C for 3 minutes, and measure the coagulation time. Replace the sample with blank solvent, and measure the coagulation time in the same way as above, which is the control.

[0023] TT test method: Pipette 100 μL of the plasma to be tested into a blood coagulation cup, add 100 μL of sample solution, add 100 μL of TT reagent, incubate at 37°C for 5 minutes, and measure the coagulation time. Replace the sample with a blank solvent and measure the coagulation time in the same manner as above, which is the control.

[0024] FIB test method: Pipette 100 μL of plasma to be tested into a blood coagulation cup, add 50 μL of sample solution, add 50 μL of thrombin solution, incubate at 37°C for 5 minutes, and measure the coagulation time. Replace the sample with a blank solvent, and measure the coagulation time in the same way as above, which is the control.

[0025] The sample solutions used in the above four detection methods were prepared by the following method: take about 1.0g of red peony root powder, weigh it accurately, put it in a stoppered conical flask, add 10mL of 40% methanol accurately, stopper it, weigh it, extract it by ultrasound (350W, 40KHz) for 45min, let it cool, weigh it again, make up the lost weight with 40% methanol, shake it well and let it stand, take an appropriate amount of solution in a centrifuge tube with a lid, centrifuge it at 12000r / min for 3min, absorb the supernatant, filter it through a microporous filter membrane (0.45μm), and obtain the sample solution. Each batch of drugs from 73 batches of red peony root was used to prepare 73 sample solutions by the above method, and each sample solution was subjected to coagulation experiments of 4 pharmacodynamic indicators (APTT, PT, TT, FIB).

[0026] Based on this experimental data, 73 sets of experimentally measured data were imported into MATLAB. The imported data is the data basis for model training and verification.

[0027] Step S102 determines the neural network model based on the prediction set and training set in the indicator data. The imported data is processed, and the processed data can ensure the stability of model training; a part of the imported data is randomly selected to form a training set and a prediction set, and the neural network model is determined based on the correlation coefficient and root mean square interpolation between the calculated value and the actual value, thereby realizing the most stable model selection with the smallest error. The following steps are used to achieve this: S1021 Standardize and normalize the imported data to form processed data with a controllable standard and unified range.

[0028] The system normalizes and standardizes the imported data. Normalization is data preprocessing. Normalization is to scale the data to a specific range, usually [0, 1] or [-1, 1]. The Min-Max Scaling method is used as an example. Normalization can prevent the scale differences between features from affecting the training of the model.

[0029] Standardization is the process of converting data into a distribution with a mean of 0 and a variance of 1. The Z-score standardization method is preferred. The processing after standardization can avoid the influence of extreme data on model training.

[0030] S1022 Construct four neural network models with the peak area of ​​Chinese medicine markers as input and pharmaceutical indicators as output layers. Before constructing the four neural network models, this embodiment screened and determined 51 samples from 73 sample data as training sets, 11 samples as prediction sets, and 11 samples as validation sets, and selected the specific structure of the model and trained the model based on the training set and the prediction set.

[0031] The 7 TCM peak areas (x) of 51 batches of training samples and 11 batches of prediction samples were used as the input layer, and the 4 pharmacodynamic indicators (y) were used as the output layer. Four models, Q-Marker-APTT, Q-Marker-PT, Q-Marker-TT and Q-Marker-FIB, were established respectively. The BP model was used as the preferred model. The correlation coefficient (γ2) and root mean square interpolation (RMSE) of the model training set and the prediction set were obtained as indicators to compare the effects of different models. For details, see Figure 4-7 .

[0032] S1023 Determine the structure of the neural network algorithm based on the correlation coefficient and the root mean square interpolation. As described in the previous step, Figure 4-7 The correlation coefficient (γ2) and root mean square error (RMSE) in each model were recorded. Hyperparameter adjustment, feature selection and model architecture optimization were performed based on the evaluation results to improve the generalization ability and prediction accuracy of the model. The number of nodes in the hidden layer of the network was screened, and the minimum value of the root mean square error (RMSE) of the training set and the prediction set was used as an indicator to determine the number of nodes in the hidden layer. A BP network with one hidden layer was established. The entire network structure adopted the 10-5-1 type, where 10 is the number of input layer nodes, 5 is the number of hidden layer nodes, and 1 is the number of output layer nodes.

[0033] S103: Train the neural network model based on the experimental data to determine the prediction accuracy of the neural network model. Use the training samples in the built BP model for training, use all samples to evaluate the accuracy of the trained samples, and after the evaluation, the maximum error and the minimum error are both within a reasonable range. The model can be used to evaluate the quality markers of red peony root efficacy.

[0034] S1031 trains the model based on the training samples in the experimental data, adjusts the parameters in the model, and generates a trained neural network model. The input features loaded during the training process are the peak area of ​​traditional Chinese medicine (x), and the target output is the pharmacodynamic index (y). The learning rate (LR) is set to 0.02 in the training software, and the network termination condition is the target error (GOAL) of 0.001, the maximum number of cycles (EPOCHS) is 2000 times, and the back propagation algorithm and the specified learning rate are used to train the model and adjust the key parameters in the model. According to the above steps, the neural network model is configured and trained, and the model parameter tuning is effectively achieved while ensuring the learning rate, target error and maximum number of cycles.

[0035] S1032 Predict all sample data using the trained neural network model, and perform error evaluation between the predicted values ​​and the measured values.

[0036] After the neural network model training is completed, we first need to ensure that the trained model can predict new sample data. The task here is to use the trained BP neural network model to predict the coagulation red peony quality markers of 73 groups of sample data and perform error evaluation with the actual measured values.

[0037] Test data set: Extract the input features (common peak area) of 73 groups of samples, ensuring that the data format is consistent with that used during training. Each group of samples still has 7 features and generates an input matrix.

[0038] Use the trained neural network model to predict all 73 groups of samples, and correspond the predicted value (Y_pred) of the BP neural network with the actual measured value (X_real) to form a result table to facilitate subsequent error analysis and visualization. Evaluate the error between the predicted result and the actual measured value, calculate the relative error, and analyze the prediction effect. The relative error calculation formula is as follows: Relative Error=∣Actual−Predicted∣ / ∣Actual∣×100% Regression analysis is performed to visualize the relationship between the predicted value and the actual value. This embodiment adopts a method including drawing a scatter plot and a regression line. Scatter plot: The scatter plot of the predicted value and the actual value is used to observe the fitting effect of the model; regression line: Draw a reference line y=x to judge the accuracy of the predicted value. Calculate and determine the maximum and minimum relative errors, and perform regression analysis on the predicted value and measured value data. The maximum relative error between the measured value X and the predicted value Y is 10.02%, and the minimum relative error is 0.1%. For specific reference Figure 8-11 The error is controlled within a controllable range and can meet the actual production needs.

[0039] S104 Prediction of Red Peony Root Efficacy Quality Markers Using Validated Neural Network Model Based on the above model training and verification steps, the peak areas of 7 red peony root anticoagulant quality markers were determined as independent variables, APTT, PT, TT, and FIB were used as dependent variables, and a red peony root spectrum-effect relationship model based on a BP neural network was constructed, the relationship between the components and the efficacy was clarified, and a simple, fast, and efficacy-related red peony root quality evaluation model was given. The above is the specific implementation steps of this embodiment. Through these steps, accurate prediction of the anticoagulant efficacy of red peony root can be achieved, and the accuracy and efficiency of drug efficacy prediction can be improved.

[0040] Fig.12 A schematic diagram of a neural network-based prediction device for red peony root efficacy quality markers provided in the disclosed embodiments. Fig.12 As shown, a neural network-based prediction device 30 for red peony root efficacy quality markers provided in an embodiment of the present disclosure includes: The data import module 310 is used to import data including multiple Chinese medicine marker peak areas and multiple pharmacodynamic indexes into the data processing system; A model determination module 320, configured to construct and determine the neural network model based on the prediction set and the training set in the indicator data; A training verification module 330 is used to train a neural network model based on experimental data and determine the prediction accuracy of the neural network model; The red peony root quality marker prediction module 340 is used to predict the red peony root efficacy quality marker using the verified neural network model.

[0041] For descriptions of the processing flow of each module in the device and the interaction flow between each module, reference may be made to the relevant descriptions in the above method embodiment, which will not be described in detail here.

[0042] The disclosed embodiment provides a prediction device for red peony root medicinal quality markers based on neural network. The data import module in the device imports data including multiple Chinese medicine marker peak areas and multiple pharmacodynamic indexes into a data processing system. The imported data is the data basis for training and verifying the neural network model. The data imported by the data import module runs in the model determination module, wherein the prediction set and the training set determine the architecture of the neural network model. Then, in the training and verification module, the model is trained and the model parameters are adjusted using the imported data in the import module to form a verified neural network model. The red peony root quality marker prediction module uses the verified neural network model to predict the red peony root medicinal quality markers.

[0043] Corresponds to Figure 1-3 The neural network-based prediction method for red peony root efficacy quality markers in the present disclosure also provides an electronic device 400, such as Fig.13 FIG. 4 is a schematic diagram of the structure of an electronic device 400 provided in an embodiment of the present disclosure, including: Processor 41, memory 42, and bus 43; memory 42 is used to store execution instructions, including memory 421 and external memory 422; memory 421 here is also called internal memory, which is used to temporarily store operation data in processor 41 and data exchanged with external memory 422 such as hard disk. Processor 41 exchanges data with external memory 422 through memory 421. When the electronic device 400 is running, the processor 41 communicates with the memory 42 through bus 43, so that the processor 41 executes Figure 1 and Figure 2 The steps of the neural network-based prediction method for red peony root efficacy quality markers.

[0044] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art disclosed in the present application. The specification and examples are to be regarded as exemplary only, and the true scope of the present application is indicated by the claims.

Claims

1. A method for predicting red peony root efficacy quality markers based on neural network, characterized in that: The following steps are involved: Importing data including multiple Chinese medicine marker peak areas and multiple pharmacodynamic indexes into a data processing system; The neural network model is determined by calculation based on the prediction set and the training set in the indicator data; Training a neural network model based on experimental data to determine the prediction accuracy of the neural network model; The validated neural network model was used to predict the efficacy quality markers of red peony root.

2. The neural network-based prediction method for red peony root efficacy quality markers according to claim 1, characterized in that: The peak areas of the multiple Chinese medicine markers are the seven most important Chinese medicine markers that affect the coagulation efficacy of red peony root; the efficacy index is the coagulation effect index of red peony root measured by 4 different measurement methods.

3. The neural network-based prediction method for red peony root efficacy quality markers according to claim 2, characterized in that: The drug efficacy indices are respectively indices determined by APTT test method, PT test method, TT test method and FIB test method.

4. The method for predicting red peony root efficacy quality markers based on neural network according to claim 1, characterized in that: The calculation and determination of the neural network model based on the prediction set and the training set in the indicator data further includes: Standardize and normalize the imported data to form processed data with a controllable standard and unified range; Construct four neural network models with the peak area of ​​Chinese medicine markers as input and pharmaceutical indicators as output layer; The structure of the neural network algorithm is determined based on the correlation coefficient and the root mean square interpolation.

5. The neural network-based prediction method for red peony root efficacy quality markers according to claim 1, characterized in that: The step of training the neural network model based on the experimental data and determining the prediction accuracy of the neural network model further comprises: Train the model based on the training samples in the experimental data, adjust the parameters in the model, and generate a trained neural network model; All sample data are predicted using the trained neural network model, and errors are evaluated between the predicted values ​​and the measured values.

6. The neural network-based prediction method for red peony root efficacy quality markers according to claim 4, characterized in that: The sample data were randomly screened to determine 51 samples as training sets, 11 samples as prediction sets, and 11 samples as validation sets.

7. The neural network-based prediction method for red peony root efficacy quality markers according to claim 6, characterized in that: The minimum value of the root mean square interpolation (RMSE) of the training set and the prediction set was used as an indicator to determine the number of nodes in the hidden layer. A BP network with one hidden layer was established. The entire network structure adopted the 10-5-1 type, where 10 is the number of input layer nodes, 5 is the number of hidden layer nodes, and 1 is the number of output layer nodes.

8. The neural network-based prediction method for red peony root efficacy quality markers according to claim 5, characterized in that: The input features loaded during the training process are the peak areas of traditional Chinese medicines, and the target outputs are pharmacodynamic indicators.

9. The neural network-based prediction method for red peony root efficacy quality markers according to claim 8, characterized in that: The BP neural network model obtained after training was used to predict the coagulation red peony root quality markers of 73 groups of sample data, and the error was evaluated with the actual measured values.

10. A neural network-based prediction device for red peony root efficacy quality markers, characterized in that: include: A data import module is used to import data including peak areas of multiple Chinese medicine markers and multiple pharmacodynamic indexes into a data processing system; A model determination module, used to construct and determine the neural network model based on the prediction set and the training set in the indicator data; A training and verification module, for training a neural network model based on experimental data and determining the prediction accuracy of the neural network model; The red peony root quality marker prediction module is used to predict the red peony root efficacy quality marker using the verified neural network model.

Images