A blood metabolic marker set for diagnosing Alzheimer's disease and its screening method and application

Abstract

The present invention discloses a blood metabolic marker set for diagnosing Alzheimer's disease and its screening method and application, belonging to the field of molecular detection technology. The present invention screens and verifies a set of specific and efficient AD blood metabolite sets, including 22 species, which makes up for the problems of low specificity and poor diagnostic accuracy of a single biomarker. The invention provides a set of metabolites with AD specificity, which has a good effect on the classification prediction of AD. At the same time, a metabolite feature selection method based on machine learning is proposed, and the identified set of metabolites is applied to the metabolic pathway level to explain the metabolic disorder of AD. It can provide a metabolic perspective explanation for the pathological mechanism of AD, and can conduct large-scale early screening for high-risk populations, which has clinical guidance significance.

Description

Technical Field

[0001] The present invention belongs to the technical field of molecular detection, and specifically relates to a blood metabolic marker set for diagnosing Alzheimer's disease and a screening method and application thereof. Background Art

[0002] Alzheimer's disease (AD) is a progressive neurodegenerative disease that causes patients to show symptoms of dementia such as memory impairment, cognitive decline, and loss of social activity ability, and eventually leads to death, making it one of the major challenges of public health and socioeconomic burden. At present, about 50 million people in the world suffer from dementia. Alzheimer's disease is the most common type of dementia. With the aging of the population, the prevalence of AD has increased year by year, and it has now become a health issue of great concern to the public. Due to the lack of treatments to prevent or effectively control the disease, there is an urgent need to understand the pathological mechanism of AD and diagnose and intervene at an early stage. Since the neuropathological changes that lead to the development of AD occur decades before the clinical manifestation of AD, and since there is currently no known treatment that can reverse the progression of AD or control the development of AD, early diagnosis is particularly important. Based on the core pathological hallmark of abnormal accumulation of β-amyloid protein (Aβ) and tau protein in the brain, the Aβ amyloid protein hierarchy hypothesis was proposed and is currently the most important theory. However, treatments aimed at clearing Aβ have failed to improve patients' cognitive function, and the failure of clinical trials has pointed out the importance of other pathological pathways, such as oxidative stress, inflammation, mitochondrial dysfunction, and abnormal metabolic pathways. Therefore, it is necessary to identify new biological pathological pathways to further improve our understanding of the pathogenesis of Alzheimer's disease and identify potential biomarkers for early diagnosis, disease detection, and prevention. The development and identification of some marker proteins (amyloid plaques and neurofibrillary tangles) provide opportunities for the use of biomarkers to study AD for early diagnosis. For example, positron emission tomography (PET) and cerebrospinal fluid (CSF)-based analysis can reveal abnormal levels of brain amyloid beta (Aβ) and phosphorylated tau protein (p-tau). The accumulation of these abnormal levels can reflect changes in the progression of AD, and these biomarkers can help with the diagnosis and prognosis of high-risk populations. However, the limited availability, high cost of acquisition, and invasiveness of these markers have hindered their routine use in clinical settings, so more easily accessible diagnostic methods, such as blood-based biomarkers, are essential for the identification and diagnosis of high-risk populations and early screening and risk assessment of large populations.

[0003] Metabolomics is a deep molecular phenotyping method that provides accurate and comprehensive molecular analysis of diagnostic phenotypes, providing a promising tool for understanding the pathogenesis pathways of AD and identifying new biomarkers. Due to the advancement of high-resolution mass spectrometry platforms, the high sensitivity and specificity of high-resolution mass spectrometry platforms make it possible to measure the metabolic status of individuals, which can strengthen research at the human population level, especially for early screening and risk assessment of large populations. Metabolomics has unique advantages. By detecting circulating metabolites in the blood, which are the final products of cell regulation and complex biochemical processes in life activities, the concentration in the blood can reflect the interaction between genetics and the environment and the metabolic changes of human tissues and organs, making it an indicator of the disease process and reflecting the disease phenotype. Recent metabolomics studies have shown the association between a large number of blood metabolites and AD risk. However, these studies are limited by small samples or small numbers of metabolic coverage, which are not enough to interpret the significant relationship between the metabolome and the disease. At the same time, there is still no consensus, and some identified metabolites cannot be well externally independently verified, and even inconsistent results appear. Blood is an easily accessible human body fluid suitable for large-scale studies at the human population level. The development of blood-based metabolomics biomarkers has provided research insights into the pathological mechanisms of AD at the metabolic level. However, due to inconsistencies in cohorts and metabolic platforms, these blood AD markers often have low reproducibility and focus on biological pathological mechanisms but fail to provide effective clinical practical significance. Although these studies have emphasized the potential of blood metabolomics in clarifying the pathogenic mechanisms of AD and identifying new biomarkers of AD, they still show poor results in early diagnosis. Although blood metabolomics has advantages such as low cost and non-invasiveness, the above limitations still prevent blood metabolomics from being widely used in clinical practice, especially in early screening and early diagnosis of the population. Summary of the invention

[0004] The technical purpose of the present invention is to provide a blood metabolic marker set for diagnosing Alzheimer's disease, which has excellent effects on the diagnosis and prediction of AD.

[0005] Another object of the present invention is to provide a method for screening a set of blood metabolic markers for diagnosing Alzheimer's disease, which can be used for the feature selection of AD-specific metabolites.

[0006] The application of a blood metabolic marker set for diagnosing Alzheimer's disease of the present invention in identifying the metabolic pathway level quantitatively verifies the metabolic disorder of AD.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A blood metabolic marker set for diagnosing Alzheimer's disease includes the following 22 types, as shown in Table 1 below:

[0009] Table 1

[0010]

[0011]

[0012] A method for screening a blood metabolic marker set for diagnosing Alzheimer's disease, for a candidate feature set, by constructing a classification model with AUC as an evaluation index and a forward stepwise feature selection method, a set of optimal feature sets is screened out.

[0013] Furthermore, the classification model selects a logistic regression model.

[0014] Furthermore, the forward stepwise feature selection method is as follows: before each cycle, the current optimal feature subset is fixed, and features are added step by step until the optimal AUC is reached. When the cycle ends, the selected features and the output AUC values ​​are the optimal feature set and the optimal AUC.

[0015] The screening method comprises the following steps:

[0016] Step 1: Select the logistic regression model as the classifier, build a classification model to classify and predict AD patients and normal people, and use AUC as the evaluation index to evaluate the characteristics of the input model;

[0017] Step 2: For the candidate feature set, that is, the full set of features, input the feature into the classification model by gradually adding features to obtain the AUC value corresponding to this set of features;

[0018] Step 3: Fix the corresponding feature of the current maximum AUC, gradually add a new feature, input it into the classification model, and obtain the AUC value under the new feature set until the AUC value reaches the maximum;

[0019] Step 4: Output the feature set corresponding to the maximum AUC value, which is the optimal feature set.

[0020] Application of a blood metabolic marker panel for Alzheimer's disease diagnosis in identifying metabolic pathway levels.

[0021] Beneficial effects of the present invention:

[0022] The present invention provides a set of AD-specific metabolites, which have a good effect on the classification and prediction of AD. At the same time, a metabolite feature selection method based on machine learning is proposed, and the identified set of metabolites is applied to the metabolic pathway level to explain the metabolic disorder of AD. It can provide a metabolic perspective explanation for the pathological mechanism of AD and can conduct large-scale early screening for high-risk populations, which has clinical guiding significance.

[0023] The AD-specific markers proposed in the present invention based on blood metabolic characteristics and the construction of a logistic regression classification model can efficiently diagnose AD patients. At the same time, the disorder in the course of AD is verified by pathway quantification. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further described below in conjunction with the accompanying drawings.

[0025] Figure 1 It is a schematic diagram of the design and analysis process in the present invention;

[0026] Figure 2 It is a schematic diagram of the feature selection method in the present invention;

[0027] Figure 3 It is a schematic diagram of the change of AUC during the feature selection process of the present invention;

[0028] Figure 4 It is a schematic diagram of model construction and generalization ability verification in the present invention;

[0029] Figure 5 is the classification accuracy classification ROC graph of the validation cohort in the present invention;

[0030] Figure 6 is a metabolic pathway enrichment map in the present invention;

[0031] Figure 7 It is a box diagram of quantitative differences at the pathway level in the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The standard substances and reagents used in the examples of the present invention are shown in Table 2:

[0034] Table 2

[0035]

[0036] 1. The sample extraction process in the embodiment of the present invention is as follows:

[0037] (1) A method for extracting a hydrophilic substance, comprising the following steps:

[0038] 1) Take out the plasma sample from the -80℃ freezer and thaw it on ice until there is no ice in the sample (all subsequent operations must be performed on ice);

[0039] 2) After the sample is thawed, vortex for 10 seconds to mix, and transfer 50 μL of the sample to the corresponding numbered centrifuge tube;

[0040] 3) Add 300 μL of 20% acetonitrile methanol internal standard extract, vortex for 3 min, and centrifuge at 12000 r / min for 10 min at 4°C;

[0041] 4) After centrifugation, transfer 200 μL of the supernatant to another corresponding numbered centrifuge tube and place it in a -20°C refrigerator for 30 min;

[0042] 5) Centrifuge again at 12000 r / min for 3 min at 4°C, and transfer 180 μL of the supernatant into the liner tube of the corresponding injection bottle for analysis.

[0043] (2) A method for extracting a hydrophobic substance, comprising the following steps:

[0044] 1) Take out the plasma sample from the -80℃ freezer and thaw it on ice until there is no ice in the sample;

[0045] 2) After the sample is thawed, vortex for 10 seconds to mix, and transfer 50 μL of the sample to the corresponding numbered centrifuge tube;

[0046] 3) Add 1 mL of internal standard lipid extract (methyl tert-butyl ether: methanol = 3:1, V / V) and vortex for 15 min;

[0047] 4) Add 200 μL of water, vortex for 1 min, and centrifuge at 12000 rpm for 10 min at 4°C;

[0048] 5) After centrifugation, transfer 200 μL of the supernatant to the corresponding numbered centrifuge tube and concentrate to complete dryness;

[0049] 6) Add 200 μL of mobile phase B, vortex for 3 min, centrifuge at 12000 r / min for 3 min, and remove the supernatant for LC-MS / MS analysis.

[0050] 2. The non-targeted detection acquisition conditions in the embodiment of the present invention are as follows:

[0051] The data acquisition instrument system mainly includes ultra-high performance liquid chromatography (UPLC) (ExionLCAD, https: / / sciex.com.cn / ) and quadrupole-time of flight mass spectrometry (TripleTOF 6600, AB SCIEX).

[0052] Liquid chromatography conditions mainly include:

[0053] Chromatographic column: ACQUITY HSS T3 (2.1×100mm, 1.8um)

[0054] Mobile phase A: 0.1% formic acid / water;

[0055] Mobile phase B: 0.1% formic acid / acetonitrile;

[0056] Column temperature: 40°C;

[0057] Flow rate: 0.35 mL / min;

[0058] Injection volume: 5uL.

[0059] The positive and negative ion mobile phase gradient conditions are shown in Table 3:

[0060] Table 3

[0061]

[0062] The mass spectrometry conditions are shown in Table 4 below:

[0063] Table 4

[0064]

[0065] III. The extensive targeted detection acquisition conditions in the embodiments of the present invention are as follows:

[0066] The data acquisition instrument system mainly includes ultra-high performance liquid chromatography (UPLC) (ExionLC AD, https: / / sciex.com.cn / ) and tandem mass spectrometry (MS / MS) ( https: / / sciex.com.cn / ).

[0067] (1) The liquid phase conditions of the hydrophilic substance are as follows:

[0068] 1) Chromatographic column: Waters ACQUITYUPLC HSS T3 C18 1.8 μm, 2.1 mm*100 mm;

[0069] 2) Mobile phase: Phase A is ultrapure water (0.1% formic acid), phase B is acetonitrile (0.1% formic acid);

[0070] 3) Elution gradient: water / acetonitrile (95:5 V / V) at 0 min, 10:90 V / V at 11.0 min, 10:90 V / V at 12.0 min, 95:5 V / V at 12.1 min, and 95:5 V / V at 14.0 min;

[0071] 4) Flow rate: 0.4 mL / min; column temperature: 40°C; injection volume: 2 μL.

[0072] (2) Liquid phase conditions for hydrophobic substances:

[0073] 1) Chromatographic column: Thermo Accucore TM C30 column, id2.1x100mm, 2.6um;

[0074] 2) Mobile phase: Phase A: acetonitrile / water (60 / 40, V / V) (containing 0.1% formic acid, 10 mmol / L ammonium formate); Phase B: acetonitrile / isopropanol (10 / 90, V / V) (containing 0.1% formic acid, 10 mmol / L ammonium formate);

[0075] 3) Mobile phase gradient: 0 min: A / B (80:20, V / V), 2 min: (70:30, V / V), 4 min: (40:60, V / V), 9 min: (15:85, V / V), 14 min: (10:90, V / V), 15.5 min: (5:95, V / V), 17.3 min: (5:95, V / V), 17.5 min: (80:20, V / V), 20 min: (80:20, V / V);

[0076] 4) Flow rate: 0.35 mL / min; column temperature: 45°C; injection volume: 2 μL.

[0077] (3) Mass spectrometry conditions for hydrophilic substances are as follows:

[0078] The electrospray ionization (ESI) temperature was 500°C, the mass spectrometer voltage was 5500V (positive), -4500V (negative), the ion source gas I (GS I) was 55psi, the ion source gas II (GS II) was 60psi, the curtain gas (CUR) was 25psi, and the collision-activated dissociation (CAD) parameter was set to high. In the triple quadrupole (Qtrap), each ion pair was scanned and detected according to the optimized declustering potential (DP) and collision energy (CE).

[0079] (4) The mass spectrometry conditions for hydrophobic substances are as follows:

[0080] The temperature of the electrospray ionization source (ESI) was 500°C, the mass spectrometer voltage was 5500 V in positive ion mode, -4500 V in negative ion mode, 45 psi for ion source gas 1 (GS1), 55 psi for ion source gas 2 (GS2), 35 psi for curtain gas (CUR), and the collision-activated dissociation (CAD) parameter was set to Medium. In the triple quadrupole, each ion pair was scanned and detected according to the optimized declustering potential (DP) and collision energy (CE).

[0081] IV. Clinical Sample Verification:

[0082] Two batches of elderly people with Alzheimer's disease were recruited in Shanghai. A total of 477 samples were selected and divided into three groups for metabolic research. The classification composition is shown in Table 5. The two batches are discovery cohort and replication cohort. Using the above-mentioned experimental reagents and experimental methods, the full spectrum metabolome PLUS test was performed by non-targeted combined with broad targeted technology. 1191 common metabolites were detected in the two batches.

[0083] Table 5

[0084]

[0085] In Table 5, n is the number of volunteers.

[0086] MCI (Mild Cognitive Impairment), AD (Alzheimer's Disease), the samples of the normal group are derived from the blood of NC (Normal condition) people.

[0087] Example 1

[0088] A method for screening a set of blood metabolic markers for diagnosis of Alzheimer's disease, such as Figure 1 The experimental design and analysis process of the study shown in the figure, in the discovery cohort, for the AD group and the NC group, a partial least squares discriminant (PLS-DA) model was constructed and statistical hypothesis testing was performed, and multiple hypothesis testing FDR correction was performed, and metabolite features with VIP>1 and FDR-corrected p value<0.05 were selected as the candidate differential metabolites set (differential metabolites).

[0089] For the AD group and the NC group, a set of AD-specific metabolites was further screened out from the differential metabolite set by using a classification model with AUC as the evaluation index and a forward stepwise feature selection method. The schematic diagram of the feature selection method is shown in Figure 2 As shown in FIG. 1 , this method is the metabolite feature selection method proposed in the present invention. The AUC change process during the feature selection process is shown in FIG. Figure 3 When the number of selected metabolites is 22, the classification model achieves the best AUC. At the same time, this set of metabolites is the AD-specific metabolite feature set (metabolite panel) identified.

[0090] Based on this set of blood metabolite characteristics, a logistic regression model for AD-NC classification was constructed. Model construction and generalization ability verification are shown in the schematic diagram Figure 4 In the discovery cohort, 5-fold cross validation was used for training and grid search was used to select hyperparameters. The optimal hyperparameters were obtained through training and a logistic regression diagnostic model was constructed. The diagnostic model was used for the validation cohort. The classification accuracy of the validation cohort is shown in the ROC diagram. Figure 5 As shown in the figure, the AUC value on the validation cohort is 0.932. It can be seen that this set of features has excellent classification and generalization prediction effects for the diagnosis and classification of AD. The diagnostic model constructed based on this set of 22 metabolic features can have a good diagnostic effect in an independent validation cohort.

[0091] Application of a blood metabolic marker panel for Alzheimer's disease diagnosis in identifying metabolic pathway levels.

[0092] For this set of 22 metabolite markers, 10 significantly enriched metabolic pathways were obtained through enrichment analysis, such as Figure 6 Among them, the main pathways affected include glycerophospholipid metabolism, arachidonic acid metabolism, and linoleic acid metabolism. These findings were then validated in an independent external dataset, confirming the relevance and robustness of these metabolic pathways in Alzheimer's disease. The expression values ​​of these 10 significantly enriched pathways in each sample can be obtained through the pathway-level quantitative calculation method proposed in the present invention. Their distribution in two cohorts and a completely independent external cohort is shown in the following figure. Figure 7 As shown, through statistical hypothesis testing, its significance is marked in the figure. It can be seen that through this quantitative method, the metabolic disorder of AD can be reflected at a higher level, that is, the pathway metabolic level.

[0093] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0094] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A blood metabolic marker set for diagnosing Alzheimer's disease, characterized in that: Including the following 22 types, as shown in the following table: 。 2. The method for screening a blood metabolic marker set for diagnosing Alzheimer's disease according to claim 1, characterized in that: The steps include: For the candidate feature sets, a set of optimal feature sets is screened out by constructing a classification model with AUC as the evaluation indicator and using the forward stepwise feature selection method.

3. The method for screening a blood metabolic marker set for diagnosing Alzheimer's disease according to claim 2, characterized in that: The classification model selects a logistic regression model.

4. The method for screening a blood metabolic marker set for diagnosing Alzheimer's disease according to claim 2, characterized in that: The forward stepwise feature selection method is as follows: before each cycle, the current optimal feature subset is fixed, and features are added step by step until the optimal AUC is reached. When the cycle ends, the selected features and the output AUC values ​​are the optimal feature set and the optimal AUC.

5. The method for screening a blood metabolic marker set for diagnosing Alzheimer's disease according to claim 2, characterized in that: The steps include: Step 1: Select the logistic regression model as the classifier, build a classification model to classify and predict AD patients and normal people, and use AUC as the evaluation index to evaluate the characteristics of the input model; Step 2: For the candidate feature set, that is, the full set of features, input the feature into the classification model by gradually adding features to obtain the AUC value corresponding to this set of features; Step 3: Fix the corresponding feature of the current maximum AUC, gradually add a new feature, input it into the classification model, and obtain the AUC value under the new feature set until the AUC value reaches the maximum; Step 4: Output the feature set corresponding to the maximum AUC value, which is the optimal feature set.

Images