Plasma protein construction precise prediction of non-small cell lung cancer meningeal metastasis risk model

Abstract

The application discloses a method for constructing a precise prediction non-small cell lung cancer meningeal metastasis risk model based on plasma proteins and application, and relates to the technical field of biological medicine. In view of the problems of NSCLC-LM late discovery and difficult diagnosis, the inventors first attempt to use Olink proteomics technology to detect specific plasma proteins of patients in a relatively non-invasive way to know whether there is a LM risk in advance, and help patients to discover and treat early. For this purpose, the inventors first collect blood samples of LM and nLM patients, then use the training set to construct a risk prediction model, and finally judge the model performance through the validation set. The results show that FGF2, CD4, IL15, CXCL12 and PDCD1LG2 5 specific proteins are involved in model construction, and the area under the curve of the cohort is above 0.8, and the sensitivity and specificity also reach a high level. The application has great significance in the field of early diagnosis and screening of NSCLC-LM.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of biological medicine, and particularly relates to a method for constructing a precise prediction model for meningeal metastasis risk of non-small cell lung cancer based on plasma proteins and application. BACKGROUND

[0002] Meningeal metastasis (LM) is one of the most serious complications of advanced non-small cell lung cancer (NSCLC), and its onset is insidious. The prognosis of patients with NSCLC-LM is extremely poor, and the survival time of patients with NSCLC-LM without treatment is only 4-6 weeks. The incidence of LM in ordinary NSCLC patients is 3%-5%, and for patients with epidermal growth factor receptor (EGFR) mutations, it is as high as 9.4%. However, autopsy series shows that the incidence of LM is about 20% or even higher, which indicates that there is a lack of early identification and diagnosis of LM.

[0003] At present, the diagnosis of LM in the clinic mainly relies on clinical symptoms, cerebrospinal fluid cytology and enhanced magnetic resonance imaging. The soft meningeal biopsy is traumatic and has high surgical difficulty, and is rarely used in clinical practice. In recent years, some new diagnostic techniques for LM have also appeared.

[0004] Clinical symptoms: The symptoms of LM generally have the following manifestations, and the involvement of the brain can have nausea and vomiting, headache, drowsiness, gait difficulty and cognitive function changes, and in the late stage, can have seizures, dysphagia and mental changes, etc. If the nerves and spinal cord are involved, specific brain nerve palsies, visual disturbances, diplopia, dysarthria, hearing loss, nerve root neck pain and cauda equina syndrome, etc. can occur. At the same time, LM often causes hydrocephalus leading to intracranial hypertension, and the clinical manifestations of increased intracranial pressure include more obvious headache when lying down and waking up, cognitive slowing, ataxia, visual decline, urinary incontinence, drowsiness, nausea and vomiting, etc. Since the symptoms of LM are not specific, it is necessary to differentiate other nervous system diseases before determining whether NSCLC-LM exists, which is one of the important reasons for the late discovery of NSCLC-LM.

[0005] Imaging examination: head enhanced magnetic resonance imaging (MRI) is an important means for diagnosing LM, and the MRI manifestations of the disease include soft meninges, dura mater, subependymal or superficial brain lesions, brain nerve enhancement and communicating hydrocephalus. Literature indicates that about 20% of LM patients have negative cerebrospinal fluid cytology, at which time the imaging results can also be used for diagnosis, but the imaging examination needs to be completed before lumbar puncture to prevent secondary inflammation caused by lumbar puncture from causing false positive soft meningeal enhancement. Compared with MRI, the sensitivity of computed tomography (CT) is lower, and it is often not used as a means to evaluate LM. It is worth noting that imaging examination cannot replace pathological results, and there are also cases where the cerebrospinal fluid cytology of some patients is positive, but the head enhanced MRI has no significant changes, and the final diagnosis still needs to be combined with the cerebrospinal fluid cytology results.

[0006] Tumor markers: Traditional serum protein markers are not specific for LM, but can be abnormally elevated in LM patients and can be used in combination for risk assessment. CYFRA21-1 (cytokeratin 19 fragment) is more sensitive in CSF than in plasma (about 47%), and elevated plasma levels are often associated with active extracranial metastases, indirectly suggesting LM risk, but the specificity is insufficient and needs to be confirmed by imaging or CSF detection. Elevated CEA (carcinoembryonic antigen) suggests high systemic metastatic load, which can be caused by LM, but false positives can occur in smokers.

[0007] Cytological analysis of CSF: CSF cytology is the gold standard for the diagnosis of LM, but the acquisition of CSF requires lumbar puncture. Although the procedure is simple, low cost, and has few serious complications, it is an invasive procedure with low popularity and success rate, and patient comfort is poor. Some literature suggests that the sensitivity of the first CSF analysis is only 50%, and the sensitivity of the second CSF analysis can be increased to 75%-85%, so it is necessary to repeat multiple sampling. Moreover, it is recommended to collect at least 10 ml of CSF for each cytological test, which makes it easier to capture malignant tumor cells. Malignant cell positivity, high protein content, low sugar concentration, and increased lymphocytes are characteristic manifestations of LM patients. It is worth noting that although CSF cytology is negative, LM cannot be ruled out if CSF tumor markers are significantly elevated. Thus, CSF cytology also has the problem of difficult sample collection and the possibility of false negative results.

[0008] Other diagnostic techniques: With the development of technology, new diagnostic techniques such as liquid biopsy have been applied to the diagnosis of NSCLC-LM. As one of the most important molecular markers in tumor prediction, circulating free DNA (cfDNA) and circulating tumor DNA (ctDNA) permeate all aspects of tumor screening, diagnosis, and treatment. There is currently no good marker for the diagnosis of LM, and LM patients are often affected by the blood-brain barrier, with low levels of ctDNA in plasma. Compared with blood, cerebrospinal fluid cell-free DNA (cfDNA) exhibits unique genetic characteristics of LM, which can more accurately determine sensitive genes and drug-resistant mutations, so the detection of cerebrospinal fluid cfDNA and ctDNA is more advantageous. Bu Hui et al. enrolled 35 NSCLC patients with LM in cerebrospinal fluid and blood samples for gene sequencing and found that ctDNA in cerebrospinal fluid was more easily detected (P = 0.003). Wu Yilong et al. found that the detection rate of ctDNA driving genes in cerebrospinal fluid of NSCLC-LM patients was 100%, significantly higher than that of other samples, and ctDNA may be an effective aid for LM diagnosis, treatment efficacy prediction, and treatment guidance. Although ctDNA technology has great value in disease diagnosis and efficacy monitoring, the application of this technology also has bottlenecks, such as in sample selection, due to the low sensitivity of ctDNA in plasma, which forces the need for lumbar puncture to collect cerebrospinal fluid specimens for detection, in addition to the inability of ctDNA to locate the source of mutations in terms of technology.

[0009] Emerging technology: Olink proteomics is a breakthrough proteomics technology that has gradually become an important part of precision medicine over the past decade with the advancement of genomics technology. Olink achieves ultra-sensitive protein detection through its unique proximity extension assay (PEA) patent technology, which enables high-throughput, low-volume, high-sensitivity, and wide dynamic range protein marker detection, especially in the detection of low-abundance proteins in body fluids. Specifically, compared to traditional enzyme-linked immunosorbent assay (ELISA), Olink technology has a sensitivity of fg / ml (such as IL-8 detection limit 30 fg / ml), which is 1000 times higher than ELISA, which has a sensitivity of only pg / ml, making it difficult to detect low-abundance proteins. Sample volume is reduced by 50-100 times, with the former requiring only 1 uL (serum / plasma) and the latter requiring 50-100 uL per target. Furthermore, Olink technology has multiple detection capabilities, with 92-3072 proteins detected at a single time, high specificity, and resistance to interference, while ELISA technology only detects a single target, has a high risk of cross-reaction, and is easily affected by enzyme activity or heterophilic antibody interference. Overall, Olink has revolutionized PEA technology, addressing the three major bottlenecks of proteomics: low abundance, multiple detection, and sample limitations, providing a foundation for precision medicine.

[0010] However, there is no report on the application of Olink proteomics technology in NSCLC-LM. SUMMARY

[0011] In view of the defects in the prior art, the inventors first attempt to use Olink proteomics technology to detect specific plasma proteins in patients to help doctors and patients know in advance whether there is a risk of LM in a relatively non-invasive way, and help patients to detect and treat early. For this purpose, the inventors collected blood samples from 34 NSCLC-LM patients and 54 nLM patients, of which 60 were used as a training set to construct a risk prediction model, and 28 were used as a validation set to judge the model performance. The results showed that 5 specific proteins were involved in model construction, and the area under the curve (AUC) of the cohort was above 0.8 (AUC = 0.84 in the training set, AUC = 0.83 in the validation set), and the sensitivity and specificity reached a high level (sensitivity = 0.78, specificity = 0.83 in the training set; sensitivity = 0.83, specificity = 0.73 in the validation set). The present application has high application value in the field of early diagnosis and screening of NSCLC-LM.

[0012] To achieve the above purpose, the present application adopts the following technical solutions:

[0013] One of the purposes of the present application is to provide a protein marker combination, which comprises FGF2, CD4, IL15, CXCL12 and PDCD1LG2.

[0014] The second purpose of the present application is to provide the use of the above-mentioned protein marker combination in the preparation of a non-small cell lung cancer meningeal metastasis diagnosis product, an early screening product, a disease monitoring product and / or a drug efficacy monitoring product.

[0015] The third purpose of the present application is to provide a screening method for the above-mentioned protein marker combination, which comprises using Olink proteomics technology.

[0016] Further, the sample source in the screening method comprises plasma.

[0017] The fourth purpose of the present application is to provide a product for the diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of non-small cell lung cancer meningeal metastasis, which comprises a reagent for detecting the above-mentioned protein marker combination.

[0018] Further, the product further comprises one or more pharmaceutically acceptable carriers or excipients.

[0019] Further, the product form comprises a kit.

[0020] The fifth object of the present application is to provide a model for diagnosing, early screening, disease monitoring and / or drug efficacy monitoring of non-small cell lung cancer meningeal metastasis, and the construction method of the model comprises using the above-mentioned protein marker combination.

[0021] Further, the model is based on Olink proteomics technology.

[0022] Further, the model is as follows:

[0023] Risk-score = -0.9988 + -0.4372 * FGF2 + 0.5168 * CD4 + -0.0111 * IL15+ 0.003 * CXCL12 + -0.0944 * PDCD1LG2.

[0024] Compared with the prior art, the present application has the following technical effects:

[0025] 1. The Olink technology is used to construct the LM diagnosis model by plasma proteins for the first time, and compared with the currently commonly used diagnosis technology, the following advantages are obtained:

[0026] Compared with clinical symptoms: from subjective experience to objective molecular early warning. Clinical symptoms have limitations, and the early symptoms of LM (headache and cranial nerve palsy) are not specific symptoms and are easily confused with brain parenchymal metastasis or paraneoplastic syndrome, with a misdiagnosis rate as high as 40%. The diagnosis model can give an objective molecular numerical evaluation of the possible size of the LM risk in the early stage of symptoms, improve the doctor's confidence in diagnosis, help patients to be discovered and treated early, and also avoid blind puncture.

[0027] Compared with head enhanced MRI: breakthrough in sensitivity and micro-metastasis detection bottleneck. The sensitivity of MRI in diagnosing LM is only 70%~80%, and it is not sensitive to micro-focal LM or diffuse dural enhancement, resulting in a high false negative rate and a bottleneck in diagnosis. The diagnosis model can detect specific molecular proteins in the blood to reflect the intracranial microenvironment state and find the possibility of LM from the microscopic molecular perspective.

[0028] Compared with tumor markers: greatly improve the sensitivity. Traditional tumor markers only provide tumor burden information and have no specificity for LM diagnosis, and the detection sensitivity is also low. The sensitivity of plasma CYFRA21-1 to LM is only 47%, which is much lower than the sensitivity of the diagnosis model.

[0029] Compared with cerebrospinal fluid cytology: the potential of non-invasive alternative gold standard. The positive rate of cerebrospinal fluid cytology in the first lumbar puncture is only 50%, and multiple punctures are often needed to improve the positive rate of detection, but multiple punctures will increase the pain, infection risk and distrust of patients. The diagnostic model only needs 1ml of plasma sample to determine the possibility of LM, although it cannot replace cerebrospinal fluid cytology as the gold standard for diagnosing LM, but it helps some patients reduce the pain of puncture.

[0030] Compared with ctDNA detection and other liquid biopsy techniques: functional complementation of genomic blind area. The positive rate of plasma ctDNA detection of LM is also low, and the dynamic range often only covers high-abundance mutations, only providing mutation profiles, but the diagnostic model can not only cover low-abundance functional proteins, but also reflect the state of tumor microenvironment, and has the potential to predict efficacy and warn of drug toxicity.

[0031] Overall, the advantages of the plasma model for traditional diagnostic models are non-invasive and high sensitivity.

[0032] For the same type of detection technology, the use of Olink technology in the diagnostic model has greatly improved the protein detection using ELISA technology, with higher sensitivity, smaller sample requirement, more detected proteins, wider dynamic range and better detection specificity, realizing the iteration of ELISA and helping the upgrade of clinical diagnostic model. The underlying technical advantage of Olink has established its core position in translational medicine.

[0033] 2. In view of the problems of NSCLC-LM discovery and diagnosis, the inventors first tried to use Olink proteomics technology to detect specific plasma proteins in patients to help doctors and patients know whether there is a risk of LM in a relatively non-invasive way, and help patients to discover and treat early. For this purpose, the inventors collected 34 NSCLC-LM patients and 54 nLM patients blood samples, of which 60 were used as a training set to construct a risk prediction model, and 28 were used as a validation set to judge the model performance. The results showed that 5 specific proteins were involved in model construction, and the area under the curve (AUC) of the cohort was above 0.8 (training set AUC = 0.84, validation set AUC = 0.83), and the sensitivity and specificity were both high (training set sensitivity = 0.78, specificity = 0.83; validation set sensitivity = 0.83, specificity = 0.73). It can be seen that the product and method in the present application have high application value in the field of early diagnosis and screening of NSCLC-LM. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 LASSO dimensionality reduction analysis chart (n=60) in Example 1 of the present application;

[0035] Figure 2 ROC curve and sensitivity, specificity level of the plasma protein prediction model in Example 1 of the present application. DETAILED DESCRIPTION

[0036] The following examples are intended to illustrate the present application but not to limit the scope of the present application. Modifications or substitutions of the method, steps or conditions of the present application, without departing from the spirit and the essence of the present application, all belong to the scope of the present application. The reagents, products and instruments used in the following examples can be obtained commercially, and the methods used in the examples are consistent with the commonly used methods unless otherwise specified.

[0037] The technical solutions of the present application are further described in detail below in combination with examples.

[0038] Example 1

[0039] I. Sample collection

[0040] 1. Sample type: plasma / serum

[0041] 2. Sample collection:

[0042] Plasma: (1) Collect at least 1 ml of whole blood in an EDTA, citric acid or heparin anticoagulant tube; (2) centrifuge at 1000-2000 x g for 10 minutes at 2-8°C, and take the supernatant; (from the collection of whole blood to centrifugation, it is performed within 1 hour at room temperature, and within 8 hours at 4°C); (3) immediately transfer the supernatant (plasma) to a clean tube (96-well plate or Eppendorf tube), and try to use sealing film to wrap on the tube to prevent sample leakage; (4) immediately detect or store at -80°C (to ensure that all samples are processed consistently, try to detect in the same batch).

[0043] Serum: (1) Collect at least 1 ml of whole blood into a serum collection tube; (2) allow the blood to clot completely at room temperature for 30-60 minutes, then centrifuge at 1000-2000 x g for 10 minutes at 2-8°C to remove the blood clot; (3) immediately transfer the serum to a clean tube (96-well plate or Eppendorf tube), and try to use sealing film to wrap on the tube to prevent sample leakage; (4) immediately detect or store at -80°C (to ensure that all samples are processed consistently, try to detect in the same batch).

[0044] 3. Sample number: 34 cases of meningeal metastasis group (LM), 54 cases of non-meningeal metastasis group (nLM), a total of 88 cases. According to the collection time sequence, the first 60 cases were used as the training set to construct the risk model, and the last 28 cases were used as the validation set for model verification (NSCLC-LM is one of the serious complications in advanced lung cancer, with low incidence and rapid disease progression. Most patients are unable to obtain timely medical intervention due to rapid disease progression, and sample collection is therefore a major challenge).

[0045] II. Detection Panel

[0046] Immuno-Oncology Panel was used: 92 kinds of immune checkpoint proteins (PD-L1, CTLA4, etc., as shown in Table 1 below).

[0047] Table 1 Details of 92 kinds of immune checkpoint proteins

[0048]

[0049] III. Sample detection

[0050] 1. Sample randomization: use the RAND function in the Excel table to output the corresponding numerical value of each sample, and then arrange the numerical values from small to large to make the samples randomly distributed in the 96-well plate.

[0051] 2. Sample sampling: according to the sample distribution table after randomization, take 10-40 μL of each sample into the 96-well plate, and add 5 μL of mixed plasma sample into wells A12 and B12, respectively, add 5 μL of Negative Control (NC) into wells C12, D12, and E12, respectively, and add 5 μL of Interplate Control (IPC) into wells F12, G12, and H12, respectively.

[0052] 3. Hybridization incubation: first prepare the hybridization mixture, thaw the reagents at room temperature, vortex and centrifuge, then take 1 nuclease-free 1.5 ml centrifuge tube, add Incubation Solution 280 μL, Incubation Stabilizer 40 μL, A-probes 40 μL, and B-probes 40 μL, vortex and centrifuge. Take out 1 8-tube, add 47 μL of hybridization mixture to each well. Take out a 96-well plate and label it as the incubation plate. Use an 8-channel pipette to add 3 μL of hybridization mixture to each well, then one by one, add 1 μL of sample to each well. After sealing with sealing film, centrifuge at 400g for 1 minute, and place in a PCR instrument. Set the infinite 4°C, record the PCR instrument insertion time, and the incubation time needs to meet 16-24 hours.

[0053] 4. Extension and amplification: Set the PCR program, 50°C for 20 minutes; 95°C for 5 minutes; 95°C for 30 seconds, 54°C for 1 minute, 60°C for 1 minute for 17 cycles; 10°C for holding. After setting, run the PCR instrument, and then press the pause key to maintain the temperature of the PCR instrument at 50°C. Prepare the extension and amplification mixture, take one 15 ml centrifuge tube, and add High Purity Water 9385 μL, PEA Solution 1100 μL, PEA Enzyme 55 μL, and PCR Polymerase 22 μL in turn, vortex and centrifuge. Take out the incubation plate from the PCR instrument and centrifuge, use an 8-channel pipettor to add 96 μL of the extension and amplification mixture to each well, vortex and centrifuge after sealing the film, and place it in the preheated PCR instrument at 50°C, and then resume the running program.

[0054] 5. Chip pretreatment: unpack, take out the chip, and check the chip and syringe for damage. Take out the syringe, press the piston hole down, and push the liquid in the syringe. Click the instrument Target 96 option, place the chip and cover plate after the instrument pushes out the sample stage, automatically return to the sample stage after clicking start, and the chip pretreatment starts.

[0055] 6. On-machine operation: take out one 1.5 ml centrifuge tube, add Detection Solution 550 μL, High Purity Water 230 μL, Detection Enzyme 7.8 μL, and PCR Polymerase 3.1 μL in turn, vortex and centrifuge. Take out one tube, and distribute 95 μL per well. Take out one 96-well plate, add 7.2 μL of the mixture to each well, and then add 2.8 μL of the product after the extension and amplification program is completed, vortex and mix, and then wait for use. Prepare the diluted sample plate, primer plate, and pretreated chip, add 5 μL of primers to the left side of the chip and 5 μL of diluted product to the right side in turn to avoid air bubbles. After the sample addition is completed, place it in the sample stage, set the running name, and click run.

[0056] 7. Data analysis and export: export the running name file after the instrument is removed, and the file suffix is “XXX.q100”. Open the Olink NPX Signature 1.5.3.0 software, import “XXX.q100”, set the reagent panel information, reagent version number, and sample name, click “OK”, and the software automatically analyzes and outputs the NPX Excel document and the QC AR PDF document.

[0057] Four, data analysis

[0058] The strategy of integrating feature selection and logistic regression modeling was adopted for data analysis. The specific analysis process is as follows:

[0059] 1. Data preprocessing: All samples were subjected to batch effect removal and standardization before modeling.

[0060] 2. Input training set and validation set: The data of the training set obtained by detection were input to construct the model, and the data of the validation set were input to validate the model.

[0061] 3. LASSO model construction and parameter optimization: In each training set, LASSO logistic regression method was used for modeling. Through 5-fold cross-validation (5-fold CV), the lambda value that minimizes the cross-validation error was selected. Then the LASSO model was trained on the entire training set using this lambda value, and applied to the validation set for prediction, the prediction probability was calculated and the AUC value on the validation set was evaluated.

[0062] 4. Model stability evaluation and feature selection: The genes selected by LASSO method in each model were counted, and their frequency was summarized. Genes with a frequency of more than 50 times (i.e. > 50%) in 100 modeling were selected as having high stability and predictive contribution.

[0063] 5. Final model construction and evaluation: With these stable feature genes as input variables, a multivariate logistic regression model was reconstructed on all samples to form the final risk score formula. Finally, the prediction probability of each sample was calculated based on the model, the ROC curve was drawn, and the AUC value was calculated to evaluate the overall predictive performance of the final model.

[0064] Five, result presentation

[0065] 1. Software and version information is shown in Table 2 below.

[0066] Table 2 Software and version information

[0067]

[0068] 2. The molecular model based on plasma proteomics can accurately predict the risk of leptomeningeal metastasis in NSCLC patients

[0069] In this study, plasma samples from non-small cell lung cancer (NSCLC) patients were subjected to proteomic detection based on the Olink platform, and a non-invasive molecular diagnostic model for predicting the risk of leptomeningeal metastasis (LM) was constructed. LASSO feature selection combined with logistic regression was used for modeling and evaluation.

[0070] 3. LASSO model construction and screening of signature proteins

[0071] The training set (divided into LM 37 cases and nLM group 23 cases) was modeled 100 times, and the gene set selected in each modeling model was found. In each modeling cycle, the features were screened by LASSO method, the genes with frequency > 50% were selected and the prediction model was established, and the optimal regularization parameter lambda (i.e. the lambda value with the minimum cross-validation error) was determined by 5-fold cross-validation, and the AUC value of the model was calculated on the corresponding validation set to evaluate the model performance.

[0072] In 100 modeling, we counted the frequency of each protein selected as a model feature to evaluate its stability under different data partitioning. According to the LASSO results, FGF2, CD4, IL15, CXCL12 and PDCD1LG2 were selected in more than 50% of the modeling, suggesting their high stability and predictive contribution in the model, and were finally included in the model to construct the final prediction formula. The protein screening process is shown in Figure 1 LASSO regression realizes feature selection and model shrinkage, and finally retains 5 key protein features to construct the prediction model.

[0073] 4. Key proteins have clear or suspected biological function basis

[0074] The five key proteins finally included in the model—FGF2, CD4, IL15, CXCL12 and PDCD1LG2—have clear or potential biological significance in tumor immune regulation, inflammatory response and blood-brain barrier crossing, supporting their feasibility as related markers of leptomeningeal metastasis. Among them:

[0075] FGF2 (fibroblast growth factor 2): has the ability to promote angiogenesis, cell migration and tissue repair, and has a clear role in vascular endothelial cell activation and neovascularization. Upregulation of FGF2 may indicate that the tumor microenvironment has stronger vascular permeability and barrier-crossing ability.

[0076] CD4: a classic helper T cell marker, plays a key role in antigen presentation and cytotoxic immune response. Decreased or disordered CD4 levels are closely related to tumor-related immune tolerance, and may reflect T cell dysfunction and failure of immune surveillance.

[0077] IL15: as an important activator of NK cells and CD8⁺ memory T cells, IL15 may reflect compensatory activation of the immune system, especially when the tumor invades the central nervous system, inducing enhanced peripheral immune response.

[0078] CXCL12: It is mainly a chemotactic factor responsible for guiding the migration of immune cells, stem cells, etc., and plays a key role in tumor metastasis, angiogenesis and shaping an immunosuppressive microenvironment.

[0079] PDCD1LG2: It is mainly an immune checkpoint inhibitory ligand that inhibits T cell function by binding to PD-1, is one of the key mechanisms for tumors to achieve immune escape, and is an important target for immunotherapy.

[0080] In summary, the above five proteins collectively depict a metastasis evolution map from peripheral immune suppression-inflammation-driven-vascular activation-barrier dissemination in biological function, providing important clues for early identification and potential mechanisms of LM.

[0081] 5. Model construction and performance evaluation

[0082] Based on the above screening results, we constructed a final logistic regression model containing five protein indicators (FGF2, CD4, IL15, CXCL12 and PDCD1LG2). The model formula is as follows:

[0083] Risk-score = -0.9988 + -0.4372 * FGF2 + 0.5168 * CD4 + -0.0111 * IL15+ 0.003 * CXCL12 + -0.0944 * PDCD1LG2

[0084] Wherein, risk-score represents the risk score of LM predicted based on the detection results of plasma samples olink proteins, and the coefficient represents the direction and size of the risk score of each protein expression. After constructing the final model, we further calculated the ROC curve based on the continuous risk score (score) output by the Logistic regression, Figure 2 A and Figure 2 C, the area under the ROC curve (AUC) is 0.844, and the best cutoff value is determined by Youden's Index (sensitivity + specificity - 1) to achieve the optimal classification effect of LM and non-LM (nLM) patients. The final results show that the best cut-off value is 0.43, and at this threshold: the sensitivity (Sensitivity) is 0.78, that is, the model can correctly identify 78% of LM patients; the specificity (Specificity) is 0.83, that is, the model can correctly exclude 83% of non-LM patients.

[0085] 6. Further verification

[0086] We further validated the risk model on a validation set (divided into LM 17 cases and nLM group 11 cases), and the results are as follows Figure 2 B and Figure 2 D. The results show that the area under the ROC curve (AUC) of the risk model in the validation set is 0.829, the sensitivity is 0.83, and the specificity is 0.73.

[0087] In summary, the model constructed in the application has high accuracy and stability in distinguishing between patients with leptomeningeal metastasis (LM) and non-metastasis (nLM).

[0088] The above-described embodiments are only preferred modes of the application and do not limit the scope of the application. Without departing from the design spirit of the application, various modifications and improvements to the technical solutions of the application made by those skilled in the art shall fall within the protection scope of the application defined by the claims.

Claims

1. A combination of protein biomarkers for the diagnosis, early screening, and / or disease monitoring of leptospirosis in non-small cell lung cancer, characterized in that, The protein biomarker combination includes FGF2, CD4, IL15, CXCL12, and PDCD1LG2.

2. The use of the protein biomarker combination of claim 1 in the preparation of diagnostic products, early screening products and / or disease monitoring products for non-small cell lung cancer leptomeningeal metastasis.

3. A method for screening combinations of protein biomarkers as described in claim 1, characterized in that, The screening method includes the use of Olink proteomics technology.

4. The screening method according to claim 3, characterized in that, The sample source in the screening method includes plasma.

5. A product for the diagnosis, early screening, and / or disease monitoring of leptomeningeal metastases in non-small cell lung cancer, characterized in that, The product contains a reagent for detecting the combination of protein biomarkers described in claim 1.

6. The product according to claim 5, characterized in that, The product also contains one or more pharmaceutically acceptable carriers or excipients.

7. The product according to claim 6, characterized in that, The product forms include reagent kits.

8. A model for the diagnosis, early screening, and / or disease monitoring of leptomeningeal metastases in non-small cell lung cancer, characterized in that, The method for constructing the model includes using the combination of protein biomarkers described in claim 1.

9. The model according to claim 8, characterized in that, The model is based on Olink proteomics technology.

10. The model according to claim 9, characterized in that, The model formula is as follows: Risk-score = -0.9988 - 0.4372 × FGF2 + 0.5168 × CD4 - 0.0111 × IL15 +0.003 × CXCL12 - 0.0944 × PDCD1LG2.

Images