Fe-MOF derived heterojunction material, preparation method thereof and application in thymic epithelial tumor diagnosis and risk stratification

Abstract

The invention relates to the technical field of biomedical detection, and discloses a Fe-MOF derived heterojunction material, a preparation method thereof and application of the Fe-MOF derived heterojunction material in thymic epithelial tumor diagnosis and risk stratification. The Fe-MOF derived heterojunction material is a Fe3O4-coated Fe-MOF nano material, and the invention further provides a thymic epithelial tumor diagnosis and / or risk stratification system based on the material, so that the thymic epithelial tumor detection performance is improved in a breakthrough manner, and the clinical diagnosis efficiency is improved. Through material innovation, method innovation and system integration, triple benefits of performance breakthrough, practical improvement and cost optimization are achieved in the field of TETs metabolism diagnosis, and good clinical application value and industrial development prospects are achieved.

Description

Technical Field

[0001] This application relates to the field of biomedical detection technology, and more specifically, to an Fe-MOF-derived heterojunction material, its preparation method, and its application in the diagnosis and risk stratification of thymic epithelial tumors. Background Technology

[0002] Thymic epithelial tumors (TETs) are a group of rare malignant tumors originating in the anterior mediastinum, including thymomas and thymic carcinomas. They are characterized by complex pathological phenotypes and insidious clinical presentations. According to the World Health Organization (WHO) 2021 classification, TETs can be divided into types A, AB, B1, B2, B3, and thymic carcinoma (type C). The 5-year survival rate for low-risk (LR) types (A, AB, B1) can reach 78%, while the 5-year survival rate for high-risk (HR) types (B2, B3, C) is only about 40% due to their high susceptibility to lymph node metastasis and distant spread. Current clinical diagnosis mainly relies on imaging techniques such as computed tomography (CT) for initial screening, but its specificity is limited (e.g., low accuracy in differentiating between thymoma, lymphoma, or thymic hyperplasia), often requiring invasive histopathological biopsy for confirmation. However, biopsies present problems such as patient resistance, sampling risks, and reliance on senior physicians' judgment for morphological results (e.g., confusion between TETs and non-tumor lymphocytes), resulting in approximately 30% of low-risk patients undergoing unnecessary surgery, leading to a waste of medical resources and overtreatment.

[0003] Metabolites, as small molecule products at the end of biological pathways, can reflect the state of tumor-driven metabolic reprogramming in real time. In TETs (tumor-induced metabolic disorders), abnormally proliferating thymocytes have endocrine activity and can trigger systemic metabolic disorders (such as paraneoplastic syndromes associated with myasthenia gravis). Therefore, metabolomics analysis is considered a powerful tool for precision diagnosis and mechanism exploration. Matrix-assisted laser desorption / ionization mass spectrometry (MALDI-MS) has become a high-performance metabolomics research platform due to its advantages of high throughput and rapid analysis. However, its application has long been limited to the detection of small molecule metabolites (molecular weight 100-1000 Da), mainly because traditional organic matrices (such as CHCA and DHB) generate high-intensity background noise in the low molecular weight range, masking the target metabolite signal. At the same time, the inhomogeneity of co-crystallization between traditional matrices and analytes leads to poor reproducibility (coefficient of variation (CV) can reach 30%-97%), making it difficult to directly apply to complex biological samples such as serum.

[0004] In recent years, nanomaterials (such as two-dimensional metal-organic framework nanosheets, 2D MOF NSs) have been developed as novel matrices to address the aforementioned problems. MOF materials, due to their high specific surface area, tunable pore structure, and weak self-decomposition properties, can reduce background interference and enhance cationization efficiency. For example, Fe-MOF materials can promote proton transfer through carboxyl ligands, thereby improving metabolite ionization efficiency. However, original MOFs still have inherent problems, including rapid electron-hole recombination in a single MOF significantly reducing metabolite ionization efficiency, and low photothermal conversion efficiency making it difficult to achieve efficient ionization.

[0005] Existing improvement strategies mainly enhance interfacial synergistic effects by constructing noble metal / metal oxide heterostructures (such as Au / MOF or Fe3O4 modification). However, these methods rely on complex multi-step modifications or weakly coupled interfaces, making it difficult to precisely control the heterostructure structure and defect distribution. While the strategy of partially pyrolyzing MOFs can introduce oxygen vacancies and heterostructure interfaces, improper temperature control can easily lead to MOF structure collapse or nanoparticle aggregation, thereby reducing performance.

[0006] Therefore, developing a novel nanomatrix material with excellent photothermal conversion efficiency and stable structure to achieve highly sensitive and reproducible detection of small molecule metabolites in the serum of patients with thymic epithelial tumors is of great significance for improving the accuracy of early diagnosis of TETs. Summary of the Invention

[0007] To address the aforementioned shortcomings in existing technologies, this application proposes a gradient pyrolysis-derived Fe3O4@Fe-MOF heterojunction (FM-450) material. This material achieves efficient charge separation and photothermal conversion through an S-shaped band structure, increasing the sensitivity of metabolic detection by 1000 times. Simultaneously, by combining machine learning algorithms, metabolic fingerprints are extracted from trace amounts of serum (1 μL), enabling non-invasive diagnosis and risk stratification of TETs. This solves the core problems of low sensitivity, cumbersome operation, and insufficient risk stratification capabilities in existing technologies. This breakthrough not only advances nanomaterial design but also provides a translatable platform for rapid clinical metabolomics diagnosis.

[0008] To achieve the above-mentioned objectives, this application adopts the following technical solution: In a first aspect, this application provides a method for preparing Fe3O4@Fe-MOF nanomaterials, comprising the following steps: Step 1: Dissolve terephthalic acid in the mixed solution, sonicate to dissolve, add FeCl2·4H2O, quickly inject triethylamine, and stir to form a colloidal suspension; Step 2: The prepared colloidal suspension is continuously sonicated, and the product is centrifuged, washed with water and vacuum dried to obtain a sheet-like Fe-MOF precursor; Step 3: Pyrolyze the Fe-MOF precursor in an inert gas atmosphere and obtain Fe3O4@Fe-MOF nanomaterials after natural cooling.

[0009] Further, in step 1, the mixed solution is a mixture of DMF, ethanol, and water; In step 1, the ratio of terephthalic acid, DMF, ethanol, water, FeCl2·4H2O and triethylamine is 0.75mmol:32mL:2mL:2mL:0.75mmol:0.8mL.

[0010] Furthermore, in step 3, the pyrolysis temperature is 350~500°C.

[0011] Furthermore, in step 3, the pyrolysis temperature is 450°C.

[0012] Secondly, this application provides a Fe3O4@Fe-MOF nanomaterial, which is prepared by the aforementioned preparation method.

[0013] Thirdly, this application provides the use of Fe3O4@Fe-MOF nanomaterials in the preparation of diagnostic and / or risk stratification products for thymic epithelial tumors.

[0014] Furthermore, the products include, but are not limited to, non-invasive diagnostic kits for TETs, mass spectrometry chips including pre-loaded FM-450, automated spotting systems, and standardized data analysis software.

[0015] Fourthly, this application provides a diagnostic system for thymic epithelial tumors, comprising: The data acquisition module is used to collect serum metabolic fingerprint data of subjects using Fe3O4@Fe-MOF nanomaterials as the matrix for laser desorption / ionization mass spectrometry. The data processing module is used to preprocess the collected serum metabolic fingerprint data to obtain a dataset containing at least 698 metabolic features; the preprocessing includes data resampling, spectral smoothing, baseline correction, index extraction, peak matching, and missing value imputation. The feature extraction module is used to extract feature intensity values ​​corresponding to at least one preset combination of metabolic markers from the preprocessed dataset; the preset combination of metabolic markers is selected from the Panel-10M combination used to distinguish thymic epithelial tumors from benign controls. The Panel-10M combination includes pyruvate, deoxyinosine, glycerol, linoleic acid, pyroglutamic acid, lactic acid, octanoylcarnitine, uracil, glycerol 3-phosphate, and nutritional bile acids. The result output module stores a trained machine learning classification model. The intelligent diagnosis module is configured to receive the feature intensity value, input it into the corresponding machine learning classification model for calculation, and output the diagnosis result. When the input feature is the intensity value corresponding to the Panel-10M combination, the result output module outputs the risk probability of the subject having thymic epithelial tumor.

[0016] Fifthly, this application provides a thymic epithelial tumor risk stratification system, comprising: The data acquisition module is used to collect serum metabolic fingerprint data of subjects using Fe3O4@Fe-MOF nanomaterials as the matrix for laser desorption / ionization mass spectrometry. The data processing module is used to preprocess the collected serum metabolic fingerprint data to obtain a dataset containing at least 698 metabolic features; the preprocessing includes data resampling, spectral smoothing, baseline correction, index extraction, peak matching, and missing value imputation. The feature extraction module is used to extract feature intensity values ​​corresponding to at least one preset combination of metabolic markers from the preprocessed dataset; the preset combination of metabolic markers is selected from the Panel-12M combination used to distinguish between low-risk and high-risk thymic epithelial tumors. The Panel-12M combination includes stearoylcarnitine, palmitic acid, squalene, oleic acid, erucic acid, deoxyadenosine, pyruvate, lactic acid, glycerol, octanoylcarnitine, deoxyinosine, and nutritional bile acids. The result output module stores a trained machine learning classification model. The intelligent diagnosis module is configured to receive the feature intensity value, input it into the corresponding machine learning classification model for calculation, and output the diagnosis result. When the input feature is the intensity value corresponding to the Panel-12M combination, the result output module outputs the risk level of the subject if they have thymic epithelial tumor.

[0017] In a sixth aspect, this application provides a combination of metabolic markers, Panel-10M, for the diagnosis of thymic epithelial tumors, the combination comprising pyruvate, deoxyinosine, glycerol, linoleic acid, pyroglutamic acid, lactic acid, octanoylcarnitine, uracil, glycerol 3-phosphate, and nutritional bile acids.

[0018] In a seventh aspect, this application provides a combination of metabolic markers, Panel-12M, for risk stratification of thymic epithelial tumors, the combination comprising stearoylcarnitine, palmitic acid, squalene, oleic acid, erucic acid, deoxyadenosine, pyruvate, lactic acid, glycerol, octanoylcarnitine, deoxyinosine, and nutritional bile acids.

[0019] Compared with the prior art, this application has the following beneficial effects: 1. Breakthrough Enhancement in Detection Performance: This application achieves a revolutionary enhancement of metabolite detection signals through the unique design of Fe3O4@Fe-MOF heterojunction material (FM-450). Compared to traditional organic matrices CHCA and DHB, FM-450 increases the signal intensity of standard metabolites (such as alanine and lysine) by up to 1000 times, reducing the detection limit to the pmol level. This effect stems directly from the material's triple synergistic mechanism: First, FM-450 exhibits strong absorption in the 300-400 nm wavelength range, perfectly matching the 355 nm laser of LDI MS, ensuring efficient energy utilization. Second, the S-shaped heterojunction structure forms an internal electric field, promoting the separation of photogenerated electron-hole pairs, with holes acting as the main charge carrier, significantly improving ionization efficiency. Finally, the local thermal effect of Fe3O4 nanoparticles promotes metabolite desorption, with a heating rate far exceeding that of single MOF materials.

[0020] In addition, the mesoporous structure of FM-450 (pore size approximately 2 nm) allows for the selective adsorption of metabolite molecules while simultaneously dispersing salts and proteins, thus maintaining clear metabolite signals even in high-salt (0.5 M NaCl) and high-protein (5 mg / mL BSA) environments. This characteristic eliminates the need for complex sample pretreatment, allowing for high-quality data to be obtained directly using only 1 μL of serum. More importantly, FM-450 exhibits excellent reproducibility: the coefficients of variation (CV) for intra-batch and inter-batch are as low as 4.1%-15.0% and 5.7%-12.8%, respectively, significantly better than CHCA (29.6%-97.0%) and DHB (31.3%-86.3%). This stability stems from the material's uniform dispersion characteristics, avoiding the problem of uneven co-crystallization between the matrix and analyte in traditional methods.

[0021] 2. Significant Advantages in Technical Practicality and Economy: This application simplifies the complex metabolic detection process into three steps: serum dilution → sample application → detection, with a total processing time of approximately 30 seconds per sample. Compared to traditional LC-MS / MS, which requires more than 30 minutes of sample preprocessing, the operational efficiency is improved by more than 60 times. This simplification is mainly due to: First, the FM-450 matrix is ​​tolerant to the salts and proteins in serum, eliminating the need for protein removal or extraction steps; second, matrix chips can be pre-prepared, and combined with an automated sample application system, high-throughput detection can be achieved. Finally, after pre-preparing the target plate, four inexperienced operators were recruited to conduct sample application tests, and the CV of the results was less than 10%, demonstrating that the technology is easy to standardize and promote.

[0022] 3. Substantial Leap in Clinical Diagnostic Efficacy: Based on serum metabolic fingerprints (SMFs) obtained from the FM-450 platform and combined with machine learning algorithms, this application has achieved a breakthrough in the diagnosis of thymomas. First, the Panel-10M diagnostic model achieved an AUC of 0.960 (95% CI: 0.931-0.988) in the validation cohort, with a sensitivity of 87.2%, specificity of 87.5%, and accuracy of 87.3%, providing a possibility for large-scale screening of thymomas. Second, the Panel-12M model achieved an AUC of 0.856 (95% CI: 0.780-0.931) when distinguishing between low-risk (LR) and high-risk (HR) TETs, significantly outperforming imaging methods (AUC=0.627), providing a new method for rapid differential diagnosis of thymoma patients at different risks.

[0023] In addition to achieving high-precision diagnosis, this application also reveals the pathological mechanisms of TETs through a metabolic biomarker panel. The metabolites in Panel-12M involve key pathways such as glycerol ester metabolism, pyruvate metabolism, and purine metabolism. Disruptions in these pathways are directly related to tumor proliferation and energy metabolism reprogramming. For example, changes in pyruvate and lactate reflect the important role of the Warburg effect in TET progression; abnormal fatty acid metabolism suggests the crucial role of lipid remodeling in malignant tumors; and differences in purine metabolites reflect the activity of nucleic acid synthesis related to cancer cell proliferation. This mechanistic insight provides a new perspective for understanding the pathogenesis of TETs and developing targeted therapies.

[0024] 4. Technological Innovation and Industrial Application Value: This application is the first to apply gradient pyrolysis engineering to the preparation of MOF heterojunctions, achieving controllable adjustment of material properties. This "one-step" synthesis strategy avoids the complex multi-step process of traditional heterojunction preparation, providing a new approach for the application of nanomaterials in biodetection. The technology successfully integrates materials science (MOF-derived materials), analytical chemistry (LDI MS), and artificial intelligence (machine learning), establishing an integrated "materials-detection-analysis" platform, providing a scalable technical framework for precision medicine. The diagnostic platform based on this technology has broad prospects for industrial transformation: First, it is suitable for routine use in hospital laboratories, and the standardized methods are easy to implement; second, the technical framework can be adapted to metabolic diagnosis of other tumor types, and the methods are highly scalable.

[0025] In summary, this application, through material innovation, methodological innovation, and system integration, achieves a triple benefit of performance breakthrough, practical improvement, and cost optimization in the field of TET metabolic diagnostics, demonstrating significant clinical application value and promising prospects for industrial development. Attached Figure Description Figure 1(1) Transmission electron microscope image and (2) Schematic diagram of the sheet thickness of the Fe-MOF precursor prepared in Example 1 of the present invention.

[0026] Figure 2 The images shown are: (1) transmission electron microscopy (TEM) image, (2) high-resolution TEM image, and (3) thermogravimetric analysis (TGA) diagram of the Fe3O4@Fe-MOF nanomaterial prepared in Example 1 of this invention.

[0027] Figure 3 The X-ray diffraction pattern, Fourier transform infrared spectrum and Fe 2p X-ray photoelectron spectrum of Fe3O4@Fe-MOF nanomaterials at different pyrolysis temperatures are shown.

[0028] Figure 4 For (1) demographic characteristics of 80 benign controls (BC, including 14 patients with thymic cysts and 66 healthy individuals) and 163 patients (70 low-risk (LR) thymic epithelial tumors and 93 high-risk (HR) thymic epithelial tumors), serum metabolic fingerprints with 698 features were generated for each sample, and (2) typical mass spectrometry and similar score distributions of benign controls (BC, green), low-risk patients (LR, orange), and high-risk patients (HR, red) in the m / z range of 80–500 Da.

[0029] Figure 5 Heatmaps were generated for the intensity of 698 metabolic features in 243 plasma samples (BC: benign control population; LR: low-risk thymic epithelial tumor patients; HR: high-risk thymic epithelial tumor patients).

[0030] Figure 6 For (1) performance evaluation of different machine learning algorithms in distinguishing thymic epithelial tumor patients from benign control populations (AUC: area under the curve; CA: attention mechanism; F1: F1 score, i.e., harmonic mean of precision and recall; Prec.: accuracy; Spec.: specificity; LASSO: minimum absolute contraction and selection operator; RR: ridge regression; NN: neural network; SVM: support vector machine algorithm); Receiver operating characteristic curve (ROC) shows the area under the curve of (2) training set and (3) validation set; (4) the situation of distinguishing thymic epithelial tumor patients and bisexual control groups by the LASSO model.

[0031] Figure 7(1) Venn diagrams of 10 m / z features selected after multiple rounds of screening, serving as a combination of diagnostic biomarkers; (2) a diagnostic model constructed using Panel-10M, used for training and validation cohorts; and (3) the diagnostic performance of differentially metabolites (including pyruvate (PA), deoxyinosine (D-ino), glycerol (Gly), linoleic acid (Lin-A), pyroglutamic acid (P-Glu), lactate (LA), octanoylcarnitine (O-ine), uracil (Ura), glycerol 3-phosphate (G 3-P), and nutritional bile acid (NA)) in patients with thymic epithelial tumors and benign controls.

[0032] Figure 8 (1) ROC curves for distinguishing between low-risk (LR) thymic epithelial tumor (TET) patients and high-risk (HR) thymic epithelial tumor (TET) patients, plotted using all 698 m / z features; (d) Venn plots show six m / z features selected based on significance (P < 0.001), fold change (FC > 1.5), LASSO score (> 0.3), and signal-to-noise ratio (S / N > 10), which were used to construct Panel-6M.

[0033] Figure 9 LASSO scores for (1) benign control (BC) and low-risk (LR) thymic epithelial tumors; and LASSO scores for (2) benign control (BC) and high-risk (HR) thymic epithelial tumors, including pyruvate (PA), deoxyinosine (D-ino), glycerol (Gly), linoleic acid (Lin-A), pyroglutamic acid (P-Glu), lactate (LA), octylcarnitine (O-ine), uracil (Ura), glycerol 3-phosphate (G 3-P), and taurocholic acid (NA)).

[0034] Figure 10(1) Schematic diagram of classifier construction process; Feature selection (green): Stage I is used to identify key metabolites to form Panel-6M, which must meet the statistical criteria (P < 0.001, FC > 1.5 and LASSO weight score > 0.3) when comparing LR-TET and HR-TET; Stage II is used to identify metabolites from Panel-10M to form a 6-score metabolite panel (Panel-6SM) with a ratio (LASSO score of BC and LR-TET divided by LASSO score of BC and HR-TET) greater than 4; Classifier construction (blue): Using the random forest algorithm, 12 metabolites from Stage I and Stage II are integrated into the final classifier of Panel-12M; (2) The optimal Panel-12M classifier is superior to the imaging model (mediastinal mass size, MMS) and other hierarchical models (Panel-6M; Panel-6SM).

[0035] Figure 11 For the stability evaluation of FM-450 reagent. (1) Mass spectrometry comparison of freshly prepared FM-450 and aged FM-450 reagent (stored for 6 months), labeled with batch-1 and batch-2 (batch-1 was synthesized 6 months ago (blue); batch-2 was recently synthesized (red), visual comparison of new and old FM-450 solutions (inset); (2) Intensity reproducibility of sodium adduct mass spectrometry peaks of 5 index metabolites (proline (Pro), glucose (Glu), sucrose (Suc), creatine (Cre)); ns: P>0.05, *: P<0.05; (3) Reproducibility of mass spectrometry data obtained at the intra-batch level (10 replicates per sample, cv: coefficient of variation) and (4) at different laser firing times (5 independent samples per metabolite), both using aged FM-450.

[0036] Figure 12 Standardization validation for automated sampling. (1) FM-450 preload (i) and random sampling in inexperienced population (ii); (2) automated spotting instrument; (3) automated spotting process and (4) coefficients of variation (CVs) of mass spectrometry signal intensity from random samples from inexperienced population (proline (Pro), glucose (Glu), sucrose (Suc) and creatine (Cre)). Detailed Implementation

[0037] The technical solutions and effects of this application will be further described in detail below with reference to embodiments and accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the invention, not the entire structure.

[0038] This application addresses the problems existing in the prior art: current screening and diagnosis of thymic epithelial tumors mainly rely on liquid biopsy and imaging. However, due to the atypical clinical manifestations, the low specificity of protein tumor markers and imaging tests, and the invasive and time-consuming nature of pathological examinations, the diagnosis of thymic epithelial tumors suffers from high false-negative rates and long diagnostic cycles, resulting in wasted disease treatment and medical costs, and seriously affecting patients' quality of life. This application constructs a biomarker combination for thymic epithelial tumor screening (including pyruvate, lactate, glycerol, uracil, pyroglutamate, glycerol-3-phosphate, deoxyinosine, octanoylcarnitine, linoleic acid, and nutritional bile acids), which features high accuracy (>90%), rapid diagnosis (less than 2 hours), and high-throughput processing (300 samples / 2 hours), and is expected to be applied to the screening and diagnosis of mediastinal tumors in clinical practice. The principle is as follows: First, a nanoparticle-enhanced laser desorption / ionization mass spectrometry platform is used to effectively extract serum metabolic fingerprints from mediastinal tumor disease groups and benign control groups (including patients with mediastinal cysts and healthy individuals). Then, through machine learning and statistical analysis, the disease group and the benign control group are accurately distinguished based on the differences in metabolic fingerprints, and a combination of biomarkers (including pyruvate, lactate, glycerol, uracil, pyroglutamate, glycerol-3-phosphate, deoxyinosine, octanylcarnitine, linoleic acid, and nutritional bile acids) is selected for further rapid screening and diagnosis. Based on the above, this application utilizes the differences in plasma metabolic fingerprints to meet the needs of accurate screening and diagnosis for patients with mediastinal tumors and healthy individuals, and constructs a combination of plasma biomarkers (including pyruvate, lactate, glycerol, uracil, pyroglutamate, glycerol-3-phosphate, deoxyinosine, octanylcarnitine, linoleic acid, and nutritional bile acids), combined with high-throughput mass spectrometry technology, to achieve accurate and convenient screening and diagnosis of thymoma.

[0039] Furthermore, low-risk and high-risk thymic epithelial tumors often have different prognoses and treatments. However, due to the high degree of overlap in clinical symptoms and imaging features, diagnosing thymic epithelial tumors of different risk levels, especially low-risk and high-risk tumors, is currently very difficult. Only pathological biopsy can differentiate between different subtypes of the disease. Diagnosis among thymic epithelial tumors remains challenging, with significant differences in diagnostic results between different doctors. Moreover, under the condition of limited pathological testing, the misclassification rate for different subtypes is as high as 80%. This application constructs a cohort of serum samples from patients with different risks of thymic epithelial tumors. With the assistance of metabolomics and machine learning algorithms, a special metabolite panel is constructed to distinguish between different risks of thymic epithelial tumors (Panel-6M: deoxyadenosine, palmitic acid, oleic acid, erucic acid, squalene, stearoylcarnitine; Panel-6SM: pyruvate, lactic acid, glycerol, octanoylcarnitine, nutritional bile acids). Based on the above two panels, we optimize the algorithm to obtain a metabolite combination that can efficiently distinguish between different risks of thymic epithelial tumors, called Panel-12M. The metabolic biomarker Panel was used to accurately differentiate between patients with thymic epithelial tumors at different risks, with accuracy far exceeding that of imaging studies. The principle is as follows: a nanoparticle-enhanced laser desorption / ionization mass spectrometry platform was used to effectively extract serum metabolic fingerprints from both low-risk and high-risk thymic epithelial tumor groups. Then, through machine learning and statistical analysis, two metabolic biomarker combinations (Panel-6M and Panel-6SM) were selected from plasma and cerebrospinal fluid metabolic fingerprints using different machine learning methods. Based on these two metabolite Panels, an optimized machine learning algorithm was used to construct a more efficient biomarker combination, Panel-12M (including deoxyadenosine, palmitic acid, oleic acid, erucic acid, squalene, stearoylcarnitine, pyruvate, lactic acid, glycerol, octanoylcarnitine, and nutritional bile acids), to accurately distinguish between low-risk and high-risk thymic epithelial tumors. Ultimately, the metabolic biomarker combination Panel-12M was extracted and constructed from the serum metabolic fingerprint profile to achieve differential diagnosis between low-risk and high-risk thymomas.

[0040] Therefore, this application provides the following examples: Example 1: Construction of Fe3O4@Fe-MOF nanomaterials using intelligent pyrolysis technology and in-situ growth technology Step 1: Fe-MOF nanosheets were synthesized using an ultrasonic-assisted water bath method. Specific procedure: 0.75 mmol of terephthalic acid was dissolved in a mixed solution of 32 mL DMF, 2 mL ethanol, and 2 mL water. After ultrasonic dissolution, 0.75 mmol of FeCl2·4H2O was added, followed by rapid injection of 0.8 mL of triethylamine. The mixture was stirred for 5 minutes to form a colloidal suspension. Step 2: The prepared colloidal suspension was continuously sonicated at 4°C for 8 hours (40 kHz). The product was centrifuged, washed with water, and vacuum dried (60°C) to obtain a sheet-like Fe-MOF precursor with a thickness of approximately 1.2 nm. Figure 1 ); Step 3: Place the Fe-MOF precursor in a tube furnace and pyrolyze it for 1 hour at a gradient temperature (350°C, 400°C, 450°C, 500°C) under an argon atmosphere; Step 4: After natural cooling, FM-X series materials are obtained (X represents the pyrolysis temperature). Among them, FM-450, the pyrolysis product at 450°C, is the optimal material. It generates Fe3O4 nanoparticles (5-20 nm in size) in situ on the MOF framework through partial pyrolysis, forming a tight heterogeneous interface. Step 5: Thermogravimetric analysis (TGA) showed that MOF decomposed in the 350-500°C range, with the Fe3O4 particles reaching their largest size and exhibiting the best interface integrity at 450°C. Figure 2 ); In terms of morphology and structure, TEM and HRTEM were used to show that Fe3O4 nanoparticles in FM-450 form an S-type heterojunction with MOF, and the interplanar spacing of 0.253 nm corresponds to the (311) crystal plane of Fe3O4; PXRD confirmed the appearance of Fe3O4 characteristic peaks (30°, 35°, etc.); In terms of chemical state studies, FTIR shows that characteristic MOF peaks (such as O=CO) are retained in FM-450; XPS technology reveals Fe... 2+ / Fe 3+ Mixed valence states and oxygen vacancies enhance charge separation. Figure 3 ).

[0041] Example 2: Acquisition of serum metabolic fingerprint data of thymic epithelial tumors and benign controls (BCs) using a nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometry platform. Step 1: Instrument and reagent preparation: matrix-assisted laser desorption / ionization time-of-flight mass spectrometry, serum sample, deionized water, matrix (Fe3O4@Fe-MOF nanoparticles). Step 2: Preprocess the plasma sample, including protein removal and metabolite extraction; Step 3: Prepare a matrix solution of 1 mg / mL by dissolving the inorganic nanoparticles in deionized water; Step 4: Sample preparation was performed on the mass spectrometry target plate. 1.0 μL of each pretreated plasma sample was spotted and dried at room temperature. Step 5: Prepare the matrix on the mass spectrometry target plate, spotting 1.0 μL of each matrix solution and drying at room temperature; Step 6: Collect serum metabolic fingerprint data using a laser desorption / ionization time-of-flight mass spectrometer; Step 7: Following Step 6, serum metabolomics data were collected from 243 serum samples (163 patients with thymic epithelial tumors and 80 benign controls). Figure 4 ); Step 8: Preprocess the metabolic fingerprint data of 243 serum samples, including data resampling, spectral smoothing, baseline correction, index extraction, peak alignment, and missing value imputation, to obtain 698 metabolic features between 80-500 Da m / z. Figure 5 ).

[0042] Example 3: Using machine learning on serum metabolomics data to achieve accurate diagnosis of thymic epithelial tumor patients and benign controls. Step 1: Divide the 243 serum samples into a training set (172 serum samples, TETs / BC=116 / 56) and a test set (71 serum samples, TETs / BC=47 / 24). Step 2: On Orange, use different machine learning methods with 5-fold cross-validation to optimize parameters and train models on the dataset, and obtain the performance of different machine learning methods on the training and validation sets. Figure 6 Among them, the LASSO algorithm has the best area under the curve result (AUC=0.975) and the best discrimination performance.

[0043] Example 4: Panel-10M, a diagnostic marker for thymic epithelial tumors Extraction of disease-specific serum metabolic biomarkers and the use of combinations of these biomarkers for timely screening and diagnosis of thymic epithelial tumors: Step 1: Ten metabolic biomarkers were selected based on the following criteria: average intensity > 500, intensity fold difference > 1.5, P-value < 0.001, and LASSO regression weight > 0.3. These biomarkers were then aggregated to form the Metabolic Panel-10M. Figure 7 ); Step 2: Compared to individual biomarkers, the combination of 10 metabolic biomarkers demonstrated better performance in screening and diagnosing thymic epithelial tumors and benign controls. Figure 7 ).

[0044] Example 5: Panel-12M, a risk stratification marker for thymic epithelial tumors Metabolic biomarkers were extracted from the serum metabolic fingerprint profiles of patients with low-risk and high-risk thymoma. A composite metabolic biomarker, Panel-12M, was used to achieve accurate differential diagnosis between these two groups. Step 1: We randomly divided 163 serum samples (LR / HR ratio: 70 / 93) into a training set (LR / HR ratio: 49 / 67) and a validation set (LR / HR ratio: 21 / 26), maintaining demographic balance between the two groups. Our analysis workflow first performed LASSO analysis on the training set, using 5-fold cross-validation and stratifying by disease risk. Based on the criteria of P < 0.001, fold change > 1.5, LASSO score > 0.3, and signal-to-noise ratio (S / N) > 10, this analysis identified six more sensitive metabolic biomarkers, forming a streamlined panel (Panel-6M). Panel-6M includes stearoylcarnitine, palmitic acid, squalene, oleic acid, erucic acid, and deoxyadenosine (DAA). Figure 8 ) Step 2: We utilized the previously established Panel-10M to differentiate risk populations, as there were significant differences in metabolic characteristics between different risk groups (metabolite score fold change range: 0.87–4.67). From this panel, we selected six metabolites with the most significant changes to create a new panel (Panel-6SM). The selection criteria were a LASSO score ratio greater than 4, i.e., comparing BC with LR-TET and BC with HR-TET, reflecting significant metabolic changes in different subtypes of patients during thymoma progression. The final Panel-6SM included pyruvate, lactate, glycerol, octanoylcarnitine, deoxyinosine, and nutritional bile acids (…). Figure 9 ).

[0045] Step 3: Through model algorithm optimization, the two initial Panel-6M and Panel-6SM are merged into a new metabolite set, Panel-12M, which has superior differential diagnostic efficacy. Compared with other initial metabolic marker combinations, the optimized metabolic marker set Panel-12M has stronger differential diagnostic performance for patients with thymoma at different risks; and compared with imaging examinations, the composite metabolic marker set Panel-12M has more stable diagnostic performance. Figure 10 ).

[0046] Example 6: Stability and reproducibility verification of Fe3O4@Fe-MOF nanomaterials (FM-450) Two batches of FM-450 material prepared according to the method of Example 1 (batch-1 stored for 6 months, batch-2 freshly prepared) were redispersed and reconstituted into solutions. Figure 11 As shown, the dispersions of the two batches of materials showed no significant visual difference (inset). Subsequently, five indicator metabolites (proline (Pro), glucose (Glu), sucrose (Suc), and creatine (Cre)) were detected using FM-450 as the matrix for both batches-1 and-2. The mass spectra obtained from the two matrices were compared.Figure 11 The relative intensities and distributions of the key metabolite signal peaks were highly consistent, indicating that the physicochemical properties of the FM-450 material did not change significantly after 6 months of storage.

[0047] Subsequently, intra-batch reproducibility and laser bombardment stability experiments were conducted. The mass spectrometry data obtained at the intra-batch level (10 replicates per sample, cv: coefficient of variation) and at different laser bombardment numbers (5 independent samples per metabolite) showed good reproducibility. Figure 11 ).

[0048] Example 7: Automated Sampling Standardization Validation Figure 12 The results show that the FM-450 was preloaded (i) and random sampling was performed in an inexperienced population (ii). The results show that the coefficients of variation (CV) of the mass spectrometry signal intensity (proline (Pro), glucose (Glu), sucrose (Suc) and creatine (Cre)) of the random samples from the inexperienced population were all less than 10%.

[0049] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

Claims

1. A method for preparing Fe3O4@Fe-MOF nanomaterials, characterized in that, Includes the following steps: Step 1: Dissolve terephthalic acid in the mixed solution, sonicate to dissolve, add FeCl2·4H2O, quickly inject triethylamine, and stir to form a colloidal suspension; Step 2: The prepared colloidal suspension is continuously sonicated, and the product is centrifuged, washed with water and vacuum dried to obtain a sheet-like Fe-MOF precursor; Step 3: Pyrolyze the Fe-MOF precursor in an inert gas atmosphere and obtain Fe3O4@Fe-MOF nanomaterials after natural cooling.

2. The preparation method according to claim 1, characterized in that, In step 1, the mixed solution is a mixture of DMF, ethanol, and water; In step 1, the ratio of terephthalic acid, DMF, ethanol, water, FeCl2·4H2O and triethylamine is 0.75mmol:32mL:2mL:2mL:0.75mmol:0.8mL.

3. The preparation method according to claim 1, characterized in that, In step 3, the pyrolysis temperature is 350~500°C.

4. The preparation method according to claim 4, characterized in that, In step 3, the pyrolysis temperature is 450°C.

5. Fe3O4@Fe-MOF nanomaterials, characterized in that, It is prepared by the preparation method according to any one of claims 1-4.

6. Use of Fe3O4@Fe-MOF nanomaterials in the preparation of diagnostic and / or risk stratification products for thymic epithelial tumors.

7. A diagnostic system for thymic epithelial tumors, characterized in that, include: The data acquisition module is used to collect serum metabolic fingerprint data of subjects using Fe3O4@Fe-MOF nanomaterials as the matrix for laser desorption / ionization mass spectrometry. The data processing module is used to preprocess the collected serum metabolic fingerprint data to obtain a dataset containing at least 698 metabolic features; the preprocessing includes data resampling, spectral smoothing, baseline correction, index extraction, peak matching, and missing value imputation. The feature extraction module is used to extract the feature intensity value corresponding to at least one preset combination of metabolic biomarkers from the preprocessed dataset. The preset combination of metabolic markers is selected from the Panel-10M combination used to distinguish thymic epithelial tumors from benign controls; The Panel-10M combination includes pyruvate, deoxyinosine, glycerol, linoleic acid, pyroglutamic acid, lactic acid, octanoylcarnitine, uracil, glycerol 3-phosphate, and nutritional bile acids. The result output module stores a trained machine learning classification model. The intelligent diagnosis module is configured to receive the feature intensity value, input it into the corresponding machine learning classification model for calculation, and output the diagnosis result. When the input feature is the intensity value corresponding to the Panel-10M combination, the result output module outputs the risk probability of the subject having thymic epithelial tumor.

8. A thymic epithelial tumor risk stratification system, characterized in that, include: The data acquisition module is used to collect serum metabolic fingerprint data of subjects using Fe3O4@Fe-MOF nanomaterials as the matrix for laser desorption / ionization mass spectrometry. The data processing module is used to preprocess the collected serum metabolic fingerprint data to obtain a dataset containing at least 698 metabolic features; the preprocessing includes data resampling, spectral smoothing, baseline correction, index extraction, peak matching, and missing value imputation. The feature extraction module is used to extract the feature intensity value corresponding to at least one preset combination of metabolic biomarkers from the preprocessed dataset. The preset combination of metabolic markers is selected from the Panel-12M combination used to distinguish between low-risk and high-risk thymic epithelial tumors. The Panel-12M combination includes stearoylcarnitine, palmitic acid, squalene, oleic acid, erucic acid, deoxyadenosine, pyruvate, lactic acid, glycerol, octanoylcarnitine, deoxyinosine, and nutritional bile acids. The result output module stores a trained machine learning classification model. The intelligent diagnosis module is configured to receive the feature intensity value, input it into the corresponding machine learning classification model for calculation, and output the diagnosis result. When the input feature is the intensity value corresponding to the Panel-12M combination, the result output module outputs the risk level of the subject if they have thymic epithelial tumor.

9. A panel-10M, a metabolic biomarker combination for diagnosing thymic epithelial tumors, characterized in that, The combination includes pyruvate, deoxyinosine, glycerol, linoleic acid, pyroglutamic acid, lactic acid, octanoylcarnitine, uracil, glycerol 3-phosphate, and nutritional bile acids.

10. A combination of metabolic biomarkers, Panel-12M, for risk stratification of thymic epithelial tumors, characterized in that, The combination includes stearoylcarnitine, palmitic acid, squalene, oleic acid, erucic acid, deoxyadenosine, pyruvic acid, lactic acid, glycerol, octanoylcarnitine, deoxyinosine, and nutritional bile acids.

Images