Method for constructing neurodegenerative disease diagnosis model based on soluble TREM2 molecules

Abstract

The invention discloses a construction method of a neurodegenerative disease diagnosis model based on soluble TREM2 molecules, and relates to the technical field of biomarker detection. The invention comprises XX. According to the method, a plurality of biomarkers and a machine learning algorithm are combined, a corresponding model is obtained, disease prediction and treatment are more accurate and comprehensive, the diagnosis accuracy is improved, compared with a traditional diagnosis method in the prior art, diseases can be recognized in the early stage, and through multi-center verification and data fusion, the diagnosis accuracy is improved. The technology has higher universality and adaptability in different regions and crowds, while the prior art mostly focuses on a single region or crowd and supports a personalized diagnosis and treatment scheme, more effective treatment measures can be formulated according to the specific conditions of patients through the design of the personalized diagnosis and treatment scheme, and the treatment effect is better. However, the prior art generally depends on a universal treatment scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of biomarker detection, and specifically to a method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules. Background Art

[0002] Neurodegenerative diseases (NDDs) include Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), etc. These diseases are usually characterized by slow development and lack obvious clinical symptoms in the early stages. Therefore, early diagnosis is crucial for timely intervention and treatment. However, existing diagnostic methods such as neuroimaging examinations (MRI, PET scans), neuropsychological tests (such as Mini-Mental State Examination, MMSE) and clinical symptom scales (such as ADCS-ADL) have certain limitations, mainly manifested as:

[0003] Delayed diagnosis: Existing imaging and clinical scoring methods can usually only make a clear diagnosis after the disease develops to a certain stage, missing the best time for early intervention.

[0004] Low sensitivity and specificity: Many diagnostic tools have poor diagnostic accuracy for early-stage diseases, especially early-stage Alzheimer's disease and mild cognitive impairment (MCI) with atypical clinical manifestations.

[0005] High cost and invasiveness: Some imaging examinations (such as PET) are expensive and require more complex equipment. The collection of cerebrospinal fluid samples is also invasive, and patient acceptance is low.

[0006] With the deepening of biomarker research, more and more molecular markers have been found to be closely related to the early diagnosis of neurodegenerative diseases. In particular, soluble TREM2 (soluble Triggering Receptor Expressed on Myeloid cells 2, sTREM2) has received widespread attention. sTREM2 is a cytokine secreted by microglia, which is involved in the regulation of brain immune response and neuroinflammation and is related to neurodegenerative diseases.

[0007] Studies have shown that TREM2 gene mutations or dysfunction of its receptors significantly increase the risk of Alzheimer's disease. More importantly, sTREM2 levels are significantly increased or changed in the plasma of AD patients, which can reflect early immune activation responses. This finding provides a new direction for the early diagnosis of neurodegenerative diseases. However, current clinical applications based on sTREM2 still face some technical challenges, especially in how to efficiently and accurately quantify sTREM2, how to combine it with other clinical indicators, and how to optimize diagnostic models through multidimensional data.

[0008] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention

[0009] The purpose of the present invention is to provide a method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules, which combines high-throughput molecular detection technology, advanced statistical analysis methods and machine learning algorithms to efficiently and accurately screen for neurodegenerative diseases in the early stage. This solution not only introduces more complex multidimensional data fusion and deep learning optimization based on the detection of a single biomarker, but also has strong innovation to solve the technical problems raised in the background technology.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules, comprising at least the following steps:

[0011] S1: Sample collection and efficient processing, including sample type and source and high-throughput molecular extraction and quality control;

[0012] S2: Extraction and quantitative detection of soluble TREM2 molecules;

[0013] S3: Multimodal biomarker detection and data fusion, including multi-omics data fusion, data cleaning and standardization, and feature extraction;

[0014] S4: Deep learning model construction and optimization;

[0015] S5: Model validation and clinical application.

[0016] Furthermore, the sample types selected were multi-center and multi-regional samples of Alzheimer's disease, Parkinson's disease, and healthy control groups to ensure the diversity and representativeness of the data;

[0017] The sample sources include plasma, cerebrospinal fluid (CSF) and peripheral blood mononuclear cells (PBMCs), and are collected and stored in a standardized manner through an intelligent sample management system.

[0018] Furthermore, the high-throughput molecular extraction and quality control in S1 at least includes the following steps:

[0019] Utilize an automated high-throughput liquid handling platform to extract soluble TREM2 molecules from plasma / cerebrospinal fluid samples to ensure the efficiency and consistency of the extraction process;

[0020] Proteomics mass spectrometry was introduced during the extraction process to systematically quantify TREM2 and other candidate biomarkers, ensuring high-precision analysis at the molecular level.

[0021] Adopt cold chain transportation and regular quality control, including antibody cross-reactivity analysis and exogenous contamination monitoring.

[0022] Furthermore, the S2 at least includes the following steps:

[0023] Antibody affinity method to extract sTREM2: high-affinity anti-TREM2 monoclonal antibodies (such as antibodies used to capture soluble TREM2) are used to extract sTREM2 from plasma or cerebrospinal fluid by immunoaffinity chromatography or immunoprecipitation. In this process, the anti-TREM2 antibody first binds to the soluble TREM2 in the sample to form an antigen-antibody complex, and then the sTREM2 molecules are enriched by affinity chromatography columns or immunoprecipitation.

[0024] Design a detection method that uses a combination of chemiluminescent immunoassay (CLIA) and electrochemiluminescent immunoassay (ECLIA) to improve the sensitivity and accuracy of sTREM2 detection, especially at low concentrations;

[0025] Introducing multiplex immunoassay technology to simultaneously measure multiple neurodegenerative disease-related biomarkers (such as Aβ, tau protein, sTREM2, etc.) on the same experimental platform to reduce sample consumption and increase detection throughput;

[0026] Each batch of experiments was verified using blank controls and high concentration standards to avoid experimental errors.

[0027] Furthermore, the multi-omics data fusion in S3 at least includes the integration of genomics (such as APOE genotype analysis), proteomics (combined detection of sTREM2 and other immune-related molecules) and metabolomics (such as fatty acid metabolites), and the use of multivariate statistical analysis methods (such as principal component analysis PCA, partial least squares regression PLS) to integrate different omics data to explore high-sensitivity and high-specificity diagnostic marker combinations.

[0028] Furthermore, the data cleaning and standardization in S3 is to perform unified standardization processing on all data, and the data cleaning and standardization at least include noise removal, outlier detection and data normalization to ensure the comparability of different data sources, and use the quantification algorithm of gene expression data to deeply process the genomic data to ensure the high accuracy of genotype analysis.

[0029] Furthermore, the feature extraction in S3 at least includes quantitative features (sTREM2 concentration), category features (clinical diagnosis typing), clinical symptom scores (such as MMSE, ADL scores, etc.) and genotype data (such as the carrying status of risk genes such as APOE4);

[0030] The features are normalized to eliminate the impact of unit variance on model training.

[0031] Furthermore, the S4 at least includes the following steps:

[0032] Using a variety of machine learning algorithms:

[0033] Support Vector Machine (SVM): used to handle classification problems with high-dimensional features, especially suitable for dealing with the complexity of multi-dimensional data;

[0034] Random Forest (RF) and XGBoost: Capture complex nonlinear relationships in data through integrated learning methods to enhance the robustness and accuracy of the model;

[0035] Deep Neural Network (DNN): Uses multi-layer neural networks for end-to-end feature learning to further improve classification accuracy;

[0036] Model training and optimization:

[0037] Use cross-validation to evaluate the performance of the model and ensure generalization ability during training;

[0038] Introduce Bayesian Optimization or Genetic Algorithm to optimize the hyperparameter configuration of machine learning models and improve the diagnostic ability of the models;

[0039] Model interpretability and visualization

[0040] Use SHAP (Shapley Additive Explanations) values ​​to analyze the interpretability of the model and clarify which features (such as sTREM2 concentration, genotype, etc.) are most critical for disease prediction;

[0041] The t-SNE (t-distributed Stochastic Neighbor Embedding) dimensionality reduction technology was used to visualize high-dimensional data to intuitively show the differences between different clinical groups.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention combines multiple biomarkers and machine learning algorithms to obtain a corresponding model, making the prediction and treatment of diseases more accurate and comprehensive, thereby improving the accuracy of diagnosis. Compared with traditional diagnostic methods in the prior art, it can identify diseases at an early stage.

[0044] 2. Through multi-center verification and data fusion, the present invention makes the technology more universal and adaptable in different regions and populations, while the existing technologies are mostly concentrated in a single region or population.

[0045] 3. The present invention supports personalized diagnosis and treatment plans. Through the design of personalized diagnosis and treatment plans, more effective treatment measures can be formulated according to the specific conditions of the patients, while the existing technology usually relies on general treatment plans. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0047] Embodiment 1:

[0048] Construction and validation of an early diagnosis model based on sTREM2

[0049] step:

[0050] Sample collection

[0051] Samples were collected from multiple regions (such as Beijing, Shanghai, Guangzhou, etc.), including AD patients, mild cognitive impairment (MCI) patients and healthy controls. There were 100 subjects in each group, and plasma, cerebrospinal fluid (CSF) and peripheral blood mononuclear cells (PBMCs) samples were collected.

[0052] Molecular detection and biomarker extraction

[0053] Soluble TREM2 in plasma was extracted using a high-throughput liquid processing platform, and sTREM2 was quantitatively detected using electrochemiluminescence immunoassay (ECLIA).

[0054] At the same time, Aβ protein and tau protein were extracted from cerebrospinal fluid samples and their levels were measured as auxiliary markers for neurodegenerative diseases.

[0055] Multimodal data fusion and feature selection

[0056] The sTREM2 concentration in plasma, Aβ and tau levels in cerebrospinal fluid, and APOE genotype information were integrated, and the partial least squares regression (PLS) method was used for data preprocessing to screen out the features most sensitive to AD diagnosis.

[0057] Machine Learning Model Building

[0058] Based on data such as sTREM2, Aβ, tau, and APOE, support vector machine (SVM) was used for training to construct an early AD diagnosis model.

[0059] The model parameters were optimized through cross-validation to improve its predictive ability for early AD.

[0060] Clinical Validation and Optimization

[0061] The model was validated in multiple centers to evaluate its accuracy and stability in different regions and populations. It was compared with traditional MMSE scores and imaging examinations (such as MRI) to verify its early diagnosis ability.

[0062] result:

[0063] The model showed good sensitivity (85%) and specificity (90%) in multicenter validation, and was able to accurately identify patients in the early stages of AD, especially in the MCI stage.

[0064] Embodiment 2:

[0065] Neurodegenerative disease risk prediction system based on multi-omics data

[0066] step:

[0067] Sample collection and multi-omics data acquisition

[0068] One hundred Parkinson's disease patients and 100 healthy controls from different hospitals were selected, and their peripheral blood and cerebrospinal fluid samples were collected.

[0069] Each patient undergoes genomic testing (such as screening for gene mutations such as APOE, LRRK2, PARK7, etc.), and proteomics technology is used to detect biomarkers related to neurodegenerative diseases in plasma (such as sTREM2, α-synuclein).

[0070] Data processing and standardization

[0071] The data of each group were standardized, the genomic data were processed using the quantification algorithm of genotype data, and the proteomic data were analyzed by mass spectrometry.

[0072] Principal component analysis (PCA) and partial least squares regression (PLS) were used to reduce the dimension and extract features of multi-omics data to identify key biomarkers associated with Parkinson's disease.

[0073] Model building and training

[0074] Based on the extracted multi-omics features, the random forest (RF) algorithm was used to construct an early prediction model for PD, and cross-validation was used to optimize hyperparameters.

[0075] The support vector machine (SVM) is combined for secondary optimization to further improve the prediction accuracy of the model.

[0076] Clinical application and real-time screening

[0077] The model is integrated into the clinical decision support system (CDSS) to provide doctors with auxiliary diagnostic tools. Patients can obtain disease risk assessment in real time through simple blood tests.

[0078] Design a smartphone APP to combine the prediction model with wearable devices (such as smart bracelets) to achieve real-time monitoring of patients' movement symptoms such as gait and tremor.

[0079] result:

[0080] The system can identify high-risk groups of the disease in the early stages of PD and has been successfully applied to clinical screening, with good clinical operability. The accuracy of the model reaches 87%, which can be used as an auxiliary diagnostic tool to help doctors identify high-risk patients in time and intervene.

[0081] Embodiment 3:

[0082] Personalized treatment plans for early diagnosis and intervention of neurodegenerative diseases

[0083] background:

[0084] Early diagnosis of neurodegenerative diseases is crucial for developing personalized treatment plans. This embodiment provides personalized diagnosis and treatment plans for Alzheimer's disease (AD) patients based on sTREM2 and other biomarkers combined with machine learning models.

[0085] step:

[0086] Sample collection and biomarker detection

[0087] Plasma and cerebrospinal fluid samples were collected from AD patients and healthy controls. ECLIA technology was used to efficiently detect sTREM2, Aβ, tau protein, and other immune-related factors (such as TNF-α, IL-6, etc.).

[0088] Data analysis and marker screening

[0089] Multivariate statistical analysis methods were used to analyze the differences in biomarkers between different patient groups and screen out marker combinations with high sensitivity and specificity.

[0090] The genomic data (such as APOE genotype) and plasma biomarker data (such as sTREM2 concentration) were combined for feature extraction to form a multidimensional feature matrix.

[0091] Machine learning model training and personalized diagnosis

[0092] Use deep neural networks (DNNs) to train multidimensional data and optimize the model to improve the ability to identify AD at an early stage.

[0093] A personalized diagnosis module is introduced into the model to generate a personalized risk assessment report for each patient based on the patient's specific biomarker characteristics.

[0094] Clinical intervention and treatment plan recommendations

[0095] Based on the prediction results provided by the machine learning model, personalized early intervention plans are recommended for high-risk patients, such as drug treatment (such as donepezil, clotixin, etc.) and lifestyle intervention (such as cognitive training, diet control, etc.).

[0096] Deliver personalized treatment plans to doctors through the electronic medical record system to assist in clinical decision-making.

[0097] result:

[0098] This personalized treatment plan can provide personalized early intervention measures for AD patients through accurate biomarker detection and machine learning models, thereby improving the effectiveness of treatment. The system has been piloted in some hospitals and received good feedback, with patients' early symptoms significantly alleviated.

[0099] Summarize

[0100] These three examples demonstrate how to use sTREM2 and other biomarkers related to neurodegenerative diseases to build early diagnosis models, risk prediction systems, and personalized treatment plans through high-throughput detection, data fusion, and machine learning technologies, thereby providing powerful auxiliary tools for clinical practice. The combination of multimodal data and the application of deep learning can effectively improve the early diagnosis rate of neurodegenerative diseases and help doctors develop more accurate treatment plans.

[0101] The following is a comparison table based on three embodiments and the prior art, which shows the advantages of the embodiment technology compared with the prior art.

[0102]

[0103]

[0104]

[0105] Through the above comparison, it can be seen that the embodiment technology is superior to the existing technology in terms of accurate diagnosis, personalized treatment, data fusion and clinical application prospects.

[0106] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

Claims

1. A method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules, characterized in that: At least the following steps are included: S1: Sample collection and efficient processing, including sample type and source and high-throughput molecular extraction and quality control; S2: Extraction and quantitative detection of soluble TREM2 molecules; S3: Multimodal biomarker detection and data fusion, including multi-omics data fusion, data cleaning and standardization, and feature extraction; S4: Deep learning model construction and optimization; S5: Model validation and clinical application.

2. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The sample types selected were multi-center and multi-regional samples of Alzheimer's disease, Parkinson's disease and healthy control groups to ensure the diversity and representativeness of the data; The sample sources include plasma, cerebrospinal fluid and peripheral blood mononuclear cells, and are collected and stored in a standardized manner through an intelligent sample management system.

3. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The high-throughput molecular extraction and quality control in S1 at least includes the following steps: Utilize an automated high-throughput liquid handling platform to extract soluble TREM2 molecules from plasma / cerebrospinal fluid samples to ensure the efficiency and consistency of the extraction process; Proteomics mass spectrometry analysis was introduced during the extraction process to systematically quantify TREM2 and other candidate biomarkers, ensuring high-precision analysis at the molecular level; Adopt cold chain transportation and regular quality control, including antibody cross-reactivity analysis and exogenous contamination monitoring.

4. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The S2 at least comprises the following steps: Antibody affinity method to extract sTREM2: high-affinity anti-TREM2 monoclonal antibodies are used to extract sTREM2 from plasma or cerebrospinal fluid by immunoaffinity chromatography or immunoprecipitation. In this process, the anti-TREM2 antibody first binds to the soluble TREM2 in the sample to form an antigen-antibody complex, and then the sTREM2 molecules are enriched by affinity chromatography columns or immunoprecipitation. Establishing a detection method that uses a combination of chemiluminescent immunoassay and electrochemiluminescent immunoassay to improve the sensitivity and accuracy of sTREM2 detection, especially at low concentrations; Introducing multiple immunochromatographic techniques to simultaneously measure multiple neurodegenerative disease-related biomarkers on the same experimental platform to reduce sample consumption and increase detection throughput; Each batch of experiments was verified using blank controls and high concentration standards to avoid experimental errors.

5. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The multi-omics data fusion in S3 at least includes the fusion of genomics, proteomics and metabolomics, and the use of multivariate statistical analysis methods to integrate different omics data to explore high-sensitivity and high-specificity diagnostic marker combinations.

6. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The data cleaning and standardization in S3 is to perform unified standardization processing on all data. The data cleaning and standardization at least include noise removal, outlier detection and data normalization to ensure the comparability of different data sources, and use the quantification algorithm of gene expression data to deeply process the genomic data to ensure the high accuracy of genotype analysis.

7. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The feature extraction in S3 at least includes quantitative features, category features, clinical symptom scores and genotype data; The features are normalized to eliminate the impact of unit variance on model training.

8. The method for constructing a neurodegenerative disease diagnostic model based on soluble TREM2 molecules according to claim 1, characterized in that: The S4 at least comprises the following steps: Using a variety of machine learning algorithms: Support vector machine: used to handle classification problems with high-dimensional features, especially suitable for dealing with the complexity of multi-dimensional data; Random Forest and XGBoost: Capture complex nonlinear relationships in data through integrated learning methods to enhance the robustness and accuracy of the model; Deep neural network: Use multi-layer neural network for end-to-end feature learning to further improve classification accuracy; Model training and optimization: Use cross-validation methods to evaluate the performance of the model and ensure generalization ability during training; Introduce Bayesian optimization or genetic algorithms to optimize the hyperparameter configuration of machine learning models and improve the diagnostic capabilities of the models; Model interpretability and visualization Use SHAP values ​​to analyze the interpretability of the model and clarify which features (such as sTREM2 concentration, genotype, etc.) are most critical for disease prediction; t-SNE dimensionality reduction technology was used to visualize high-dimensional data to intuitively show the differences between different clinical groups.

Images