Mass spectrum-based plasma lipid detection method for early diagnosis of Alzheimer's disease
Through the combination of shotgun lipoomics and LC-MRM/MS technology and machine learning algorithms, the accuracy and classification accuracy of lipid molecule identification in early diagnosis of Alzheimer's disease were solved, and a biological value of diagnosis was discovered Markers that support early diagnosis and understanding of disease mechanisms.
Patent Information
- Application Number
- CN202411410562.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2025-01-07
AI Technical Summary
The prior art is difficult to accurately identify complex and diverse lipid molecules in the early diagnosis of Alzheimer's disease, and traditional lipomics analysis methods have limited ionization and fragmentation conditions, and cannot fully reveal the changing characteristics of lipids in brain tissue.
The expression spectrum of plasma lipids was determined and quantitatively analyzed by combining shotgun lipoomics and LC-MRM/MS technology, and LIPID MSPredict Tool and Skyline Software assists in data processing, and screens biomarkers through partial least squares discriminant analysis and support vector machine algorithm.
It improves the identification accuracy and classification accuracy of lipid molecules, can accurately classify healthy people and Alzheimer's patients, and discover biomarkers with diagnostic value, supporting the understanding of early diagnosis and disease mechanisms.
Smart Images
Figure CN119269677A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of biomarkers, and in particular relates to a plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry. Background Art
[0002] Lipids are not only components of cell membranes, but also important executors of cell functions that participate in regulating a variety of biological processes, such as cell proliferation and signal transduction, and play an important role in human health and brain function. The brain is rich in lipids, and the disruption of lipid balance is associated with neurodegenerative diseases such as Alzheimer's disease (AD).
[0003] Studies have found that lipid metabolism in the brain tissue of AD patients changes significantly, especially in phospholipids and cholesterol related to neuronal membrane composition and function. For example, lipid raft changes and lipid peroxidation may participate in the pathological process of AD by affecting cell membrane fluidity, signal transduction, energy metabolism and other pathways. Therefore, analyzing the connection between lipids and AD is crucial to revealing the metabolic problems of the disease.
[0004] Lipid structures are complex and diverse, usually composed of a backbone structure, fatty acyl chains, and head groups. Because of their combined nature, the number of carbon atoms, degree of unsaturation, fatty acid content, and changes in head groups all pose challenges to lipid identification. In actual detection, researchers usually only detect a precursor ion and the fragment ions of the corresponding backbone structure, ignoring the impact of its structural heterogeneity on identification accuracy. Therefore, establishing a method for accurately identifying lipid molecules is of great significance for fully exploring and understanding the functions of heterogeneous lipids.
[0005] In recent years, lipid analysis methods have made great progress, especially lipidomics based on mass spectrometry (MS), which can perform high-throughput analysis of lipids in biological samples, not only with high accuracy and sensitivity, but also can provide structural information.
[0006] Nowadays, the commonly used technology for lipidomics analysis is LC-ESI-MS / MS, which can distinguish isomers while reducing ionization suppression, but its limited ionization and fragmentation conditions cannot be applied to all lipids. "Shotgun lipidomics" is another complementary method. During the direct infusion of samples, the ion information corresponding to the lipid molecules of interest can be obtained from a constant solution concentration to reveal the structure and type of lipid molecules. In addition, in order to more accurately characterize and quantify the molecules in the sample, high-resolution / high-mass accuracy parallel reaction monitoring / multiple reaction monitoring (Multiple reaction monitoring, MRM) methods must be used for targeted quantitative analysis of lipids. Therefore, the use of LC-MRM / MS technology can systematically analyze the characteristics of lipid molecule changes in AD, which will help us improve our understanding of the pathological mechanisms of AD and support the development of blood-based biomarkers. Summary of the invention
[0007] The object of the present invention is to provide a plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry to solve the problems raised in the above background technology.
[0008] To achieve the above object, the present invention provides the following technical solutions:
[0009] The plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry comprises the following steps:
[0010] Step 1: Shotgun lipidomics to determine plasma lipid profiles:
[0011] The lipid extract was pushed into a triple quadrupole linear ion trap mass spectrometer at a flow rate of 8 μL / min using a syringe pump. Precursor ion and neutral loss scans were performed on the backbone structure of lipids, including fatty acids, glycerophospholipids, glycerides, sphingolipids, and sterols, in positive and negative electrospray ionization modes, respectively. The determination of lipid molecules was based on mass spectrometry data and manually annotated with the assistance of the LIPID MSPredict Tool.
[0012] Step 2: LC-MRM / MS for lipid quantitative analysis:
[0013] Based on the m / z of the lipid molecule obtained in step 1, the ion information of the backbone structure and the fatty acid chain structure, an LC-MRM / MS scanning method is established, that is, the lipid molecule name, the precursor ion m / z, the m / z of two characteristic fragment ions, the declustering voltage (CE), the injection voltage (EP), the collision energy (CE), the collision chamber emission potential (CXP), and the scanning time (Time) are set, and data are collected using a triple quadrupole liquid chromatography-mass spectrometer QTRAP 6500 system, and the same lipid molecule is determined by monitoring the co-elution of two characteristic ions corresponding to the backbone structure and the fatty acid chain structure at the same retention time;
[0014] Step 3: Data Analysis
[0015] Import the raw data into Skyline, perform peak calibration based on internal standards or known standards, and then derive the peak area of each fragment ion. The content of each lipid molecule can be expressed by the sum of the peak areas of the two fragment ions. To further eliminate the influence of individual differences or injection volume, the data are normalized by peak area to obtain the relative content of each lipid molecule, clarify the lipid change characteristics between the two groups, filter and preprocess the data of the relative content of each lipid molecule, and use Metaboanalyst 5.0 for multivariate statistical analysis.
[0016] In addition, the partial least squares discriminant analysis (PLS-DA) algorithm was used to classify and screen the features to obtain biomarkers;
[0017] Step 4: Machine Learning
[0018] Support vector machine (SVM) was used to classify and rank the lipid molecules, and the most discriminative lipid molecules were selected as biomarkers.
[0019] Preferably, in step 1, a triple quadrupole linear ion trap mass spectrometer system (QTRAP 4000, SCIEX) is used to scan the lipid skeleton structure.
[0020] Preferably, in step 2, a triple quadrupole linear ion trap mass spectrometer QTRAP 6500 system is used to perform MRM scanning.
[0021] Compared with the prior art, the present invention has the following beneficial effects:
[0022] 1. The LC-MRM / MS method, which simultaneously monitors the co-elution of two characteristic fragment ions corresponding to the precursor ions of lipid molecules at the same retention time, can increase the accuracy of lipid molecule identification.
[0023] 2. Based on the results of mass spectrometry quantitative data and multivariate statistical analysis, machine learning algorithms can be used to accurately classify healthy people and Alzheimer's patients and improve classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is a work flow chart of the present invention;
[0025] Figure 2 This is a validation diagram based on the LC-MRM / MS method of the present invention;
[0026] Figure 3 .It is a diagram of the significant changes and key characteristic molecules of the differential lipid molecules between the two groups of the present invention.
[0027] Figure 4 This is a diagram showing the model performance of the diagnostic markers of the present invention.
[0028] Figure 5 The figure is a graph showing the relative expression changes of 10 lipid molecules that are potential diagnostic markers of the present invention between the two groups. DETAILED DESCRIPTION
[0029] The following will be combined with the drawings in 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.
[0030] Embodiment 1:
[0031] The plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry comprises the following steps:
[0032] Step 1: Extract lipids
[0033] The lipid extraction method of the present invention can be found in the literature [Bligh EG, Dyer WJ. A rapid method of total lipid extraction and purification. Can J Biochem Physiol. 1959. 37(8): 911–917. DOI: 10.1139 / o59-099. PMID: 13671378.]. 100 μL of plasma sample was used for lipid extraction.
[0034] Step 2: Shotgun lipidomics determines the plasma lipid expression profile.
[0035] The sample was pushed with a syringe pump at a stable flow rate of 8 μL / min to ensure uniform injection. Precursor ion and neutral loss scans were performed in positive and negative electrospray ionization modes, covering key lipid categories such as fatty acids, glycerophospholipids, glycerides, sphingolipids, and sterols. Based on a triple quadrupole linear ion trap mass spectrometer (QTRAP 4000 system), shotgun lipidomics technology was used to collect data for the pool samples of the healthy control group and the pool samples of AD patient plasma, respectively, to obtain characteristic ion pair information of five lipids (fatty acids, glycerophospholipids, glycerides, sphingolipids, and sterols). The determination of lipid molecules was based on mass spectrometry data and manually annotated with the assistance of the LIPID MS Predict Tool.
[0036] The mass spectrometry data were preliminarily screened to remove low-quality or abnormal data points (e.g., intensity > 5000, signal-to-noise ratio > 5). The identification of each lipid molecule was verified by at least two independent experiments.
[0037] Step 3: LC-MRM / MS for lipid quantitative analysis
[0038] Based on the m / z of the lipid molecule obtained in step 2, the ion information of the backbone structure and the fatty acid chain structure, the LC-MRM / MS scanning method was established, that is, the lipid molecule name, the precursor ion m / z, the m / z of the two characteristic fragment ions, the declustering voltage (CE), the injection voltage (EP), the collision energy (CE), the collision chamber emission potential (CXP), and the scanning time (Time) were set, and the triple quadrupole liquid chromatography-mass spectrometer QTRAP 6500 system was used for data acquisition. The same lipid molecule ( Figure 2 ). The collected data were quality checked (intensity greater than 2%, signal-to-noise ratio>5) to ensure the reliability and consistency of the data, and abnormal data were reviewed or eliminated.
[0039] Step 4: Data Analysis
[0040] The original mass spectrometry data was opened with Skyline software. To eliminate the influence of instrument stability on identification, internal standards or known standards were used for peak calibration to ensure the accuracy of the data, and the peak areas corresponding to the fragment ions were derived. The sum of the peak areas of the two fragment ions represented the content of the corresponding lipid molecules. To further eliminate the influence of individual differences or injection volume, the data were normalized by peak area to obtain the relative content of each lipid molecule.
[0041] In order to clarify the characteristics of lipid changes between the two groups, the data of the relative content of each lipid molecule in step 4 were filtered and preprocessed (the data with more than 50% sample content of null values in each group were filtered), and bioinformatics analysis was performed using Metaboanalyst 5.0 to obtain differential lipid molecule changes, that is, fold change >1.5 or <0.667, P <0.05, representing that the lipids in AD samples were significantly up-regulated or down-regulated compared with healthy controls, respectively).
[0042] Partial least squares discriminant analysis (PLS-DA) was used to screen the lipid molecules obtained in step 4 again according to the variable importance index (VIP>1) to obtain potential lipid markers.
[0043] Step 5: Machine learning algorithm to screen potential lipid markers
[0044] The characteristic lipid molecules screened out in step 4 are used to mine potential lipid markers through support vector machine (SVM), and its accuracy in classifying diseases is verified.
[0045] Support vector machine (SVM) is a powerful classifier that distinguishes different categories by finding the best hyperplane in feature space. The advantages of SVM are that it performs well in high-dimensional space, can handle nonlinear problems, and has good generalization ability for small sample data. In the screening of lipid biomarkers, SVM can effectively identify lipid markers that distinguish different categories. Comprehensive lipidomics analysis and machine learning algorithms are conducive to finding biomarkers for disease prediction, diagnosis and prevention, and ultimately discovering new therapeutic targets and strategies.
[0046] Use support vector machine (SVM) to establish a classification model to classify lipid molecules and rank their importance. Optimize model parameters to improve the accuracy and generalization ability of the model. Biomarker selection: Select the most discriminative lipid molecules from the model as biomarkers, and verify and confirm the selected biomarkers to ensure that they have biological significance and clinical application value.
[0047] The model was cross-validated and validated with an independent test set to evaluate the performance and stability of the model.
[0048] Optimize and adjust the model based on the verification results to improve the accuracy and reliability of the model.
[0049] Compared with traditional biochemical detection methods, lipidomics analysis based on mass spectrometry technology has the advantages of high throughput, high sensitivity, and high accuracy. Combined with machine learning algorithms, it can effectively discover biomarkers in diseases, and is expected to provide a new perspective for the early diagnosis, disease stratification, and treatment of AD.
[0050] Based on the above, the present invention utilizes the LC-MRM / MS system to monitor the co-elution of two characteristic fragments corresponding to the same lipid precursor ion at the same retention time to accurately identify lipid molecules, and combines it with a machine learning algorithm to screen potential lipid markers, providing a basis for the early diagnosis and treatment of AD.
[0051] Embodiment 2:
[0052] See also Figure 1 - Figure 5 As shown,
[0053] The lipid extraction method of the present invention can be found in the literature [Bligh EG, Dyer WJ. A rapid method of total lipid extraction and purification. Can J Biochem Physiol. 1959. 37(8): 911–917. DOI: 10.1139 / o59-099. PMID: 13671378.]. 100 μL of plasma sample was used for lipid extraction.
[0054] Plasma samples were systematically scanned using a shotgun lipidomics approach.
[0055] The lipid samples of the control group and AD group formed by pooling 5 samples in each group were pushed into the triple quadrupole linear ion trap mass spectrometer system (QTRAP 4000, SCIEX) at a flow rate of 8 μL / min by a syringe pump. In the positive and negative electrospray ionization modes, the precursor ion (Precursor Ion) and neutral loss (Neutral Loss) scans were performed for the backbone structures of five lipids (fatty acids, glycerophospholipids, glycerides, sphingolipids, and sterols) to obtain shotgun non-targeted lipidomics data. The instrument parameters were set as follows: the electrospray ion source was ESI, the ion spray voltage was 5500 eV, and the mass scan range was 400 to 1000 m / z.
[0056] Identification of lipid molecules
[0057] In the experiment, shotgun mass spectrometry data were collected and peak extracted using Analyst software. Lipid molecules were identified based on mass spectrometry data and manually annotated with the assistance of LIPID MS Predict Tool. When extracting mass spectrometry peak information, we set a threshold>5000 for annotation of lipid molecules. In the end, a total of 292 lipid molecules were identified, including 29 fatty acids, 180 glycerophospholipids, 32 glycerides, 41 sphingolipids, and 10 sterols, and their corresponding ion pair information was obtained.
[0058] Establishment of an accurate lipid identification method using LC-MRM / MS
[0059] The plasma lipids were identified and quantified using the LC-MRM / MS method in positive and negative ion modes using the QTRAP 6500 system (SCIEX). The sample injection volume was 10 μL, and the chromatographic separation was performed using Column (4.6×250mm, 5μm, Welch Materials). The instrument parameters were set as follows: mobile phase solution A was isopropanol: hexane: 100 mM ammonium acetate (v / v / v 58:40:2), and solution B was isopropanol: hexane: 100 mM ammonium acetate (v / v / v 50:40:10). The sample was eluted with a solution gradient at 600 nL / min: 50% mobile phase B from 0-5 minutes, the mobile phase linearly increased to 100% mobile phase B from 6-10 minutes and maintained for 5 minutes, and gradually returned to 50% mobile phase B from 16-30 minutes until the next analysis.
[0060] The data were processed using Skyline software (SCIEX), and only peaks with intensities greater than 2% and signal-to-noise ratios greater than 5 were identified. Two characteristic daughter ions co-eluted at the same retention time were used to determine lipid molecules. The sum of the peak areas of each pair of daughter ions represented the relative content of the corresponding substance and was used for further quantitative analysis of lipids.
[0061] The results showed that in the detection of multiple lipid molecules, the same lipid molecule was successfully identified by monitoring the co-elution of two daughter ions at the same retention time, and 270 lipids (such as Figure 2 shown).
[0062] Bioinformatics analysis
[0063] The mass spectrometry peak area data (representing the relative content of each lipid molecule) were normalized (peak area normalization) and bioinformatics analysis was performed using Metaboanalyst (https: / / www.metaboanalyst.ca / ).
[0064] After data were filtered (data with more than 50% sample content of null values in each group were filtered) and filled with blank values (KNN method-feature specificity), a total of 80 differential lipids (such as Figure 3 A).
[0065] In addition, the partial least squares discriminant analysis (PLS-DA) algorithm was used to classify and screen the features to obtain biomarkers. Here, the variable importance index (VIP value) was selected as the basis for screening biomarkers, and the data with a VIP value greater than 1 were used as candidate biomarkers (e.g. Figure 3 The remaining 72 differentially expressed lipids were selected as potential biomarkers for further screening.
[0066] Machine learning algorithm for mining potential lipid biomarkers of AD
[0067] The lipid marker panel was developed based on the support vector machine (SVM) model. The above 72 differential lipids were used for biomarker discovery, and the model was evaluated using the receiver operating characteristic (ROC) curve.
[0068] With ROC area > 0.7, 10 lipids were screened as potential biomarkers for AD diagnosis, including LPE 18:0 (AUC = 0.936), LPE 22:6 (AUC = 0.740), PS 38:4 (AUC = 0.91), PS (20:3 / 22:1) (AUC = 0.837), PG (18:1 / 18:1) (AUC = 0.823), PE (18:0 / 22:3) (AUC = 0.783), SM (d18:1 / 16:0) (AUC = 0.912), SM (d18:1 / 18:1) (AUC = 0.826), Cer (d22:0 / 14:1) (AUC = 0.824), and LacCer (d19:3 / 21:3) (AUC = 0.734). Further analysis showed that these 10 lipid molecules formed a potential diagnostic panel with good diagnostic accuracy for AD (AUC = 0.995, Figure 4 A, 4B). Compared with healthy controls, the changes of these 10 lipid molecules in AD patients are shown in Figure 5 .
[0069] The above experimental results show that lipids play an important role in maintaining body homeostasis, and their changes can be used for clinical auxiliary diagnosis of AD and show good diagnostic accuracy.
[0070] Therefore, the lipid identification method of the present invention can increase the accuracy of lipid identification, and combined with the machine learning algorithm, has a high application prospect in mining potential biomarkers, and can assist clinicians in early diagnosis and treatment.
[0071] 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 plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry, characterized in that: The following steps are involved: Step 1: Shotgun lipidomics to determine plasma lipid profiles: The lipid extract was pushed into the triple quadrupole linear ion trap mass spectrometer QTRAP4000 system at a flow rate of 8 μL / min using a syringe pump. Precursor ion or neutral loss scanning was performed on the backbone structure of lipids, including fatty acids, glycerophospholipids, glycerides, sphingolipids, and sterols, in positive and negative electrospray ionization modes. The determination of lipid molecules was based on mass spectrometry data and manually annotated with the assistance of the LIPID MS Predict Tool. Step 2: LC-MRM / MS for lipid quantitative analysis: Based on the m / z of the lipid molecule obtained in step 1, the ion information of the backbone structure and the fatty acid chain structure, a LC-MRM / MS scanning method is established, that is, the lipid molecule name, the precursor ion m / z, the m / z of two characteristic fragment ions, the declustering voltage (CE), the injection voltage (EP), the collision energy (CE), the collision chamber emission potential (CXP), and the scanning time (Time) are set, and a triple quadrupole liquid chromatography-mass spectrometer QTRAP 6500 system is used for data acquisition, and the same lipid molecule is determined by monitoring the co-elution of two characteristic fragment ions corresponding to the backbone structure and the fatty acid chain structure at the same retention time; Step 3: Data analysis: The raw data were imported into Skyline, and the peak areas of the fragment ions were derived after peak calibration based on the internal standard or known standard. The content of each lipid molecule can be expressed by the sum of the peak areas of the two fragment ions. To further eliminate the influence of individual differences or injection volume, the data were normalized by peak area to obtain the relative content of each lipid molecule, clarify the lipid change characteristics between the two groups, filter and preprocess the data of the relative content of each lipid molecule, and use Metaboanalyst5.0 for multivariate statistical analysis; in addition, the partial least squares discriminant analysis (PLS-DA) algorithm was used to classify and screen the features to obtain biomarkers; Step 4: Machine Learning: Support vector machine (SVM) was used to classify and rank the lipid molecules, and the most discriminative lipid molecules were selected as biomarkers.
2. The plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry according to claim 1, characterized in that: In the step 1, a triple quadrupole linear ion trap mass spectrometer system (QTRAP 4000, SCIEX) is used to scan the lipid skeleton structure.
3. The plasma lipid detection method for early diagnosis of Alzheimer's disease based on mass spectrometry according to claim 1, characterized in that: In the step 2, a triple quadrupole linear ion trap mass spectrometer QTRAP 6500 system is used to perform multiple reaction monitoring (MRM) based on two characteristic ion information.