A biomarker panel for early prediction of bronchopulmonary dysplasia in preterm infants and a method of detecting the same

Abstract

The application belongs to the technical field of biomedical detection, and particularly relates to a biomarker combination for early prediction of bronchopulmonary dysplasia of premature infants and a detection method thereof, the biomarker combination comprising: (1) lysophosphatidylcholine LPC 14:0; (2) phosphatidylcholine PC 32:1|16:0_16:1; (3) phosphatidylethanolamine PE 38:5|18:1_20:4; (4) triacylglycerol TG 58:7|TG 18:1_20:2_20:4; and (5) oxidized fatty acid FA 22:4;2O. The screened lipid biomarker combination is closely related to the core pathological mechanism of alveolar development and surfactant metabolism, can reflect the molecular pathological changes of BPD, and can be used for risk prediction within one week after birth of the premature infant, which is much earlier than traditional diagnosis methods, and provides a valuable time window for clinical intervention.

Description

Technical Field

[0001] This invention belongs to the field of biomedical detection technology, specifically relating to a combination of biomarkers for early prediction of bronchopulmonary dysplasia in premature infants and their detection methods, especially a combination of lipid biomarkers for early prediction of bronchopulmonary dysplasia in premature infants and their non-targeted lipidomics detection methods. Background Technology

[0002] Prematurity is one of the most pressing challenges in modern neonatal medicine, significantly impacting neonatal survival and long-term health outcomes. Infants born before 32 weeks of gestation are at higher risk of respiratory complications. Lipids play a fundamental role in neonatal development, particularly in lung function and energy metabolism; however, the lipidomic changes and molecular mechanisms in preterm neonates remain poorly understood.

[0003] Bronchopulmonary dysplasia (BPD) is one of the most common and serious chronic pulmonary complications in extremely preterm infants, with an incidence rate approaching 30% in infants born before 29 weeks of gestation. Its core pathological features include impaired alveolar development, abnormal pulmonary vascular development, and chronic inflammation. BPD is an early high-risk factor for adverse pulmonary developmental outcomes in preterm infants.

[0004] Currently, there are no effective treatments for BPD. Some children with severe BPD may also have pulmonary hypertension, and may even require tracheotomy or long-term respiratory oxygen therapy. Surviving children may still face the risk of recurrent respiratory infections, abnormal lung function, and long-term respiratory diseases during infancy and school age.

[0005] Currently, the diagnosis of BPD is mainly based on the need for continuous oxygen therapy and the duration of respiratory support after birth. This diagnostic method is essentially a delayed diagnosis of clinical phenotype, usually only becoming clear several weeks after birth or even at a corrected gestational age of 36 weeks, failing to achieve early prediction and precise intervention. Furthermore, differences exist between different versions of diagnostic criteria, leading to a lack of uniformity in clinical grading and prognostic assessment, affecting the comparability of research results and the development of intervention strategies. Summary of the Invention

[0006] This invention, based on a non-targeted lipidomics LC-MS / MS system, develops a combination of early biomarkers and their detection methods for lung developmental abnormalities in preterm infants. The developed non-targeted lipidomics research workflow is efficiently applied to clinical plasma samples from preterm infants for lipid profile analysis. This establishes a high-throughput, high-sensitivity, high-stability, low-error, and highly reproducible non-targeted lipidomics method for analyzing plasma samples from preterm infants, enabling early prediction of lung developmental abnormalities in preterm infants. This provides early auxiliary diagnosis of bronchopulmonary dysplasia (BPD) in preterm infants, facilitating intervention and treatment, and improving long-term prognosis. It also provides a basis for exploring the mechanisms, detection, and treatment of BPD in clinical practice.

[0007] The purpose of this invention is to provide a combination of lipid biomarkers for predicting the risk of bronchopulmonary dysplasia in preterm infants, in order to solve the technical problem of the lack of specific early molecular biomarkers in the prior art.

[0008] Another object of the present invention is to provide a kit for detecting the above-mentioned combination of lipid biomarkers.

[0009] Another object of the present invention is to provide a non-targeted lipidomics detection method based on liquid chromatography-mass spectrometry for the detection of the above-mentioned combination of lipid biomarkers.

[0010] Another object of the present invention is to provide the application of the above-mentioned lipid biomarker combination in the preparation of a risk prediction kit for bronchopulmonary dysplasia in preterm infants.

[0011] To achieve the above objectives, the present invention adopts the following technical solution: a lipid biomarker combination for predicting the risk of bronchopulmonary dysplasia in preterm infants, the biomarker combination comprising:

[0012] (1) Lysophosphatidylcholine (LPC) 14:0;

[0013] (2) Phosphatidylcholine PC 32:1 | 16:0_16:1;

[0014] (3) Phosphatidylethanolamine PE 38:5 | 18:1-20:4;

[0015] (4) Triglycerides TG 58:7 | TG 18:1_20:2_20:4;

[0016] (5) Oxidized fatty acids FA 22:4;2O.

[0017] LPC 14:0 was downregulated in high-risk samples of bronchopulmonary dysplasia, while PC 32:1 | 16:0_16:1, PE 38:5 | 18:1_20:4, TG 58:7 | TG 18:1_20:2_20:4 and FA 22:4;2O were upregulated in high-risk samples of bronchopulmonary dysplasia.

[0018] A kit for detecting the above-mentioned combination of lipid biomarkers, comprising:

[0019] (1) Standards of the lipid biomarker combination;

[0020] (2) Internal standard mixture, wherein the internal standard mixture comprises EquiSPLASH LipidoMIX, deuterated arachidonic acid and deuterated palmitoylcarnitine;

[0021] (3) Chromatographic columns, mobile phase reagents and mass spectrometry detection reagents used for liquid chromatography-mass spectrometry detection.

[0022] Preferably, the standard comprises a series of standard solutions with a concentration gradient of 10-1000 ng / mL.

[0023] Preferably, the concentration of EquiSPLASH LipidoMIX in the internal standard mixture is 5 μg / mL, and the concentrations of deuterated arachidonic acid and deuterated palmitoylcarnitine are both 500 ng / mL.

[0024] A non-targeted lipidomics detection method based on liquid chromatography-mass spectrometry (LC-MS) for the detection of the above-mentioned combination of lipid biomarkers includes the following steps:

[0025] (1) Sample pretreatment: Take plasma samples and add isopropanol solution containing internal standard mixture for protein precipitation. The volume ratio of plasma sample to isopropanol is 1:4. The internal standard mixture contains EquiSPLASH LipidoMIX, deuterated arachidonic acid (20:4 AA-d11) and deuterated palmitoyl carnitine (Palmitoyl-L-carnitine-d3), with a final concentration of 500 ng / mL. After vortex mixing, protein precipitation is carried out for 1 hour. Centrifuge at 13500g and 4°C for 10 minutes and take the supernatant.

[0026] (2) Chromatographic separation: Separation was performed using an ACQUITY UPLC CSH C18 reversed-phase column (100 mm × 2.1 mm, particle size 1.7 μm) and a pre-column (5 mm × 2.1 mm, particle size 1.7 μm). Mobile phase A consisted of 60% acetonitrile (v / v), 40% water (v / v), 0.1% formic acid (v / v), and 10 mM ammonium formate. Mobile phase B consisted of 9% acetonitrile (v / v), 90% isopropanol (v / v), 1% water (v / v), 0.1% formic acid (v / v), and 10 mM ammonium formate. A gradient elution program was used: 0.0 min 15% B, 4.0 min 30% B, 5 min 48% B, 22.00 min 82% B, 25 min 99% B, 26 min 99% B, 26.10 min 15% B, 30 min 15% B; flow rate 0.4 mL / min, column temperature 65°C, positive ion mode injection volume 3 μL, negative ion mode injection volume 6 μL;

[0027] (3) Mass spectrometry detection: A Q-Exactive mass spectrometer was used for detection in data independent acquisition (DIA) mode. 20 SWATH scanning windows were set with a window width of 26 m / z and an overlap of 1 m / z between adjacent windows, covering the mass range of 85-1250 m / z. The full scan accumulation time was 50 ms, the DIA window accumulation time was 30 ms, and the total cycle time was 650 ms. The stepped normalized collision energy (NCE) was 20-40 eV. The MS horizontal resolution was over 30000, and the MS / MS horizontal resolution was 17500. The spray voltage for positive ion mode was 3500V, the spray voltage for negative ion mode was 3200V, the S lens RF voltage was 55, the capillary tube temperature was 320℃, and the auxiliary gas heating temperature was 300℃.

[0028] (4) Data processing: MS-DIAL ver4.60 software was used for peak detection, alignment and lipid identification. Peak alignment was performed by retention time correction. Data normalization was performed by retention time correction based on internal standard (T_R-IS) combined with QC-LOWESS normalization method to obtain the quantitative detection results of the lipid biomarker combination.

[0029] Preferably, the data processing in step (4) uses peak height as the quantitative basis.

[0030] The application of the above-mentioned lipid biomarker combination in the preparation of a risk prediction kit for bronchopulmonary dysplasia in preterm infants.

[0031] The application of the above detection methods in lipidomics data analysis.

[0032] A quality control method for plasma samples, wherein the plasma samples are tested using the above-mentioned detection method, and the batch-to-batch repeatability is assessed by the relative standard deviation of the QC samples.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This invention provides a highly specific combination of lipid biomarkers: the five lipid molecule combinations screened in this invention are closely related to the pathological mechanism of BPD, can reflect early molecular changes such as abnormal pulmonary surfactant metabolism and oxidative stress, and can predict risk within one week after birth.

[0035] High-throughput and high-coverage detection was achieved: using the DIA-SWATH mode with 20 scanning windows, 1229 lipids could be detected in positive ion mode and 1090 lipids could be detected in negative ion mode, which greatly improved the number of lipids covered.

[0036] Improved detection stability and repeatability: By adopting a 4-fold isopropanol dilution scheme and a T_R-IS-based internal standard combined with QC-LOWESS dual normalization process, the relative standard deviation of QC samples was controlled within 12.5%.

[0037] It has the potential to be transformed into a clinical reagent kit: The reagent kit provided by this invention has a well-defined composition and standardized operation, and is suitable for large-scale clinical sample testing. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the 20 scanning windows of the DIA method used in this embodiment of the invention.

[0039] Figure 2 A schematic diagram showing the grouping and numbering of neonatal plasma samples.

[0040] Figure 3 This is a diagram showing the specific lipid composition of the EquiSPLASH LipidoMIX internal standard mixture.

[0041] Figure 4 This is a flowchart of the sample pretreatment process.

[0042] Figure 5 A comparison chart showing the number of lipids identified at different sample dilution factors.

[0043] Figure 6 This is a schematic diagram of the DIA NEW positive and negative ion mode detection window.

[0044] Figure 7 This is a comparison chart of the results of two DIA detection methods (DIA NEW and DIA OLD).

[0045] Figure 8A comparison of the number of ions identified at different collision energies (CE) in DIA positive mode.

[0046] Figure 9 A comparison of the number of ions identified under different collision energies (CE) in the DIA negative mode.

[0047] Figure 10 This is the first batch of lipid recognition maps in positive / negative ion modes.

[0048] Figure 11 To compile a heatmap of lipid metabolism differences from samples at three time points.

[0049] Figure 12 Box plot / statistic showing that LPC 14:0 was significantly downregulated in children with BPD.

[0050] Figure 13 Statistical graph showing the significant upregulation of four biomarkers (PC 32:1, PE 38:5, TG 58:7, FA 22:4;2O) in children with BPD.

[0051] Figure 14 To compare the AUC values ​​of single lipid and five-lipid combined models in predicting the occurrence of BPD;

[0052] Figure 15 To compare the overall performance of single lipid and five-lipid combined models in predicting the occurrence of BPD. Detailed Implementation

[0053] The present invention will be further described below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the description of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0054] Example

[0055] Establishment of a non-targeted lipidomics detection method based on liquid chromatography-mass spectrometry (LC-MS) and screening of biomarkers

[0056] 1. Research subjects and sample collection

[0057] This study included 178 newborns. Neonatal plasma samples were numbered and grouped for cohort construction. Newborn samples were divided into three time periods: within 1 week after birth, within 2-4 weeks after birth, and at a corrected gestational age of 36 weeks (see [link]). Figure 2For each child, 2 mL of EDTA-anticoagulated blood was collected, centrifuged at 3000g for 10 minutes to separate the plasma, and stored at -80℃ for later use. All plasma samples were to be divided into three batches for separate processing and loading, with each batch containing an equal amount of complete samples from the three time points.

[0058] 2. Pretreatment of plasma samples from preterm infants

[0059] This invention aims to compare dilution methods using chloroform-methanol versus isopropanol (IPA). Plasma samples were stored at −80°C and thawed on ice on the day of sample preparation. 20 μL aliquots were used for non-targeted lipid extraction via isopropanol-based protein precipitation. To normalize lipid molecules in clinical samples, 5 μg / ml of EquiSPLASH LipidoMIX (see composition details) was added before precipitation. Figure 3 The sample was mixed with fatty acids (arachidonic acid (20:4 AA-d11) and palmitoyl-L-carnitine (d3)) in the isopropanol fraction, resulting in a final IS internal standard mixture concentration of 500 ng / mL in IPA. After vortexing, the protein was precipitated for 1 hour and then centrifuged at 13500 g, 4 °C for 10 minutes. The supernatant was transferred to a sealed glass vial and stored in the dark. The sample was analyzed by liquid chromatography-Q-Exactive-MS / MS as soon as possible after preparation.

[0060] This study first selected four sample dilution factors (3 / 4 / 5 / 6 times diluted with IPA) to enhance the IS detection effect and improve the stability and repeatability of batch processing of a large number of clinical trial samples. The aim was to find the optimal sample dilution factor among them.

[0061] We compared the total number of lipids it identified (total lipid count is approximately positively correlated with IS signal intensity). After validation with mouse serum, the final lipid counts detected after processing four different sample dilution factors are shown below. Figure 5 .

[0062] Of the four dilution strategies employed, the 4-fold dilution yielded the highest number of lipids (1224) compared to a 3-fold dilution yielding 1139 lipids, a 5-fold dilution yielding 507 lipids, and a 6-fold dilution yielding 411 lipids in the same sample. Therefore, the 4-fold dilution strategy was ultimately selected for this study.

[0063] 3. Mass spectrometry detection method

[0064] 3.1 Detection Scanning Mode and Scanning Window

[0065] To address the challenge of detecting a large number of differentially expressed lipids in complex clinical samples, this embodiment employs DIA technology for a full-scan mode, followed by fragmentation and analysis across multiple windows throughout the entire sample. This allows for the collection of comprehensive lipid fragmentation information in a single run, improving the stability and reproducibility of clinical trial sample detection and significantly enhancing the overall coverage of identified lipids. The study aims to establish a standardized, high-throughput lipidomics research model applicable to clinical biological samples.

[0066] The DIA detection window was optimized, and two DIA methods were used for scanning. These two DIA detection methods include DIA NEW (with 9 detection scanning windows each for positive and negative ions). Figure 6 ), containing 20 SWATH (sequential window acquisition of all theoretical fragment ion mass spectra) windows, experimental DIA OLD method ( Figure 1 ).

[0067] The lipid counts displayed in MS-DIAL ver4.60 are shown below. Figure 7 This allows for a clear and intuitive observation of the more comprehensive detection advantages of the DIA OLD method. Furthermore, this SWATH method boasts 20 positive and negative ion scanning windows, resulting in a more comprehensive ion scan. Therefore, we optimized our mass spectrometry detection method to the DIA OLD method, which includes 20 detection scanning windows.

[0068] This DIA method can detect low-abundance precursor ions, thus improving the sensitivity and accuracy of detecting large batches of clinical trial samples.

[0069] ① Use the Q-Exactive instrument to set up the DIA method:

[0070] Drag the DIA module into the method flow from the template (if a first-level scan is added, drag the Full Mass module in front of the DIA module);

[0071] ② Modify the resolution and window width in the DIA module as needed (if the actual window step size is 25 m / z, then set it to 26 m / z, with the extra 1 m / z representing the overlap between windows);

[0072] ③ Generate the median value (m / z) for each window in the Inclusion List as the target. After completing these three steps, save the method and you can begin sample data acquisition.

[0073] 3.2 Optimize the pyrolysis voltage

[0074] Collision energy (CE) is a parameter in mass spectrometry that directly controls the energy of collisions between ions and neutral molecules. Specifically, it refers to the energy applied by the instrument when ions collide with neutral gas molecules in the collision chamber during collision-induced dissociation, and the different collision modes involved. Since the magnitude of CE theoretically affects the fragmentation of precursor ions and the generation of daughter ions, it may influence the abundance of detected ions. Its unit is typically eV (electron volt).

[0075] Adjusting the CE (Collision Energy) can optimize ion dissociation efficiency, affecting fragment ion generation and detection sensitivity. NCE (Normalized Collision Energy) is the normalized form of CE and is related to CE by the following formula:

[0076] NCE = CE / (m / z) × 100

[0077] NCE eliminates the influence of ion mass differences on collision energy, allowing ions of different masses to dissociate under the same energy conditions, improving data comparability. NCE is more commonly used for non-targeted analysis. Stepped normalized collision energy (Stepped NCE), on the other hand, is a technique that adjusts the collision energy in stages, sequentially acquiring fragment ion spectra through multiple levels of collision energy.

[0078] During mass spectrometry scanning, it was found that different collision energies (CE) affect the number of lipids detected. The study first compared the ion detection performance with four different CE values ​​and one NCE value in both positive and negative modes. It was found that NCE 20-40 eV could detect a greater and more precise number of ions, as shown below. Figure 8 , 9 .

[0079] Therefore, this invention initially selected a mass spectrometry method with the above-mentioned 20 detection windows and a collision energy of NCE 20-40 eV for analysis.

[0080] 4. Chromatographic detection methods

[0081] Lipid chromatographic separation was performed using an ACQUITY UPLC CSH C18 reversed-phase nonpolar column (100 mm × 2.1 mm; particle size 1.7 μm) and a pre-column (5 mm × 2.1 mm; particle size 1.7 μm). Mobile phase A consisted of 60% acetonitrile (ACN) (v / v), 40% water (H2O) (v / v), 0.1% formic acid (HCOOH) (v / v), and 10 mM ammonium formate. Mobile phase B consisted of 9% acetonitrile (ACN) (v / v), 90% isopropanol (IPA) (v / v), 1% water (H2O), 0.1% formic acid (HCOOH) (v / v), and 10 mM ammonium formate. Gradient elution (0.0 min, 15% B; 4.0 min, 30% B; 5 min, 48% B; 22.00 min, 82% B; 25 min, 99% B; 26 min, 99% B; 26.10 min, 15% B; 30 min, 15% B) was run at a flow rate of 0.4 mL / min and a constant oven temperature of 65 °C. The injection volume was set to 3 μL in positive ion mode and 6 μL in negative ion mode.

[0082] 5. Data Processing and Normalization Methods

[0083] Data processing for non-targeted lipidome screening was performed using MS-DIAL ver4.60. This software can process SWATH data, including peak detection, alignment, and lipid identification, relying on retention time (TR), mass-to-charge ratio (m / z) values, isotopic proportions, and MS / MS similarity based on the included computer simulation library. Raw area and raw height datasets were derived from the aligned peak results and normalized separately. LowESS normalization was performed using the LowESS normalization tool. mTIC-based normalization (based on the sum of the heights of identified metabolite peaks) was performed using MS-DIAL's built-in functionality.

[0084] The TR-IS method, based on internal standards (IS), is used to process and analyze volcano plots, box plots, and cluster plots of normalized joint datasets in R Studio 1.3.1073.0, R 4.4.1, and Excel (Office 2021).

[0085] Raw data from non-targeted lipidomics can be extracted in the form of peak area or peak height. Generally, using peak height offers greater stability against signal interference and has less impact on integration errors, thus improving precision in cases of low abundance features.

[0086] This correction method has been used to complete the first round of screening and fine identification, identifying a total of 959 lipids (the second and third rounds of screening are underway). Figure 10 It is the analysis result obtained from software processing.

[0087] 6. Data analysis results and screening of lipid biomarkers for early prediction of BPD

[0088] Using the aforementioned non-targeted lipidomics platform, a total of 1229 lipids were included in the data analysis under positive ion mode, including lipids detected only in a few samples. A total of 1090 lipids were included in the data analysis under negative ion mode, including lipids detected only in a few samples. There is some overlap between positive and negative ion modes. Lipid data are reported as relative concentrations in ng / mL. Concentrations were calculated using a one-to-one internal standard method.

[0089] Lipid molecules significantly associated with BPD were screened from plasma samples of preterm infants within 1 week of birth and 2-4 weeks after birth.

[0090] I. Screening methods for lipid biomarkers

[0091] In this invention, a combination of lipid biomarkers significantly associated with the occurrence of bronchopulmonary dysplasia (BPD) is obtained through a non-targeted lipidomics detection platform, based on a systematic analysis of plasma samples from premature infants and a multi-step statistical screening strategy.

[0092] (a) Data preprocessing

[0093] Raw data obtained from liquid chromatography-mass spectrometry (LC-MS / MS) were processed using MS-DIAL (ver 4.60) software, including peak extraction, retention time correction, peak alignment, and lipid identification. To improve data stability and comparability, the following normalization strategy was adopted:

[0094] 1. Retention time correction method based on internal standard (T) R -IS);

[0095] 2. Signal drift correction is performed by combining the LOWESS (Locally Weighted Scatterplot Smoothing) normalization method of QC samples.

[0096] Peak height was used as the quantitative basis for subsequent analysis.

[0097] Simultaneously, the data undergoes quality control screening, including:

[0098] • Remove lipid features that are missing in more than 50% of the samples;

[0099] • Remove lipid features with a relative standard deviation (RSD) greater than 30% in the QC sample.

[0100] The above processing yielded high-quality lipid quantification data for subsequent analysis.

[0101] (II) Criteria for Initial Screening of Differential Lipids

[0102] A comparative analysis of lipid expression levels between the BPD group and the non-BPD group was conducted, and the criteria for screening differentially expressed lipids were as follows:

[0103] 1. Fold Change (FC):

[0104] Upregulation of lipids: FC ≥ 1.5;

[0105] Downregulation of lipids: FC ≤ 0.67;

[0106] 2. Statistical significance:

[0107] The Wilcoxon rank-sum test was used.

[0108] A p-value < 0.05 was considered statistically significant.

[0109] 3. Multiple comparison correction:

[0110] The Benjamini-Hochberg method was used to correct for the false discovery rate (FDR).

[0111] FDR < 0.1 is used as the screening threshold.

[0112] (III) Multivariate Statistical Analysis

[0113] To further improve the reliability of the screening results, a partial least squares discriminant analysis (PLS-DA) model was used to model and analyze the samples, and the variable importance projection values ​​(VIP) were calculated. The screening criteria are as follows:

[0114] VIP > 1.0.

[0115] (iv) Comprehensive screening strategy for lipid biomarkers

[0116] Based on meeting the above univariate and multivariate screening criteria, candidate biomarkers are further screened using the following standards:

[0117] 1. It exhibits a consistent trend of change at different time points (1 week and 2–4 weeks after birth);

[0118] 2. It exhibits good repeatability across different batches of samples;

[0119] 3. It has a clear association with biological processes related to BPD, such as lung development, surfactant metabolism, or oxidative stress.

[0120] (v) Final screening results

[0121] Through the above screening process, five lipid molecules significantly associated with BPD were ultimately identified as a biomarker combination (see upregulation and downregulation for details). Figures 11-13 ),include:

[0122] 1. Lysophosphatidylcholine (LPC) 14:0 (downregulated);

[0123] 2. Phosphatidylcholine PC 32:1 | 16:0_16:1 (upregulation);

[0124] 3. Phosphatidylethanolamine PE 38:5 | 18:1-20:4 (upward adjustment);

[0125] 4. Triglycerides (TG) 58:7 | 18:1_20:2_20:4 (up-adjusted);

[0126] 5. Oxidized fatty acids FA 22:4;2O (upregulated).

[0127] The lipid profile described above can reflect abnormalities in pulmonary surfactant metabolism and changes in oxidative stress levels during the development of BPD.

[0128] 7. Evaluation of the predictive efficacy of lipid biomarkers

[0129] To evaluate the clinical application value of the screened lipid biomarkers, this invention uses Receiver Operating Characteristic Curve (ROC) to assess their predictive efficacy.

[0130] (a) Predictive efficacy of single lipid biomarkers

[0131] The predictive power of each lipid molecule for BPD is as follows:

[0132] • LPC 14:0: AUC is 0.75–0.80, indicating moderate predictive ability;

[0133] • PC 32:1: AUC is 0.78–0.83;

[0134] • PE 38:5: AUC is 0.80–0.85;

[0135] ·TG 58:7: AUC is 0.78–0.82;

[0136] ·FA 22:4;2O: AUC was 0.82–0.87.

[0137] The corresponding sensitivity and specificity are both in the range of 70%–85%.

[0138] (II) Construction and Evaluation of Joint Prediction Model

[0139] Incorporating the above five lipid biomarkers into a multivariate logistic regression model to construct a joint prediction model significantly improved its predictive performance.

[0140] • Area under the curve (AUC): 0.90–0.95;

[0141] • Sensitivity: 85%–90%;

[0142] Specificity: 83%–88%.

[0143] Compared to single lipid models, combined models demonstrate higher accuracy and stability in predicting BPD (see [link]). Figure 14 and Figure 15 ).

[0144] (III) Determination of the optimal cutoff value

[0145] The Youden index (Youden Index = Sensitivity + Specificity − 1) is used to determine the optimal discrimination threshold in order to achieve the best balance between sensitivity and specificity.

Claims

1. A combination of lipid biomarkers for predicting the risk of bronchopulmonary dysplasia in preterm infants, characterized in that, The biomarker combination includes: (1) Lysophosphatidylcholine (LPC) 14:0; (2) Phosphatidylcholine PC 32:1 | 16:0_16:1; (3) Phosphatidylethanolamine PE 38:5 | 18:1-20:4; (4) Triglycerides TG 58:7 | TG 18:1_20:2_20:4; (5) Oxidized fatty acids FA 22:4;2O; LPC 14:0 was downregulated in high-risk samples of bronchopulmonary dysplasia, while PC 32:1 | 16:0_16:1, PE 38:5 | 18:1_20:4, TG 58:7 | TG 18:1_20:2_20:4 and FA 22:4;2O were upregulated in high-risk samples of bronchopulmonary dysplasia.

2. A kit for detecting the lipid biomarker combination of claim 1, characterized in that, Include: (1) Standards of the lipid biomarker combination; (2) Internal standard mixture, wherein the internal standard mixture comprises EquiSPLASH LipidoMIX, fatty acids and palmitoylcarnitine; (3) Chromatographic columns, mobile phase reagents and mass spectrometry detection reagents used for liquid chromatography-mass spectrometry detection.

3. The reagent kit according to claim 2, characterized in that, The standard comprises a series of standard solutions with a concentration gradient of 10-1000 ng / mL.

4. The reagent kit according to claim 2, characterized in that, The internal standard mixture contained EquiSPLASHLipidoMIX at a concentration of 5 μg / mL, and fatty acids and palmitoylcarnitine at a concentration of 500 ng / mL.

5. The reagent kit according to claim 2, characterized in that, In the internal standard mixture, the fatty acids and palmitocarnitine are deuterated arachidonic acid and deuterated palmitocarnitine, respectively.

6. A non-targeted lipidomics detection method based on liquid chromatography-mass spectrometry, used for the detection of the lipid biomarker combination described in claim 1, characterized in that, Includes the following steps: (1) Sample pretreatment: Take plasma sample, add isopropanol solution containing internal standard mixture for protein precipitation, centrifuge, and take supernatant; (2) Chromatographic separation: Separation was performed using an ACQUITY UPLC CSH C18 reversed-phase column; (3) Mass spectrometry detection: A Q-Exactive mass spectrometer was used for detection in independent data acquisition (DIA) mode, with 20 SWATH scanning windows set. (4) Data processing: MS-DIAL software was used for peak detection, alignment and lipid identification. Peak alignment was performed by retention time correction. Data normalization was performed by retention time correction based on internal standard combined with QC-LOWESS normalization method to obtain the quantitative detection results of the lipid biomarker combination.

7. The method according to claim 6, characterized in that, In step (3), the positive ion mode spray voltage for mass spectrometry detection is 3500V, the negative ion mode spray voltage is 3200V, the S-lens RF voltage is 55, the capillary tube temperature is 320℃, and the auxiliary gas heating temperature is 300℃.

8. The method according to claim 6, characterized in that, The MS horizontal resolution mentioned in step (3) exceeds 30000, and the MS / MS horizontal resolution is 17500.

9. The use of the lipid biomarker combination according to claim 1 in the preparation of a risk prediction reagent for bronchopulmonary dysplasia in preterm infants.

10. The application of the method according to any one of claims 6-7 in lipidomics data analysis.

Images