Marker and kit for diagnosis or prediction of drug efficacy for schizophrenia

Abstract

This invention provides the application of reagents for detecting the expression levels of biomarkers in peripheral blood in the preparation of diagnostic systems or drug efficacy prediction systems for schizophrenia. The biomarkers include C3 protein, C4BPA protein, C4 protein in plasma extracellular vesicles, and C4 protein in plasma. This invention also provides a kit for the diagnosis of schizophrenia or for predicting drug efficacy, the kit comprising antibodies for measuring the levels of the biomarkers, including C3 protein, C4BPA protein, C4 protein in plasma extracellular vesicles, and C4 protein in plasma. This invention represents the first discovery of biomarkers with high value for personalized and accurate diagnosis of schizophrenia and prediction of drug efficacy, overcoming the predicament of lacking convenient, stable, and highly accurate objective peripheral blood biomarkers for the diagnosis and prediction of schizophrenia, and contributing to improved diagnostic accuracy and the adoption of optimal treatment strategies for schizophrenia.

Description

Technical Field

[0001] This invention belongs to the field of medical technology, and particularly relates to the field of schizophrenia technology, specifically to biomarkers and reagent kits for the diagnosis of schizophrenia or for predicting the efficacy of drugs. Background Technology

[0002] Schizophrenia is a severe mental illness that affects an individual's thinking, emotions, and behavior, primarily manifesting as positive and negative symptoms and cognitive impairment. Schizophrenia mainly develops in adolescence or early adulthood, with similar incidence rates between men and women. Recent epidemiological surveys indicate that the lifetime prevalence of schizophrenia in China has risen to 0.6%. With China's aging population and accelerated urbanization, the prevalence of schizophrenia is likely to gradually increase. Schizophrenia is characterized by its long-term persistence and recurrent nature, requiring long-term treatment and attention, thus having a significant negative impact on patients, families, society, and the economy. The medical and nursing burden of schizophrenia is extremely heavy, with high direct and indirect medical costs, and is considered one of the major social health burdens. For example, the average annual direct medical cost per schizophrenia patient in the United States ranges from several thousand to hundreds of thousands of dollars. Some estimates suggest that the total economic burden of schizophrenia annually may reach tens of billions of dollars. Therefore, actively promoting early diagnosis and treatment of schizophrenia is a pressing medical, health, and socioeconomic issue that must be addressed.

[0003] Currently, the diagnosis of schizophrenia primarily relies on the subjective assessment of a patient's symptoms and medical history by clinicians. However, the patient's subjective feelings and expressions can easily influence the clinician's judgment, leading to different diagnoses based on individual experience and experience. Furthermore, the clinical symptoms of schizophrenia are complex and highly heterogeneous, overlapping with other severe mental disorders such as bipolar disorder and major depressive disorder, potentially resulting in misclassification. Additionally, patients with mild or early symptoms may not be easily identified. Epidemiological studies show that nearly 25% of schizophrenia patients are misdiagnosed or missed, leading to delayed or inappropriate treatment. Simultaneously, the highly heterogeneous responses to antipsychotic medications make it impossible for physicians to predict patient sensitivities to drug efficacy and side effects, further complicating schizophrenia treatment. Therefore, reliable and objective biomarkers represent a solution for improving the accuracy of schizophrenia diagnosis and treatment.

[0004] In recent years, significant progress has been made in the diagnosis and prognostic studies of schizophrenia. Many research teams have developed relevant biomarkers based on brain imaging (MRI, EEG), cerebrospinal fluid, and peripheral blood. However, brain imaging is very expensive and requires specialized personnel for operation and analysis, and related biomarkers have no clinical application in China. Cerebrospinal fluid examination is invasive and cannot be practically applied in clinical practice. Blood is readily available, simple to operate, and minimally invasive, making it an ideal source for biomarker development. However, although many teams have proposed relevant biomarkers based on blood proteins or metabolites, the performance of these biomarkers is inconsistent among different researchers, making it impossible to draw consistent conclusions. This may be because blood-derived proteins or metabolites may not exist in a very stable form in vitro, and the complex testing process may disrupt their original physiological state and content, leading to significant heterogeneity in the results.

[0005] As an emerging form of liquid biopsy, the billions of extracellular vesicles (EVs) in peripheral blood provide a valuable resource for developing biomarkers based on blood samples. EVs are nanoparticles enclosed in a double-membrane structure secreted by cells and tissues, capable of transporting intracellular biomolecules under both normal and pathological conditions to mediate intracellular signaling and intercellular communication. Studies have found that they are associated with the pathogenesis of neurodegenerative diseases. Furthermore, there is evidence that EVs in the central nervous system can cross the blood-brain barrier and diffuse into the peripheral blood. EVs are rich in various proteins, some of which undergo quantitative changes during pathogenesis. Drug treatment may also alter the protein composition of EVs. Therefore, mounting evidence suggests that EV proteins are effective diagnostic and prognostic biomarkers for neurological diseases, including Parkinson's disease and Alzheimer's disease. Current research on EV-derived proteins in schizophrenia is limited. Previous studies have primarily focused on proteins specifically related to mitochondrial activity and insulin signaling. Summary of the Invention

[0006] The purpose of this invention is to overcome at least one of the shortcomings of the prior art and provide a biomarker and kit for the diagnosis of schizophrenia or the prediction of drug efficacy, which can accurately diagnose schizophrenia at the individualized level and optimize clinical diagnostic strategies.

[0007] To address the aforementioned technical problems, the first aspect of this invention provides the application of a reagent for detecting the expression level of biomarkers in peripheral blood in the preparation of a diagnostic system for schizophrenia or a drug efficacy prediction system. Its main feature is that the biomarkers include C3 protein, C4BPA protein (C4b-binding protein alpha chain), and C4 protein.

[0008] Preferably, the biomarker also includes PROS1 protein (Vitamin K-dependent protein S).

[0009] Preferably, the biomarker also includes the PROS1 protein in plasma extracellular vesicles.

[0010] Preferably, the biomarkers include C3 protein, C4 protein and C4BPA protein in plasma extracellular vesicles, and C4 protein in plasma; or, the biomarkers include C3 protein and C4BPA protein in plasma extracellular vesicles, and C4 protein in plasma.

[0011] Specifically, this invention provides two combinations of biomarkers: one combination consists of C3, C4, C4BPA, and PROS1 proteins in plasma extracellular vesicles, and C4 protein in plasma; the other combination consists of C3 and C4BPA proteins in plasma extracellular vesicles, and C4 protein in plasma.

[0012] A second aspect of the present invention provides a kit for the diagnosis of schizophrenia or for predicting the efficacy of medications, characterized in that the kit includes antibodies for measuring the levels of biomarkers, including C3 protein, C4BPA protein, and C4 protein. Specifically, the kit includes a C3 antibody for measuring C3 protein levels, a C4BPA antibody for measuring C4BPA protein levels, and a C4 antibody for measuring C4 protein levels.

[0013] Preferably, the biomarker also includes the PROS1 protein.

[0014] Preferably, the biomarker also includes the PROS1 protein in plasma extracellular vesicles.

[0015] Preferably, the biomarkers include C3 protein, C4 protein and C4BPA protein in plasma extracellular vesicles, and C4 protein in plasma; or, the biomarkers include C3 protein and C4BPA protein in plasma extracellular vesicles, and C4 protein in plasma.

[0016] Preferably, the measured biomarker content is substituted into the model established by the XGboost algorithm.

[0017] A third aspect of the present invention provides a method for diagnosing schizophrenia or predicting the efficacy of medication, specifically including the following steps:

[0018] (1) Extract extracellular vesicles from the plasma sample to be tested;

[0019] (2) Extracting proteins from extracellular vesicles using cell lysis buffer;

[0020] (3) The content of the obtained extracellular vesicle proteins was detected using the corresponding antibodies;

[0021] (4) The content of C4 protein in plasma was detected using C4 antibody;

[0022] (5) Input the obtained protein content into the model and diagnose it as "schizophrenia" or "other"; the drug efficacy prediction result is "3-month estimated PANSS score reduction rate **" or "Not Available".

[0023] The proteins in extracellular vesicles include C3 protein and C4BPA protein, or C3 protein, C4 protein, PROS1 protein and C4BPA protein.

[0024] The biomarkers and kits for the diagnosis or prediction of drug efficacy in schizophrenia of this invention are the first to discover biomarkers with high value for personalized and accurate diagnosis of schizophrenia and prediction of drug efficacy: C3 and C4BPA proteins in plasma extracellular vesicles and C4 protein in plasma, or C3, C4, C4BPA, PROS1 proteins in plasma extracellular vesicles and C4 protein in plasma. This breakthrough overcomes the predicament of lacking convenient, stable, and highly accurate peripheral blood biomarkers for objective diagnosis and prediction of schizophrenia, which is conducive to improving the diagnostic accuracy of schizophrenia and adopting the best treatment strategy. At the same time, it is expected to provide scientific basis and clinical support for the discovery of novel antipsychotic drug targets with potential therapeutic value. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the experimental design for screening, training, and validating extracellular vesicle proteins and plasma biomarkers for the diagnosis and drug efficacy prediction of schizophrenia.

[0026] Figure 2 The main methods and steps for diagnosing and predicting extracellular vesicle proteins and plasma proteins in schizophrenia according to the present invention are illustrated in the diagram.

[0027] Figure 3 A and 3B are schematic diagrams illustrating the biological processes involved in the extracellular vesicle proteome in schizophrenia.

[0028] Figure 4 A and 4B represent the AUC curves, accuracy, sensitivity, specificity, true positive predictive value, and false positive predictive value (schizophrenia vs. healthy controls) for the first and second models, respectively.

[0029] Figure 5A and 5B represent the AUC curves, accuracy, sensitivity, specificity, true positive predictive value, and false positive predictive value of the first and second models, respectively (schizophrenia vs. bipolar disorder).

[0030] Figure 6 A and 6B represent the AUC curves, accuracy, sensitivity, specificity, true positive predictive value, and false positive predictive value (schizophrenia vs. major depressive disorder) of the first and second models, respectively.

[0031] Figure 7 An individualized diagnostic scoring tool was developed for the XGBoost algorithm model.

[0032] Figure 8 Linear regression for predicting drug efficacy using personalized diagnostic scoring tools.

[0033] Figure 9 This is the input and result reporting interface for a personalized scoring app. Detailed Implementation

[0034] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to specific embodiments.

[0035] Figure 1 This is a flowchart illustrating the experimental design for screening, training, and validating extracellular vesicle proteins and plasma biomarkers for the diagnosis and drug efficacy prediction of schizophrenia. Figure 2 This invention provides a diagram illustrating the main methods and steps for diagnosing and predicting extracellular vesicle proteins and plasma proteins in schizophrenia. Specific details are provided below with reference to embodiments. Unless otherwise specified, the experimental methods described in the following embodiments are conventional methods; the reagents or consumables involved are commercially available unless otherwise specified.

[0036] Example 1

[0037] Proteomics screening

[0038] The subjects were divided into a disease group and a control group. The disease group consisted of 20 patients with schizophrenia collected from the Shanghai Mental Health Center, while the control group consisted of 28 age- and sex-matched healthy individuals.

[0039] The specific method includes the following steps:

[0040] (1) Collect plasma from each subject and enrich extracellular vesicles in the plasma using ultracentrifugation.

[0041] (2) Proteins in extracellular vesicles were extracted using cell lysis buffer and prepared into peptide samples;

[0042] (3) Use liquid chromatography-tandem mass spectrometry to perform proteomics mass spectrometry detection on peptide samples.

[0043] The results are as follows Figure 3 As shown in A, 3B and Table 1 below, the extracellular vesicle proteome of patients with schizophrenia is mainly involved in complement activation. In this complement activation network, the expression levels of plasma extracellular vesicle proteins C3, C4, C4BPA and PROS1 in patients with schizophrenia are significantly higher than those in the normal control group and are related to the course of schizophrenia. The expression of other proteins in the network did not differ significantly between the two groups.

[0044] Table 1: Differential expression levels of extracellular vesicle proteins between the proteomics screening control group and the disease group

[0045] Extracellular vesicle proteins Control group (n=28) Disease group (n=20) multiple Corrected P value C3(mean±SD) 1.96E+06 (5.15E+05) 2.67E+06 (1.19E+06) 1.37 0.039 C4(mean±SD) 9.72E+04 (4.64E+04) 1.93E+05 (8.80E+04) 1.99 <0.001 C4BPA(mean±SD) 8.52E+05(2.61E+05) 1.66E+06 (9.48E+05) 1.95 0.003 PROS1(mean±SD) 1.83E+05 (5.40E+04) 2.72E+05 (9.05E+04) 1.49 <0.001

[0046] Example 2

[0047] First validation of supersensitive electroluminescence (MSD) technology

[0048] Following the proteomics screening in Example 1, the optimal proteins were selected from extracellular vesicle proteins that were significantly upregulated compared to the control group and were most associated with the pathological process of schizophrenia. Specifically, proteins C3, C4, C4BPA, and PROS1 were selected and quantitatively verified using ultrasensitive electroluminescence technology.

[0049] Meanwhile, plasma C4 protein, which has been shown to be significantly upregulated in schizophrenia patients in previous studies, was added for validation.

[0050] The specific implementation method is as follows:

[0051] (1) A new batch of subjects were collected and divided into a disease group and a control group. The disease group consisted of 26 patients with schizophrenia collected from the Shanghai Mental Health Center, and the control group consisted of 26 healthy individuals matched for age and gender.

[0052] (2) Extracellular vesicles in the plasma of each subject were enriched by ultracentrifugation and proteins in the vesicles were extracted with cell lysis buffer.

[0053] (3) Take 0.0625ug of protein sample and add it to the MSD multiarray high binding plate; dry the plate at room temperature for 90 minutes, and then incubate it at 37°C for 30 minutes; pre-incubate the antigen-coated plate with 3% BSA at room temperature for 60 minutes, and store it at 4°C overnight; add the verification antibodies: anti-rabbit Complement C3, anti-rabbit Complement C4, anti-rabbit Complement C4BPA, and anti-rabbit PROS1 to the plate and incubate it at room temperature for 1 hour; wash with PBST and incubate the plate with goat anti-rabbit SULFO-TAG™ antibody for 1 hour; add MSD-T read buffer to the plate and read the detection results on the MSD Sector Imager 2400 reader.

[0054] (4) The plasma C4 protein content was detected simultaneously using an ELISA kit.

[0055] The results are shown in Table 2. It can be seen that in the first validation phase, the expression levels of extracellular vesicle proteins C3, C4, C4BPA and PROS1, as well as plasma C4 protein, were significantly higher in patients with schizophrenia than in the normal control group.

[0056] Table 2: Differential expression levels of histones between the control group and the disease group during the first validation phase.

[0057]

[0058]

[0059] Example 3

[0060] Second ELISA verification

[0061] Following the initial validation in Example 2, and considering the feasibility of clinical use, vesicles were enriched using a commercial extracellular vesicle enrichment kit, with plasma C4 protein added, and the validation was performed again in a third population group.

[0062] The specific implementation method is as follows:

[0063] (1) A new batch of subjects were collected and divided into a disease group and a control group. The disease group consisted of 24 patients with schizophrenia collected from the Shanghai Mental Health Center, and the control group consisted of 24 healthy individuals matched for age and gender.

[0064] (2) Extracellular vesicles in plasma were enriched using a commercial extracellular vesicle enrichment kit.

[0065] (3) Extract proteins from vesicles using cell lysis buffer.

[0066] (4) Take 100 μg of total protein and use an ELISA kit to detect the contents of C3, C4, C4PBA and PROS1 in extracellular vesicles; at the same time, use an ELISA kit to detect the contents of C4 in plasma.

[0067] The results are shown in Table 3. It can be seen that in the second validation phase, the expression levels of extracellular vesicle proteins C3 and C4BPA and plasma C4 protein in patients with schizophrenia were significantly higher than those in the normal control group, and the average levels of extracellular vesicle proteins C4 and PROS1 were higher than those in the normal control group.

[0068] Table 3: Differential protein expression levels between the normal group and the disease group during the second validation phase

[0069] Protein name (mean ± SD) Control group (n=24) Disease group (n=24) multiple Corrected P value Extracellular vesicle protein C3 12.71E+02(2.84E+02) 17.07E+02(2.76E+02) 1.34 0.006 Extracellular vesicle protein C4 17.30E+04 (12.18E+04) 24.77E+04(17.05E+04) 1.45 0.16 Extracellular vesicle protein C4BPA 76.88E+04(25.28E+04) 104.51E+04(37.2E+04) 1.36 0.037 extracellular vesicle protein PROS1 7.03E+02(2.80E+02) 9.08E+02(2.73E+02) 1.30 0.18 Plasma protein C4 25.34(12.27) 52.10(51.40) 2.05 0.015

[0070] The data analysis methods in Examples 1 to 3 are as follows: Data analysis is performed using the R software package. For comparisons between two groups, normality is first assessed. If a normal distribution is satisfied, homogeneity of variance is tested. For two groups with homogeneous variance, Student's t-test is used for comparison; for two groups with unequal variance, Welch correction analysis is used. If a normal distribution is not satisfied, the Mann-Whitney U test is used. P < 0.05 is considered statistically significant.

[0071] Example 4

[0072] The first model of the marker was built using the XGboost algorithm.

[0073] Using the XGBoost algorithm and based on 15-fold cross-validation with 5-fold folding, a diagnostic model for the above five proteins (i.e., extracellular vesicle proteins C3, C4, C4BPA, PROS1, and plasma protein C4) was established as the first model, with the following feature parameters:

[0074] The weak learner booster is set to gbtree.

[0075] Define a loss function objectiveiv that minimizes the loss function, with the parameter set to logistic.

[0076] The learning rate parameter, learning_rate, is set to 0.05.

[0077] The maximum depth of the specified tree is max_depth is 6.

[0078] The number of model iterations, n_estimators, is 700.

[0079] The Gamma parameter is 0.5

[0080] The minimum sum of the weights of the leaf nodes is 1.

[0081] Subsample is 0.8

[0082] Reg lamba is 0.7

[0083] The evaluation metric is eval_metri, with the parameter set to auc.

[0084] Among them, ROC curve analysis is used to evaluate the value of protein biomarkers in the diagnosis of schizophrenia. The closer the area under the curve (AUC) is to 1, the higher the diagnostic value of the indicator.

[0085] like Figure 4 As shown in Figure A, the first model had an AUC of 89.9% in distinguishing between schizophrenia and healthy controls, an accuracy of 89.89%, a sensitivity of 90.35%, and a specificity of 89.62%.

[0086] like Figure 5 As shown in Figure A, the first model had an AUC of 93.42% in distinguishing between patients with schizophrenia and bipolar disorder, an accuracy of 86.58%, a sensitivity of 83.31%, and a specificity of 89.67%.

[0087] like Figure 6 As shown in Figure A, the first model achieved an AUC of 92.8%, an accuracy of 87.0%, a sensitivity of 88.0%, and a specificity of 87.0% when differentiating between patients with schizophrenia and major depressive disorder (MDD).

[0088] The results showed that the biomarkers based on these five proteins have high diagnostic value for schizophrenia and can be well distinguished from easily confused BD and MDD.

[0089] Example 5

[0090] A second model of the markers was built using the XGboost algorithm.

[0091] Based on the feasibility of clinical use, and according to the verification results of the extracellular vesicle extraction kit in Example 3, the extracellular vesicle proteins C3 and C4PBA were significantly elevated in patients with schizophrenia.

[0092] Using the XGBoost algorithm and based on 15-fold cross-validation with 5-fold folding, a diagnostic model for three proteins (extracellular vesicle proteins C3 and C4BPA, and plasma protein C4) was established as a second model. The characteristic parameters are as follows:

[0093] The weak learner is booste, with parameters set to gbtree.

[0094] Define a loss function objectiveiv that minimizes the loss function, with the parameter set to logistic.

[0095] The learning rate parameter, learning_rate, is set to 0.1.

[0096] The maximum depth of the specified tree is max_depth = 3

[0097] The number of model iterations, n_estimators, is 500.

[0098] The Gamma parameter is 0.5

[0099] The minimum sum of the weights of the leaf nodes is 1.

[0100] Subsample is 0.8

[0101] Reg lamba is 0.7

[0102] The evaluation metric is eval_metri, with the parameter set to auc.

[0103] Among them, ROC curve analysis is used to evaluate the value of protein biomarkers in the diagnosis of schizophrenia. The closer the area under the curve (AUC) is to 1, the higher the diagnostic value of the indicator.

[0104] like Figure 4 As shown in B, the second model had an AUC of 87.6% in distinguishing between schizophrenia and healthy controls, with a 95% confidence interval of 86.1%–89.1%, an accuracy of 87.63%, a sensitivity of 87.16%, and a specificity of 88.45%.

[0105] like Figure 5 As shown in B, the second model had an AUC of 90.9%, accuracy of 85.2%, sensitivity of 84.5%, and specificity of 86.3% in distinguishing between patients with schizophrenia and bipolar disorder.

[0106] like Figure 6 As shown in B, the second model achieved an AUC of 82%, an accuracy of 76.3%, a sensitivity of 75.4%, and a specificity of 77.4% when distinguishing between patients with schizophrenia and major depressive disorder (MDD).

[0107] The results showed that biomarkers based on three proteins (i.e., extracellular vesicle proteins C3, C4BPA, and plasma protein C4) also have high diagnostic value for schizophrenia and can be well distinguished from easily confused BD and MDD.

[0108] In addition, biomarkers of three proteins (i.e., extracellular vesicle proteins C3, C4BPA, and plasma protein C4) are better able to meet the cost-effectiveness requirements of clinical use, while also achieving better diagnostic results.

[0109] Example 6

[0110] Develop personalized scoring tools

[0111] Using the XGboost algorithm, diagnostic models developed based on Examples 4 and 5 can be developed as follows: Figure 7 The individualized scoring tool shown is used to assess the disease status of subjects at the individual level.

[0112] For example, by inputting the expression levels of the above five proteins (extracellular vesicle proteins C3, C4, C4BPA, PROS1, and plasma protein C4) or three proteins (extracellular vesicle proteins C3, C4BPA, and plasma protein C4) of each subject into the corresponding model, a specific value can be reported. When the value is positive, the subject is predicted to "have schizophrenia", and when the value is negative, the subject is predicted to "other".

[0113] Example 7

[0114] Establish a regression model between individualized patient scores and drug efficacy.

[0115] Using the individualized scoring system developed in Example 6, a regression model was established between patient individualized scores and 3-month drug efficacy (e.g.) Figure 8 (As shown).

[0116] The regression model can be input as: individualized patient scores at baseline, and output as: predicted PANSS score reduction rate after treatment, with a reduction rate ≥25% indicating effective drug treatment.

[0117] The aforementioned personalized scoring tools can be created as follows: Figure 9 The software shown is in this form.

[0118] In this invention, based on the study of EV proteomics in schizophrenia, all EV proteins associated with schizophrenia were screened more comprehensively. The study found that EV proteins related to the complement system were most relevant to the pathological process of schizophrenia and showed significant differences between patients and healthy individuals.

[0119] Meanwhile, these complement-related EVs exhibit distinct patterns of behavior in schizophrenia, bipolar disorder, and major depressive disorder, effectively distinguishing schizophrenia from the latter.

[0120] Because EVs are encapsulated in a lipid bilayer, their internal proteins are protected from degradation by proteases, nucleases, and other degrading enzymes. Therefore, EV proteins are more stable than plasma proteins, exhibiting greater consistency across different populations and experimental batches, demonstrating their excellent potential as stable diagnostic biomarkers for schizophrenia. Consequently, detecting the expression levels of complement system-related proteins derived from blood EVs can be used as a routine diagnostic method for predicting drug efficacy, thus optimizing clinical diagnostic strategies.

[0121] Furthermore, past research on genetic blood biomarkers has primarily relied on simple analytical methods such as linear regression and logistic regression for model building in diagnosis and drug efficacy prediction. These methods can only establish linear relationships or certain forms of nonlinear relationships, making it difficult to handle more complex relationships. They are also sensitive to data noise; when outliers or noisy data are present, the models are prone to overfitting or underfitting. In modeling, these simple methods typically consider only a single feature variable, failing to handle the interactions between multiple feature variables. Moreover, these models do not consider the biomarker's ability to differentiate from other easily confused mental illnesses (such as bipolar disorder and major depressive disorder), and cannot provide individualized diagnosis and prediction.

[0122] This invention utilizes the XGBoost algorithm, an ensemble method that leverages gradient-boosting trees and is a powerful machine learning algorithm. It can handle high-dimensional and sparse data, establish more complex nonlinear relationships, and offers greater flexibility. XGBoost can combine multiple weak models to generate accurate classifications and predictions, exhibits robustness to noisy data, and can achieve relatively accurate results even with small sample sizes. XGBoost leverages the combinatorial capabilities of tree models to automatically uncover interactions between features and perform feature selection. Therefore, machine learning based on XGBoost can be used to develop novel biomarkers based on extracellular vesicle proteins in blood cells and generate scoring systems for the accurate diagnosis of schizophrenia and the prediction of antipsychotic drug responses at a personalized level.

[0123] In this specification, the invention has been described with reference to specific embodiments thereof. However, it will be apparent that various modifications and variations can be made without departing from the spirit and scope of the invention. Therefore, the specification and drawings should be considered illustrative rather than restrictive.

Claims

1. The application of a reagent for detecting the expression level of biomarkers in peripheral blood in the preparation of diagnostic systems or drug efficacy prediction systems for schizophrenia, characterized in that, The biomarkers include C3 protein in plasma extracellular vesicles, C4 protein in plasma extracellular vesicles, C4BPA protein in plasma extracellular vesicles, PROS1 protein in plasma extracellular vesicles, and C4 protein in plasma; or, the biomarkers include C3 protein in plasma extracellular vesicles, C4BPA protein in plasma extracellular vesicles, and C4 protein in plasma.

Images