Quantum dot labeling detection method for human serum cerebrospinal fluid GFAP antibody

Abstract

The invention relates to the technical field of antibody detection, and provides a quantum dot labeling detection method for a human serum cerebrospinal fluid GFAP antibody. The method comprises the following steps: firstly, collecting a sample and performing pretreatment, including centrifugal separation, dilution treatment and low-temperature preservation; respectively combining the two antibodies of the GFAP antibody with different quantum dot fluorescent dyes to prepare markers, and storing in a dark place; then mixing the pretreated sample with a marker, and performing incubation and washing treatment to obtain a precipitation sample for detection; then, placing the precipitation sample in dual-wavelength fluorescence detection equipment, collecting and recording a fluorescence signal, and obtaining fluorescence intensity data; performing signal denoising and nonlinear correction on the collected fluorescence data, establishing a concentration distribution model of the GFAP antibody by calculating a signal ratio, and outputting probability distribution of the concentration distribution model; and finally, performing optimization analysis on detection data by using the trained deep reinforcement learning model, and outputting a predicted concentration value of the GFAP antibody.

Description

Technical Field

[0001] The present invention relates to the technical field of antibody detection, and more specifically, to a quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid. Background Art

[0002] Glial Fibrillary Acidic Protein (GFAP) antibodies in human serum and cerebrospinal fluid are one of the important biomarkers for evaluating central nervous system damage. GFAP, as a structural protein, is mainly distributed in astrocytes and neural stem cells. When the central nervous system is damaged or inflamed, GFAP will be released into the peripheral blood or cerebrospinal fluid. Therefore, quantitative detection of GFAP antibodies can not only assist in the diagnosis of brain tumors such as glioblastoma, but also has important significance in the study of neural damage, brain trauma and neurodegenerative diseases. Its test results can provide an important reference for the diagnosis, grading and prognosis of the disease.

[0003] Related studies have shown that GFAP has significant value in the diagnosis and evaluation of a variety of central nervous system diseases. Through the analysis of serum samples of patients with different pathological types and healthy controls, researchers found that changes in GFAP levels can help distinguish certain specific brain diseases and have a certain correlation with the pathological characteristics of the disease. Reference 1 (Jung CS, et al., Serum GFAP as a diagnostic marker for glioblastoma multiforme. Brain, 2007) mentioned that GFAP, as a brain-specific marker, shows high sensitivity and specificity in the early detection and typing diagnosis of diseases such as glioblastoma.

[0004] In recent years, a variety of new detection methods have been developed for sensitive detection of GFAP antibodies, including label-free detection methods based on electrochemical technology. These methods improve the sensitivity and efficiency of detection by combining nanomaterials and immunosensor technology. Reference 2 (Ozcelikay G, et al., Ultrasensitive determination of GFAP in human serum-matrix with a label-free impedimetric immunosensor. Biosensors, 2022) proposed a label-free impedance immunosensor, which achieved ultrasensitive detection of GFAP by optimizing electrode surface modification technology, providing a new development direction for GFAP detection technology.

[0005] However, although the existing GFAP antibody detection methods have been significantly improved in terms of sensitivity and detection range, there are still some limitations. For example, traditional detection methods are easily interfered in complex biological sample environments, and the stability and specificity of the test results are difficult to guarantee. At the same time, these methods lack in-depth analysis of the test data and result optimization mechanisms, and cannot fully reflect the concentration distribution and clinical relevance of GFAP antibodies. Therefore, there is an urgent need for a GFAP antibody detection method that can improve detection sensitivity, adapt to complex sample environments, and have the ability to optimize results. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid, and designs a double antibody labeling strategy for GFAP antibodies in quantum dot fluorescent labeling to enhance detection sensitivity; through a deep learning algorithm, the fluorescence signal data is denoised, corrected and the concentration distribution modeling analysis is performed, and the concentration of the GFAP antibody is predicted, which effectively improves the detection sensitivity and data reliability.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid comprises the following steps:

[0009] Step S1, collecting human serum or cerebrospinal fluid samples, centrifuging, diluting, adding protease inhibitors and then storing at low temperature;

[0010] Step S2, respectively combining two antibodies against GFAP antibody with two quantum dot fluorescent dyes to generate quantum dot fluorescent labeling reaction, and after forming the label, the label is stored away from light for later use;

[0011] Step S3, mixing the pretreated sample with the two markers, incubating and shaking, and then washing with a buffer and retaining the precipitated sample;

[0012] Step S4, placing the precipitation sample in a dual-wavelength fluorescence detection device, collecting the first and second fluorescence signals respectively, and recording the fluorescence intensity value;

[0013] Step S5, denoising and nonlinear correction are performed on the collected fluorescence signal, signal ratio is calculated, and a concentration distribution model of GFAP antibody is established, and a probability distribution of GFAP antibody concentration is output;

[0014] Step S6, inputting the generated concentration data into the trained deep reinforcement learning model, optimizing the detection results, and outputting the GFAP antibody prediction value, wherein the reward function of the deep reinforcement learning model is:

[0015]

[0016] In the formula, R is the reward value, C p The predicted value of GFAP antibody concentration generated by the deep reinforcement learning model, C r is the experimental reference concentration value, C MAP is the maximum posterior probability concentration value in the concentration probability distribution, C upper is the upper limit of the concentration confidence interval, C lower is the lower bound of the credible interval of concentration.

[0017] As a further solution of the present invention, step S2 comprises the following steps:

[0018] Step S21, combining the first antibody against GFAP antibody with quantum dot Qdot 525 fluorescent dye to form a first marker through N-hydroxysuccinimide ester coupling reaction;

[0019] Step S22, combining the second antibody against GFAP antibody with quantum dot Qdot 655 fluorescent dye to form a second marker through maleimide coupling reaction;

[0020] Step S23, the labeled antibodies are stored in a dark environment at 4°C, and the storage solution is PBS containing 50 mM triethanolamine buffer.

[0021] As a further embodiment of the present invention, the first antibody is an anti-GFAP monoclonal antibody, and the second antibody is an anti-GFAP polyclonal antibody.

[0022] As a further solution of the present invention, step S3 includes the following steps:

[0023] Step S31, mixing the pretreated sample with the first marker and the second marker in a volume ratio of 1:1:1;

[0024] Step S32, incubating at 37°C for 30 minutes, with slight shaking at intervals of 5 minutes during the incubation process;

[0025] Step S33, washing three times with PBS buffer containing 0.05% Tween-20, each washing condition is centrifugation at 3000 rpm for 5 minutes, discarding the supernatant and retaining the precipitate.

[0026] As a further solution of the present invention, step S4 includes the following steps:

[0027] Step S41, transferring the incubated precipitated sample to a fluorescence detection device with dual-wavelength excitation function;

[0028] Step S42, using an excitation light with a wavelength of 470 nm to excite the sample in the device, collecting a first fluorescence signal, and obtaining a first fluorescence intensity value;

[0029] Step S43: Use excitation light with a wavelength of 555 nm to excite the sample in the device, collect a second fluorescence signal, and obtain a second fluorescence intensity value.

[0030] As a further solution of the present invention, step S5 includes the following steps:

[0031] Step S51, inputting the first fluorescence intensity value and the second fluorescence intensity value obtained in step S4 into a deep learning noise reduction model to perform noise reduction processing on the fluorescence data;

[0032] Step S52, using a third-order polynomial fitting model to quantitatively calibrate the fluorescence intensity data after noise reduction, and establish a nonlinear corresponding relationship between the fluorescence signal intensity and the GFAP antibody concentration, wherein the third-order polynomial fitting formula is:

[0033] S=aC 3 +bC 2 +cC+d;

[0034] In the formula, S is the corrected fluorescence intensity value, C is the GFAP antibody concentration value, a, b, c, d are fitting parameters, which are optimized by gradient descent method;

[0035] Step S53, calculating the ratio of the first fluorescence intensity value and the second fluorescence intensity value after the above calibration to obtain corresponding ratio data;

[0036] Step S54, using a Bayesian inference model to analyze the comparison value data and output a probability density function of the concentration distribution;

[0037] Step S55 , based on the GFAP antibody concentration probability distribution obtained in step S54 , analyzing the probability distribution to determine the concentration credible interval and the maximum posterior probability concentration value in the concentration probability distribution.

[0038] As a further solution of the present invention, the deep learning model is a three-layer convolutional neural network, using a 3×3 convolution kernel and a ReLU activation function, which is used to improve the signal-to-noise ratio of the fluorescence signal.

[0039] As a further solution of the present invention, training the deep reinforcement learning model comprises the following steps:

[0040] Step A1, the concentration probability distribution data generated in step S5 and the experimental reference concentration value together constitute a training data set, wherein the concentration probability distribution data includes the maximum posterior probability concentration value and the concentration credible interval in the concentration probability distribution, which are respectively used as state inputs of the deep reinforcement learning model;

[0041] Step A2, inputting the fluorescence signal ratio data and the GFAP antibody concentration value after nonlinear correction into the deep reinforcement learning model as environmental feedback data, and setting the experimental reference concentration value as the supervised learning benchmark target;

[0042] Step A3, constructing a deep reinforcement learning model to generate a predicted value of GFAP antibody concentration;

[0043] Step A4, defining the reward function of the model, initializing the parameters of the deep reinforcement learning model, setting the learning rate, reward threshold and number of iterations, and training the model through the policy gradient optimization algorithm; during the training process, the model generates concentration prediction values ​​according to the input data, and optimizes the model parameters according to the reward function until the model converges;

[0044] Step A5, verifying the deep reinforcement learning model using test data that has not participated in training;

[0045] Step A6, applying the optimized deep reinforcement learning model to the detection task and outputting the predicted value of GFAP antibody concentration.

[0046] As a further scheme of the present invention, step S1 includes placing the collected human serum or cerebrospinal fluid sample in a 4°C environment for centrifugation at 3000 rpm for 10 minutes to separate the supernatant; adding the separated supernatant to a phosphate buffer containing a protease inhibitor at a mass volume ratio of 0.1% for dilution at a multiple of 1:5; and storing the diluted sample in a -20°C environment for subsequent testing.

[0047] Compared with the prior art, the quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid of the present invention has the following beneficial effects:

[0048] The present invention combines GFAP antibody with two quantum dot fluorescent dyes respectively by designing a dual antibody labeling strategy, thereby enhancing the signal intensity and specificity of the detection. At the same time, the dual-wavelength fluorescence detection technology is used to accurately capture the fluorescence signal, avoiding the interference of background noise, and ensuring stable detection results in complex sample environments. Compared with the label-free detection method that relies on impedance changes in Document 2, the quantum dot labeling strategy of the present invention not only improves the detection sensitivity, but also reduces the influence of nonspecific reactions through dual-channel acquisition of fluorescence signals, making the detection results more stable and reliable.

[0049] The present invention combines the deep reinforcement learning model to denoise and nonlinearly correct the fluorescence signal, and predicts the concentration of GFAP antibody by constructing a concentration distribution model. Specifically, the present invention uses fluorescence intensity ratio calculation and Bayesian inference model in the data analysis stage to provide the probability distribution and credible interval of concentration, and further optimizes the concentration prediction results through the deep reinforcement learning model. Although the method in document 2 has a high initial sensitivity, it lacks the ability of data analysis and result optimization, and cannot accurately reflect the overall picture of concentration distribution. The present invention effectively improves the accuracy of detection and the depth of data analysis through the optimization mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure shows the distribution of serum GFAP values ​​in patients with a single supratentorial space-occupying lesion and the healthy control group.

[0051] Figure 2 Schematic diagram of the construction of label-free GFAP immunosensor.

[0052] Figure 3 The present invention is a schematic flow chart of a quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid.

[0053] Figure 4 A schematic diagram of the process of training a deep reinforcement learning model according to the present invention.

[0054] In the figure, Serum GFAP [μg / l]: serum GFAP [μg / l]; Astro II: astrocytoma grade II; AstroIII: astrocytoma grade III; anapl.Oligo: anaplastic oligodendroglioma; GBM: glioblastoma; singleMetastasis: single metastasis; multiple Metastasis: multiple metastasis; Normal Subjects: normal population; n = 13: sample size = 13; n = 14: sample size = 14; n = 4: sample size = 4; n = 50: sample size = 50; n = 17: sample size = 17; n = 6: sample size = 6; 0.5mM H2SO4: 0.5mmol / L sulfuric acid; CV-0.2–1.2V: cyclic voltammetry scanning voltage range -0.2 to 1.2V; 5μL AuNps: 5 microliters of gold nanoparticles; 5mM L-cysteine: 5 mmol / L L-cysteine; CV-0.6–1.6 V: cyclic voltammetry scanning voltage range -0.6 to 1.6 V; EDC / NHS: 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride / N-hydroxysuccinimide; Anti-GFAP: anti-GFAP antibody; BSA: bovine serum albumin; GFAP: astrocyte fibrillary acidic protein. DETAILED DESCRIPTION

[0055] 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. 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.

[0056] Example 1

[0057] like Figure 1 The distribution diagram of serum GFAP values ​​in patients with a single supratentorial space-occupying lesion and healthy controls is shown, showing the frequency distribution of serum GFAP values ​​measured in patients with a single supratentorial space-occupying lesion and healthy controls. Each point represents the test result of a single GFAP value. The figure shows that GFAP levels are significantly increased in patients with glioblastoma (GBM), with significant differences compared with patients with other types of brain lesions and healthy controls. Specifically, the GFAP values ​​of GBM patients are significantly higher than those of patients with WHO grade II diffuse astrocytoma (Astro II), grade III anaplastic astrocytoma (Astro III), and grade III anapl.Oligo. In addition, the GFAP values ​​of GBM patients are much higher than those of patients with single or multiple brain metastases and normal subjects, who usually do not show detectable GFAP levels in the test. This shows that the significant increase of GFAP in GBM patients can serve as a powerful biomarker for its specific diagnosis.

[0058] A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid, characterized in that it comprises the following steps:

[0059] Step S1, collecting human serum or cerebrospinal fluid samples, centrifuging, diluting, adding protease inhibitors and then storing at low temperature for subsequent testing;

[0060] Step S2, respectively combining two antibodies against GFAP antibody with two quantum dot fluorescent dyes to generate quantum dot fluorescent labeling reaction, and after forming the label, the label is stored away from light for later use;

[0061] Step S3, mixing the pretreated sample with the two markers, incubating and shaking, washing with a buffer and retaining the precipitated sample for subsequent detection;

[0062] Step S4, placing the precipitation sample in a dual-wavelength fluorescence detection device, collecting the first and second fluorescence signals respectively, and recording the fluorescence intensity value;

[0063] Step S5, denoising and nonlinear correction are performed on the collected fluorescence signal, signal ratio is calculated, and a concentration distribution model of GFAP antibody is established, and a probability distribution of GFAP antibody concentration is output;

[0064] Step S6, input the generated concentration data into the trained deep reinforcement learning model, optimize the test results, and output the GFAP antibody prediction value.

[0065] Step S1 in the embodiment of the present invention includes placing the collected human serum or cerebrospinal fluid sample in a 4°C environment for centrifugation at 3000 rpm for 10 minutes to separate the supernatant; adding the separated supernatant to a phosphate buffer containing a protease inhibitor at a mass volume ratio of 0.1% for dilution at a ratio of 1:5; and storing the diluted sample in a -20°C environment for subsequent testing.

[0066] Step S2 in the embodiment of the present invention includes the following steps:

[0067] Step S21, combining the first antibody against GFAP antibody with quantum dot Qdot 525 fluorescent dye to form a first marker through N-hydroxysuccinimide ester coupling reaction;

[0068] Step S22, combining the second antibody against GFAP antibody with quantum dot Qdot 655 fluorescent dye to form a second marker through maleimide coupling reaction;

[0069] Step S23, the labeled antibodies are stored in a dark environment at 4°C, and the storage solution is PBS containing 50 mM triethanolamine buffer.

[0070] In the embodiment of the present invention, the first antibody is an anti-GFAP monoclonal antibody, and the second antibody is an anti-GFAP polyclonal antibody.

[0071] Step S3 in the embodiment of the present invention includes the following steps:

[0072] Step S31, mixing the pretreated sample with the first marker and the second marker in a volume ratio of 1:1:1;

[0073] Step S32, incubating at 37°C for 30 minutes, with slight shaking at intervals of 5 minutes during the incubation process;

[0074] Step S33, washing three times with PBS buffer containing 0.05% Tween-20, each washing condition is centrifugation at 3000 rpm for 5 minutes, discarding the supernatant and retaining the precipitate.

[0075] Step S4 in the embodiment of the present invention includes the following steps:

[0076] Step S41, transferring the incubated precipitated sample to a fluorescence detection device with dual-wavelength excitation function;

[0077] Step S42, using an excitation light with a wavelength of 470 nm to excite the sample in the device, collecting a first fluorescence signal, and obtaining a first fluorescence intensity value;

[0078] Step S43: Use excitation light with a wavelength of 555 nm to excite the sample in the device, collect a second fluorescence signal, and obtain a second fluorescence intensity value.

[0079] Step S5 in the embodiment of the present invention includes the following steps:

[0080] Step S51, inputting the first fluorescence intensity value and the second fluorescence intensity value obtained in step S4 into a deep learning noise reduction model, performing noise reduction processing on the fluorescence data, so that the signal-to-noise ratio of the noise-reduced fluorescence data is at least doubled relative to the original data;

[0081] Step S52, using a third-order polynomial fitting model to quantitatively calibrate the fluorescence intensity data after noise reduction, and establish a nonlinear corresponding relationship between the fluorescence signal intensity and the GFAP antibody concentration, wherein the third-order polynomial fitting formula is:

[0082] S=aC 3 +bC 2 +cC+d;

[0083] In the formula, S is the corrected fluorescence intensity value, C is the GFAP antibody concentration value, a, b, c, d are fitting parameters, which are optimized by gradient descent method;

[0084] Step S53, calculating the ratio of the first fluorescence intensity value and the second fluorescence intensity value after the above calibration to obtain corresponding ratio data;

[0085] Step S54, using the Bayesian inference model to analyze the comparison value data, and outputting the probability density function of the concentration distribution, the expression of the function is:

[0086]

[0087] Wherein, C is the concentration of GFAP antibody, D is the ratio of the first fluorescence intensity value to the second fluorescence intensity value after calibration, p(C|D) is the posterior probability density function of the GFAP antibody concentration C under the condition of given observation data D, that is, the probability density function of the required concentration distribution, p(D|C) is the likelihood function, which indicates the probability of observing data D at a specific concentration C, p(C) is the prior distribution, which indicates the prior knowledge of C when there is no observation data, ∫p(D|C)·p(C)dC is the normalization constant, which ensures that the integral of the posterior distribution over C is 1;

[0088] Step S55 , based on the GFAP antibody concentration probability distribution obtained in step S54 , analyzing the probability distribution to determine the concentration credible interval and the maximum posterior probability concentration value in the concentration probability distribution.

[0089] The deep learning model in the embodiment of the present invention is a three-layer convolutional neural network, using a 3×3 convolution kernel and a ReLU activation function, which is used to improve the signal-to-noise ratio of the fluorescence signal.

[0090] Training the deep reinforcement learning model in an embodiment of the present invention includes the following steps:

[0091] Step A1, the concentration probability distribution data generated in step S5 and the experimental reference concentration value together constitute a training data set, wherein the concentration probability distribution data includes the maximum posterior probability concentration value and the concentration credible interval in the concentration probability distribution, which are respectively used as state inputs of the deep reinforcement learning model;

[0092] Step A2, inputting the fluorescence signal ratio data and the GFAP antibody concentration value after nonlinear correction into the deep reinforcement learning model as environmental feedback data, and setting the experimental reference concentration value as the supervised learning benchmark target;

[0093] Step A3, constructing a deep reinforcement learning model to generate a predicted value of GFAP antibody concentration;

[0094] Step A4, define the reward function of the model, initialize the parameters of the deep reinforcement learning model, set the learning rate, reward threshold and number of iterations, and train the model through the policy gradient optimization algorithm; during the training process, the model generates concentration prediction values ​​according to the input data, and optimizes the model parameters according to the reward function until the model converges, wherein the reward function of the deep reinforcement learning model is:

[0095]

[0096] In the formula, R is the reward value, C p The predicted value of GFAP antibody concentration generated by the deep reinforcement learning model, C r is the experimental reference concentration value, CMAP is the maximum posterior probability concentration value in the concentration probability distribution, C upper is the upper limit of the concentration confidence interval, C lower is the lower limit of the concentration confidence interval;

[0097] Step A5, verifying the deep reinforcement learning model using test data that has not participated in training;

[0098] Step A6, applying the optimized deep reinforcement learning model to the detection task and outputting the predicted value of GFAP antibody concentration.

[0099] Example 2

[0100] like Figure 2 The figure shows the schematic diagram of the construction of the label-free GFAP immunosensor. First, the screen-printed carbon electrode (SPCE) is activated by an electrochemical method to generate hydroxyl groups (-OH) on the surface, laying the foundation for subsequent modification. Subsequently, gold nanoparticles (AuNPs) are added dropwise to the surface of the activated electrode to form a uniform modification layer, thereby enhancing the conductivity and electron transfer ability of the electrode. On this basis, L-cysteine ​​(L-cysteine) is electropolymerized by cyclic voltammetry (CV) to form a self-assembled monolayer (SAM), and its surface carboxyl groups (-COOH) provide active sites for the stable fixation of anti-GFAP antibodies. Next, the carboxyl groups on the surface of L-cysteine ​​are activated using EDC / NHS solution, and the anti-GFAP antibodies are covalently fixed on the electrode surface to achieve high specific recognition of GFAP. Finally, bovine serum albumin (BSA) is used to block the unreacted groups on the electrode surface to reduce nonspecific adsorption and improve the selectivity of the sensor. This design gives the sensor high conductivity and excellent antibody fixation ability, which can achieve sensitive detection and accurate quantification of GFAP concentration.

[0101] However, although the label-free impedance immunosensor method shows high sensitivity in GFAP antibody detection, it is susceptible to interference in complex sample environments, and the stability and specificity of detection are still insufficient.

[0102] like Figure 3 The figure shows a schematic flow diagram of a quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid of the present invention, which provides a quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid, combining quantum dot fluorescence labeling, dual-wavelength fluorescence detection and deep learning optimization analysis, which significantly improves the sensitivity of detection and data reliability. Specifically, the method includes the following steps:

[0103] Step S1, collecting human serum or cerebrospinal fluid samples, centrifuging, diluting, adding protease inhibitors and then storing at low temperature for subsequent testing;

[0104] Step S2, respectively combining two antibodies against GFAP antibody with two quantum dot fluorescent dyes to generate quantum dot fluorescent labeling reaction, and after forming the label, the label is stored away from light for later use;

[0105] Step S3, mixing the pretreated sample with the two markers, incubating and shaking, washing with a buffer and retaining the precipitated sample for subsequent detection;

[0106] Step S4, placing the precipitation sample in a dual-wavelength fluorescence detection device, collecting the first and second fluorescence signals respectively, and recording the fluorescence intensity value;

[0107] Step S5, denoising and nonlinear correction are performed on the collected fluorescence signal, signal ratio is calculated, and a concentration distribution model of GFAP antibody is established, and a probability distribution of GFAP antibody concentration is output;

[0108] Step S6, input the generated concentration data into the trained deep reinforcement learning model, optimize the test results, and output the GFAP antibody prediction value.

[0109] Example 3

[0110] like Figure 4 The flowchart of training a deep reinforcement learning model in the present invention is as follows. The training of a deep reinforcement learning model in an embodiment of the present invention includes the following steps:

[0111] Step A1, the generated concentration probability distribution data and the experimental reference concentration value together constitute a training data set, wherein the concentration probability distribution data includes the maximum posterior probability concentration value and the concentration credible interval in the concentration probability distribution, which are respectively used as state inputs of the deep reinforcement learning model;

[0112] Step A2, inputting the fluorescence signal ratio data and the GFAP antibody concentration value after nonlinear correction into the deep reinforcement learning model as environmental feedback data, and setting the experimental reference concentration value as the supervised learning benchmark target;

[0113] Step A3, constructing a deep reinforcement learning model, the model includes a three-layer convolutional neural network, each layer has a convolution kernel size of 3×3, an activation function of ReLU, and an output layer is a fully connected layer, for generating a predicted value of GFAP antibody concentration;

[0114] Step A4, define the reward function of the model, initialize the parameters of the deep reinforcement learning model, set the learning rate, reward threshold and number of iterations, and train the model through the policy gradient optimization algorithm; during the training process, the model generates concentration prediction values ​​according to the input data, and optimizes the model parameters according to the reward function until the model converges, wherein the reward function of the deep reinforcement learning model is:

[0115]

[0116] In the formula, R is the reward value, C p GFAP antibody concentration prediction value generated by the deep learning reinforcement model, C r is the experimental reference concentration value, C MAP is the maximum posterior probability concentration value in the concentration probability distribution, C upper is the upper limit of the concentration confidence interval, C lower is the lower limit of the concentration confidence interval;

[0117] Step A5, using test data that has not participated in training to verify the deep reinforcement learning model, evaluate the mean absolute error, mean square error and credible interval coverage of the model-predicted concentration value, and ensure that the prediction performance of the model meets the set threshold;

[0118] Step A6, applying the optimized deep reinforcement learning model to the detection task and outputting the predicted value of GFAP antibody concentration.

[0119] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0120] Finally: 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 in the protection scope of the present invention.

Claims

1. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid, characterized in that: The following steps are involved: Step S1, collecting human serum or cerebrospinal fluid samples, centrifuging, diluting, adding protease inhibitors and then storing at low temperature; Step S2, combining two antibodies against GFAP antibody with two quantum dot fluorescent dyes to generate quantum dot fluorescent labeling reaction, and after forming the label, the label is stored away from light for later use; Step S3, mixing the pretreated sample with the two markers, incubating and shaking, washing with a buffer and retaining the precipitated sample; Step S4, placing the precipitation sample in a dual-wavelength fluorescence detection device, collecting the first and second fluorescence signals respectively, and recording the fluorescence intensity value; Step S5, denoising and nonlinear correction are performed on the collected fluorescence signal, signal ratio is calculated, and a concentration distribution model of GFAP antibody is established, and a probability distribution of GFAP antibody concentration is output; Step S6, inputting the generated concentration data into the trained deep reinforcement learning model, optimizing the detection results, and outputting the GFAP antibody prediction value, wherein the reward function of the deep reinforcement learning model is: In the formula, R is the reward value, C p The predicted value of GFAP antibody concentration generated by the deep learning reinforcement model, C r is the experimental reference concentration value, C MAP is the maximum posterior probability concentration value in the concentration probability distribution, C upper is the upper limit of the concentration confidence interval, C lower is the lower bound of the credible interval of concentration.

2. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Step S2 includes the following steps: Step S21, combining the first antibody against GFAP antibody with quantum dot Qdot 525 fluorescent dye to form a first marker through N-hydroxysuccinimide ester coupling reaction; Step S22, combining the second antibody against GFAP antibody with quantum dot Qdot 655 fluorescent dye to form a second marker through maleimide coupling reaction; Step S23, the labeled antibodies are stored in a dark environment at 4°C, and the storage solution is PBS containing 50 mM triethanolamine buffer.

3. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1 or 2, characterized in that: The first antibody is an anti-GFAP monoclonal antibody, and the second antibody is an anti-GFAP polyclonal antibody.

4. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Step S3 includes the following steps: Step S31, mixing the pretreated sample with the first marker and the second marker in a volume ratio of 1:1:1; Step S32, incubating at 37°C for 30 minutes, with slight shaking at intervals of 5 minutes during the incubation process; Step S33, washing three times with PBS buffer containing 0.05% Tween-20, each washing condition is centrifugation at 3000 rpm for 5 minutes, discarding the supernatant and retaining the precipitate.

5. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Step S4 includes the following steps: Step S41, transferring the incubated precipitated sample to a fluorescence detection device with dual-wavelength excitation function; Step S42, using an excitation light with a wavelength of 470 nm to excite the sample in the device, collecting a first fluorescence signal, and obtaining a first fluorescence intensity value; Step S43: Use excitation light with a wavelength of 555 nm to excite the sample in the device, collect a second fluorescence signal, and obtain a second fluorescence intensity value.

6. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Step S5 includes the following steps: Step S51, inputting the first fluorescence intensity value and the second fluorescence intensity value obtained in step S4 into a deep learning model to perform noise reduction processing on the fluorescence data; Step S52, using a third-order polynomial fitting model to quantitatively calibrate the fluorescence intensity data after noise reduction, and establish a nonlinear corresponding relationship between the fluorescence signal intensity and the GFAP antibody concentration, wherein the third-order polynomial fitting formula is: S=aC 3 +bC 2 +cC+d; In the formula, S is the corrected fluorescence intensity value, C is the GFAP antibody concentration value, a, b, c, d are fitting parameters, which are optimized by gradient descent method; Step S53, calculating the ratio of the first fluorescence intensity value and the second fluorescence intensity value after the above calibration to obtain corresponding ratio data; Step S54, using a Bayesian inference model to analyze the comparison value data and output a probability density function of the concentration distribution; Step S55 , based on the GFAP antibody concentration probability distribution obtained in step S54 , analyzing the probability distribution to determine the concentration credible interval and the maximum posterior probability concentration value in the concentration probability distribution.

7. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1 or 6, characterized in that: The deep learning model is a three-layer convolutional neural network, which uses a 3×3 convolution kernel and a ReLU activation function to perform noise reduction on the fluorescence data.

8. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Training the deep reinforcement learning model includes the following steps: Step A1, the concentration probability distribution data generated in step S5 and the experimental reference concentration value together constitute a training data set, wherein the concentration probability distribution data includes the maximum posterior probability concentration value and the concentration credible interval in the concentration probability distribution, which are respectively used as state inputs of the deep reinforcement learning model; Step A2, inputting the fluorescence signal ratio data and the GFAP antibody concentration value after nonlinear correction into the deep reinforcement learning model as environmental feedback data, and setting the experimental reference concentration value as the supervised learning benchmark target; Step A3, constructing a deep reinforcement learning model to generate a predicted value of GFAP antibody concentration; Step A4, defining the reward function of the model, initializing the parameters of the deep reinforcement learning model, setting the learning rate, reward threshold and number of iterations, and training the model through the policy gradient optimization algorithm; during the training process, the model generates concentration prediction values ​​according to the input data, and optimizes the model parameters according to the reward function until the model converges; Step A5, verifying the deep reinforcement learning model using test data that has not participated in training; Step A6, applying the optimized deep reinforcement learning model to the detection task and outputting the predicted value of GFAP antibody concentration.

9. A quantum dot labeling detection method for GFAP antibodies in human serum and cerebrospinal fluid according to claim 1, characterized in that: Step S1 includes placing the collected human serum or cerebrospinal fluid sample in a 4°C environment for centrifugation at 3000 rpm for 10 minutes to separate the supernatant; adding the separated supernatant to a phosphate buffer containing a protease inhibitor at a mass volume ratio of 0.1% to dilute it at a dilution factor of 1:5; and storing the diluted sample in a -20°C environment for subsequent testing.

Images