Serum metabolism marker combination for diagnosing and classifying infectious endocarditis and screening method thereof

Abstract

The invention discloses a serum metabolism marker combination for diagnosing and classifying infectious endocarditis and a screening method thereof, and relates to the technical field of biological science. According to the method, a serum metabolism fingerprint spectrum is obtained through a matrix-assisted laser desorption / ionization mass spectrometry (MALDI MS) technology, and feature screening and model construction are carried out in combination with multiple machine learning models (including ridge regression, random forest, LASSO regression and the like). Through machine learning, five-fold cross validation is adopted to optimize model performance, and independent queue verification is utilized to evaluate generalization ability and overfitting risk. According to the technical scheme, rapid diagnosis and classification of IE can be achieved within 10 minutes, the pretreatment process is simple (an existing IE diagnosis method is complex in process and consumes 2-5 days), the requirements of clinical IE rapid screening and large-scale sample analysis can be met, and the method is suitable for industrial production. The detection flux is obviously improved; the cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of bioscience and technology, and in particular to a combination of serum metabolic markers for the diagnosis and classification of infective endocarditis and a screening method thereof. Background Technology

[0002] Infective endocarditis (IE) is a serious systemic infectious disease that threatens human health. Its lesions primarily affect the heart valves (including natural and artificial valves), the endocardial wall, congenital heart defects, or implanted cardiac devices. The disease is mainly caused by pathogens such as Staphylococcus, Streptococcus, and Enterococcus. IE has a high mortality rate; more than one-third of patients die within one year of diagnosis. Therefore, establishing a rapid, highly sensitive, and highly specific diagnostic method is crucial for optimizing treatment, improving prognosis, and reducing mortality.

[0003] Currently, the clinical diagnosis of IE (extravascular coagulation) mainly relies on the modified Duke criteria, which uses blood culture results and imaging examinations (such as vegetation detection) as the core judgment criteria. However, existing diagnostic methods have many limitations: the diagnostic process is complex, the testing cycle is long, and there are many negative blood culture samples, which leads to diagnostic delays and difficulties in choosing a treatment strategy.

[0004] Metabolomics studies have shown that disease states can cause alterations in cellular or tissue metabolic pathways, which can be reflected in dynamic changes in the concentration of metabolites in bodily fluids such as blood and urine. In the development of infectious diseases (IE), metabolic abnormalities often precede the appearance of clinical symptoms; therefore, identifying relevant metabolic biomarkers is helpful for early diagnosis. Serum, as a common source of metabolites, offers advantages such as easy access, minimal invasiveness, and high patient acceptance, making it an ideal sample for biomarker screening.

[0005] Mass spectrometry is a core tool for metabolic biomarker research. While traditional methods such as gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) can achieve precise quantification, they still suffer from problems such as complex sample pretreatment, the need for enrichment steps, long analysis times, easy omission of low-abundance metabolites, and difficulty in achieving high-throughput automation. Furthermore, complex matrices in biological samples can easily lead to column and ion source contamination, further affecting analytical stability and reproducibility. In contrast, matrix-assisted laser desorption / ionization mass spectrometry (MALDI-MS), as a soft ionization mass spectrometry technique, can be used not only for the detection of large biomolecules (such as proteins, peptides, and polysaccharides) but also for the rapid screening of small molecule metabolites. This method features simple pretreatment, requiring only simple dilution of serum samples followed by co-crystallization with the matrix, enabling single-sample detection speeds down to the second. It boasts high repeatability, high throughput, and automation potential, making it ideal for non-targeted screening of IE-related specific metabolic biomarkers, accurately reflecting the patient's pathological state.

[0006] Furthermore, the treatment strategy for interstitial bleeding (IE) is closely related to the type of pathogen, and rapidly identifying the pathogen type is crucial for improving treatment outcomes. For example, approximately 25–42% of IE patients require heart valve surgery, and the timing of the surgery directly impacts prognosis. Streptococcus is one of the common pathogens causing IE, and developing rapid identification and differentiation methods for IE caused by streptococcal infection can assist clinicians in developing targeted anti-infection regimens or adjusting surgical decisions as early as possible.

[0007] To address the aforementioned issues, this invention aims to provide a combination of serum metabolic biomarkers and a screening method based on MALDI MS technology for the diagnosis and classification of infective endocarditis. By identifying specific metabolic biomarkers in serum associated with IE, the invention achieves early, rapid, and accurate diagnosis and classification of IE, providing a reliable basis for clinical treatment decisions, and ultimately improving patient prognosis and reducing mortality. Summary of the Invention

[0008] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to develop a combination of serum metabolic markers and a screening method for the diagnosis and classification of infective endocarditis.

[0009] To achieve the above objectives, the present invention provides a combination of serum metabolic markers for the diagnosis and classification of infective endocarditis and a screening method thereof.

[0010] Furthermore, the serum metabolic marker combination includes 2-ketobutyric acid, urea, aminoacetone, pyroglutamic acid, and ornithine.

[0011] Furthermore, a method for screening serum metabolic markers for the diagnosis of infective endocarditis includes the following steps: Step 1: Metabolic detection was performed on serum samples from the IE group and the healthy control group using matrix-assisted laser desorption / ionization mass spectrometry to obtain serum metabolic fingerprint profiles; The method for obtaining the serum metabolic fingerprint is as follows: 1) Iron nanoparticles were synthesized using a hydrothermal method and dissolved in water to prepare a matrix solution with a concentration of 1 mg / mL; 2) Spot 1 μL of serum sample diluted 10 times onto the mass spectrometry target plate, let it air dry at room temperature, add 1 μL of matrix solution, and let it air dry again at room temperature to obtain the serum sample for mass spectrometry analysis; 3) Serum samples were analyzed by matrix-assisted laser desorption / ionization time-of-flight mass spectrometry to obtain serum metabolic fingerprints; Step 2: Perform data preprocessing on the serum metabolic fingerprint spectrum, and collect the m / z signal intensity matrix corresponding to each serum sample to obtain serum metabolic characteristics; Step 3: Divide the serum metabolic characteristics collected from the IE group and the healthy control group into training set and test set. On the training set, use a variety of machine learning algorithms to train the model. Evaluate each model through five-fold cross-validation and select the full spectrum diagnostic model. Step 4: Perform a t-test on all 146 features of the training set samples from Step 3 to screen them. p Forty-four m / z features with a value < 0.05 were selected. A greedy algorithm was used to optimize and select the five most discriminative metabolic features, which were then compared with human metabolome databases to identify corresponding serum metabolic biomarkers. The resulting serum metabolic biomarkers are: 2-ketobutyric acid (2-KA), urea (Ure), aminoacetone (Ami), pyroglutamic acid (PA), and ornithine (Orn).

[0012] Furthermore, in step two, the data preprocessing includes spectral smoothing, baseline correction, and peak alignment.

[0013] Furthermore, in step three, the machine learning algorithm includes one or more of ridge regression, random forest, and LASSO regression.

[0014] Furthermore, in step three, the machine learning algorithm is ridge regression.

[0015] Furthermore, in step four, the m / z values ​​of the five metabolic biomarkers are as follows: 2-ketobutyric acid: [2-KA+H] + The m / z is 103.04, and the reaction is: [Ure + 2Na - H] + The m / z value is 105.00. Aminoacetone: [Ami + 2Na-H] + The m / z value is 118.02. Pyroglutamic acid: [PA+H] + The m / z value is 130.05. Ornithine: [Orn + 2Na- H] + The m / z value is 177.06.

[0016] Furthermore, a rapid detection method for infective endocarditis based on the method described above establishes an IE standard detection model using machine learning algorithms based on the five serum metabolic markers, inputs the serum metabolic fingerprint profile to be tested into the IE standard detection model, and outputs the detection results.

[0017] Furthermore, the machine learning algorithm is LASSO regression.

[0018] Furthermore, a method for identifying streptococcal endocarditis based on the method described includes the following steps: S1. Using all 146 serum metabolic features of streptococcal and non-streptococcal IE samples as input data, an initial classification model was constructed using various machine learning algorithms. S2. Based on the best-performing LASSO regression model, extract the contribution scores of each serum metabolic feature output by the model; set the screening threshold as LASSO regression contribution score > 0.3, screen 146 initial serum metabolic features, remove features with scores below the threshold, and retain high contribution features. S3. Construct a classifier based on the top 10 serum metabolic features after screening; S4. Input the metabolic fingerprint of the serum to be tested into the classifier and output the detection results.

[0019] Furthermore, in step S1, the machine learning algorithm is ridge regression, random forest, LASSO regression, or neural network.

[0020] In a preferred embodiment 1 of the present invention, a method for extracting the metabolic fingerprint of serum from IE patients using iron nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometry is described in detail.

[0021] In another preferred embodiment 2 of the present invention, a screening method for screening metabolic biomarkers based on serum metabolic fingerprints is described in detail.

[0022] In another preferred embodiment 3 of the present invention, a method for constructing an IE classification model for streptococcal infection is described in detail.

[0023] Technical effects: 1. This invention uses iron nanoparticles as the inorganic nanomatrix of MALDI MS. This material has a high specific surface area and a stable porous structure, which can significantly improve photoelectric performance and surface dissociation / ionization efficiency, effectively enhance the desorption and ionization ability of IE-related metabolites, and achieve high-throughput (384 samples per chip) and rapid (30 seconds / sample) detection of low concentration metabolites in serum without the need for culture.

[0024] 2. This invention obtains serum metabolic fingerprints using MALDI MS and combines various machine learning models (including ridge regression, random forest, LASSO regression, etc.) for feature selection and model construction. Using Orange software (3.36.2), five-fold cross-validation is employed to optimize model performance, and independent cohort validation is used to assess generalization ability and overfitting risk. Five candidate metabolic biomarkers (2-ketobutyric acid, urea, aminoacetone, pyroglutamate, and ornithine) are ultimately selected. Based on this technical solution, rapid diagnosis and classification of IE can be achieved within 10 minutes. The preprocessing procedure is simple (existing diagnostic methods for IE are complex and take 2-5 days), meeting the needs of rapid clinical IE screening and large-scale sample analysis, significantly improving detection throughput and reducing costs.

[0025] 3. Verification showed that the AUC of the metabolic biomarker combination of this invention for diagnosing IE reached 0.882. Furthermore, the AUC of the classification and discrimination of streptococcal IE further constructed by this invention reached 0.945, providing a reliable technical solution for achieving rapid diagnosis and accurate subtyping of IE.

[0026] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0027] Figure 1 This is a representative metabolic fingerprint of the IE group and the healthy control group in a preferred embodiment 1 of the present invention; Figure 2 This is a comparative diagram of the distribution of five specific metabolites in serum samples of the IE group and the healthy control group, according to a preferred embodiment 2 of the present invention. Figure 3 This is a receiver operating characteristic (ROC) curve of an IE standard diagnostic model constructed using five specific metabolites, which is a preferred embodiment 2 of the present invention; Figure 4 This is a performance diagram of the classifier-146 for streptococcal infection IE detection, which is a preferred embodiment 3 of the present invention; Figure 5 This is a performance diagram of the classifier-10 for streptococcal infection IE detection, which is a preferred embodiment 3 of the present invention. Detailed Implementation

[0028] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0029] Example 1: Extraction of metabolic fingerprints from serum of IE patients using iron nanoparticle-enhanced laser desorption / ionization time-of-flight mass spectrometry

[0030] Experimental steps: Step 1: Iron nanoparticles were synthesized using a modified hydrothermal method (CHEN W, YU H, HAO Y, et al. ComprehensiveMetabolic Fingerprints Characterize Neuromyelitis Optica Spectrum Disorder by Nanoparticle-Enhanced Laser Desorption / Ionization Mass Spectrometry [J]. ACSNano, 2023, 17(20): 19779-92.), and a 1 mg / mL iron nanoparticle matrix solution was prepared by dissolving the nanoparticles in deionized water.

[0031] Step 2: Spot 1 μL of serum sample (only the serum sample needs to be diluted 10 times) onto the mass spectrometry target plate and allow it to air dry at room temperature to complete the preparation of the serum sample on the mass spectrometry target plate.

[0032] Step 3: Add 1 μL of matrix solution to the air-dried serum sample and allow it to air dry at room temperature.

[0033] Step 4: Using MALDI MS technology, serum metabolic fingerprints were collected from IE patients and healthy individuals to obtain serum metabolic fingerprint profiles. Figure 1 ).

[0034] Example 2: Screening of metabolic biomarkers based on serum metabolic fingerprints

[0035] Step 1: Perform data preprocessing on the serum metabolic fingerprint obtained in Example 1, including spectral smoothing, baseline correction and peak matching, etc., and collect the m / z signal of each sample. Finally, 146 m / z signals were collected for each sample. Step 2: Divide the 72 serum samples into a training set and a test set. The training set includes 18 healthy control serum samples and 23 IE group serum samples, while the test set includes 12 healthy control serum samples and 19 IE group serum samples. Step 3: Train the model on the training set using machine learning algorithms such as Ridge Regression, Random Forest and LASSO Regression respectively. Use the Ridge Regression function of Orange software (version 3.36.2) to evaluate the model performance and select the full-spectrum diagnostic model.

[0036] Step 4: Perform a t-test on all 146 features of the training set samples from Step 3, and select 44 m / z features with P < 0.05. Optimize and select the 5 most discriminative metabolic features using a greedy algorithm, and compare them with the Human Metabolome Database (HMDB). Five metabolic biomarkers were identified: 2-ketobutyric acid ([2-KA+H]). + The m / z is 103.04), urea ([Ure + 2Na- H]). + (m / z is 105.00), aminoacetone ([Ami + 2Na-H]) + (m / z is 118.02), pyroglutamic acid ([PA+H]) + (m / z is 130.05) and ornithine ([Orn+ 2Na- H]) + (The m / z is 177.06) Figure 2 ).

[0037] Step 5: Based on the five selected specific metabolic biomarkers, a standard diagnostic model for IE is constructed using the LASSO regression machine learning algorithm; the diagnostic performance of this model is then validated on the test set. Figure 3 It can be seen that the diagnostic model has excellent sensitivity and specificity, and performs well.

[0038] Example 3: Construction of an IE classification model for streptococcal infection

[0039] Step 1: The serum metabolic fingerprint of the serum samples to be tested (including streptococcal IE samples and non-streptococcal IE samples) was obtained using the method in Example 1. The data was preprocessed, including spectral smoothing, baseline correction and peak matching, etc. The m / z signal of each sample was collected, and finally 146 m / z signals were collected for each sample. Step 2: Using Orange software (version 3.36.2), with all 146 serum metabolic features obtained in Step 1 (generated from the m / z signals) as input data, we constructed an initial classification model using four machine learning algorithms: Ridge Regression, Random Forest, LASSO Regression, and Neural Network.

[0040] The results show that among the four algorithms, the LASSO regression model (named Classifier-146) has the best performance, with the highest AUC value of 0.655. Figure 4 ).

[0041] Step 3: Based on the LASSO regression model constructed in Step 2, extract the contribution scores of each serum metabolic feature output by the model; set the screening threshold as "LASSO contribution score > 0.3", screen the 146 initial serum metabolic features, remove features with scores below the threshold, and retain features with high contribution.

[0042] Step 4: Construct a streptococcal infection IE classifier-10 based on the top 10 serum metabolic features obtained in Step 3 (Table 1): The AUC value of classifier-10 reached 0.945 ( Figure 5 This enables efficient detection of streptococcal IE infection.

[0043] Table 1. Ranking of the top 10 metabolic features by contribution

[0044] The technical advantages of this invention are as follows: 1. Excellent detection performance Currently, diagnostic methods for IE (extracorporeal membrane oxygenation) are complex and time-consuming (2-5 days). This invention uses iron nanoparticles as the inorganic nanomatrix for MALDI MS. This material possesses a high specific surface area and a stable porous structure, which can significantly improve photoelectric properties and surface dissociation / ionization efficiency, effectively enhancing the desorption and ionization capabilities of IE-related metabolites. It can achieve high-throughput (384 samples per chip) and rapid (30 seconds / sample) detection of low concentrations of metabolites in serum without the need for culture.

[0045] This invention acquires serum metabolic fingerprints using matrix-assisted laser desorption / ionization mass spectrometry (MALDI-MS) and combines various machine learning models (including ridge regression, random forest, and LASSO regression) for feature selection and model construction. Using Orange software (3.36.2), five-fold cross-validation is employed to optimize model performance, and independent cohort validation is used to assess generalization ability and overfitting risk. Five candidate metabolic biomarkers were ultimately selected. Based on this technical approach, rapid diagnosis and classification of IE can be achieved within 10 minutes. The preprocessing workflow is simple, meeting the needs of rapid clinical IE screening and large-scale sample analysis, significantly improving throughput and reducing costs.

[0046] Furthermore, in IE diagnosis, blood culture-negative samples account for as high as 2.5%-31.0%. These samples cannot obtain pathogen information through traditional culture methods, severely impacting diagnostic accuracy. This invention constructs a unique metabolic fingerprint profile based on serum samples from 72 IE patients, and then combines this with machine learning algorithms to establish a diagnostic model. Validation shows that the model achieves a receiver operating characteristic (AUC) of 0.882 for IE diagnosis accuracy and an even higher AUC of 0.945 for IE type classification accuracy, effectively solving the diagnostic challenge of blood culture-negative samples and providing dual technical support for accurate IE diagnosis.

[0047] 2. Outstanding clinical diagnostic value

[0048] The metabolic fingerprinting diagnostic technology established in this invention demonstrates outstanding performance in the diagnosis and classification of IE (infectious leukemia): it not only solves the pain points of traditional methods being "time-consuming and culture-dependent," achieving rapid detection per sample in seconds, but also overcomes the diagnostic challenge of blood culture-negative samples. It shows great potential in diagnosing IE and differentiating streptococcal IE, which helps clinicians achieve early diagnosis and accurate classification of IE, and provides key evidence for timely formulation of anti-infection plans or adjustment of surgical decisions.

[0049] 3. Broad prospects for industrial application

[0050] This invention not only provides a MALDI MS solution based on inorganic nanomatrix for the early diagnosis and classification of IE, but also boasts strong compatibility—it is directly compatible with existing mass spectrometry diagnostic platforms and can be implemented in clinical settings without large-scale equipment modifications. With the increasing demand for rapid diagnostic technologies in precision medicine, this invention can significantly improve the diagnostic efficiency and accuracy of IE, possessing excellent clinical translational potential and industrialization prospects. It is expected to provide key technical support for the early intervention and precision treatment of IE, and will play a significant role in promoting future clinical diagnostic practices.

[0051] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A combination of serum metabolic markers for diagnosing infective endocarditis, characterized in that, The serum metabolic marker combination includes 2-ketobutyric acid, urea, aminoacetone, pyroglutamic acid, and ornithine.

2. A method for screening serum metabolic markers for the diagnosis of infective endocarditis, characterized in that, Includes the following steps: Step 1: Metabolic detection was performed on serum samples from the IE group and the healthy control group using matrix-assisted laser desorption / ionization mass spectrometry to obtain serum metabolic fingerprint profiles; The method for obtaining the serum metabolic fingerprint is as follows: 1) Iron nanoparticles were synthesized using a hydrothermal method and dissolved in water to prepare a matrix solution with a concentration of 1 mg / mL; 2) Spot 1 μL of serum sample diluted 10 times onto the mass spectrometry target plate, let it air dry at room temperature, add 1 μL of matrix solution, and let it air dry again at room temperature to obtain the serum sample for mass spectrometry analysis; 3) Serum samples were analyzed by matrix-assisted laser desorption / ionization time-of-flight mass spectrometry to obtain serum metabolic fingerprints; Step 2: Perform data preprocessing on the serum metabolic fingerprint spectrum, and collect the m / z signal intensity matrix corresponding to each serum sample to obtain serum metabolic characteristics; Step 3: Divide the serum metabolic characteristics collected from the IE group and the healthy control group into training set and test set. On the training set, use a variety of machine learning algorithms to train the model. Evaluate each model through five-fold cross-validation and select the full spectrum diagnostic model. Step 4: Perform a t-test on all 146 features of the training set samples from Step 3 to screen them. p Forty-four m / z features with an m / z value < 0.05 were selected. Five of the most discriminative metabolic features were then optimized and selected using a greedy algorithm. These features were compared with those in the human metabolome database to identify the corresponding serum metabolic biomarkers. The resulting serum metabolic biomarkers are: 2-ketobutyric acid, urea, aminoacetone, pyroglutamic acid, and ornithine.

3. The method for screening serum metabolic markers for the diagnosis of infective endocarditis as described in claim 2, characterized in that, In step two, the data preprocessing includes spectral smoothing, baseline correction, and peak alignment.

4. The method for screening serum metabolic markers for the diagnosis of infective endocarditis as described in claim 2, characterized in that, In step three, the machine learning algorithm includes one or more of ridge regression, random forest, and LASSO regression.

5. The method for screening serum metabolic markers for the diagnosis of infective endocarditis as described in claim 2, characterized in that, In step three, the machine learning algorithm used is ridge regression.

6. The method for screening serum metabolic markers for the diagnosis of infective endocarditis as described in claim 2, characterized in that, In step four, the m / z values ​​corresponding to the five metabolic biomarkers are as follows: 2-ketobutyric acid was detected in the form of hydrogen ions with an m / z of 103.04; urea was detected in the form of disodium hydrogen ions with an m / z of 105.00; aminoacetone was detected in the form of disodium hydrogen ions with an m / z of 118.02; pyroglutamic acid was detected in the form of hydrogen ions with an m / z of 130.05; and ornithine was detected in the form of disodium hydrogen ions with an m / z of 177.

06.

7. A rapid detection method for infective endocarditis based on the method of claim 2, characterized in that, An IE standard detection model is established using machine learning algorithms based on the five serum metabolic biomarkers. The serum metabolic fingerprint of the target serum is input into the IE standard detection model to output the detection results.

8. The rapid detection method for infective endocarditis as described in claim 7, characterized in that, The machine learning algorithm used is LASSO regression.

9. A method for identifying streptococcal infective endocarditis based on the method of claim 2, characterized in that, Includes the following steps: S1. Using all 146 serum metabolic features of streptococcal and non-streptococcal IE samples as input data, an initial classification model was constructed using various machine learning algorithms. S2. Based on the best-performing LASSO regression model, extract the contribution scores of each serum metabolic feature output by the model. The screening threshold was set to LASSO regression contribution score > 0.

3. 146 initial serum metabolic features were screened, and features with scores below the threshold were removed, while features with high contribution were retained. S3. Based on the top 10 serum metabolic features after screening, construct a classifier; S4. Input the metabolic fingerprint of the serum to be tested into the classifier and output the detection results.

10. The method for identifying streptococcal infective endocarditis as described in claim 9, characterized in that, In step S1, the machine learning algorithm is ridge regression, random forest, LASSO regression, or neural network.

Images