Electrochemical immunoassay method based on artificial intelligence

By forming an efficient immune sensor on the electrode and combining artificial intelligence technology, using cyclic voltammetry and multiple linear regression models, the problem of insufficient operational complexity and sensitivity of traditional immune detection methods is solved, and fast and accurate detection of antigen concentration is achieved , suitable for the efficient, convenient and economical testing needs of modern biomedical fields.

CN119044291APending Publication Date: 2024-11-29NORTHEASTERN UNIV AT QINHUANGDAO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411162861.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Traditional immune detection methods have shortcomings in operational complexity, detection time, equipment requirements, cost control and detection sensitivity, and are difficult to meet the needs of modern biomedical fields for efficient, convenient and economical immune detection.

Method used

Using an electrochemical immunodetection method based on artificial intelligence, an efficient immune sensor is formed by chemical modification on the working electrode, and the electrochemical signal is captured by cyclic voltammetry, combined with a random forest algorithm and a multivariate linear regression model to predict antigen concentrations, optimized Data processing and feature selection enable fast and accurate detection.

Benefits of technology

It significantly improves the accuracy and sensitivity of the test, reduces the consistency and repetition of the experimental environment on the results, reduces the detection cost, improves the detection efficiency and reliability of the results, and is suitable for clinical diagnosis, environmental monitoring and food safety testing. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure HDA0005006944710000011
    Figure HDA0005006944710000011
  • Figure HDA0005006944710000012
    Figure HDA0005006944710000012
  • Figure HDA0005006944710000021
    Figure HDA0005006944710000021
Patent Text Reader

Abstract

The invention discloses an electrochemical immunoassay method based on artificial intelligence, relates to the technical fields of disease diagnosis, food safety, environmental monitoring and the like, and aims to realize rapid and accurate automatic detection of a target analyte by utilizing electrochemical data. According to the invention, artificial intelligence and electrochemical immunodetection are innovatively combined to form a detection system capable of efficiently and accurately analyzing electrochemical signals. The method comprises the following steps: firstly, constructing an electrochemical detection system based on an immunosensor, and capturing a target analyte through an antibody fixed on the surface of an electrode plate; then measuring a current curve of the target analyte by using cyclic voltammetry, and extracting eight characteristic parameters of the cyclic voltammetry curve; and finally, training and modeling the extracted characteristic parameters by utilizing a machine learning algorithm so as to predict the concentration of the analyte. By introducing a machine learning algorithm, full-automatic detection is realized, the accuracy of prediction is remarkably improved through accurate data model fitting, the problem of detection inconsistency caused by antibody specificity is effectively reduced, and the reliability and repeatability of detection are enhanced. The method overcomes the defects that a traditional immunoassay method cannot give consideration to sensitivity, operation convenience, detection time and cost, expands the linear range of detection, achieves the effect of improving the detection efficiency and accuracy, is low in cost, and provides wide application prospects for the fields of clinical diagnosis, environmental monitoring, food safety detection and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method for quantitatively detecting the concentration of an immunoanalyte, and in particular to a method for accurately and rapidly detecting an immunoanalyte by establishing a multivariate regression model using artificial intelligence technology and fitting electrochemical signals. Background Art

[0002] Immunoassay technology uses the specific binding between antigens and antibodies to detect the presence and concentration of target molecules and is widely used in medical diagnosis, food safety and environmental monitoring. Enzyme-linked immunosorbent assay (ELISA) is a commonly used immunoassay technology in laboratories. It is performed by fixing antibodies or antigens on a solid phase carrier, using antibodies labeled with enzymes to bind to target molecules, and then quantitatively detecting the target through an enzyme-catalyzed color reaction. Although ELISA has high sensitivity and specificity, its operation is complex and the steps are cumbersome. It depends on the quality of reagents, and its linear detection range is narrow, which limits its applicability in certain applications. Lateral flow immunoassay (LFIA) is a fast, portable and simple technology that is often used for rapid on-site screening. This method detects by allowing the sample to flow laterally on a membrane containing antibodies, and the signal is visualized by a marker. LFIA is suitable for occasions where rapid screening is required, but its sensitivity is low and it is mainly used for qualitative or semi-quantitative detection. The accuracy of the results depends on manual readings and there may be batch-to-batch differences. Electrochemical immunoassay technology detects by monitoring the changes in the electrochemical signal generated after the target analyte binds to the fixed antibody. This technology is easy to operate, has a short detection time and low cost. However, electrochemical immunoassay has low sensitivity in the detection of low-concentration target analytes, and the experimental conditions have a great impact on the consistency and repeatability of the results, which may lead to unstable test results. The challenges faced by existing technologies include how to improve accuracy and sensitivity while taking into account the ease of operation, detection time and cost, and reducing the impact of the experimental environment on the consistency and repeatability of the results to meet more demanding application scenarios. Summary of the invention

[0003] The technical problem to be solved by the present invention is that the traditional immunoassay method has many deficiencies in operation complexity, detection time, equipment requirements, cost control and detection sensitivity, and it is difficult to meet the demand for efficient, convenient and economical immunoassay in the modern biomedical field. Therefore, the present invention aims to provide a fast, accurate and efficient immunoanalyte concentration detection scheme through a new electrochemical immunoassay method based on artificial intelligence technology.

[0004] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0005] An artificial intelligence-based immunoassay concentration detection method, comprising the following steps:

[0006] S1: Preparation of specific immunosensors. The specific operation includes chemical modification on the working electrode. By dripping chemical reagents such as cysteamine (β-mercaptoethylamine) on the electrode surface, amino functional groups are introduced. Subsequently, the electrode is immersed in a glutaraldehyde solution for further chemical cross-linking treatment to form an aldehyde activation layer on the electrode surface. Next, the specific antibody is dripped onto the electrode surface and covalently bound to the electrode through a chemical reaction to form a sensor surface with efficient immunoreactivity. In order to prevent nonspecific adsorption, an ethanolamine solution is then added to the electrode for sealing treatment. Ethanolamine can effectively block the unreacted aldehyde groups on the electrode surface, further reducing background noise and nonspecific binding, thereby improving the sensitivity and specificity of the sensor. The innovation of this step is that the sensitivity and specificity of the immunosensor are significantly improved by optimizing the electrode surface modification process.

[0007] S2: dropping an antigen solution of a specific concentration onto the prepared specific immunosensor to form an antigen-antibody complex. Subsequently, a number of identical specific immunosensors are prepared, and antigen solutions of different concentration gradients are sequentially dropped onto these sensors to form antigen-antibody immune complexes of different antigen concentrations.

[0008] S3: Electrochemical measurement of antigen-antibody complexes is performed using cyclic voltammetry. The immunosensor containing the antigen-antibody complex is placed in a potassium ferrocyanide solution for electrochemical measurement. The current-potential curve is recorded by cyclic voltammetry, the electrochemical signal is captured, and a characteristic cyclic voltammogram is generated. Since the formation of the antigen-antibody complex affects the electrochemical reaction on the electrode surface, the concentration of the antigen can be inferred by detecting the change in the current signal.

[0009] S4: Extract characteristic parameters of cyclic voltammetry curve and perform characteristic processing, extract multiple characteristic parameters from the voltammetry curve, these parameters include maximum current, shortest straight-line distance of positive and negative scanning curves, area of ​​closed region of scanning curve, integral of oxidation current curve, minimum potential, slope between maximum current point and minimum current point, minimum current, end current, maximum potential, initial current and voltage of zero current, 11 characteristic parameters.

[0010] S5: Preprocess the extracted feature parameters and sample concentration label data. The extracted features were ranked by importance using the random forest algorithm, and the features with significant impact on the prediction performance were screened out, thereby simplifying the complexity of the model and improving the accuracy of the prediction. The three features with the lowest importance, namely the maximum potential, initial current, and zero-point current, were removed. In order to further optimize the data processing, principal component analysis (PCA) was used to reduce the dimension of the selected features, reduce the data dimension, and visualize the data to evaluate the quality of the data. Label data preprocessing significantly improved the performance of the model by unifying the original concentration unit of the antigen to pg / mL and converting it to a logarithmic form with a base of 10. This conversion narrowed the numerical range of the concentration data, avoided numerical instability in model training, and strengthened the linear relationship between the concentration data and the feature data, thereby improving the fitting effect of the model. In addition, the logarithmic transformation also reduced the skewness of the data, improved the uniformity of the data distribution, and improved the prediction accuracy. This preprocessing strategy not only improved the prediction performance of the model in the low and high concentration ranges, but also enhanced its robustness and reliability.

[0011] S6: Establish and evaluate the machine learning model. In this invention, we use a multivariate linear regression model to predict antigen concentration. The model learns the linear relationship between input features and antigen concentration to achieve efficient concentration prediction. The core principle of the multivariate linear regression model is to establish a linear equation:

[0012] y=β0+β1x1+β2x2+...+β p x p +∈ (1)

[0013] In this equation, y is the dependent variable (output variable) to be predicted, x1, x2, x3, …, x p are independent variables (input features), β1, β2, β3, …β p are the parameters (slopes) of the model, and ∈ is the error term, which represents the random error that cannot be explained by the model. The model estimates these parameters by learning from the training data, thereby achieving accurate prediction of new observations. During the training process, statistical methods such as ordinary least squares are used to fit the model, and the regression loss function is optimized to improve the prediction ability.

[0014] The implementation process of the model includes data preprocessing, feature extraction and selection, model training and testing. The data preprocessing stage involves standardization and cleaning of experimental data to ensure data consistency and quality. The feature extraction stage extracts key features reflecting the essence of electrochemical reactions from the cyclic voltammetry curve. The feature selection stage screens out the most influential features for antigen concentration prediction to improve the accuracy and robustness of the model. In the model training stage, the preprocessed feature parameters are input into the model, and the regression coefficients are optimized using the training set and hyperparameters are tuned to ensure optimal performance. In the testing stage, the model is evaluated by the mean relative percentage error (MAPE) and leave-one-out cross validation (LOOCV) to verify the accuracy and stability of its prediction. Leave-one-out cross validation ensures that each sample is used for training and testing by using each sample in the data set as a test set and the remaining samples as training sets. This method can comprehensively evaluate model performance, effectively avoid overfitting, and maximize the use of limited data resources, and is particularly suitable for small sample data sets. The main advantage of the present invention is that it uses a multivariate linear regression model to efficiently analyze complex electrochemical data and provide accurate antigen concentration predictions. This method has high computational efficiency and model interpretability, and is suitable for processing small and medium-sized data sets. It significantly improves the accuracy and reliability of the prediction model by optimizing the data processing and feature selection processes, thereby meeting the practical application requirements of high precision and high stability and improving the effect of electrochemical detection.

[0015] S7: Detect unknown target analyte concentration. For target analytes of unknown concentration, use the same electrochemical measurement method to record cyclic voltammetry curves and extract corresponding characteristic parameters. The extracted characteristic parameters are input into the trained machine learning model for prediction, and the model will output the specific concentration of the target analyte in the sample. Through this method, rapid, accurate and efficient detection of immune analytes can be achieved.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] First, the present invention combines artificial intelligence with electrochemical immunoassay technology to achieve fully automated data processing and analysis, greatly improving detection efficiency and accuracy. Secondly, a machine learning algorithm is used to model multi-feature parameters of electrochemical signals, which significantly improves the fitting accuracy of the data model compared to the traditional linear regression method. In addition, the present invention effectively solves the problems of instability and poor repeatability of antibody-antigen binding signals in traditional immunoassay methods, thereby improving the consistency and reliability of the test results. Finally, the detection process is optimized and the detection cost is reduced, making the present invention have broad application prospects in the fields of clinical diagnosis, environmental monitoring, and food safety testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1A flow chart of the electrochemical immunoassay method based on artificial intelligence provided by the present invention.

[0019] Figure 2 The present invention provides an operation step diagram of a method for constructing an electrochemical sensor.

[0020] Figure 3 The present invention provides a schematic diagram for extracting characteristics of a cyclic voltammetry curve. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to explain the present invention and are not used to limit the scope of the present invention.

[0024] The present invention is implemented using Staphylococcal enterotoxin B (SEB) as an example. First, a specific immunosensor is prepared. In this process, the working electrode is first cleaned with deionized water and ethanol and dried at room temperature. Then, 20 microliters of a 10 mmol per liter cysteamine solution is added to the electrode surface and incubated at room temperature for 2 hours to form an amino self-assembled monolayer. After completion, the electrode surface is rinsed with phosphate buffered saline (PBS buffer) to remove unbound cysteamine and dry. Next, 20 microliters of a 2.5% glutaraldehyde solution is added and incubated at room temperature for 15 minutes to promote a cross-linking reaction between glutaraldehyde and the amino group of cysteamine to form an aldehyde activation layer. Subsequently, the electrode surface is rinsed with PBS buffer and dried. Next, 20 microliters of a 100 microgram per milliliter SEB antibody solution is added and incubated at room temperature for 2 hours to allow the SEB antibody to covalently bind to the electrode surface through the aldehyde group to form an immunoreactive layer. Finally, the electrode was washed with PBS buffer to remove unbound SEB antibody, and 20 μL of 0.5 mol ethanolamine solution was dropped onto the electrode surface and incubated at 37 °C for 15 min to block unreacted aldehyde groups, thereby preparing a specific immunosensor.

[0025] In terms of forming antigen-antibody complexes, SEB solutions of different concentrations were prepared, with concentrations of 0, 1 ng / ml, 10 ng / ml, 100 ng / ml, 1 μg / ml, and 10 μg / ml, respectively. 20 μl of SEB antigen solutions of different concentrations were added dropwise to the working electrode surface of the immunosensor and incubated at 37°C for 15 minutes to ensure that the SEB antigen was fully bound to the SEB antibody on the electrode surface to form an antigen-antibody complex. After incubation, the electrode was rinsed with PBS buffer to remove unbound antigens, and the electrode surface was dried to ensure the stability of the complex.

[0026] Next, the antigen-antibody complex was electrochemically measured by cyclic voltammetry. The immunosensor containing the antigen-antibody complex was placed in a 5 mmol per liter potassium ferrocyanide electrolyte solution and measured using an electrochemical workstation. The scanning voltage range was set to -0.3 volts to 0.5 volts, the scanning rate was 100 millivolts per second, the sampling interval was 1 millivolt, the sensitivity was 1e-4A / V, and the current-potential curve was recorded. The experimental results show that as the SEB antigen concentration increases, the redox peak current of the cyclic voltammetry curve gradually decreases, indicating that the formation of the antigen-antibody complex hinders the redox reaction of potassium ferrocyanide at the electrode interface.

[0027] The characteristic parameters extracted from the cyclic voltammetry curve include the maximum current, the shortest straight-line distance between the maximum current point and the minimum current point of the positive and negative scanning curves, the area of ​​the closed region of the scanning curve, the integral of the oxidation current curve, the minimum potential, the slope between the maximum current point and the minimum current point, the minimum current, the end current, the maximum potential, the initial current, and the voltage of the zero-point current. These characteristic parameters were preprocessed, including feature selection and dimensionality reduction, and the three characteristic parameters of the maximum potential, the initial current, and the voltage of the zero-point current were removed. The concentration label was also preprocessed, the concentration unit was converted to picograms per milliliter (pg / ml), and logarithmic transformation was performed to standardize the data.

[0028] The processed feature parameters and label data were input into the multivariate linear regression model to establish and optimize the regression model to predict the concentration of the target antigen. The model was evaluated by mean relative percentage error and leave-one-out cross-validation method to verify its accuracy and stability.

[0029] Finally, a cyclic voltammetry experiment is performed on a SEB sample of unknown concentration to obtain the electrochemical response curve of the sample and extract key characteristic parameters. These characteristic parameters are input into the established and optimized multiple linear regression model, which uses these characteristic parameters to predict the concentration and finally calculates the target analyte concentration in the SEB sample to be tested. In this way, the concentration of the target analyte in the unknown SEB sample is accurately determined, thereby effectively meeting the actual detection needs. The method of the present invention is not only applicable to the detection of Staphylococcal enterotoxin B (SEB), but can also be extended to the detection of other biomarkers.

Claims

1. A method for performing an immunoassay based on artificial intelligence to determine the concentration of a target analyte, comprising: preparing a specific immunosensor by chemically modifying an electrode of an electrochemical sensor and then modifying an antibody to prepare a specific immunosensor; Forming an antigen-antibody complex, adding a sample liquid of the antigen to be detected into the specific immunosensor to allow the antigen to bind to the antibody; Electrochemically measuring the antigen-antibody complex by cyclic voltammetry, placing the immunosensor containing the antigen-antibody complex in an electrolyte solution, measuring the signal of the reaction between the immunosensor complex and the electrolyte solution, and recording the cyclic voltammetry curve; Extract characteristic parameters of the cyclic voltammetry curve, and extract multiple characteristic parameters from the cyclic voltammetry curve, wherein the parameters include maximum current (maxI), the shortest straight line distance between the maximum current point and the minimum current point of the positive and negative scanning curves (distance), the closed area of ​​the scanning curve (area), the integral of the oxidation current curve (OID), the minimum potential (minV), the slope between the maximum current point and the minimum current point (k), the minimum current (minI), and the end current (endI); Preprocessing characteristic parameters of the cyclic voltammetry curve, the preprocessing comprising performing feature selection and dimensionality reduction processing on the extracted characteristic parameters of the cyclic voltammetry curve; Preprocessing the sample concentration label data, the preprocessing comprising performing a logarithmic transformation on the sample concentration label; Establish and evaluate machine learning models for predicting target analytes, input selected feature parameters into a multivariate linear regression model for training, establish and optimize regression loss functions based on sample concentration labels to establish a prediction model for target antigen concentration, and validate the model using mean relative percentage error and leave-one-out cross-validation to evaluate its accuracy and stability; The concentration of the target analyte of unknown concentration on the specific immunosensor is measured, the cyclic voltammetry curve is measured, the characteristic parameters of the cyclic voltammetry curve are extracted and input into the machine learning model for detection, and the concentration of the analyte in the sample to be tested is determined.

2. The method of claim 1, wherein the step of chemically modifying the electrode of the electrochemical sensor comprises: 20 μL of 10 mmol / L cysteamine solution was dropped onto the surface of the working electrode, and the working electrode was incubated at room temperature for 2 hours, and then the electrode was rinsed with phosphate buffered saline (PBS buffer) and dried; 20 μl of 2.5% glutaraldehyde solution was added dropwise to the cysteamine-modified electrode surface. After incubation for 15 minutes, the electrode surface was rinsed with PBS buffer and dried.

3. The method of claim 1, wherein the step of modifying the antibody comprises: Add 20 μl of 100 μg / ml antibody solution to the pretreated working electrode surface and incubate at room temperature for 2 hours. Rinse the electrode surface with PBS buffer and dry. 20 μl of 0.5 mol ethanolamine was applied to the working electrode area and incubated at 37° C. for 15 minutes. The electrode surface was rinsed with PBS buffer and dried, wherein the ethanolamine reacted with unreacted glutaraldehyde to inactivate unreacted electrophilic reagent.

4. The method of claim 1, wherein the step of forming an antigen-antibody complex comprises: Antigen solutions of different concentrations were prepared, 20 μl of the antigen solution to be tested was added dropwise to the surface of the working electrode, and incubated at 37°C for 15 minutes. After incubation, the electrode was rinsed with PBS buffer and completely dried.

5. The method of claim 4, wherein the different concentrations of the antigen solutions are in the range of 0 to 10 micrograms per milliliter.

6. The method of claim 1, wherein the electrolyte solution is a potassium ferrocyanide solution, and the antigen-antibody complex hinders the redox reaction of potassium ferrocyanide due to the lack of significant electroactivity, resulting in the current varying with the antigen concentration, and this characteristic is utilized to analyze the current-potential curve by cyclic voltammetry to detect the antigen concentration.

7. The method of claim 1, wherein the characteristic parameter preprocessing of the cyclic voltammetry curve comprises: Use the random forest algorithm to rank the extracted features and select the features that have a significant impact on the prediction performance; The selected features were processed using principal component analysis (PCA) to reduce the data dimension for visualization and to judge the data quality.

8. The method of claim 1, wherein the sample concentration label data preprocessing includes unifying the original antigen concentration unit into picograms per milliliter and then converting it into a logarithm with a base of 10 as a label for the training model.