Optimization of Safflower Seed Extraction Process Based on Ant Colony Algorithm-Backpropagation Neural Network

Abstract

This invention discloses an optimization method for safflower seed extraction based on ant colony algorithm backpropagation neural network, relating to the field of traditional Chinese medicine extraction technology. The method for optimizing the safflower seed extraction process includes: (1) ultrasonic extraction of safflower seeds, conducting single-factor experiments, and selecting ultrasonic power, ethanol concentration, material-to-liquid ratio, and extraction temperature as influencing factors; (2) calculating the extraction rates of CS and FS using HPLC, and determining the optimal values ​​of the influencing factors based on the extraction rates; (3) calculating the comprehensive evaluation value using the entropy weight method, and obtaining the optimal process parameters using an ant colony algorithm backpropagation neural network model; the specific steps of the ant colony algorithm backpropagation neural network include: (1) constructing a backpropagation neural network model; (2) constructing an ant colony algorithm optimization neural network model. The safflower seed extract of this invention has significant antioxidant stress effects, providing a reference for large-scale extraction of safflower seeds.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine extraction technology, specifically to the optimization of safflower seed extraction process based on ant colony algorithm backpropagation neural network. Background Technology

[0002] Safflower (Carthamus tinctorius L.) belongs to the genus Carthamus in the family Asteraceae. Also known as safflower, it originated in India and is distributed in Central Asia, Southwest Asia, and the Mediterranean region. Safflower is a crop used for both medicinal and edible purposes. Its corolla, seeds, stems, and leaves are rich in various active substances. Safflower contains nutrients such as fats, proteins, various vitamins, and trace elements, and has effects such as lowering blood lipids, inhibiting bacteria, preventing cardiovascular diseases, and enhancing cell metabolism. Safflower seeds are white, obovate achenes with similar effects to the flowers and can be used both as food and medicine. Safflower seed extract contains various components, the most common being N-(p-coumaroyl)-hydroxytryptamine (CS) and N-ferulin hydroxytryptamine (FS). CS has been shown to have therapeutic effects on cardiovascular diseases, gliomas, and cholesterol. FS can treat atherosclerosis by reducing related inflammation.

[0003] Ultrasonic extraction primarily increases the contact between the solvent and the medicinal material through vibration, enhancing solvent penetration into cells and improving extraction efficiency. Ultrasonic extraction requires no heating device and offers protection for thermally unstable and easily decomposed medicinal substances. Compared to traditional methods, ultrasonic extraction time is significantly reduced, typically achieving the desired extraction rate in 24-40 minutes. A major advantage of ultrasonic extraction is its ability to extract from most types of herbs and various components. Ultrasonic-assisted extraction offers advantages such as short extraction time, high efficiency, high yield, high product purity, and environmental safety.

[0004] Artificial neural networks (ANNs) are models that use machines to train and predict data, capable of handling various complex or uncomplex problems. Ant colony optimization (ACO) is a multivariate clustering algorithm that obtains the optimal solution for data through positive and negative feedback. ACO-BPNN is a novel model that combines ant colony optimization and backpropagation neural networks. Artificial neural networks have been increasingly applied in fields such as plant science, medicine, and information transmission.

[0005] Safflower seed oil has been extensively studied in food, cosmetics, and animal husbandry. However, research on safflower seed extract is scarce. Currently, there is a lack of research on obtaining the optimal extraction process for safflower seeds through model optimization. Summary of the Invention

[0006] The purpose of this invention is to provide a method for optimizing the safflower seed extraction process based on the ant colony algorithm and backpropagation neural network. The safflower seed extract under optimized conditions has a significant antioxidant effect, providing theoretical support for the large-scale extraction of safflower seeds.

[0007] To achieve the above-mentioned objectives, the technical solution of the present invention is as follows:

[0008] On one hand, this invention provides a method for optimizing safflower seed extraction process based on ant colony algorithm backpropagation neural network, comprising the following steps:

[0009] (1) Safflower seeds were extracted by ultrasound. Single-factor experiments were conducted, and ultrasound power, ethanol concentration, material-liquid ratio and extraction temperature were selected as the influencing factors of the single-factor experiments.

[0010] (2) The extraction rates of N-(p-coumaroyl)-hydroxytryptamine (CS) and N-ferulin hydroxytryptamine (FS) were calculated by HPLC, and the optimal values ​​of influencing factors were determined based on the extraction rates.

[0011] (3) The comprehensive evaluation value is calculated by the entropy weight method, and the optimal process parameters are obtained by backpropagating the neural network model using the ant colony algorithm.

[0012] The specific steps of the ant colony algorithm backpropagation neural network include:

[0013] (1) Constructing a backpropagation neural network model

[0014] ① Construct a neural network model with n hidden layer neurons, where n = 2, 3, 4, 5, 6;

[0015] ②The hidden layer activation function is the Sigmoid function. a The activation function of the output layer is a linear function;

[0016] ③ Backpropagation optimizer settings: Optimizer: Adam, learning step size Lr: 0.01;

[0017] ④ Loss function: MSEloss b ;

[0018] ⑤ Use the backpropagation optimizer to optimize the neural network model with 400 epochs based on the initial weights, and find the minimum value of the loss function;

[0019] (2) Constructing an ant colony algorithm to optimize the neural network model

[0020] ① Randomly generate ants distributed in different spatial locations and decode the spatial locations, i.e., the initial weights;

[0021] ② Using the backpropagation neural network model constructed above, the spatial positions of different ants are optimized to find the optimal positions for each ant and record the pheromone F. c ;

[0022] ③ Encode the weights of each ant and record the optimal position among them. d ;

[0023] ④ Normalize the ant pheromone F. e ;

[0024] ⑤ Based on the relationship between the normalized pheromone F and the transition probability p0 f Global update of the spatial location of ants g Or partial update h ;

[0025] ⑥ Update the ant pheromones based on the evaporation rate lf i ;

[0026] ⑦ Repeat steps ①-⑥ above until the upper limit of the ant colony algorithm is reached, and output the optimal model result.

[0027] Specifically, the Sigmoid function at point a is:

[0028]

[0029] In the formula, x is the weight;

[0030] The loss function MSEloss at point b is:

[0031]

[0032] y i It is the true value, y i ' is the predicted value, and i represents different samples.

[0033] The function at point c is:

[0034] F k =-MSEloss k ;

[0035] In the formula, k represents different ants;

[0036] The function at point d is: min{MSEloss};

[0037] The function at point e is:

[0038] The function at point f is:

[0039] The function at point g is:

[0040] Global update: W(i,j)=W(i,j)+(Sigmoid(R(i,j)-0.5))×lr×W(i,j);

[0041] In the above formula, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix;

[0042] The function at point h is:

[0043] Local update: W(i,j)=W(i,j)+(arctan((R(i,j)-0.5)×2))×lr×W(i,j);

[0044] In the formula, arctan is the arctangent function, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix;

[0045] The function at position i is: F k (t+1)=F k (t)×(1-lf);

[0046] In the formula, lf is the evaporation rate of the ant colony algorithm, and t represents the t-th ant colony algorithm cycle.

[0047] Specifically, the material-liquid ratio mentioned in step (1) is the mass ratio of safflower seed powder to the volume ratio of ethanol.

[0048] Preferably, the HPLC column used in step (2) is an Agilent Eclipse XDB-C. 18 (4.6×250mm, 5μm).

[0049] Preferably, the mobile phase of the HPLC method in step (2) is an aqueous solution (A) of 0.4% formic acid and a methanol solution (B).

[0050] Preferably, the flow rate of the HPLC method in step (2) is 1 mL / min.

[0051] Preferably, the gradient elution method of HPLC in step (2) is: 0 min, 90% A; 3 min, 90% A; 20 min, 50% A; 30 min, 10% A.

[0052] Preferably, in step (2), the injection volume of HPLC is 10 μL, the column temperature is 35℃, and the detection wavelength is 310 nm.

[0053] Preferably, the extraction rate (%) in step (2) is calculated as: compound concentration in extract (mg / mL) / [crude drug weight (mg) / crude drug extraction solvent volume (mL)] × 100%.

[0054] Preferably, the optimal values ​​of the influencing factors determined in step (2) are: ultrasonic power 160-320W, ethanol concentration 70-90%, liquid-to-solid ratio 15-25mL / g, and extraction temperature 60-80℃.

[0055] Preferably, the comprehensive evaluation value (Y) in step (3) = (extraction rate) CS )×0.5036+(extraction rate) FS )×0.4964.

[0056] Preferably, the optimal process parameters in step (3) are: ultrasonic power 320W, ethanol concentration 80%, liquid-to-solid ratio 15mL / g, and extraction temperature 80℃.

[0057] In another aspect, the present invention provides a safflower seed extract prepared by the above-described method.

[0058] In another aspect, the present invention provides the safflower seed extract prepared by the above method or the application of the above safflower seed extract in the preparation of antioxidants.

[0059] Specifically, the antioxidant can reduce the content of MDA and increase the content of SOD.

[0060] Specifically, the antioxidant can increase HO-1 mRNA levels and decrease NOX2 mRNA levels.

[0061] The beneficial effects of this invention are as follows:

[0062] This invention provides a method for optimizing the safflower seed extraction process based on ant colony algorithm-backpropagation neural network. The safflower seed extract under optimized conditions has significant antioxidant stress effects, providing a theoretical basis for the clinical development of safflower seeds. This invention also broadens the application field of the ACO-BPNN model and provides a reference for the large-scale extraction of safflower seeds. Attached Figure Description

[0063] Figure 1 The figures are HPLC chromatograms; where (a) is the HPLC chromatogram of the sample; and (b) is the HPLC chromatogram of the mixed reference.

[0064] Figure 2 The values ​​are the response values ​​of CS and FS; where (a) is the ultrasonic frequency response value of CS and FS; (b) is the ethanol concentration response value of CS and FS; (c) is the feed-to-liquid ratio response value of CS and FS; and (d) is the extraction temperature response value of CS and FS.

[0065] Figure 3The three-dimensional response surface plots are shown below: (a) is the three-dimensional response surface plot of ultrasonic frequency and ethanol concentration; (b) is the three-dimensional response surface plot of ultrasonic frequency and solid-liquid ratio; (c) is the three-dimensional response surface plot of ultrasonic frequency and solid-liquid ratio; (d) is the three-dimensional response surface plot of ethanol concentration and solid-liquid ratio; (e) is the three-dimensional response surface plot of ethanol concentration and extraction temperature; and (f) is the three-dimensional response surface plot of solid-liquid ratio and extraction temperature.

[0066] Figure 4 The value is the comprehensive evaluation value under the ACO-BPNN model, where (a) is the relationship between the predicted value and the actual value; and (b) is the residual plot.

[0067] Figure 5 The results of in vivo antioxidant activity are shown in the figure, where (a) serum SOD activity; (b) serum MDA activity; (c) HO-1 mRNA level in brain tissue; (d) NOX2 mRNA level in brain tissue; n=6 for each group. For the same letter ns, adjacent letters have P<0.05, and alternating letters have P<0.01. Detailed Implementation

[0068] To make the technical means, creative features, and achieved objectives and effects of this invention easier to understand, the invention is further illustrated below with specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention. Unless otherwise specified, the operating methods and equipment used in the following embodiments are conventional operating methods, and the materials and equipment used in each embodiment are the same.

[0069] Example 1

[0070] 1. Materials and Methods

[0071] 1.1 Medicines, reagents and instruments

[0072] Dried safflower seeds (batch number: 20221010) were purchased from Jimsar County, Changji Hui Autonomous Prefecture, Xinjiang, China. Standards N-(p-coumaroyl)-hydroxytryptamine, N-ferulin hydroxytryptamine, and 2,2-diphenyl-1-trinitrophenylhydrazine (DPPH) were obtained from Shanghai Yuanye Biotechnology Co., Ltd., China. The purity of the standards was ≥98%. The FRAP kit was obtained from Shanghai Beyotime Biotechnology Co., Ltd., China. The solvent used in the mobile phase was HPLC grade. Invitrogen was purchased from Thermo Fisher Scientific, China. 5× All-In-One RT MasterMix was purchased from Applied Biological Materials Inc., Canada. 2× Color SYBR Green qPCR Master Mix was purchased from Titan Science, Shanghai, China.

[0073] Ezra multi-functional grinder, Ezela Electric Co., Ltd. Agilent 1200 high-performance liquid chromatograph, Agilent Technologies Inc. xP105DR analytical balance, Mettler Toledo Instruments Ltd. xM-400ULF LCD low-frequency ultrasonic cleaner, from Xiaomei Ultrasonic Instruments Co., Ltd.

[0074] 1.2 Preparation of Sample and Reference Solutions

[0075] The dried safflower seeds were pulverized using a pulverizer. The pulverized safflower seed powder was mixed with different volumes of solvent and extracted using different ultrasonic frequencies. After extraction, the sample was allowed to stand and cool to room temperature. The supernatant was collected by centrifugation and filtered through a 0.45 μm microporous membrane to obtain the sample solution.

[0076] Accurately weigh N-(p-coumaroyl)-hydroxytryptamine (CS) and N-ferulin hydroxytryptamine (FS) into 1.5 mL centrifuge tubes, dissolve them in methanol to obtain a reference solution. Accurately weigh the CS and FS reference solutions, dilute them with methanol to different concentrations, and obtain five reference solutions of different concentrations.

[0077] 1.3 High Performance Liquid Chromatography Analysis

[0078] The chromatographic column was an Agilent Eclipse XDB-C. 18 (4.6 × 250 mm, 5 μm). The mobile phase consisted of an aqueous solution of 0.4% formic acid (A) and methanol (B), with a flow rate of 1 mL / min. The gradient elution method was as follows: 0 min, 90% A; 3 min, 90% A; 20 min, 50% A; 30 min, 10% A. The injection volume was 10 μL, the column temperature was 35 °C, and the detection wavelength was 310 nm.

[0079] 1.4 Methodological Research

[0080] 1.4.1 Specificity

[0081] The test solution and the mixed reference solution of the two components were injected into the liquid chromatograph for analysis. The specificity of the method was determined according to the chromatographic conditions described above.

[0082] 1.4.2 Linear Relationship

[0083] The concentrations of CS were 0.024, 0.034, 0.048, 0.068, and 0.08 mg / mL, and the concentrations of FS were 0.068, 0.09, 0.12, 0.15, and 0.2 mg / mL. A linear regression equation was established for CS and FS with concentration on the x-axis and peak area on the y-axis.

[0084] 1.4.3 Precision

[0085] Take the mixed reference solution and inject it 6 times under the chromatographic conditions described above. Record the peak area of ​​each component and calculate the relative standard deviation (RSD).

[0086] Preparation of mixed reference solution: Take appropriate amounts of 1 mg / mL CS and FS standard stock solutions, add methanol, and dilute to CS and FS concentrations of 0.048 mg / mL and 0.15 mg / mL, respectively.

[0087] 1.4.4 Stability

[0088] Take freshly prepared sample solutions and test them at 0, 2, 4, 6, 8, 12, and 24 hours. Record the peak area of ​​each component and calculate the relative standard deviation (RSD).

[0089] 1.4.5 Repeatability

[0090] Six sample solutions were prepared in parallel under the same conditions from the same batch of medicinal materials for testing. The peak areas of each component were recorded and the relative standard deviation (RSD) was calculated.

[0091] 1.4.6 Sample Recovery

[0092] Under the extraction conditions described above, rediscover safflower seed samples of known concentrations and add reference solution. Perform six parallel determinations under the chromatographic conditions described above. Calculate the relative standard deviation (RSD) for each component.

[0093] 1.5 Single-factor experimental design

[0094] Safflower seeds were weighed, and extraction was performed under the following conditions: ultrasonic frequency (80-400W), ethanol concentration (60-100%), liquid-to-solid ratio (5-25mL / g), and extraction temperature (40-80℃). Single variables were used for single-factor investigations. When investigating one factor, the other factors were: ultrasonic frequency 240W, ethanol concentration 100%, liquid-to-solid ratio 15mL / g, extraction temperature 60℃, and extraction time 60min. The supernatant was analyzed by HPLC.

[0095] The concentrations of each component were calculated based on the standard curve. The yield calculation formula is as follows:

[0096] Extraction rate (%) = concentration of compound in extract (mg / mL) / [weight of crude drug (mg) / volume of crude drug extraction solvent (mL)] × 100% (Equation 1).

[0097] 1.6 Optimization of Safflower Seed Extraction Process

[0098] 1.6.1 Calculation of Comprehensive Evaluation Value using Entropy Weight Method

[0099] To assess the overall extraction rate, the entropy weight method was used to assign coefficients to the two components. The overall estimate is calculated as follows:

[0100] Overall evaluation value (Y) = (extraction rate) CS )×0.5036+(extraction rate) FS )×0.4964 (Equation 2).

[0101] Entropy weight method: The two compounds are normalized, and the weight of each compound is calculated using the entropy weight method. Finally, the comprehensive evaluation result of the extract is calculated based on the weight.

[0102] In Equation 2 above, 0.5036 is the weighting coefficient of CS and 0.4964 is the weighting coefficient of FS.

[0103] 1.6.2 RSM Model

[0104] Based on the single-factor experiments, the comprehensive evaluation value of safflower seeds was selected as the dependent variable. Response surface methodology experiments were designed using Design Expert 13 software according to the principles of Box-Benhnken center composite design (BBD). There were 30 experimental groups with a total of 6 center points. Conditions and levels are shown in Table 1, and the detailed design of the 30 experimental groups is shown in Table 5.

[0105] Table 1 Experimental conditions and levels

[0106]

[0107] 1.6.3 ACO-BPNN Model

[0108] The BBD-designed 30 sets of data results were used as the input and output layers for training and prediction. The relevant settings and methods were as follows: A neural network model with 4 hidden layer neurons was constructed, using mean squared error (MSE) as the loss function, and gradient descent was used to optimize the model (optimizer: Adam, learning rate lr: 0.01, number of iterations: 300). Since the initial weights of the neural network have a significant impact on the results, an ant colony optimization algorithm was used to optimize the initial weights.

[0109] 1.6.3.1 Constructing the Backpropagation Neural Network Model

[0110] (1) Construct a neural network model with n hidden layer neurons, n = 2, 3, 4, 5, 6;

[0111] (2) The activation function of the hidden layer is the Sigmoid function. a The activation function of the output layer is a linear function;

[0112] (3) Backpropagation optimizer settings: Optimizer: Adam, learning step size Lr: 0.01;

[0113] (4) Loss function: MSE loss b ;

[0114] (5) Using the backpropagation optimizer, the neural network model is optimized for 400 epochs based on the initial weights to find the minimum value of the loss function.

[0115] 1.6.3.2 Constructing an Ant Colony Algorithm to Optimize the Neural Network Model

[0116] (1) Randomly generate ants distributed in different spatial locations and decode the spatial locations, i.e., the initial weights;

[0117] (2) Using the backpropagation neural network model constructed above, the spatial positions of different ants are optimized to find the optimal positions of each ant and record the pheromone F. c ;

[0118] (3) Encode the weights of each ant and record the optimal position among the ants. d ;

[0119] (4) Normalize the pheromone F of the ants e ;

[0120] (5) Based on the relationship between the normalized pheromone F and the transition probability p0 f Global update of the spatial location of ants g Or partial update h ;

[0121] (6) Update the pheromones of the ants based on the evaporation rate lf. i ;

[0122] (7) Repeat steps (1)-(6) above until the upper limit of the ant colony algorithm is reached, and output the optimal model result.

[0123] The functions of a, b, c, d, e, f, g, h, i are as follows:

[0124] The Sigmoid function at point a is:

[0125]

[0126] In the formula, x is the weight;

[0127] The loss function MSEloss at point b is:

[0128]

[0129] y i It is the true value, y i ' is the predicted value, and i represents different samples.

[0130] The function at point c is:

[0131] F k =-MSEloss k ;

[0132] In the formula, k represents different ants;

[0133] The function at point d is: min{MSEloss};

[0134] The function at point e is:

[0135] The function at point f is:

[0136] The function at point g is:

[0137] Global update: W(i,j)=W(i,j)+(Sigmoid(R(i,j)-0.5))×lr×W(i,j);

[0138] In the above formula, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix;

[0139] The function at point h is:

[0140] Local update: W(i,j)=W(i,j)+(arctan((R(i,j)-0.5)×2))×lr×W(i,j);

[0141] In the formula, arctan is the arctangent function, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix;

[0142] The function at position i is: F k (t+1)=F k (t)×(1-lf);

[0143] In the formula, lf is the evaporation rate of the ant colony algorithm, and t represents the t-th ant colony algorithm cycle.

[0144] 1.7 Model Validation

[0145] The extraction conditions for safflower seeds obtained from the two optimized models were validated through experiments, and the concentrations of the two components were calculated using the peak areas of the chromatograms. Each set of results was repeated six times. The two models were evaluated using relative error. The relative error was calculated using the following formula:

[0146] Relative error (%) = (predicted value - experimental value) / predicted value × 100% (Formula 3).

[0147] 1.8 In vitro antioxidant experiment of safflower seed extract

[0148] The antioxidant capacity of safflower seed extract under two extraction conditions was evaluated using the DPPH and FRAP methods. The DPPH method involved accurately weighing an appropriate amount of DPPH powder and adding anhydrous ethanol to prepare a 0.3 mM / L DPPH working solution. 180 μL of the DPPH working solution and 20 μL of the sample solution, Trolox solution, or solvent were pipetted into a 96-well plate and thoroughly mixed. After incubation at room temperature in the dark for 30 min, the absorbance was measured at 517 nm. The calculation formula is as follows:

[0149]

[0150] Abs DPPH This is the absorbance of the mixture of DPPH and anhydrous ethanol. Abss and Absc are the absorbances of the sample solutions containing DPPH or anhydrous ethanol, respectively. Refer to the kit instructions for the FRAP method. Trolox is used as a positive reference.

[0151] 1.9 In vivo antioxidant experiment of safflower seed extract

[0152] 1.9.1 Animal Experiments and Design

[0153] Eighteen adult male Sprague-Dawley rats (200±20g) were provided by the Experimental Animal Center of Zhejiang University of Traditional Chinese Medicine (ethics number: IACUC-20220516-15). After one week of acclimatization feeding, the rats were divided into three groups: sham-operated group, model group, and safflower seed extract treatment group, with six rats in each group.

[0154] The MCAO model was prepared using a modified Longa wire embolization method. The modeling method was as follows:

[0155] After weighing, rats were injected intramuscularly with atropine (0.04 mg / kg), followed by an intraperitoneal injection of 40 mg / kg of acetaminophen 50 5 min later. After successful anesthesia, the rats were fixed in a supine position, and their neck hair was shaved. After disinfection with alcohol, the subcutaneous tissue and muscle were dissected along the middle of the neck to expose the blood vessels on the right side of the neck (by left and right sides of the rat). The right common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA) were then isolated. The ECA and CCA were ligated, and the ICA was closed with an arterial clamp. A small incision was then made in the common carotid artery, and a fine silicone-coated surgical nylon monofilament was inserted into the internal carotid artery. After 60 minutes of ischemia, the filament was removed, and the wound was sutured. A Longa score of 1-3 was used as the standard for successful model establishment after the rats recovered. Three days after administration, the animals were sacrificed, and tissue samples were harvested. Rats in the treatment group were administered the drug by gavage (1.3 g / kg, the dosage was based on the amount of raw drug, i.e., the weight of safflower seeds), while rats in the other groups were administered the corresponding physiological saline by gavage.

[0156] The specific preparation method of the safflower seed extract for the treatment group was as follows: extraction was carried out according to the optimized method, namely, ultrasonic frequency of 320W, ethanol volume concentration of 80%, liquid-to-solid ratio of 15mL / g, extraction temperature of 80℃, and extraction time of 60min. After extraction, the supernatant was filtered, rotary evaporated until no alcohol odor was detected, and water was added to the required concentration.

[0157] The criteria for Longa's rating are:

[0158] 0 points: Normal, no neurological deficits;

[0159] 1 point: The contralateral forepaw cannot be fully extended, indicating mild neurological deficit;

[0160] 2 points: When walking, the rat circled to the opposite side (the paralyzed side), indicating moderate neurological deficit;

[0161] 3 points: When walking, the rat's body leans to the opposite side (the paralyzed side), indicating severe neurological deficit;

[0162] 4 points: Unable to walk spontaneously, showing loss of consciousness.

[0163] 1.9.2 Antioxidant Levels

[0164] Peripheral blood was collected from the abdominal aorta of rats. After 30 minutes at room temperature, the blood was centrifuged at 4500 rpm for 15 minutes to obtain the supernatant. Then, SOD and MDA were detected according to the kit instructions (SOD kit purchased from Nanjing Jiancheng Bioengineering Institute, catalog number A001-3; MDA kit purchased from Solarbio Biotechnology Co., Ltd., catalog number BC0025).

[0165] 1.9.3 HO-1 and NOX2 mRNA levels

[0166] use Total RNA was extracted from the brain tissue of rats in each group. Reverse transcription was performed according to the manufacturer's instructions. Real-time qPCR amplification was performed using Universal SYBR Green qPCR Master Mix. HO-1 and NOX2 mRNA levels were detected by real-time quantitative PCR. GAPDH was used to quantify HO-1 and NOX2 mRNA levels. Primer sequences used for RT-qPCR analysis are shown in Table 2.

[0167] Table 2.

[0168]

[0169] The reaction system for real-time quantitative PCR includes: final reaction volume (10 μL); 2x SYBR Qpcr MIX 5 μL; cDNA 1 μL; forward primer (10 μM) 0.2 μL; reverse primer (10 μM) 0.2 μL; and sterile water added to 10 μL.

[0170] The PCR reaction program was as follows: 95℃, 5 min; 95℃, 10 s; 60℃, 30 s; 95℃, 10 s; 60℃, 1 min; 95℃, 30 s, for a total of 40 cycles.

[0171] 1.10 Statistical Analysis

[0172] The software used was GraphPad Prism 9 and Design Expert 13, with Visual Studio Code running Python. One-way ANOVA was used to evaluate the effects on each group. P < 0.05 was considered significant. Significance was expressed as *P < 0.05, **P < 0.01.

[0173] 2. Results

[0174] 2.1. Methodological Survey

[0175] 2.1.1 Specificity

[0176] The chromatographic peaks of the test solution and the two-component mixed reference solution are shown in the figure. Figure 1 The chromatographic peak resolution of both quantitative components at 310 nm was greater than 1.5, indicating ideal chromatographic separation. No interfering peaks were observed in the extraction solvent, demonstrating the good specificity of this method.

[0177] 2.1.2 Linear Relationship

[0178] The linear regression equations for the two components are shown in Table 3. The regression equation for each component is R0. 2 The value >0.999 indicates a positive linear relationship, which can be used to determine the content of each component.

[0179] Table 3 Linear Relationships

[0180] Element Linear relationship between concentration and response value <![CDATA[R 2 ]]> Linear range (μg / mL) CS y = 59325x + 21.96 0.9994 2.4-80 FS y = 34254x + 243.31 0.9994 6.8-200

[0181] 2.1.3 Precision, stability, repeatability, and sample recovery rate

[0182] The method validation results are shown in Table 4. The RSDs for precision and repeatability were both below 2%, and the RSDs for stability were both below 1%. The results indicate that the instrument is stable, the method is reliable, and the samples are stable within 24 hours.

[0183] Table 4. Methodological Survey Results

[0184]

[0185] 2.2 Single-factor experiment

[0186] 2.2.1 Ultrasonic frequency

[0187] The yield results of the two components at different ultrasonic frequencies show that Figure 2 (a) The yield of CS remained stable from an ultrasonic frequency of 80-160W, increased from 160-240W, and remained stable from 240-400W. The yield of FS first increased and then slowly decreased in the range of 80-240W. The reason is that when the ultrasonic frequency is within a certain range, the ultrasonic waves disrupt the plant cell walls, and the exchange of substances between the cells and the solvent tends to be completed as the frequency increases. Therefore, this invention selected a higher yield of 160-320W for subsequent research.

[0188] 2.2.2 Ethanol concentration

[0189] The yield results of the two components at different ethanol concentrations are shown in the figure. Figure 2 (b) In this study, the yield of CS gradually increases in the ethanol concentration range of 60-80% and gradually decreases in the ethanol concentration range of 80-100%. The yield of FS increases in the ethanol concentration range of 60-70% and decreases in the ethanol concentration range of 70-100%. When the concentration is 90-100%, the yields of both CS and FS decrease. Based on the principle of "like dissolves like," CS and FS exhibit better yields and greater solubility at ethanol concentrations of 70-90%, therefore this range was chosen for further investigation in this invention.

[0190] 2.2.3 Liquid-to-solid ratio

[0191] The yield results of the two components at different liquid-to-solid ratios are shown in the figure. Figure 2(c) The CS yield increased in the range of 5-20 mL / g and decreased in the range of 20-25 mL / g. The FS yield increased at 5-10 mL / g, decreased at 10-15 mL / g, increased at 15-20 mL / g, and decreased at 20-25 mL / g. The yield was higher at 15-25 mL / g. Therefore, this invention selected 15-25 mL / g for further research.

[0192] 2.2.4 Extraction Temperature

[0193] The yield results of the two components at different extraction temperatures are shown in the figure. Figure 2 (d) The yield of CS increased in the temperature range of 40-70℃ and tended to stabilize in the temperature range of 70-80℃. The yield of FS showed an increasing trend in the temperature range of 40-80℃. As the temperature increases, the movement of solvent molecules accelerates, promoting the dissolution of components. When the temperature rises to a certain value, heat will disturb the stability of the components and reduce the extraction rate. Therefore, this invention selected the temperature range of 60-80℃ for subsequent research.

[0194] 3.3 RSM Model Results

[0195] The range of factors was selected based on the univariate results. The conditions and results of the RSM experiment are shown in Table 5. Multiple linear regression analysis was performed on the 30 data sets, and the overall evaluation of the relationships with each factor is as follows:

[0196] Y 综合评价价值 =0.185956+0.00365422A+0.00288556B-0.00324564C+0.000973157

[0197] D-0.00154132AB-0.00391808AC-0.0012893AD+0.00995856BC+0.00123389BD+0.00071382CD-0.00988253A 2 -0.00611679B 2 -0.00162696C 2 -0.00686053D 2 .

[0198] The analysis of variance for the comprehensive evaluation value under the RSM model is shown in Table 6. The results show good model fit (P<0.01). From the significance of the P-value and F-value, the influence of the four factors on the safflower seed extraction rate is in the order A>C>B>D, indicating that the ultrasonic frequency has the greatest impact. The interactions among the four factors are as follows: Figure 3As shown, the steepness and contour plot of the three-dimensional response surface can be visualized to reflect the impact of the interactions between various factors on the comprehensive evaluation value of safflower seeds. The steeper the arch of the response surface, the stronger the interaction between the two factors. The optimal extraction results and conditions predicted by the RSM method are shown in Table 7.

[0199] Table 5. Experimental conditions and extraction results of 30 RSM groups

[0200]

[0201]

[0202] Table 6 shows the comprehensive evaluation values ​​of the RSM design based on analysis of variance.

[0203]

[0204] Table 7. Prediction conditions and predicted values ​​for RSM and ACO-BPNN

[0205]

[0206] 2.4 Results of the ACO-BPNN Model

[0207] The ACO-BPNN model was optimized using 30 sets of data from the RSM experiment. The optimization results of the ACO-BPNN model are shown in Table 7. Figure 4 The simulation process shows that the training and predicted values ​​are evenly distributed. This indicates that the ACO-BPNN model has a good fit.

[0208] Extraction rate validation under 2.5RSM and ACO-BPNN optimization conditions

[0209] Two different modeling and calculation methods were used to derive two different sets of optimal extraction conditions, which were then validated (validation conditions are shown in Table 8). The R-values ​​of the two models are... 2 The values ​​were RSM: 0.7915 and ACO-BPNN: 0.9154. This indicates that the ACO-BPNN model has good applicability and high predictive value. Considering practical considerations, all factors were approximated as integers in the validation experiments. The actual conditions for the RSM validation experiment were: ultrasonic frequency 280W, ethanol concentration 74%, liquid-to-solid ratio 15mL / g, and extraction temperature 70℃. The actual conditions for the ACO-BPNN validation experiment were: ultrasonic frequency 320W, ethanol concentration 80%, liquid-to-solid ratio 15ml / g, and extraction temperature 80℃. The prediction results and actual results of the two models are shown in Table 8. Furthermore, based on the actual extraction results, the comprehensive evaluation value of ACO-BPNN is higher than that of RSM, and the relative error of ACO-BPNN is lower than that of RSM, indicating that ACO-BPNN has a stronger predictive effect.

[0210] The reason why the actual value is lower than the predicted value may be due to limitations of the instrument and conditions, or inconsistencies between the actual extraction conditions and the predicted conditions, which caused the error.

[0211] Table 8 Actual and predicted values ​​under RSM and ACO-BPNN prediction conditions

[0212]

[0213]

[0214] 2.6. In vitro antioxidant activity

[0215] Extraction conditions were determined by comparing the actual values ​​under the two models, and the total antioxidant index of the extract was measured (total antioxidant index consists of DPPH and FRAP). The procedures were performed according to the kit instructions (DPPH kit purchased from Shanghai Yuanye Biotechnology Co., Ltd., catalog number S30629; FRAP kit purchased from Shanghai Beyotime Biotechnology Co., Ltd., catalog number S0116), with tricyclic oxide used as a positive control. The DPPH and FRAP results under the RSM model were 1.9889±0.2418 and 1.4643±0.0440 mmol Trolox / g, respectively. The DPPH and FRAP results under the ACO-BPNN model were 2.2819±0.4199 and...

[0216] 1.6367 ± 0.1173 mmol Trolox / g. All calculations were performed based on the dry weight of safflower seeds. Based on the antioxidant results of the two models, the antioxidant activity of safflower seed extract under the ACO-BPNN model was higher than that under the RSM model.

[0217] 2.7 Antioxidant activity in MCAO model rats

[0218] Three days after cerebral ischemia-reperfusion, the serum SOD and MDA levels in each group were as follows: Figure 5 As shown in (a)-(b), this reflects the varying degrees of antioxidant stress resistance in rats after cerebral ischemia-reperfusion. Compared with the sham-operated group, the model group showed a significant decrease in serum SOD levels (P<0.01) and a significant increase in MDA levels (P<0.01). Compared with the model group, the treatment group showed a significant increase in serum SOD levels (P<0.01) and a significant decrease in MDA levels (P<0.05).

[0219] The mRNA expression levels of HO-1 and NOX2 in brain tissue of each group are as follows: Figure 5As shown in (c)-(d) in the figure. Compared with the sham-operated group, the mRNA expression levels of HO-1 and NOX2 in the model group were significantly increased (P<0.01). The HO-1 level in the treatment group was significantly higher than that in the model group (P<0.01), and the NOX2 level was significantly lower than that in the model group (P<0.05).

[0220] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing safflower seed extraction process based on ant colony algorithm and backpropagation neural network, characterized in that, Includes the following steps: (1) Safflower seeds were extracted by ultrasound. Single-factor experiments were conducted, and the ultrasonic power, ethanol concentration, material-liquid ratio and extraction temperature were selected as the influencing factors of the single-factor experiments. (2) The extraction rates of N-(p-coumaroyl)-hydroxytryptamine and N-ferulin hydroxytryptamine were calculated by HPLC, and the optimal values ​​of influencing factors were determined based on the extraction rates; (3) The comprehensive evaluation value is calculated by the entropy weight method, and the optimal process parameters are obtained by backpropagating the neural network model using the ant colony algorithm; The specific steps of the ant colony algorithm backpropagation neural network include: (I) Constructing a backpropagation neural network model 1) Construct a neural network model with n hidden layer neurons, where n = 2, 3, 4, 5, 6; 2) the activation function of the hidden layer is a sigmoid function a and the activation function of the output layer is a linear function. 3) Backpropagation optimizer settings: Optimizer: Adam, learning step size Lr: 0.01; 4) Loss function: MSEloss b ; 5) Use the backpropagation optimizer to optimize the neural network model with 400 epochs based on the initial weights, and find the minimum value of the loss function; (II) Constructing an ant colony algorithm to optimize the neural network model ① Randomly generate ants distributed in different spatial locations and decode the spatial locations, i.e., the initial weights; ② Using the backpropagation neural network model constructed above, the spatial positions of different ants are optimized to find the optimal positions for each ant and record the pheromone F. c ; ③ Encode the weights of each ant and record the optimal position among them. d ; ④ Normalize the ant pheromone F. e ; ⑤ Based on the relationship between the normalized pheromone F and the transition probability p0 f Global update of the spatial location of ants g Or partial update h ; ⑥ Update the ant pheromones based on the evaporation rate lf i ; ⑦ Repeat steps ①-⑥ above until the upper limit of the ant colony algorithm is reached, and output the optimal model result.

2. The method according to claim 1, characterized in that, The Sigmoid function at point a is: ； In the formula, x is the weight; The loss function MSEloss at point b is: ； It is the actual value. These are predicted values, where i represents different samples.

3. The method according to claim 1, characterized in that, The function at point c is: ； In the formula, k represents different ants; The function at point d is: ; The function at point e is: ; The function at point f is: ; The function at point g is: Global update: ; In the formula, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix; The function at point h is: Partial update: ; In the formula, arctan is the arctangent function, lr is the learning rate of the ant colony algorithm, and R(i,j) is a 0-1 random matrix; The function at position i is: ; In the formula, lf is the evaporation rate of the ant colony algorithm, and t represents the t-th ant colony algorithm cycle.

4. The method according to claim 1, characterized in that, The extraction rate (%) mentioned in step (2) = concentration of compound in extract (mg / mL) / [weight of crude drug (mg) / volume of crude drug extraction solvent (mL)] × 100%.

5. The method according to claim 1, characterized in that, The optimal values ​​for the influencing factors determined in step (2) are: ultrasonic power 160-320 W, ethanol concentration 70-90%, liquid-to-solid ratio 15-25 mL / g, and extraction temperature 60-80℃.

6. The method according to claim 1, characterized in that, The comprehensive evaluation value (Y) in step (3) = (extraction rate) CS )×0.5036+(extraction rate) FS )×0.4964.

7. The method according to claim 1, characterized in that, The optimal process parameters in step (3) are: ultrasonic power 320 W, ethanol concentration 80%, liquid-to-solid ratio 15 mL / g, and extraction temperature 80 °C.

Images