A method for quantitative determination of a neurotransmitter mixture based on artificial neural networks
By constructing a method based on fluorescent probes and artificial neural networks, the problem of difficulty in quantitatively detecting multiple catechol neurotransmitters in existing technologies has been solved, enabling accurate detection of the concentration of neurotransmitter mixtures and expanding the scope of applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-29
- Publication Date
- 2026-04-07
AI Technical Summary
Existing technologies are difficult to rapidly and accurately quantify multiple catechol neurotransmitters simultaneously. In particular, array methods cannot predict the concentration ratio of unknown samples, and the total concentration of the sample must be known before detection.
Using an artificial neural network-based approach, a four-layer fully connected neural network model was constructed by reacting a fluorescent probe solution with a neurotransmitter at a specific pH value and combining it with fluorescence spectroscopy analysis. The neural network was then trained using fluorescence intensity data to achieve the concentration detection of a mixture of neurotransmitters.
It enables accurate detection of the concentration of each component in a neurotransmitter mixture, avoiding the sample dependence limitation of array methods and has broader application potential.
Smart Images
Figure CN116359189B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of analytical chemistry technology and relates to a method for the quantitative determination of neurotransmitter mixtures based on artificial neural networks. Background Technology
[0002] Neurotransmitters play a crucial role in transmitting impulses between cells in the cardiovascular, endocrine, renal, and central nervous systems. Within the human body, neurotransmitters, as substances that transmit information from the brain, influence our thinking, movement, work, and other behaviors. The pathogenesis of many diseases is often related to decreased neurotransmitter concentrations in the bloodstream, such as heart failure, neuromuscular disorders, Parkinson's disease, and schizophrenia. Emotions, such as happiness and excitement, are caused by increased dopamine levels in the body. Since abnormal neurotransmitter levels are associated with various diseases, the ability to quickly and accurately detect neurotransmitters is of significant practical importance.
[0003] However, selective detection of neurotransmitter mixtures has always been a difficult challenge, especially the simultaneous quantitative detection of multiple catechol neurotransmitters. In early analytical chemistry, array-based sensors, also known as "chemical noses / tongues," were commonly used for multianalyte sensing. This method utilizes the classification capabilities of arrays to distinguish neurotransmitters in different proportions. However, this array method has significant disadvantages. First, array methods can only classify and cannot predict. That is, if a new mixture sample is to be detected, the concentration ratio of that sample must exist in a database, which inevitably leads to a large amount of preliminary work and poor versatility. Second, this method requires a prerequisite: the total concentration of the sample must be known. However, this is impossible when processing blind samples. Therefore, using array methods to solve the problem of predicting the concentration ratio of neurotransmitter mixtures without separation is not practical.
[0004] Artificial Neural Networks (ANNs) have been a hot research topic in the field of machine learning within artificial intelligence since the 1980s. They simulate the human brain's neural network processing of external information, creating a simple model that combines different networks using various connection methods. An ANN is a computational model composed of a large number of interconnected neurons. Each node represents a specific output function. The connection between any two nodes represents a weighted value for the signal passing through that connection, which is equivalent to the memory of the artificial neural network. The output of an ANN relies on the neural network's analysis and fitting of past data to optimize its weights (a process called training). This trained neural network is then used to fit unfamiliar data to obtain the output data. Summary of the Invention
[0005] The purpose of this invention is to provide a method for quantitative determination of neurotransmitter mixtures based on artificial neural networks.
[0006] The method for detecting and analyzing neurotransmitters based on artificial neural networks is as follows:
[0007] (1) Prepare a 3,5-dihydroxybenzoic acid solution with a concentration of 200 μM-1 mM and a pH of 8-11 as a probe solution. Then, add the same volume of neurotransmitter solutions with different total concentrations to the probe solution and react for 120±5 s. Measure the fluorescence intensity. Select an excitation wavelength within the range of 340±10 nm to obtain the highest fluorescence intensity of the emission spectrum and plot a standard curve with the total concentration of the neurotransmitter solution. Add the neurotransmitter solution to be tested to the probe solution and measure the highest fluorescence intensity under the same conditions. Calculate the total concentration of neurotransmitters in the neurotransmitter solution to be tested using the standard curve.
[0008] (2) Prepare a group-modified monobenzene ring fluorescent molecular probe solution with a concentration of 200 μM-1 mM and a pH of 8-11; prepare neurotransmitter solution samples composed of dopamine, levodopa, levodopa methyl ester and norepinephrine in different concentration ratios; add the same volume of each neurotransmitter solution sample to the group-modified monobenzene ring fluorescent molecular probe solution and react for 120±5s before performing fluorescence spectroscopy testing. Select the excitation wavelength in the range of 250-450nm to obtain the highest fluorescence intensity of the emission spectrum; input the concentration ratio of dopamine, levodopa, levodopa methyl ester and norepinephrine and the corresponding highest fluorescence intensity of the emission spectrum into the artificial neural network for training to obtain the neural network model;
[0009] (3) Add the neurotransmitter solution to be tested to the group-modified monobenzene ring fluorescent molecular probe solution prepared in step (2), and obtain the highest fluorescence intensity of the emission spectrum under the same conditions. Then input the solution into the neural network model to calculate the concentration ratio of each neurotransmitter in the neurotransmitter solution to be tested.
[0010] (4) The concentration of each neurotransmitter is calculated based on the total concentration of the neurotransmitter solution obtained in step (1) and the concentration ratio of each neurotransmitter obtained in step (3).
[0011] The neurotransmitter is one or more of dopamine, levodopa, levodopa methyl ester, and norepinephrine.
[0012] In step (1), the concentration range of neurotransmitters after the probe solution is added to the neurotransmitter solution is 100-1000 nM.
[0013] The structural formula of the group-modified monobenzene ring fluorescent molecular probe is one of the following:
[0014]
[0015] Where X = H, CH3, COOH, NO2, Cl, Br, I, or CH2CH3.
[0016] Step (2) Prepare at least two different group-modified monobenzene ring fluorescent molecular probe solutions and react them with neurotransmitter solution samples respectively.
[0017] In step (2), the sum of the number of excitation wavelengths selected for each group-modified monobenzene ring fluorescent molecular probe solution is not less than 7; the interval between two adjacent excitation wavelengths selected for each group-modified monobenzene ring fluorescent molecular probe solution is not less than 10 nm.
[0018] Step (2) Prepare at least 400 neurotransmitter solution samples with different concentration ratios, wherein the contents of dopamine, levodopa, levodopa methyl ester and norepinephrine are uniformly selected in the range of 0-100%.
[0019] The concentration of neurotransmitters in the neurotransmitter solution sample after adding a group-modified monobenzene ring fluorescent molecular probe solution ranged from 100 to 1000 nM.
[0020] The neurotransmitter solution sample also includes one or more of the following: solutions of different concentrations prepared from one, two, or three of dopamine, levodopa, levodopa methyl ester, and norepinephrine.
[0021] The artificial neural network was created using the TensorFlow framework in Python, with the following specific parameters: 4 layers, with 20 neurons in the first hidden layer, 12 neurons in the second hidden layer, and 5 neurons in the third hidden layer, and the ReLU activation function for the hidden layers.
[0022] The quantitative determination method for neurotransmitter mixtures based on artificial neural networks of the present invention can perfectly detect the concentration of each component of the neurotransmitter mixture. Compared with array sensor methods, artificial neural networks do not have many limitations on the test samples and can be more widely used in the medical and chemical fields. Attached Figure Description
[0023] Figure 1 The fluorescence spectra of twelve fluorescent products resulting from the cross-reaction of three types of resorcinol derivatives and four types of catechol neurotransmitters in Example 1 are shown.
[0024] Figure 2 The principle and fitting curve of using the monobenzene ring fluorescent probe 3,5-dihydroxybenzoic acid to detect the total concentration of a neurotransmitter mixture solution.
[0025] Figure 3 This serves as the basis for selecting the input layer of an artificial neural network obtained by reacting three types of resorcinol derivatives and four types of catechol neurotransmitters at different excitation wavelengths.
[0026] Figure 4 This is a diagram of an artificial neural network structure used to detect mixtures of neurotransmitters.
[0027] Figure 5 This is a performance characterization diagram of an artificial neural network used to detect mixtures of neurotransmitters. Detailed Implementation
[0028] To illustrate the present invention more clearly, the following embodiments are provided, but the scope of protection of the present invention is not limited to the following embodiments.
[0029] Example 1
[0030] (I) Fluorescence Properties Study of Resorcinol Derivatives and Catechol Neurotransmitters: 200 μM solutions of 3,5-dihydroxybenzoic acid (DHBA), resorcinol, and 3,5-dihydroxybenzane (DHPM) were prepared. Na₂CO₃ was added to adjust the pH to 10. Then, 30 μM solutions of dopamine, levonorgestrel, levonorgestrel methyl ester, and norepinephrine were added, respectively. The volume ratio of the resorcinol derivative solution to the neurotransmitter solution was 1:1. The reaction was carried out by stirring at room temperature for 120 s to obtain the product solution. The structural formulas of the products are shown in Table 1. The fluorescence spectra were measured using a fluorescence spectrophotometer, and the results are as follows: Figure 1 As shown in the figure, the results indicate that there are significant optical differences among the twelve fluorescent products of the cross-reaction of three types of resorcinol derivatives and four types of catechol neurotransmitters. It is feasible to use these optical differences to construct an artificial neural network to detect the concentration ratio of neurotransmitter mixtures.
[0031] Table 1
[0032]
[0033] (II) Detection of the total concentration of the neurotransmitter mixture: A 200 μM solution of 3,5-dihydroxybenzoic acid was prepared, and the pH was adjusted to 10 with Na2CO3. Then, 30 μM solutions of dopamine, levonorgestrel, levonorgestrel methyl ester, and norepinephrine were added separately. The volume ratio of the 3,5-dihydroxybenzoic acid solution to the neurotransmitter solution was 1:1. The reaction was carried out by stirring at room temperature for 120 s to obtain the product solution. The 3,5-dihydroxybenzoic acid solution itself has a fluorescence peak at 411 nm, and this fluorescence peak decreases with the reaction of the probe and the neurotransmitter. The degree of decrease does not change with the type of neurotransmitter. Figure 2(a, b) Using this property, we designed an experiment to detect the total concentration of a neurotransmitter mixture solution.
[0034] To obtain the total neurotransmitter concentration ratio equation: Prepare a 200 μM solution of 3,5-dihydroxybenzoic acid, adjust the pH to 10 with Na2CO3, and add neurotransmitter solutions of concentrations of 0 nM, 100 nM, 200 nM, 300 nM, 400 nM, and 500 nM. Figure 2 As shown in (b), the fluorescence intensity of DHBA reacting with different neurotransmitters decreased by the same amount when the neurotransmitter concentrations were equal. Therefore, dopamine solution was used directly instead of neurotransmitter solution. The volume ratio of 3,5-dihydroxybenzoic acid solution to dopamine solution was 1:1. After stirring at room temperature for 120 s, the fluorescence emission spectrum was measured using an excitation wavelength of 340 nm. A standard curve was plotted between the strongest peak of the reaction product of 3,5-dihydroxybenzoic acid and dopamine in the obtained fluorescence emission spectrum and the concentration of dopamine solution. Figure 2 (c, d) yields the low-concentration dopamine ratio equation F. l = -0.628C(nM) + 3025.58.
[0035] (III) Construction of an artificial neural network for predicting the proportion of neurotransmitter mixtures: Different excitation wavelengths were selected using excitation spectra with different fluorescence spectra. The excitation wavelengths for the reaction between resorcinol and neurotransmitters were 260 nm, 350 nm, 395 nm, and 417 nm; the excitation wavelength for the reaction between DHBA and neurotransmitters was 417 nm; and the excitation wavelengths for the reaction between DHPM and neurotransmitters were 320 nm and 417 nm, respectively. Figure 3 The fluorescence intensity of the emission spectra corresponding to these excitation wavelengths was used as the input layer of the neural network, successfully converting the two-dimensional spectral data into one-dimensional digital data. Using this data as the input layer, a four-layer fully connected neural network was constructed. The input layer contained seven-dimensional data, and the output layer contained three-dimensional data. The first hidden layer had 20 neurons, the second had 12 neurons, and the third had 5 neurons. The neural network used the ReLU function as the activation function for the hidden layers, such as... Figure 4 All analyses were performed in a Python environment.
[0036] Database construction for artificial neural networks: A 200 μM resorcinol solution was prepared, and the pH was adjusted to 10 with Na2CO3. Then, a mixed solution of dopamine, levodopa, levodopa methyl ester, and norepinephrine with different concentrations and ratios (totaling 100 nM, totaling 485 groups; the contents of dopamine, levodopa, levodopa methyl ester, and norepinephrine were uniformly selected within the range of 0-100%) was added. The volume ratio of resorcinol solution to neurotransmitter mixture solution was 1:1, and the reaction was carried out by stirring at room temperature for 120 s to obtain the product solution. Fluorescence emission spectra at excitation wavelengths of 260 nm, 350 nm, 395 nm, and 417 nm were collected using an F-7000 fluorescence spectrophotometer. Prepare 200 μM DHBA and DHPM solutions, repeat the above steps, and collect fluorescence emission spectra using an F-7000 fluorescence spectrophotometer at excitation wavelengths of 417 nm and 320 nm, respectively. Integrate the above data to obtain a 7*485 Excel spreadsheet containing peak fluorescence emission spectra data of different concentration ratios and different excitation wavelengths of different resorcinol analogs.
[0037] Performance evaluation of artificial neural networks: The 485 columns of data were divided, with 90% selected as the training set and 10% as the test set. The training set was used for training and calibration, and then validated using a new test set with unknown samples, as shown in Table 2.
[0038] Table 2
[0039]
[0040] Using the loss function (Loss) and correlation coefficient (R) 2 () is used as an indicator of the quality of the predictive model. For example... Figure 5 As the number of iterations increases, the loss function gradually approaches 0, while R... 2 A value close to 1 indicates that the model fits as expected. The loss function represents the standard deviation between the predicted and observed values, R0. 2 This represents the variance of the dependent variable explained by the model; in other words, it reflects the goodness of fit between the model and the data (R²). 2 ).
[0041] Application Example 1
[0042] Dopamine detection in PC12 cells: Rat adrenal pheochromocytoma PC12 cells were maintained in CM2-1 medium at 37°C with humidified air containing 5% CO2. The medium was changed daily throughout the cell lifespan. 4mM K +PC12 cells were stimulated for 30 minutes to promote the release of neurotransmitters. PC12 cells were digested with trypsin to obtain a cell suspension. The cell suspension was centrifuged at 1000g for 5 minutes to obtain a cell pellet. The cells were resuspended and centrifuged again to obtain clear PC12 cell lysate. The centrifuged cell lysate sample was diluted 100-fold (PC12 cells) as the test sample.
[0043] First, the PC12 neurotransmitter sample was added to a 200 μM solution of 3,5-dihydroxybenzoic acid, and the pH was adjusted to 10 with Na2CO3. The volume ratio of the 3,5-dihydroxybenzoic acid solution to the diluted PC12 cell lysate was 1:1. After stirring at room temperature for 120 s, the fluorescence emission spectrum was measured using an excitation wavelength of 340 nm. The fluorescence intensity at 411 nm in the obtained fluorescence emission spectrum was substituted into the ratio equation obtained earlier to calculate the total concentration of the neurotransmitter mixture as 168.18 nm. The previously diluted PC12 cell lysate was then diluted to 100 nm for use.
[0044] A 100 nm PC12 neurotransmitter sample was added to a resorcinol probe (pH=10), with a final probe concentration of 200 μM. After reacting at 25 °C for 10 min, fluorescence emission spectra at excitation wavelengths of 260 nm, 350 nm, 395 nm, and 417 nm were collected using an F-7000 fluorescence spectrophotometer. This process was repeated. DHBA and DHPM probes were added to the same neurotransmitter sample (pH=10), with probe concentrations greater than 200 μM. After reacting at 25 °C for 10 min, fluorescence emission spectra were collected using an F-7000 fluorescence spectrophotometer at excitation wavelengths of 417 nm and 320 nm, and 417 nm, respectively. The above-mentioned test samples were taken, and DA and DOPA, with concentrations equal to the total neurotransmitter concentration, were added to the test samples. These were designated as the second and third groups of test samples. The above operation was repeated, and good detection results were still obtained. Moreover, the detection results obtained by this method were basically the same as those obtained by high-performance liquid chromatography, indicating that the artificial neural network sensor of this invention has good potential for detecting intracellular neurotransmitter mixtures. This experiment was repeated three times for the final calculation of RSD. The data are shown in Table 3.
[0045] Table 3
[0046]
Claims
1. A method for detecting and analyzing neurotransmitters based on artificial neural networks, characterized in that, The specific steps of the method are as follows: (1) Prepare a 3,5-dihydroxybenzoic acid solution with a concentration of 200 μM-1 mM and a pH of 8-11 as a probe solution. Then, add the same volume of neurotransmitter solutions with different total concentrations to the probe solution and react for 120±5 s. Measure the fluorescence intensity. Select an excitation wavelength within the range of 340±10 nm to obtain the highest fluorescence intensity of the emission spectrum and plot a standard curve with the total concentration of the neurotransmitter solution. Add the neurotransmitter solution to be tested to the probe solution and measure the highest fluorescence intensity under the same conditions. Calculate the total concentration of neurotransmitters in the neurotransmitter solution to be tested using the standard curve. (2) Prepare a group-modified monobenzene ring fluorescent molecular probe solution with a concentration of 200 μM-1 mM and a pH of 8-11; prepare neurotransmitter solution samples composed of dopamine, levodopa, levodopa methyl ester and norepinephrine in different concentration ratios; add the same volume of each neurotransmitter solution sample to the group-modified monobenzene ring fluorescent molecular probe solution and react for 120±5s before performing fluorescence spectroscopy testing. Select the excitation wavelength in the range of 250-450nm to obtain the highest fluorescence intensity of the emission spectrum; input the concentration ratio of dopamine, levodopa, levodopa methyl ester and norepinephrine and the corresponding highest fluorescence intensity of the emission spectrum into the artificial neural network for training to obtain the neural network model; (3) Add the neurotransmitter solution to be tested to the group-modified monobenzene ring fluorescent molecular probe solution prepared in step (2), and obtain the highest fluorescence intensity of the emission spectrum under the same conditions. Then input the solution into the neural network model to calculate the concentration ratio of each neurotransmitter in the neurotransmitter solution to be tested. (4) The concentration of each neurotransmitter is calculated based on the total concentration of the neurotransmitter solution obtained in step (1) and the concentration ratio of each neurotransmitter obtained in step (3).
2. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, The neurotransmitter is one or more of dopamine, levodopa, levodopa methyl ester, and norepinephrine.
3. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, In step (1), the concentration range of neurotransmitters after the probe solution is added to the neurotransmitter solution is 100-1000 nM.
4. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, The structural formula of the group-modified monobenzene ring fluorescent molecular probe is one of the following: Where X = H, CH3, COOH, NO2, Cl, Br, I, or CH2CH3.
5. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 4, characterized in that, Step (2) Prepare at least two different group-modified monobenzene ring fluorescent molecular probe solutions and react them with neurotransmitter solution samples respectively.
6. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 5, characterized in that, The sum of the number of excitation wavelengths corresponding to the monobenzene ring fluorescent molecular probe solutions modified by each group is not less than 7; the interval between two adjacent excitation wavelengths corresponding to the monobenzene ring fluorescent molecular probe solutions modified by each group is not less than 10 nm.
7. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, Step (2) Prepare at least 400 neurotransmitter solution samples with different concentration ratios, wherein the contents of dopamine, levodopa, levodopa methyl ester and norepinephrine are uniformly selected in the range of 0-100%.
8. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, The concentration of neurotransmitters in the neurotransmitter solution sample after adding a group-modified monobenzene ring fluorescent molecular probe solution ranged from 100 to 1000 nM.
9. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, The neurotransmitter solution sample also includes one or more of the following: solutions of different concentrations prepared from one, two, or three of dopamine, levodopa, levodopa methyl ester, and norepinephrine.
10. The method for detecting and analyzing neurotransmitters based on artificial neural networks according to claim 1, characterized in that, The artificial neural network was created using the TensorFlow framework in Python, with the following specific parameters: 4 layers, with 20 neurons in the first hidden layer, 12 neurons in the second hidden layer, and 5 neurons in the third hidden layer, and the ReLU activation function for the hidden layers.