Biomarker and model for predicting postoperative cognitive impairment risk of elderly patients and application of biomarker and model

Abstract

The invention discloses a biomarker and a model for predicting postoperative cognitive impairment risk of an elderly patient and application of the biomarker and the model, and belongs to the technical field of medical diagnosis. Protein related to postoperative cognitive impairment is obtained through Olink proteomics identification, a biomarker group composed of GPC5 protein, NEP protein, SPOCK1 protein, LXN protein, IL12 protein, CLM-1 protein, NTRK3 protein and CD200R1 protein is obtained through optimization, and a POCD risk prediction model is constructed according to the marker group. Through verification, the POCD risk prediction model has high clinical applicability, and a reliable tool is provided for preoperative POCD high-risk population screening of elderly patients.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical diagnosis, and particularly relates to a biomarker, a model for predicting the risk of postoperative cognitive dysfunction in elderly patients, and their applications. Background Art

[0002] Postoperative cognitive dysfunction (POCD) refers to a complication in which patients experience impairments in aspects such as memory, executive function, and orientation after anesthesia and surgery, accompanied by a decline in social activity ability. So far, its exact cause and pathogenesis are not clear, and there is a lack of a unified diagnostic standard. The occurrence of POCD will prolong the hospital stay of patients, reduce the quality of life of patients, increase postoperative mortality, and cause a serious burden on individuals and society.

[0003] With the aggravation of social aging, the number of surgeries for elderly patients is increasing, and POCD has attracted increasing attention. The degenerative changes in the central nervous system of the elderly have their special features, including a decrease in brain mass, the number of neurons, a reduction in blood flow velocity due to decreased blood vessel elasticity, and an expansion of the extracellular space, etc. These changes will cause a decline in the cognitive function of the elderly.

[0004] Normal aging will lead to changes in the functional connectivity of the main networks of the default mode network (DMN) and some other networks, affecting the reaction time, dual-task performance, executive ability, etc. of the elderly. Existing clinical studies have shown that anesthetic drugs may mainly interfere with the higher-order brain networks related to patient cognition, thus leading to cognitive function changes. In short, networks related to high-level neuropsychological functions (such as DMN, executive control network, salience network) are more sensitive to anesthesia; damage to the DMN in elderly patients will affect cognitive function. When under the action of various perioperative factors such as anesthesia and surgery, the further deterioration of cognitive function is manifested as POCD.

[0005] The incidence of POCD varies greatly among different studies. An international multicenter study on POCD (ISPOCD) showed that among 1,218 patients over 60 years old who underwent non-cardiac surgery, the incidence of POCD was 25.8% at 1 week after surgery and 9.9% at 3 months after surgery. Among cardiac surgery patients, the incidence of POCD was 30% - 80% within several weeks after surgery and 10% - 60% 3 - 6 months later. The main reasons for these differences are mainly related to factors such as different behavioral testing methods, surgical types, testing time limits, and patient ages used in different studies. Although there is currently no unified neuropsychological testing method for POCD, the international academic community generally uses the ISPOCD method and the neuropsychological test battery (NP tests) modified by Newman as templates. Domestic studies generally use the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment Scale (MoCA) to detect POCD.

[0006] It is worth noting that current neuropsychological methods mainly focus on testing cognitive functions such as learning and memory, and rarely involve changes in patients' emotions such as anxiety and depression. In addition, whether the pain, reduced sleep quality, drug use, anxiety state, and environmental factor changes that commonly exist during the perioperative period have an impact on the test still needs further study. Another research report shows that the neuropsychological test time (test time window) is crucial for the diagnosis of POCD. Moreover, how to consider patient loss to follow-up and exclude the learning effect of patients are all methodological issues worthy of consideration. Neuropsychological tests in clinical diagnosis and treatment cannot meet the needs of early diagnosis of POCD. Therefore, identifying potential POCD patients and making early diagnoses has important clinical significance.

[0007] After conducting a large number of investigations on the incidence of POCD, the current research focus has shifted to the selection of POCD warning indicators. Currently, most biomarkers for POCD are discovered based on its pathophysiological processes, mainly including the following categories: (1) Central inflammation reaction-related biomarkers such as IL-1, IL-6, TNF-α, etc.; (2) Amyloid-β (Aβ) and tau protein; (3) Nerve injury-related markers such as S-100β, neuron-specific enolase, and glial fibrillary acidic protein, etc.; (4) Neurotransmitters and neurotrophic factors such as acetylcholine and brain-derived neurotrophic factor, etc.; (5) APOEε4 gene. However, there is currently a lack of biological targets with high specificity, high sensitivity, and convenient detection in clinical practice.

[0008] Exploring specific POCD molecular markers and developing convenient, efficient, and non-invasive prediction and early diagnosis new technologies have become urgent clinical problems to be solved. However, due to various factors, relevant studies all have their respective limitations and there are no high-quality and large-sample clinical studies. Currently, there is still a lack of self-developed non-invasive and efficient technologies for early prediction, early diagnosis, or screening of postoperative cognitive dysfunction applicable to Chinese residents. Summary of the Invention

[0009] Based on the above background, the present invention provides a group of neuroprotein-related biomarkers that can be used for predicting the risk of POCD occurrence and early screening in elderly patients. Further, the present invention constructs a POCD risk prediction model according to the biomarkers. Through verification, the POCD risk prediction model has high clinical applicability, can provide a reliable tool for screening high-risk populations of POCD before surgery in elderly patients, and is expected to become a standardized tool for POCD risk prediction and early diagnosis.

[0010] The present invention includes the following technical solutions:

[0011] In the first aspect, the present invention provides the application of a group of biomarkers and / or reagents for detecting the biomarkers in the preparation and / or screening of products for predicting the risk of postoperative cognitive dysfunction (POCD) in elderly patients, wherein the biomarkers are a combination of proteins or their encoding genes shown in 1)-8):

[0012] 1) GPC5 protein or its encoding gene;

[0013] 2) NEP protein or its encoding gene;

[0014] 3) SPOCK1 protein or its encoding gene;

[0015] 4) LXN protein or its encoding gene;

[0016] 5) IL12 protein or its encoding gene;

[0017] 6) CLM-1 protein or its encoding gene;

[0018] 7) NTRK3 protein or its encoding gene;

[0019] 8) CD200R1 protein or its encoding gene.

[0020] The products include but are not limited to models, reagents, kits, chips, test strips, membrane strips, or detection platforms.

[0021] The reagents for detecting the biomarker include any reagents required for detecting the expression level of the proteins shown in 1)-8) or their encoding genes by RT-PCR, RT-qPCR, biochip detection, Southern blotting, in situ hybridization, flow cytometry, enzyme-linked immunosorbent assay, immunoblotting, immunohistochemistry, immunofluorescence, or fluorescent dye method.

[0022] In some embodiments of the present invention, the reagents for detecting the biomarker are selected from at least one of the following:

[0023] Ⅰ) Antibodies specifically against the proteins shown in 1)-8);

[0024] Ⅱ) Primers specifically amplifying the genes encoding the proteins shown in 1)-8);

[0025] Ⅲ) Probes specifically recognizing the genes encoding the proteins shown in 1)-8) or their transcripts.

[0026] Preferably, the reagents for detecting the biomarker include any reagents required for detecting the expression level of the proteins shown in 1)-8) by flow cytometry, enzyme-linked immunosorbent assay, immunoblotting, or immunohistochemistry.

[0027] In some embodiments of the present invention, the reagents for detecting the biomarker are selected from antibodies specifically against the proteins shown in 1)-8), and the antibodies can be conjugated with one of acridinium ester, HRP, ALP, or FITC.

[0028] Most preferably, the reagents for detecting the biomarker include any reagents required for detecting the expression level of the proteins shown in 1)-8) by enzyme-linked immunosorbent assay.

[0029] Unless otherwise specifically explained, the prediction in the present invention includes predicting the risk of postoperative cognitive dysfunction in patients before, during, and after surgery, and the surgery is a conventional surgery for treating diseases known to those skilled in the art, including but not limited to orthopedic surgery, and the anesthesia state during the surgery is selected from general anesthesia or spinal anesthesia.

[0030] Unless otherwise specifically explained, the elderly patients in the present invention are selected from men and / or women aged ≥ 60 years old, and in the specific embodiments of the present invention, the elderly patients are selected from men and / or women aged ≥ 65 years old.

[0031] In a second aspect, the present invention provides a prediction model for predicting the risk of postoperative cognitive dysfunction in elderly patients, and the model includes an information acquisition module and a prediction module.

[0032] The information acquisition module is used to perform the step of acquiring the biomarker expression level information of the subject, and the biomarkers are selected from the combination of GPC5 protein, NEP protein, SPOCK1 protein, LXN protein, IL12 protein, CLM-1 protein, NTRK3 protein and CD200R1 protein.

[0033] The prediction module is used to perform the step of predicting the risk of the subject having cognitive dysfunction according to the biomarker expression level information.

[0034] Specifically, the model predicts the probability of the subject having postoperative cognitive dysfunction by detecting the expression level of the biomarker protein.

[0035] Those skilled in the art can obtain the above-mentioned biomarker protein expression level information in a large number of samples and build the prediction model described in the present invention according to conventional machine learning methods. The prediction model includes, but is not limited to, random forest, kNN or SVM.

[0036] In the third aspect, the present invention provides an application of the prediction model described in the second aspect of the present invention in the preparation of a product for predicting the risk of postoperative cognitive dysfunction in elderly patients. The application includes the following steps:

[0037] S1: Quantitatively detect the expression level of the biomarker in a batch of samples, and the biomarker is selected from the combination of GPC5 protein, NEP protein, SPOCK1 protein, LXN protein, IL12 protein, CLM-1 protein, NTRK3 protein and CD200R1 protein;

[0038] S2: Perform data analysis and machine learning algorithms on the biomarker expression level data;

[0039] S3: Quantitatively detect the expression level of the biomarker in the subject sample, and predict the risk of the subject having cognitive dysfunction based on the probability score in the machine learning algorithm.

[0040] The machine learning algorithm described in the present invention is a conventional technical means in the art, including: inputting training set sample data, training the model, evaluating the model, inputting test set sample data for blind testing, and evaluating the blind testing results of the test set.

[0041] The technical solution provided by the present invention has the following technical advantages:

[0042] (1) The present invention successfully constructs a biomarker library covering a variety of proteins with significant correlation differences in POCD, provides basic data support for in-depth understanding of the pathological mechanism, risk prediction and early diagnosis of POCD, and at the same time enhances the ability of research physicians in the field of basic research.

[0043] (2) Based on the screened differential proteins and their functional interaction networks, the present invention successfully constructed a POCD risk prediction model and evaluated its prediction performance through logistic regression analysis, obtaining good results. The verification results showed that the model constructed by the present invention has high clinical applicability, can provide a reliable tool for screening high-risk populations of preoperative POCD in elderly patients, and further improve the practical ability of research physicians in model development and clinical application.

[0044] (3) The POCD risk prediction model provided by the present invention has the advantages of non-invasive, convenient and strong operability. Through peripheral blood detection, accurate preoperative risk assessment can be achieved, providing personalized solutions for targeted intervention and perioperative management of elderly patients. With multi-center clinical verification and further optimization of the model, it is expected to become a standardized tool for POCD risk prediction and early diagnosis, opening up a new path for clinical decision-making and scientific research achievement transformation of physicians in the field of disease prediction and prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Research design and workflow involved in the present invention.

[0046] Figure 2 Differentially expressed proteins in non-POCD and POCD patients before (left) and after (right) surgery.

[0047] Figure 3 ROC curve analysis of the prediction model constructed based on 4 proteins in Olink data.

[0048] Figure 4 ROC curve analysis of the prediction model constructed with 8 proteins in the discovery cohort.

[0049] Figure 5 Plasma levels of 8 proteins were measured by ELISA in the validation cohort, *p<0.05, ns: no statistical significance.

[0050] Figure 6 ROC curve analysis of the prediction model constructed with 8 proteins in the validation cohort. DETAILED DESCRIPTION OF THE INVENTION

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work belong to the scope of protection of the present invention.

[0052] Terms and abbreviations involved in the present invention are explained as follows:

[0053] GPC5: Glypican Proteoglycan 5, a cell surface heparan sulfate proteoglycan, is a member of the glypican family and an important synaptogenic factor that regulates synapse maturation and refinement during development.

[0054] SPOCK1: SPARC (Osteonectin), Cwcv And Kazal Like Domains Proteoglycan 1, also known as testican-1, is currently widely studied in the context of cancer. SPOCK is mainly located in the central nervous system, predominantly in the postsynaptic part of hippocampal pyramidal cells, and is involved in the regulation of midbrain dopamine neurons.

[0055] CLM-1: MRF35-like molecule-1, also known as a member of the CD300 molecular-like family, is an inhibitory receptor that regulates immune responses.

[0056] EZR: Ezrin, is one of the specific proteins that connect the cytoskeleton to the cell membrane, and it contributes to the pathways of endocytosis, exocytosis, and transmembrane signaling.

[0057] IL12: Interleukin 12 is a pro-inflammatory cytokine that promotes Th1 cell differentiation.

[0058] LXN: Latexin, a metal carboxypeptidase inhibitor, is expressed in a subset of glutamatergic projection neurons in the mammalian lateral neocortex, and LXN plays an important role in specific networks of the developing brain.

[0059] NEP: Neprilysin, a type II membrane metalloendopeptidase, is currently found to be one of the most critical metalloproteinases involved in the degradation and clearance of β-amyloid protein, which is an important pathological mechanism in the progression of Alzheimer's disease.

[0060] NTRK3: Neurotrophic tyrosine receptor kinase 3, a membrane-bound receptor, phosphorylates the NTRK3 protein and the MAPK signaling pathway upon binding to neurotrophins. Studies have found that NTRK3 gene mutations are associated with mood and anxiety disorders.

[0061] The CD200-CD200Ra pairing system inhibits the activation of microglia in the central nervous system under physiological conditions. CD200 is a transmembrane protein, also known as OX-2 protein, which is widely distributed on the surface of various cells such as glial cells, neurons, epithelial cells, etc., and is an important immune regulatory protein. CD200R1 is a subtype of the CD200 receptor.

[0062] Olink: A newly developed technology that plays an important role in precision proteomics research and provides new options for multi-omics research. Olink technology can detect low-concentration biomarkers in a small volume and simultaneously detect multiple important biomarkers efficiently and accurately. Olink has been applied to serum samples of cancer, cardiovascular diseases, aging, multiple sclerosis, psoriasis and hidradenitis suppurativa, primary biliary cholangitis, Alzheimer's disease, diabetes, diabetic nephropathy, and clonal hematopoiesis. In this study, the Olink target 96neuro exploratory panel (Olink Proteomics AB, Uppsala, Sweden) was used, and the analysis readings were expressed as standardized protein expression values. Details of the analysis validation information, such as the limit of detection, within-assay and between-assay precision data, can be found at the link www.olink.com.

[0063] The research protocols and results involved in this invention are as follows:

[0064] 1. Inclusion of research subjects

[0065] This study was a prospective multi-center prediction study. A total of 285 patients over 65 years old who underwent orthopedic surgery were included. The inclusion criteria included at least 9 years of compulsory education, research compliance, no history of severe acute infection or malignant disease, and a baseline MMSE score higher than or equal to 27 points.

[0066] The research protocol of this study was reviewed and approved by the Ethics Committee of Zhongshan Hospital, Fudan University (B2022 - 476R). All analyses complied with local and international regulations on the ethics of human subject research, and written informed consent was provided. The study conformed to the principles of the Declaration of Helsinki.

[0067] The G*Power software was used to calculate the sample size. 88 patients (non-POCD:POCD = 58:30) were included in the discovery cohort, and 40 patients (non-POCD:POCD = 20:20) were included in the validation cohort.

[0068] 2. Research methods

[0069] In this study, the mini-mental state examination (MMSE) was used to assess the cognitive function of patients. MMSE examinations were conducted by nurses on the patients before surgery (baseline), 1 day after surgery (Day 1), and 3 days after surgery (Day 3). Patients with a baseline MMSE examination score higher than or equal to 27 points were included in this study. Patients with a postoperative MMSE score lower than 26 points were diagnosed with POCD. Plasma samples of patients were collected, aliquoted, and frozen at -80°C for future research. The study design and process are as Figure 1 shown.

[0070] Prediction model construction: Based on Olink proteomics data, a logistic regression model was constructed using the stat R software package. The ROC curve was plotted using the pROC R software package, and the area under the curve (AUC) was calculated.

[0071] 3. Data analysis

[0072] The baseline characteristics and biochemical indexes of POCD patients were expressed as the mean ± standard deviation of the measurement data with normal distribution and the median (25% and 75%) of the measurement data with non-normal distribution. Count data were described using proportions. Independent sample t-tests were used for measurement data with normal distribution, and non-parametric tests were used for measurement data with non-normal distribution. Chi-square tests or Fisher's exact tests were used for the analysis of count data. Statistical analysis was performed using IBM SPSS Statistics 25 and R software (version 4.3.1). When the p-value < 0.05, it was considered statistically significant.

[0073] 4. Research results

[0074] A total of 285 patients were included in this study. The cognitive function of the patients was evaluated using the MMSE examination before surgery (28.53 ± 1.02). On the first day after surgery, the MMSE scores of 50 patients were lower than 27 points, and 16.78% of the patients developed postoperative cognitive dysfunction. To explore the potential risks of these patients, they were divided into the non-POCD group and the POCD group according to the MMSE scores. In this study, there were more female patients (male:female = 118:167, p < 0.03), but the incidence of POCD in female patients was less (male:female = 28 / 118:22 / 167, p < 0.04). The baseline and intraoperative characteristics of the two groups of patients, including age, height, weight, BMI, incidence, fasting blood glucose level, resting blood pressure, and anesthesia status, are shown in Table 1.

[0075] Table 1 Clinical characteristics of the patients

[0076]

[0077]

[0078] Fifty-eight non-POCD patients and 30 POCD patients were included in the discovery cohort. The baseline and intraoperative characteristics of the patients, including age, gender, height, weight, BMI, incidence, fasting blood glucose level, resting blood pressure, and anesthesia status, are shown in Table 2.

[0079] Table 2 Clinical characteristics of the patients in the discovery cohort

[0080]

[0081]

[0082] In the discovery cohort, the Olink neuroproteomics panel was used to detect the patient blood samples, and potential biomarkers in the plasma of POCD patients were detected, specifically as Figure 2 shown. Compared with the non-POCD group, the plasma levels of GPC5, LXN, NEP, Sparc, Cwcv, and SPOCK1 in POCD patients were lower at the preoperative baseline. Importantly, the GPC5 protein further decreased in both postoperative POCD and non-POCD patients, the LXN and NEP proteins decreased in non-POCD patients during surgical stimulation, and SPOCK1 did not respond to surgery. The postoperative GPC5 protein level in POCD patients was lower than that in non-POCD patients, and the EZR and IL12 protein levels were higher than those in non-POCD patients, with both groups showing a decrease in IL12 postoperatively and an increase in EZR in non-POCD patients postoperatively.

[0083] Binary logistic regression was used to predict the potential biomarkers in POCD patients, and four proteins significantly decreased in POCD patients were screened as biomarkers, namely the combination of GPC5, LXN, NEP, and SPOCK1. Using these biomarkers as variables, a prediction model was established, the receiver operating characteristic (ROC) curve was plotted, and the area under the curve (AUC) value was calculated to be 0.74 (95% CI: 0.62 - 0.86), as Figure 3 shown.

[0084] Furthermore, other proteins with differential expression trends (p < 0.10) were included, and the AUC value of each protein was calculated. Eight proteins with the highest AUC values were selected for prediction, namely the combination of GPC5, NEP, SPOCK1, LXN, IL12, CLM-1, NTRK3, and CD200R1. Using these eight proteins as variables, a prediction model was established, the receiver operating characteristic (ROC) curve was plotted, and the AUC value under the ROC curve was 0.84 (95% CI: 0.74 - 0.93), as Figure 4 shown.

[0085] There were 20 non-POCD patients and 20 POCD patients in the validation cohort. The patient baseline and intraoperative characteristics, including age, gender, height, weight, BMI, morbidity, fasting blood glucose level, resting blood pressure, and anesthesia status, are shown in Table 3.

[0086] Table 3 Clinical characteristics of patients in the validation cohort

[0087]

[0088]

[0089] To verify the circulating changes in POCD patients and their neurobiological markers, plasma samples from the validation cohort were detected by ELISA. Specifically, ELISA kits for CD200R1 (ab272196, Abcam), CLM-1 (CSB-EL004921HU, Cusabio), GPC5 (ab313967, Abcam), IL12 (KE00240, Proteintech), LXN (ab313989, Abcam), NEP (RDRNEP Hu, Reddot Biotech), NTRK3 (KA5476, Abnova), and Spock1 (KA5960, Abnova) were used, and the specific operation methods can be found in the kit instructions.

[0090] Consistent with the Olink results, as Figure 5 shown, the plasma levels of GPC5, NEP, and SPOCK1 in the POCD group were significantly lower than those in the non-POCD group. The changes in the proteins NTRK3, CD200R1, CLM-1, IL12, and LXN in the plasma of POCD were similar to the Olink detection results but not statistically significant. Further verification showed that the AUC value of the ROC curve for the eight proteins was 0.92 (95% CI: 0.84 - 0.92), as Figure 6 shown.

[0091] The research of this invention found that 16.78% of elderly orthopedic surgery patients had cognitive dysfunction. Therefore, obtaining highly specific biomarkers is particularly important for early predicting the risk of POCD in elderly patients. This invention identified a series of neurocyte biomarkers through Olink proteomics, and further optimized to obtain a combination of eight proteins as biomarkers for predicting POCD, and successfully constructed a POCD risk prediction model. The prediction performance was evaluated by logistic regression analysis, and good results were obtained, providing a new idea for future surgical risk assessment.

[0092] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. Use of biomarkers and / or reagents for detecting said biomarkers in the preparation and / or screening of products for predicting the risk of postoperative cognitive dysfunction in elderly patients, characterized in that: The biomarkers are a combination of proteins or genes encoding them as shown in 1)-8): 1) GPC5 protein or its encoding gene; 2) NEP protein or its encoding gene; 3) SPOCK1 protein or its encoding gene; 4) LXN protein or its encoding gene; 5) IL12 protein or its encoding gene; 6) CLM-1 protein or its encoding gene; 7) NTRK3 protein or its encoding gene; 8) CD200R1 protein or its encoding gene.

2. The use according to claim 1, characterized in that: The products include models, reagents, test kits, chips, test strips, membrane strips or detection platforms.

3. The use according to claim 1, characterized in that: The reagents for detecting biomarkers include any reagents required for detecting the expression level of the proteins shown in 1)-8) or their encoding genes by RT-PCR, RT-qPCR, biochip detection, Southern blotting, in situ hybridization, flow cytometry, enzyme-linked immunosorbent assay, immunoblotting, immunohistochemistry, immunofluorescence or fluorescent dye method.

4. The use according to claim 3, characterized in that: The reagent for detecting biomarkers is selected from at least one of the following: Ⅰ) Antibodies specific for proteins shown in 1)-8); II) Primers for specifically amplifying genes encoding proteins shown in 1)-8); III) A probe that specifically recognizes a gene encoding a protein shown in 1) to 8) or its transcript.

5. The use according to claim 4, characterized in that: The reagents for detecting biomarkers include any reagents required for detecting the expression levels of the proteins shown in 1)-8) by flow cytometry, enzyme-linked immunosorbent assay, immunoblotting or immunohistochemistry.

6. The use according to claim 1, characterized in that: The prediction includes predicting the risk of a patient developing cognitive dysfunction before, during and after surgery.

7. The use according to claim 1, characterized in that: The elderly patients are selected from males and / or females aged ≥60 years.

8. A prediction model, which is used to predict the risk of postoperative cognitive dysfunction in elderly patients, and includes an information acquisition module and a prediction module; The information acquisition module is used to perform the step of acquiring the biomarker expression level information of the subject, wherein the biomarker is selected from the combination of GPC5 protein, NEP protein, SPOCK1 protein, LXN protein, IL12 protein, CLM-1 protein, NTRK3 protein and CD200R1 protein; The prediction module is used to execute the step of predicting the risk of cognitive dysfunction in the subject based on the biomarker expression level information.

9. The prediction model according to claim 8, characterized in that The prediction model is constructed by the following method: obtaining the biomarker protein expression level information in a large number of samples, and modeling according to a conventional machine learning method, wherein the prediction model includes random forest, kNN or SVM.

10. Use of the prediction model according to claim 8 or 9 in the preparation of a product for predicting the risk of postoperative cognitive dysfunction in elderly patients, the application comprising the following steps: S1: quantitatively detecting the expression level of biomarkers in batch samples, wherein the biomarkers are selected from a combination of GPC5 protein, NEP protein, SPOCK1 protein, LXN protein, IL12 protein, CLM-1 protein, NTRK3 protein and CD200R1 protein; S2: Perform data analysis and machine learning algorithms on biomarker expression level data; S3: Quantitatively detect the expression level of biomarkers in the subject's sample, and predict the risk of cognitive dysfunction in the subject based on the probability score in the machine learning algorithm.

Images