Biomarker for diagnosing Alzheimer disease and application thereof

Abstract

The invention relates to the field of biomedicine, in particular to a biomarker for Alzheimer's disease and application. The invention specifically discloses a protein biomarker capable of diagnosing Alzheimer's disease (AD). The protein biomarker is selected from one or more of YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1. According to the protein biomarkers obtained through screening and used for diagnosing the Alzheimer disease (AD), the AUC value of each marker is larger than or equal to 0.6; the combined diagnosis AUC value of the composition of the protein biomarker is greater than or equal to 0.95.

Description

Technical Field

[0001] The present invention relates to the field of biomedicine, and in particular to biomarkers of Alzheimer's disease and applications thereof. Background Art

[0002] Alzheimer's disease (AD) is a progressive neurodegenerative disease with an insidious onset. AD pathological changes occur at least 10-20 years before the onset of clinical symptoms. When patients present with typical clinical symptoms and go to the hospital for treatment, the disease has often progressed to the middle and late stages, and the best time for intervention and treatment has been missed. Previous drugs for the treatment of AD can only temporarily improve symptoms but cannot change the course of the disease. Recently, new drugs targeting amyloid beta (Aβ) have been fully approved by the U.S. Food and Drug Administration, but these new drugs are mainly effective in early AD patients, and early identification of early patients requires accurate diagnosis based on biomarkers.

[0003] Despite the important progress made in current biomarker research, the pathophysiological characteristics of Alzheimer's disease (AD) remain largely unknown. Amyloid-β (Aβ) plaques and tau neurofibrillary tangles are recognized as the two important pathological signs of AD, but they only reflect part of the complex pathophysiological process of AD. Since there is currently no effective means to treat AD or delay the progression of AD, especially when targeted therapies against Aβ and tau proteins have failed to observe significant therapeutic benefits, it is particularly important to expand the understanding of AD biomarkers. This is of great significance for the discovery of new and efficient biomarkers for the accurate diagnosis of AD.

[0004] Cerebrospinal fluid (CSF) proteomics methods provide an indispensable systematic tool for exploring the complex underlying pathological changes of AD. As the final product of gene expression, proteins can integrate the effects of genetic and non-genetic factors on the disease and directly reveal the pathophysiological characteristics of the disease. Although some CSF biomarkers associated with AD have been discovered before, there are some limitations. First, most existing studies usually adopt a hypothesis-driven approach and only include a few candidate proteins in the study. This not only ignores the valuable information of the entire proteomics and fails to discover new markers, but also fails to prove which biomarker performs best among the large number of proteins. Second, some studies have provided in-depth insights into AD-related CSF proteomics using unbiased "omics" methods. However, to date, the ideal CSF biomarker that can excellently distinguish AD from normal controls has not been established. The accuracy of this ideal marker for AD diagnosis needs to be comparable to or even better than the classic AT(N) biomarker. This unmet clinical need urgently requires further expansion of the coverage of proteomics testing.

[0005] Patent CN107576805A discloses the level of PKR in the cerebrospinal fluid of an individual, especially beta-amyloid protein (Aβ), total tau (T-tau) and phosphorylated tau (ptau) markers, to infer whether the individual is at risk of developing a neurodegenerative disease; Patent CN107636468A discloses a method for reducing the progression rate of cognitive deterioration in patients by reducing the patient's brain iron level or reducing the patient's CSF ferritin level; CN114174830A discloses the simultaneous detection of CD69 protein, tau and phosphorylated tau protein in mitogenic lymphocytes, and beta-amyloid peptide in cerebrospinal fluid, which can replace or supplement the traditional Alzheimer's disease detection method. And the reference level of each AD diagnostic biomarker includes the standardized measurement level of the AD diagnostic biomarker from one or more samples from confirmed AD patients, and the level of tau protein in the subject's sample is lower than the reference level, and the levels of p-tau and β-amyloid peptide in the subject's sample are higher than the reference level; CN111796102B discloses growth-associated protein 43 (GAP43), neurogranulin, synaptosome-associated protein 25 (SNAP25) and synaptotagmin 1, which can detect asymptomatic Alzheimer's disease (AD) and even predict the risk of AD in subjects 5-7 years before the onset of AD, and have the advantages of almost no trauma and low risk. WO2023216293A1 discloses a system for predicting dementia or mild cognitive impairment, which includes: a data acquisition module for obtaining data on the subject's miRNA level, the subject's apolipoprotein E4 genotype, the subject's age, and the subject's gender; and a module for calculating the probability of suffering from dementia or mild cognitive impairment (MCI), which is used to calculate the data obtained in the data acquisition module to calculate the probability (p) of the subject suffering from dementia or MCI; as a tool for diagnosing MCI and other dementias, the above-mentioned markers are all based on detection after combining with other proteins and cerebrospinal fluid (CSF) proteins.

[0006] Recent advances in detection technology have greatly increased the number of CSF proteins that can be quantified simultaneously to more than 6,000. Based on this, the present invention uses an innovative data-driven approach to comprehensively reveal AD-related proteomic changes and CSF biomarkers in an unprecedented large cohort with complete information, which consists of cognitively normal (CN), mild cognitive impairment (MCI) and clinically diagnosed AD dementia subjects, all of whom have undergone AT(N) core biomarker testing, which can further support the biological diagnosis of AD. Summary of the invention

[0007] The present application, through strict control design, screened and obtained multiple protein biomarkers with AD diagnostic function in biologically defined AD / normal populations and clinically diagnosed AD / normal populations, thereby completing the present invention.

[0008] In a first aspect, the present invention provides a protein biomarker for diagnosing Alzheimer's disease (AD), wherein the protein biomarker is selected from one or more of YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1.

[0009] Furthermore, the protein biomarker is preferably YWHAG.

[0010] Furthermore, the protein biomarker is preferably SMOC1.

[0011] Furthermore, the protein biomarker is preferably a combination of YWHAG, SMOC1, PIGR and TMOD2.

[0012] Furthermore, the protein biomarker is preferably a combination of YWHAG, ACHE, PCSK1, MMP10 and IRF1.

[0013] Furthermore, the protein biomarker may be derived from blood or cerebrospinal fluid.

[0014] Furthermore, the protein biomarkers that can diagnose Alzheimer's disease (AD) can also be combined with demographic characteristics.

[0015] Furthermore, the demographic characteristics may include age, gender, years of education and / or APOEε4 allele.

[0016] In a second aspect, the present invention provides a use of a protein biomarker in the preparation of a preparation for diagnosing Alzheimer's disease (AD).

[0017] Furthermore, the protein biomarker is selected from one or more of YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1.

[0018] Furthermore, the protein biomarker is preferably YWHAG.

[0019] Furthermore, the protein biomarker is preferably SMOC1.

[0020] Furthermore, the protein biomarker is preferably a combination of YWHAG, SMOC1, PIGR and TMOD2.

[0021] Furthermore, the protein biomarker is preferably a combination of YWHAG, ACHE, PCSK1, MMP10 and IRF1.

[0022] Furthermore, the protein biomarker may be derived from blood or cerebrospinal fluid.

[0023] Furthermore, the protein biomarkers that can diagnose Alzheimer's disease (AD) can also be combined with demographic characteristics.

[0024] Furthermore, the demographic characteristics may include age, gender, years of education and / or APOEε4 allele.

[0025] In a third aspect, a kit for diagnosing Alzheimer's disease (AD) comprises reagents for detecting genes, mRNA, cDNA or proteins of protein biomarkers, instructions, etc.

[0026] Furthermore, the protein biomarker is selected from one or more of YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1.

[0027] Furthermore, the protein biomarker is preferably YWHAG.

[0028] Furthermore, the protein biomarker is preferably SMOC1.

[0029] Furthermore, the protein biomarker is preferably a combination of YWHAG, SMOC1, PIGR and TMOD2.

[0030] Furthermore, the protein biomarker is preferably a combination of YWHAG, ACHE, PCSK1, MMP10 and IRF1.

[0031] Furthermore, the protein biomarker may be derived from blood or cerebrospinal fluid.

[0032] Furthermore, the protein biomarkers that can diagnose Alzheimer's disease (AD) can also be combined with demographic characteristics.

[0033] Furthermore, the demographic characteristics may include age, gender, years of education and / or APOEε4 allele.

[0034] Beneficial Effects

[0035] The protein biomarkers for diagnosing Alzheimer's disease (AD) screened in the present application have an AUC value of each marker ≥ 0.6; the combined diagnosis AUC value of the combination of protein biomarkers is ≥ 0.95. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Differences in protein biomarker expression between groups at different disease stages. (Note: The volcano plot in Figure a shows the results of differential expression analysis of protein biomarkers between the biologically defined AD group and the CNA–T–control group; the volcano plot in Figure b shows the results of differential expression analysis of protein biomarkers between the clinically diagnosed AD dementia group and the CN group.)

[0037] Figure 2 Importance ranking and SHAP plot visualization of biologically defined AD dysregulated proteins.

[0038] Figure 3 Importance ranking and SHAP plot visualization of dysregulated proteins in clinically diagnosed AD dementia.

[0039] Figure 4 ROC curve of CSF protein alone or in combination with other variables for distinguishing different biological stages of AD (ROC curve shows the identification accuracy results of the protein biomarker of the present invention alone or in combination with other variables. The identification accuracy can be quantified as AUC and is marked in numerical form in the figure. Demographic characteristics refer to age, gender, years of education and APOEε4 allele.).

[0040] Figure 5 ROC curves for differentiating CN from clinical AD dementia using the protein biomarkers of the present invention alone or in combination with other variables. (The ROC curves show the results of the discrimination accuracy of the protein biomarkers of the present invention alone or in combination with other variables. The discrimination accuracy can be quantified as AUC and is indicated in numerical form in the figure. Demographic characteristics refer to age, gender, years of education, and APOEε4 allele.)

[0041] Figure 6 Validation results of the external independent dataset PPMI. (The cloud-rain graph shows the differential expression results of CSF proteins between the CNA–T– and preclinical AD groups; the bar graph shows the discrimination accuracy results of different model variables for distinguishing CNA–T– from preclinical AD, and the discrimination accuracy is quantified as AUC and marked in numerical form in the figure. Demographic characteristics refer to age, gender, years of education, and APOEε4 allele.) DETAILED DESCRIPTION

[0042] The specific embodiments of the present invention are further described below. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in the embodiments described below can be combined with each other as long as they do not conflict with each other.

[0043] The AD Neuroimaging Project (ADNI) database described in this article is an ongoing large-scale prospective multicenter longitudinal population cohort study that has recruited participants aged 55 to 90 years in 57 regions in the United States and Canada since 2003. The cohort collects detailed clinical, genetic, imaging, biological samples, and neuropsychological assessment data of the subjects, aiming to provide support for the study of biomarkers and risk factors related to mild cognitive impairment (MCI) and AD. The population recruited by ADNI includes AD patients, MCI patients, and elderly controls, and more than 2,000 subjects have been successfully recruited to date. In addition to collecting and evaluating the baseline characteristics of the subjects, ADNI also conducts long-term follow-up of the participants and updates the follow-up data such as disease diagnosis, cognitive assessment, and biological sample testing in real time. For more information, please visit the official website: http: / / adni.loni.usc.edu / . The ADNI study was approved by the local institutional ethics review committees of all participating centers. All participants or their legal guardians signed written informed consent.

[0044] All ADNI participants underwent standardized assessments and were diagnosed with CN, MCI, or AD dementia according to standardized diagnostic criteria. Detailed inclusion and exclusion criteria for the ADNI cohort population are available at www.adni-info.org.

[0045] The diagnosis of Alzheimer's disease (AD) described in this article includes two types of diagnosis: one is biological AD diagnosis, and the other is clinical AD diagnosis. The "biological AD diagnosis" mentioned here refers to the diagnosis of AD based on the two indicators of Aβ and tau. It is one of the currently recognized AD diagnostic criteria and is also referred to as "biologically defined AD" in this article.

[0046] Herein, "A" in the terms "AT-", "A+T-", "A-T+" and "A+T+" refers to β-amyloid deposition; "T" refers to pathological tau protein. Accordingly, the "AT-" group is a subject with negative β-amyloid deposition and tau protein expression; the "A+T-" group is a subject with positive β-amyloid deposition and negative tau protein expression; the "A+T+" group is a subject with positive β-amyloid deposition and tau protein expression; the "A-T+" group is a subject with negative β-amyloid deposition and positive tau protein expression. Among the above groups, the "A+T-" and "A+T+" groups are "with AD pathology"; the "A-T+" and "AT-" groups are "without AD pathology".

[0047] The term "clinically diagnosed AD" refers to the diagnosis of AD based on clinical symptoms such as the subject's cognitive function level and / or dementia degree.

[0048] In this paper, the subjects were divided into a cognitive normal group ("CN" or "CN group"), a mild cognitive impairment ("MCI" or "MCI group") group, and a clinically diagnosed AD dementia group ("dementia" or "dementia group") according to the cognitive test scores (including the Mini-Mental State Examination (MMSE), the Montreal Cognitive Assessment (MoCA), and the Clinical Dementia Rating Scale (CDRSB) scores).

[0049] It should be noted that the results of biological AD diagnosis and clinical AD diagnosis may not be consistent. For example, people who appear to be cognitively normal clinically (CN) may still appear as A+T- or A+T+, that is, they have AD pathology in the biological sense, or they may appear as A-T+ or AT-, that is, they do not have AD pathology in the biological sense. Therefore, although CN can be used as a control group, it is not an ideal control group in the strict sense. In order to conduct strict control, the present invention uses a population of subjects with normal cognition and no AD pathology, namely "CNA-T-", as a true control group.

[0050] The group terms used in this article are as follows:

[0051] The CN group “CNA–T–” refers to subjects without AD pathology in the CN population; “preclinical AD” refers to subjects with AD pathology in the CN population.

[0052] In the MCI group: “MCIA–T–” refers to subjects without AD pathology in the MCI population; “AD-derived MCI” refers to subjects with AD pathology in the MCI population.

[0053] In the dementia group: “Dementia A–T–” refers to subjects without AD pathology in the dementia population; “AD-induced dementia” refers to subjects with AD pathology in the dementia population.

[0054] Although the AD diagnostic protein biomarkers discovered in the present invention are derived from cerebrospinal fluid samples, according to previous literature (Differences Between Plasma and Cerebrospinal Fluid Glial Fibrillary Acidic Protein Levels Across the Alzheimer Disease Continuum, JAMA Neurol. 2021 Dec 1; 78(12): 1471-1483.), blood biomarkers related to neurological diseases can generally be used as substitutes for cerebrospinal fluid biomarkers. These protein biomarkers in the blood and their combinations also have the ability to diagnose AD.

[0055] The experimental methods in the following examples are conventional methods unless otherwise specified, and the experimental materials used in the following examples are commercially available unless otherwise specified.

[0056] Example 1 Research subjects and methods

[0057] 1.1 Selection of research subjects and population

[0058] The subjects in this part of the study came from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, the Parkinson's Progression Markers Initiative (PPMI) and the CABLE cohort. The ADNI cohort was used as a discovery cohort, and the PPMI cohort and CABLE cohort were used as validation cohorts.

[0059] Subjects with a Mini-Mental State Examination (MMSE) score of ≥24 and a Clinical Dementia Rating (CDR) score of 0 were classified as the CN control group. Subjects with an MMSE score between 24 and 30, objective memory loss confirmed by the delayed recall test of the Wechsler Memory Scale Logical Memory II, a CDR score of 0.5, preserved daily living ability, and no dementia were classified as the MCI group. AD dementia was diagnosed according to the criteria of the NINCDS-ADRDA consortium, with an MMSE score between 20 and 26 and a CDR score of 0.5 or 1.0. Only ADNI subjects with baseline CSF proteomic measurements, CSF Aβ42, and CSF phosphorylated-tau181 (p-tau181) data were included in this part of the study, and the rest of the participants were excluded. A total of 707 ADNI participants (including 138 patients with a clinical diagnosis of AD dementia, 403 patients with MCI, and 166 CN controls) were included, and 51 of these participants underwent postmortem neuropathological assessment.

[0060] The external validation cohort included CN subjects from the Parkinson's ProgressionMarkers Initiative (PPMI) cohort. The population inclusion and exclusion criteria are detailed at http: / / ppmi-info.org / study-design. The PPMI study was approved by the local institutional review boards of all participating centers. All participants or their legal guardians signed written informed consent. PPMI subjects included in the study had CSF proteomic measurement data, CSF Aβ42, and CSF p-tau181 data at baseline, and the remaining participants were excluded. A total of 172 subjects met the inclusion criteria.

[0061] The external validation cohort also included 242 non-demented participants from the Chinese Alzheimer's Biomarker and Lifestyle (CABLE cohort). These participants also had CSF proteome measurement data, CSF Aβ42, and CSF p-tau181 data. They were aged between 40 and 90 years old and were all Han Chinese. Specifically, CABLE is a large prospective cohort established for the Han population in northern China. It started in 2017 and has recruited more than 2,000 non-demented (including CN and MCI) participants to date. All participants came from Qingdao Municipal Hospital. Through comprehensive and unified questionnaire assessments, neuropsychological tests, physical examinations, and biological sample collection (including blood and cerebrospinal fluid) of these participants, the CABLE cohort aims to identify biomarkers and risk factors related to AD. The following patients were excluded: (1) patients with central nervous system infection, head trauma, other neurodegenerative diseases (such as Parkinson's disease), or other major neurological diseases; (2) patients with major psychological illness; (3) patients with serious systemic diseases (such as cancer); (4) patients with a family history of genetic diseases. CSF and blood samples were collected by professional clinicians and frozen in a -80°C refrigerator and then transported to Huashan Hospital Affiliated to Fudan University for subsequent quantitative testing. The CABLE study was approved by the local institutional ethics review committees of all participating centers. All participants or their legal guardians signed written informed consent.

[0062] 1.2 CSF proteomics detection method

[0063] CSF samples were collected at baseline and frozen at -80°C from each ADNI evaluation center. CSF proteomics analysis was performed on 758 baseline samples of the ADNI cohort using aptamer-based technology on the SomaScan 7K platform. 2. ADNI’s 7K SomaScan assay is able to simultaneously detect levels of more than 7,000 human proteins, covering more than one-third of the number of proteins in the human body (approximately 20,000 proteins). The SomaScan method can quantitatively detect proteins in a dynamic range from fM to μM, spanning 10 orders of magnitude. Its detection limit is lower than that of traditional immunoassays. Quantified CSF protein concentrations are expressed as relative fluorescence units (RFU) values. Data normalization was performed by SomaLogic. The normalization operation was performed at the sample level. In order to avoid combining different levels of protein signal features in an additional normalization step, the aptamers were divided into three normalization groups based on the signal-to-noise ratio observed in technical replicates and samples. Median normalization was then performed to eliminate other detection biases such as protein concentration, pipetting changes, reagent concentration changes, and detection time. Finally, reference normalization was performed for individual samples to account for additional technical and biological differences. This normalization step was performed by iterative adaptive normalization using the maximum likelihood method, which is a modification of median normalization.

[0064] Following a strict process, quality control was performed on the normalized SomaScan7k data provided by SomaLogic. Specifically, aptamers were removed according to the following criteria:

[0065] (1) The maximum absolute difference between the aptamer scale factor and the median scale factor on any plate was ≥ 0.5;

[0066] (2) the median coefficient of variation across plates was ≥ 0.15;

[0067] (3) According to log 10 After conversion, the aptamer levels were calculated, and if the aptamer value was outside the 1.5-fold interquartile range, it was identified as an outlier;

[0068] (4) analytes and subjects with response rates <65% were excluded;

[0069] (5) The response rates of the aptamers were recalculated and the aptamers and subjects with response rates <85% were eliminated;

[0070] (6) The response rates of the subjects were recalculated and subjects with response rates <85% were eliminated;

[0071] (7) Non-human analytes and analytes without protein targets were eliminated.

[0072] After normalization and quality control procedures, the ADNI website generated data for 7,008 analytes measured in RFU, which could be mapped to 6,361 unique proteins.

[0073] In this part of the study, all 7008 analytes were included in the analysis to select the analyte most closely associated with AD. For proteins tested for ≥2 analytes, we only selected one analyte most closely associated with AD for subsequent analysis.

[0074] Baseline CSF proteomics of subjects in the PPMI cohort was measured using Novartis' proprietary SomaScan platform. The standardization and normalization procedures for PPMI proteomics data, detailed information on quality control, and specific SomaScan detection methods are the same as above. In this part of the study, we only selected some CSF proteins in the PPMI database as a validation of the proteins important for AD diagnosis screened in the ADNI cohort.

[0075] Baseline CSF proteomics of subjects in the CABLE cohort was measured using 4d-dia technology. diaPASEF analysis was performed using Bruker timsTOF Pro proprietary trapped ion mobility spectrometry following standardized assay protocols and standards. The mass spectrometer was coupled to a nano-HPLC system (Bruker Daltonics nanoElute) and chromatographic separation was performed using a C18 capillary column (250 mm × 75 μm, 1.6 μm particle size, IonOptiks). Samples were separated in an LC gradient from 2% to 80% acetonitrile in 10 mM ammonia (pH = 10.0) at a flow rate of 200 nL / min. The diaPASEF analysis ion mobility range was 0.6–1.6 Vs / cm2 (1 / K0). Both the TIMS accumulation time and the ion mobility separation time were fixed at 50 ms. To cover the mass spectral range of 400–1200 m / z, the isolation window was set to an interval of 25 m / z. Hela protein digest standards were used for quality control. All samples were transported to Fudan University for unified testing. In this part of the study, only some CSF proteins in the CABLE database were selected as validation of the proteins important for AD diagnosis screened by the ADNI cohort.

[0076] 1.3AD core biomarker detection

[0077] In ADNI and PPMI, AD core biomarkers including CSF Aβ42, p-tau181, and total tau (t-tau) protein levels were measured using Roche Elecsys and cobas e 601 automated immunoassay analysis systems. In the CABLE cohort, AD core biomarkers including CSF Aβ42, p-tau181, and t-tau protein levels were measured using ELISA. ADNI participants were divided into A–T–, A+T–, A+T+, and A–T+ groups based on CSF Aβ42 and CSF p-tau181 levels. According to the report (Schindler, SE, et al. Cerebrospinal fluid biomarkers measured by Elecsys assays compared to amyloid imaging. Alzheimers Dement 14, 1460-1469 (2018).), the cutoff values ​​of CSF Aβ42 and CSF p-tau181 were 1,098 pg / ml and 26.64 pg / ml, respectively. In the PPMI cohort, the cutoff values ​​of CSF Aβ42 and CSF p-tau181 were 683 pg / ml and 24 pg / ml, respectively. In the CABLE cohort, the cutoff values ​​of CSF Aβ42 and CSF p-tau181 were 128 pg / ml and 44 pg / ml, respectively. 1 According to the 2018 NIA-AA research framework, CNA+T+, MCIA+T+, and dementia A+T+ participants can be divided into preclinical AD, AD-derived MCI, and AD-derived dementia groups, respectively. These three groups are also included in the AD continuum. All CSF biomarkers were tested in duplicate and averaged. The coefficient of variation of proteins within and between batches was less than 15%.

[0078] 1.4 Postmortem neuropathological evaluation

[0079] Postmortem neuropathology processing and evaluation were performed according to the NIA-AA guidelines. There are three histopathological scoring standards used to describe the core neuropathological lesions of AD, including the Thal stage for describing the distribution of Aβ plaques (0-5 points), the Braak stage for describing the distribution and density of tau neurofibrillary tangles (0-6 points), and the CERAD scoring system for describing the density of Aβ plaques (0-3 points). Based on these three aspects, a 4-point scale was converted to comprehensively evaluate the neuropathological changes of AD, including "none" (0 points), "low" (1 point), "medium" (2 points), and "high" (3 points). A score of ≥2 points for AD neuropathological changes indicates that the autopsy is confirmed to be AD, and <2 points indicates that the autopsy is not confirmed to be AD.

[0080] 1.5 Statistical analysis

[0081] All protein biomarker levels were log-transformed and then Z-transformed for all subsequent analyses.

[0082] First, a multivariate linear regression model adjusted for age, sex, years of education, and APOEε4 carrier status was used to identify proteins with differential expression between different diagnostic groups (e.g., biologically defined AD group vs CNA–T– control group, clinically diagnosed AD dementia group vs CN group). Considering the number of proteins tested, the Bonferroni method was used to correct for multiple testing to assess the significance of the association.

[0083] We then identified CSF proteins that are important for AD biological and clinical diagnosis:

[0084] Step 1: Protein importance ranking: The dysregulated proteins associated with biologically defined AD obtained after Bonferroni correction in the multivariate linear regression model were placed into the initially trained light gradient boosting machine (LGBM) classifier and ranked according to the importance of each protein's contribution to the model's diagnostic performance. The contribution of each protein to the model's performance can be quantified as information gain, which represents the ability of CSF proteins to distinguish biologically defined AD from CNA–T– controls. The top-ranked proteins contributed more than 90% to the overall information gain and were therefore retained.

[0085] Step 2: Hierarchical clustering was performed according to the Spearman rank correlation test to further remove redundant proteins with multicollinearity. Proteins that clustered into one category were highly correlated, and only the protein with the largest AUC was selected from each cluster to enter the next sequential forward selection analysis. According to the updated protein importance for AD diagnosis, the proteins were sorted from high to low, and the sequential forward selection strategy was used to add the proteins to the newly trained LGBM classifier, two proteins at a time, until no significant improvement in the area under the curve (AUC) was detected in two consecutive DeLong tests, which was considered to have achieved the best diagnostic performance (i.e., no significant improvement in diagnostic performance was observed when additional proteins were added to the model), and the addition of proteins to the model was manually terminated. Repeating the above steps, a few CSF proteins that are most important for the clinical diagnosis of AD dementia were screened out by testing the ability of the dysregulated proteins associated with clinically diagnosed AD dementia obtained after Bonferroni correction in the multivariate linear regression model to distinguish AD (clinical diagnosis) from CN. The selected important proteins were visualized using Shapley Additive exPlanations (SHAP) plots.

[0086] In the third step, ROC analysis was performed to evaluate the accuracy of the few important proteins screened above for AD diagnosis (including biological and clinical diagnosis) alone or in combination with demographic indicators (age, gender, years of education and APOEε4 allele). The performance of the model was evaluated by nested internal leave-one-out cross-validation, in which only one individual was used as the test set and all the remaining individuals were used as the training set. The procedure was repeated until all participants were used as the test set. The estimated risk probability of each individual was obtained based on nested cross-validation, and bootstrap repeated sampling was performed for 1000 iterations to evaluate the discrimination performance of the model and report the median AUC and 95% confidence interval (CI).

[0087] In order to evaluate the comprehensive diagnostic ability of proteins, a protein panel, i.e., a combination of multiple proteins screened in the above steps, was used to repeat the above ROC analysis instead of a single CSF protein. In this part of the study, the ADNI cohort was used as a discovery cohort, and the PPMI cohort and CABLE cohort were used as validation cohorts. The DeLong test was used to compare whether the AUC values ​​output by different models were significantly different.

[0088] Data analysis and visualization were performed using Python (version 3.9) and R (version 4.1.2). The Python software packages used included lifelines (version 0.27.4), LightGBM (version 3.3.2), scikit-learn (version 1.0.2), pyMannKendall (version 1.4.2), and Shap (version 0.40.0). The R software packages used included ggplot2 and survival.

[0089] Example 2 Screening and discovery of AD diagnostic protein biomarkers

[0090] 2.1 Basic characteristics of the population

[0091] This study included 707 ADNI participants, with 166, 403, and 138 people with a baseline clinical diagnosis of CN, MCI, and AD dementia, respectively (Table 1). The three groups differed in age and APOEε4 carrier status (referring to whether they carried the APOEε4 allele), but there were no significant differences in gender distribution and education level. All participants included in the study not only underwent comprehensive CSF proteomic testing, but also had complete CSF Aβ42 and p-tau181 data. Based on this, the population was divided into different AT groups (A–T– group, A+T– group, A+T+ group, and A–T+ group). According to the 2018 NIA-AA research framework, CNA+T+, MCIA+T+, and dementia A+T+ participants can be referred to as preclinical AD, AD-derived MCI, and AD-derived dementia participants, respectively, and 269 A+T+ (including 21 preclinical AD, 153 AD-derived MCI, and 95 AD-derived dementia) participants were biologically defined as AD. In addition, of the 53 participants who underwent postmortem neuropathology evaluation, 41 were confirmed as AD by postmortem pathology.

[0092] The detailed demographic characteristics of the CN participants in the PPMI cohort included in the validation analysis are shown in Table 2 , including 108 CNA–T– controls and 7 preclinical AD participants. Continuous variables are expressed as median [lower quartile, upper quartile], and dichotomous variables are expressed as number (percentage).

[0093] Detailed demographic characteristics of the participants in the CABLE cohort included in the validation analysis are shown in Table 3 , including 75 CNA–T– controls, 49 preclinical AD participants, and 60 biologically defined AD participants. Continuous variables are expressed as median [lower quartile, upper quartile], and dichotomous variables are expressed as number (percentage).

[0094] Table 1 Basic characteristics of the population included in ADNI

[0095]

[0096]

[0097] Note: In the descriptive analysis of variables, the chi-square test was used to compare the differences between the groups for categorical variables, and the Kruskal-Wallis test was used to compare the differences between the groups for continuous variables.

[0098] Table 2 Basic characteristics of the population included in PPMI

[0099]

[0100] Note: In the descriptive analysis of variables, Fisher's exact test was used to compare the differences between groups for categorical variables, and Kruskal-Wallis test was used to compare the differences between groups for continuous variables.

[0101] Table 3 Basic characteristics of the population included in CABLE

[0102]

[0103]

[0104] Note: In the descriptive analysis of variables, the chi-square test was used to compare the differences between the groups for categorical variables, and the Kruskal-Wallis test was used to compare the differences between the groups for continuous variables.

[0105] 2.2 CSF proteomic differential expression analysis

[0106] Among the 6361 proteins detected by CSF proteomics, 262 proteins were differentially expressed between the biologically defined AD group and the CNA–T–control (cognitively normal) group after correction for age, sex, years of education, and APOEε4 carrier status (Bonferroni-corrected P < 0.05; Figure 1 a).

[0107] These CSF proteins were upregulated in AD patients, and the most significantly differentially expressed proteins included 14-3-3 protein gamma (YWHAG) (β = 1.56, P = 8.97 × 10 -42 ), SPARC-related modular calcium-binding protein 1 (SMOC1) (β = 1.38, P = 1.44 × 10 -30), 14-3-3 protein zeta / delta (YWHAZ) (β=1.31, P=1.67×10 -28 ), calcineurin subunit B type 1 (PPP3R1) (β = 1.29, P = 2.58 × 10 -25 ) and 14-3-3 protein beta / alpha (YWHAB) (β = 1.29, P = 6.07 × 10 -25 In addition, polymeric immunoglobulin receptor (PIGR) (β = 0.64, P = 0.03) and tropomyosin-2 (TMOD2) (β = 1.32, P = 1.85 × 10 -16 ) was significantly upregulated in AD patients.

[0108] Among the 6361 proteins, 50 proteins were differentially expressed between the control group with a baseline diagnosis of CN and the group with a baseline clinical diagnosis of AD dementia after correction for age, sex, years of education, and APOEε4 carrier status (P < 0.05 after Bonferroni correction; Figure 1 b). 41 CSF proteins were upregulated in clinically diagnosed AD dementia patients; namely, 14-3-3 protein zeta / delta (YWHAZ) (β = 1.03, P = 6.55 × 10 -18 ) was the most significantly differentially expressed protein, followed by stromelysin-2 (MMP10) (β = 1.01, P = 1.61 × 10 -13 ), acetylcholinesterase (ACHE) (β = 1.02, P = 3.61 × 10 -10 ) and neurofilament light polypeptide (NEFL) (β=0.77, P=9.67×10 -10 ).

[0109] Another nine CSF proteins were downregulated in clinically diagnosed AD dementia patients, such as interferon regulatory factor 1 (IRF1) (β=-0.59, P=0.013) and neuroendocrine convertase 1 (PCSK1) (β=-0.60, P=0.017).

[0110] 2.3 Protein importance ranking for AD diagnosis

[0111] For the 262 proteins found in the previous step to be differentially expressed in the biologically defined AD group compared with the CNA–T– group, we ranked them according to their importance for the diagnosis of AD (biologically defined).

[0112] The four important proteins screened out for AD biological definition include YWHAG, SMOC1, PIGR and TMOD2. The results are visualized in the form of bar graphs, line graphs and SHAP graphs. Figure 2 As shown in the bar graph in Figure a, YWHAG is the strongest indicator for diagnosing AD (biological definition), followed by SMOC1 and PIGR. When the first few proteins are included in the diagnostic model, the accuracy of AD dementia diagnosis increases rapidly, such as Figure 2 As shown in the AUC on the right axis, as more proteins are included in the model, the diagnostic ability gradually tends to fluctuate smoothly. Using this sequential forward selection scheme, the top four important proteins finally selected for the clinical diagnosis of AD dementia are marked in red font in Figure a and displayed in the SHAP diagram in Figure b.

[0113] The 50 proteins found in the previous step to be differentially expressed in the clinically diagnosed AD group compared with the CN group were ranked according to their importance for the diagnosis of AD (clinical definition). Five important proteins for the clinical diagnosis of AD dementia were screened, namely ACHE, YWHAG, PCSK1, MMP10 and IRF1. The results were visualized in the form of bar graphs, line graphs and SHAP graphs, see Figure 3 As shown in the bar graph in Figure a, ACHE is the strongest indicator for clinical diagnosis of AD dementia, followed by YWHAG and PCSK1.

[0114] When the first few proteins were included in the diagnostic model, the accuracy of AD dementia diagnosis increased rapidly, but as more proteins were included in the model, the diagnostic ability gradually became flat and fluctuated. Using this sequential forward selection scheme, the top five important proteins finally selected for clinical diagnosis of AD dementia are marked in red in Figure a and displayed in the SHAP diagram in Figure b.

[0115] Example 3 Analysis of the diagnostic accuracy of AD diagnostic protein biomarkers

[0116] For the important proteins screened in Example 2, we further explored their accuracy in AD diagnosis (including biological definition and clinical diagnosis) when used alone or in combination with other indicators.

[0117] 3.1 Diagnostic efficacy of AD diagnostic protein biomarkers for biological AD

[0118] CSF YWHAG alone (AUC = 0.969 [0.946-0.987]) was able to discriminate CNA–T– from biologically defined AD very well, with performance comparable to that of CSF SMOC1 (AUC = 0.943 [0.914-0.965], P DeLongtest =0.062). CSF PIGR (AUC = 0.663 [0.594-0.728]) and TMOD2 (AUC = 0.844 [0.797-0.887]) also have a certain degree of identification accuracy. The AUC values ​​of the above four proteins in diagnosing CNA-T- and biologically defined AD were as low as 0.663 and as high as 0.969, which revealed that the four protein biomarkers screened out can be used to diagnose AD.

[0119] To achieve higher diagnostic accuracy, the diagnostic performance of CSF protein in combination with other simple demographic characteristics (including age, sex, years of education, and APOE ε4 allele) was investigated. Compared with CSF YWHAG alone, the combination of YWHAG and demographic indicators (AUC = 0.973 [0.954-0.986]) further improved the discrimination performance (P DeLongtest =0.541)( Figure 4 ); however, when SMOC1 (AUC = 0.968), PIGR (AUC = 0.872), or TMOD2 (AUC = 0.923) was combined with demographic characteristics, a significant improvement in diagnostic accuracy was observed compared with CSF proteins alone (P DeLongtest <0.01).

[0120] In order to evaluate the comprehensive efficacy of the selected markers for the proteins important for AD biological diagnosis (including YWHAG, SMOC1, PIGR and TMOD2), these four proteins were combined into a protein panel for ROC analysis. The AUC value of this protein panel was as high as 0.987 [0.977-0.994], which greatly improved the accuracy of AD diagnosis. Combining this protein panel with demographic characteristics did not further improve the diagnostic accuracy (AUC = 0.991 [0.983-0.996], P DeLongtest=0.109).

[0121] The above results show that when CSF proteins are combined with other simple demographic characteristics, the AUC values ​​of their diagnostic performance are all above 0.8; at the same time, the combined diagnostic efficacy of the four proteins reaches 0.987; the combined diagnostic efficacy of the four proteins combined with demographic characteristics reaches 0.991.

[0122] When the same variable model was applied to differentiate CNA–T– from preclinical AD, MCIA–T– from AD-derived MCI, and dementia A–T– from AD-derived dementia ( Table 4 ), CSF YWHAG and SMOC1 performed well alone, with AUCs of 0.960, 0.950, and 0.864 for YWHAG and 0.947, 0.923, and 0.947, respectively, for SMOC1 ( Figure 4 ), the identification AUCs of PIGR were 0.646, 0.703 and 0.890, and those of TMOD2 were 0.808, 0.812 and 0.947, respectively.

[0123] The accuracy of YWHAG alone in differentiating CNA–T– from preclinical AD was comparable to that of YWHAG combined with demographic characteristics (AUC = 0.955 [0.885-0.991]) or the four-protein panel (AUC = 0.979 [0.946-0.996]).

[0124] The discrimination accuracy of SMOC1 alone was comparable to that of SMOC1 combined with demographic features (AUC = 0.965 [0.900-0.997]) or the accuracy of a protein panel composed of the four proteins.

[0125] In differentiating MCIA–T– from AD-induced MCI, the accuracy of YWHAG alone was comparable to that of YWHAG combined with demographic characteristics (AUC=0.955[0.925-0.976]) or the protein panel composed of four proteins (AUC=0.972[0.953-0.986]), and the accuracy of SMOC1 alone was comparable to that of SMOC1 combined with demographic characteristics (AUC=0.934[0.897-0.962]).

[0126] In differentiating A–T– dementia from AD, the accuracy of YWHAG alone was comparable to that of YWHAG combined with demographic characteristics (AUC=0.964[0.907-1.000]), but the accuracy of the protein panel composed of the four proteins (AUC=0.970[0.889-1.000]) was significantly improved compared with that of CSF YWHAG alone. The accuracy of SMOC1 alone was comparable to that of SMOC1 combined with demographic characteristics (AUC=0.953[0.844-1.000]) or the protein panel composed of the four proteins.

[0127] Table 4 ROC analysis results of protein biomarkers alone or in combination with other variables to distinguish different biological stages of AD

[0128]

[0129]

[0130] Note: The 4-protein panel refers to the combination of 4 proteins (including YWHAG, SMOC1, PIGR and TMOD2) that distinguish CNA–T– from biologically defined AD.

[0131] 3.2 Diagnostic efficacy of AD diagnostic protein biomarkers for clinical AD

[0132] Regarding the differentiation between CN and clinically diagnosed AD dementia, CSF YWHAG alone (AUC = 0.857 [0.809-0.900]) had a high accuracy ( Figure 5 ), and its accuracy was comparable to that of ACHE (AUC = 0.781 [0.711-0.846], P DeLongtest =0.389) or MMP10 (AUC = 0.819 [0.768-0.868], P DeLongtest =0.238), but superior to PCSK1 (AUC = 0.661 [0.598-0.720]) or IRF1 (AUC = 0.661 [0.599-0.722]).

[0133] In order to visually observe the discriminative ability of these important proteins screened out, they were compared with the established AD CSF core biomarkers.

[0134] In differentiating CN from clinically diagnosed AD dementia, the AUCs of CSF Aβ42, p-tau181, and t-tau were 0.844, 0.822, and 0.802, respectively; the AUC of YWHAG was comparable to that of CSF Aβ42, but significantly higher than that of CSF p-tau181 and t-tau (PDeLongtest <0.01).

[0135] Compared with CSF proteins alone, the discrimination performance of the model was not significantly improved when CSF YWHAG, ACHE or MMP10 was combined with demographic characteristics (including age, sex, years of education and APOEε4 allele). However, the discrimination accuracy was significantly improved when demographic characteristics were combined with PCSK1 (AUC = 0.803) or IRF1 (AUC = 0.819) compared with CSF proteins alone.

[0136] In differentiating CN from clinically diagnosed AD dementia, the model with the highest discrimination accuracy was a protein panel composed of five proteins (including ACHE, YWHAG, PCSK1, MMP10, and IRF1) (AUC = 0.975 [0.954-0.990]) ( Figure 5 ), and its discrimination accuracy was even better than that of CSF Aβ42 / 40 (AUC = 0.911), p-tau / Aβ42 (AUC = 0.897), t-tau / Aβ42 (AUC = 0.900), and the combination of three classic AD CSF core biomarkers (Aβ42, p-tau181, and t-tau) (AUC = 0.871) (all P DeLongtest <0.05).

[0137] Combining this protein panel with demographic characteristics did not further improve the diagnostic accuracy (AUC = 0.978 [0.959-0.991], P DeLongtest =0.102).

[0138] When the same variable model was applied to discriminate CN from MCI (Table 5), the discrimination accuracy of YWHAG (AUC = 0.645) was comparable to that of CSF Aβ42 (AUC = 0.659), p-tau181 (AUC = 0.625), or t-tau (AUC = 0.614).

[0139] For the protein panel composed of 5 proteins (AUC=0.731), its identification accuracy was comparable to that of CSF Aβ42 / 40 (AUC=0.715) or the combination of three classic AD CSF core biomarkers (Aβ42, p-tau181 and t-tau) (AUC=0.693), but better than CSF p-tau / Aβ42 (AUC=0.684) or t-tau / Aβ42 (AUC=0.688).

[0140] In differentiating MCI from clinically diagnosed AD dementia, the discrimination accuracy of YWHAG (AUC = 0.721) was comparable to that of CSF Aβ42 (AUC = 0.718), but much higher than that of CSF p-tau181 (AUC = 0.680) or t-tau (AUC = 0.674). The discrimination accuracy of the protein panel composed of the five proteins (AUC = 0.840) was better than that of CSF Aβ42 / 40 (AUC = 0.743), p-tau / Aβ42 (AUC = 0.736), t-tau / Aβ42 (AUC = 0.742), or the combination of CSF Aβ42, p-tau181, and t-tau (AUC = 0.690).

[0141] Furthermore, to test the universality of the obtained protein panel, we used a protein panel consisting of four proteins (including YWHAG, SMOC1, PIGR and TMOD2) that distinguish CNA–T– from biologically defined AD to discriminate CN from clinically diagnosed AD dementia, and obtained an AUC of 0.831 [0.781-0.878].

[0142] When a protein panel consisting of five proteins (including ACHE, YWHAG, PCSK1, MMP10, and IRF1) was used to distinguish CNA–T– from biologically defined AD, an AUC of 0.934 [0.904-0.960] was obtained when differentiating CN from clinically diagnosed AD dementia.

[0143] Table 5 ROC analysis results of protein biomarkers alone or in combination with other variables to distinguish different clinical stages of AD

[0144]

[0145]

[0146] Note: The 5-protein panel refers to the combination of 5 proteins (including ACHE, YWHAG, PCSK1, MMP10 and IRF1) that distinguish CN from clinically diagnosed AD dementia.

[0147] Example 4 Autopsy neuropathology verifies the diagnostic efficacy of AD diagnostic protein biomarkers for AD

[0148] The union of four proteins used to distinguish CNA–T– from biologically defined AD (including YWHAG, SMOC1, PIGR, and TMOD2) and five proteins used to distinguish CN from clinically diagnosed AD dementia (including ACHE, YWHAG, PCSK1, MMP10, and IRF1) resulted in eight proteins, namely YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10, and IRF1. The expression of these eight proteins in subjects undergoing ADNI postmortem neuropathology was studied.

[0149] The results showed that after adjusting for age, sex, education level and APOEε4 carrier status, the CSF YWHAG (β=0.87, P=0.006), SMOC1 (β=0.75, P=0.019) and TMOD2 (β=0.95, P=0.003) levels in autopsy-confirmed AD patients were upregulated compared with those in autopsy-confirmed non-AD population, while the levels of CSF PIGR, ACHE, PCSK1, MMP10 and IRF1 were not significantly different between the autopsy-confirmed AD and non-AD groups.

[0150] YWHAG (AUC = 0.834 [0.667-0.963]), SMOC1 (AUC = 0.839 [0.707-0.943]), and TMOD2 (AUC = 0.795 [0.650-0.905]) showed good performance in distinguishing autopsy-confirmed AD from non-AD subjects, which was comparable to the performance of CSF Aβ42 (AUC = 0.757), p-tau181 (AUC = 0.872), t-tau (AUC = 0.843), and the combination of these three (AUC = 0.852), as well as the performance of a panel composed of four proteins (AUC = 0.884) and a panel composed of five proteins (AUC = 0.842) (all P < 0.05). DeLongtest >0.05); while the performance of CSF PIGR, ACHE, PCSK1, MMP10 and IRF1 was relatively poor (Table 6).

[0151] Table 6 ROC analysis results for distinguishing autopsy-confirmed AD from non-AD

[0152]

[0153] Note: The 4-protein panel refers to the combination of 4 proteins (including YWHG, SMOC1, PIGR and TMOD2) that distinguish CNA–T– from biologically defined AD; the 5-protein panel refers to the combination of 5 proteins (including ACHE, YWHAG, PCSK1, MMP10 and IRF1) that distinguish CN from clinically diagnosed AD dementia.

[0154] Example 5 External independent data set PPMI verifies the diagnostic efficacy of AD diagnostic protein biomarkers for AD

[0155] The performance of the eight screened proteins (including YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1) in the external independent dataset PPMI cohort was studied.

[0156] The results showed that after adjusting for age, sex, education level, and APOEε4 carrier status, CSF YWHAG (β = 1.31, P = 9.71 × 10 -5 ), SMOC1 (β=1.83, P=4.29×10 -8 ) and TMOD2 (β = 1.52, P = 2.64 × 10 -7 ) levels were upregulated, while the levels of CSF PIGR, ACHE, PCSK1, MMP10, and IRF1 were not significantly different between the CNA–T– group and the preclinical AD group ( Figure 6 a). YWHAG (AUC = 0.934 [0.835-1.000]), SMOC1 (AUC = 0.997 [0.985-1.000]), TMOD2 (AUC = 0.974 [0.938-0.997]), the panel composed of four proteins (AUC = 0.988 [0.958-1.000]) and the panel composed of five proteins (AUC = 0.913 [0.795-1.000]) all showed high accuracy in identification ( Figure 6 b), indicating that they have certain potential in early diagnosis of AD. Example 6 External independent data set CABLE verifies the diagnostic efficacy of AD diagnostic protein biomarkers for AD

[0157] The performance of the eight screened proteins (including YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1) in the external independent dataset CABLE cohort was studied.

[0158] The results showed that after adjusting for age, sex, education level, and APOEε4 carrier status, CSFYWHAG levels were significantly higher in the preclinical AD group (β = 0.76, P = 3.31 × 10 -5 ) and the biologically defined AD group (β = 0.80, P = 2.53 × 10 -6 ) were significantly upregulated. YWHAG (AUC = 0.754 [0.669-0.839]), the four-protein panel (AUC = 0.8051 [0.7222-0.8879]), and the five-protein panel (AUC = 0.7652 [0.6798-0.8506]) not only had high accuracy in distinguishing CNA-T- from biologically defined AD, but also YWHAG (AUC = 0.734 [0.641-0.827]), the four-protein panel (AUC = 0.7964 [0.7062-0.8865]), and the five-protein panel (AUC = 0.7521 [0.6591-0.8451]) had high accuracy in distinguishing CNA-T- from preclinical AD (Table 7). These results all indicate their potential in AD diagnosis.

[0159] Table 7 ROC analysis results of the external independent dataset CABLE cohort

[0160]

[0161]

[0162] Note: The 4-protein panel refers to the combination of 4 proteins (including YWHG, SMOC1, PIGR and TMOD2) that distinguish CNA–T– from biologically defined AD; the 5-protein panel refers to the combination of 5 proteins (including ACHE, YWHAG, PCSK1, MMP10 and IRF1) that distinguish CN from clinically diagnosed AD dementia.

Claims

1. A protein biomarker for diagnosing Alzheimer's disease (AD), wherein the protein biomarker is selected from one or more of YWHAG, SMOC1, PIGR, TMOD2, ACHE, PCSK1, MMP10 and IRF1.

2. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The protein biomarker is selected from a combination of YWHAG, ACHE, PCSK1, MMP10 and IRF1.

3. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The protein biomarker is selected from a combination of YWHAG, SMOC1, PIGR and TMOD2.

4. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The protein biomarker is selected from YWHAG or SMOC1.

5. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The protein biomarkers may be derived from blood or cerebrospinal fluid.

6. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The protein biomarkers that can diagnose Alzheimer's disease (AD) can also be combined with demographic characteristics for diagnosis.

7. The protein biomarker for diagnosing Alzheimer's disease (AD) according to claim 1, characterized in that: The demographic characteristics may include age, gender, years of education and / or APOEε4 allele.

8. Use of the protein biomarker according to claim 1 in preparing a preparation for diagnosing Alzheimer's disease (AD).

9. A kit for diagnosing Alzheimer's disease (AD), the kit comprising a reagent for detecting a gene, mRNA, cDNA or protein of a protein biomarker, instructions, etc., wherein the protein biomarker is the protein biomarker according to claim 1.

Images