The application of biomarkers in the assessment of pancreatic cancer

Abstract

The present invention relates to the application of biomarkers in the assessment of pancreatic cancer, and specifically to the application of biomarker detection reagents in the preparation of compositions or kits for the assessment, diagnosis or monitoring of pancreatic cancer, wherein the biomarkers include microRNAs, and the microRNAs include any one or a combination of at least two of 15 microRNAs such as hsa-let-7b-5p. The biomarkers provided by the present invention have high sensitivity and high specificity in distinguishing healthy people, pancreatitis patients and pancreatic cancer patients, reduce the false positive rate of pancreatic cancer diagnosis, and have broad application prospects.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, and in particular to the application of biomarkers in the assessment of pancreatic cancer, and specifically to the application of biomarker detection reagents in the preparation of compositions or kits for the assessment, diagnosis or monitoring of pancreatic cancer. Background Art

[0002] Pancreatic cancer is a malignant tumor of the digestive tract with high morbidity and mortality, of which more than 90% are pancreatic ductal adenocarcinoma. Due to the special anatomical location of the pancreas, the early onset is asymptomatic and the patient is already in the advanced stage when diagnosed. However, only less than 20% of advanced patients can be treated with surgical intervention. Therefore, the prognosis of pancreatic cancer is poor, and the five-year survival rate of patients is less than 9%. If it can be detected early, the five-year survival rate of pancreatic cancer patients can be increased to 30% to 40%. At the same time, some pancreatic masses are difficult to distinguish from benign pancreatic cysts or chronic pancreatitis, and chronic pancreatitis can increase the risk of pancreatic ductal adenocarcinoma. Therefore, how to accurately and efficiently differentiate pancreatic cancer from the interference of healthy people and pancreatitis populations is of great clinical significance.

[0003] At present, the diagnostic methods for pancreatic ductal adenocarcinoma include chest imaging, endoscopic ultrasound-guided fine needle aspiration biopsy (EUS-FNA), and carbohydrate antigen 19-9 (CA19-9) detection, but the detection effects of these methods are all limited. Chest imaging examination methods include CT, magnetic resonance imaging (MRI), and PET-CT, but imaging detection methods have poor sensitivity for detecting substantial lesions. For example, the sensitivity of MRI for detecting substantial lesions is only 22%, and the sensitivity for detecting cystic lesions less than 1 cm is only 82%. EUS-FNA is an invasive detection method that is prone to infection, bleeding, and tumor spread. At the same time, the false positive rate is high, which will cause the resected histological examination results to show benign lesions. CA19-9 is a blood-based tumor marker that is widely used for recurrence monitoring of pancreatic cancer. Studies have shown that the sensitivity and specificity of CA19-9 for the diagnosis of pancreatic cancer and benign pancreatic diseases are poor, at 78.2% and 82.8%, respectively, and cannot be used as an effective early diagnosis tumor marker. Therefore, there is an urgent need to develop minimally invasive early diagnostic tumor markers to assist in the differential diagnosis of pancreatic cancer.

[0004] Small extracellular vesicles (sEVs) are small vesicles with a diameter of about 30-150nm, also known as exosomes. Exosomes have received extensive attention and in-depth research as an important form of liquid biopsy due to their rich contents and ability to reflect their normal and abnormal physiological states. Studies have shown that miRNA derived from blood small extracellular vesicles is a potential tumor marker for early diagnosis of pancreatic cancer. The study by Shiwei Guo et al. showed that there were some differences in the expression of plasma small extracellular vesicle miRNA in patients with pancreatic ductal adenocarcinoma vs chronic pancreatitis. The average sensitivity and specificity of the pancreatic cancer diagnostic model constructed based on hsa-miR-95-3p / hsa-miR-26b-5p were 84.1% and 96.6%, respectively. Combining the diagnostic model with serum CA19-9 can increase the average sensitivity of the pancreatic cancer diagnostic model to 96.5%. However, there are few studies using healthy people and people with chronic pancreatitis as control groups, and the diagnostic models for pancreatic cancer and non-pancreatic cancer discovered based on chips or RT-PCR technology are ineffective, and are unable to detect unknown miRNAs in small extracellular vesicles, resulting in a lack of efficient blood small extracellular vesicle miRNA biomarkers for early diagnosis of pancreatic cancer.

[0005] CN103361424A discloses a method for diagnosing whether a subject suffers from pancreatic cancer or is at risk of developing pancreatic cancer, comprising measuring the level of at least one miR gene product in a test sample from the subject, wherein a change in the level of the miR gene product in the test sample compared to the level of the corresponding miR gene product in the control sample indicates that the subject suffers from pancreatic cancer or is at risk of developing pancreatic cancer. However, the patent is mainly based on the difference in expression of specific miRNAs in pancreatic cancer cells for identification, and the diagnostic effect is unknown.

[0006] In summary, how to develop a highly sensitive and specific diagnostic biomarker for diagnosing pancreatic cancer has become an urgent problem to be solved in this field. Summary of the invention

[0007] To solve the above technical problems, the present invention provides a biomarker for the diagnosis of pancreatic cancer and its application, as well as the application of the biomarker detection reagent in the preparation of a composition or kit for evaluating, diagnosing or monitoring pancreatic cancer. By discovering and verifying a highly efficient miRNA biomarker and establishing a highly efficient diagnostic model in combination with serum CA19-9 levels, the accuracy of early diagnosis of pancreatic cancer is improved, the false positive rate of pancreatic cancer diagnosis is reduced, and the survival time of pancreatic cancer patients is prolonged.

[0008] To achieve this object, the present invention adopts the following technical solutions:

[0009] In a first aspect, the present invention provides the use of a biomarker detection reagent in the preparation of a composition or kit for evaluating, diagnosing or monitoring pancreatic cancer, wherein the biomarker includes microRNA; the microRNA includes any one of hsa-let-7b-5p, hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-140-3p, hsa-miR-142-5p, hsa-miR-144-3p, hsa-miR-150-5p, hsa-miR-200a-3p, hsa-miR-223-3p, hsa-miR-24-2-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p or hsa-miR-95-3p, or a combination of at least two thereof.

[0010] The present invention provides a combination of microRNAs as a high-efficiency biomarker for diagnosing pancreatic cancer, chronic pancreatitis and healthy individuals, thereby improving the accuracy of pancreatic cancer diagnosis.

[0011] Preferably, the microRNA includes any one of hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p and hsa-miR-95-3p, or a combination of at least two of them.

[0012] Preferably, the microRNA includes hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p.

[0013] Preferably, the biomarkers also include CA19-9.

[0014] The carbohydrate antigen CA19-9 is a mucin-type carbohydrate protein tumor marker, a glycolipid on the cell membrane, a Lewis blood group antigen tumor marker, and exists in the form of salivary mucin in serum. The present invention further improves the accuracy of early diagnosis of pancreatic cancer and reduces the false positive rate of pancreatic cancer diagnosis by combining microRNA markers with CA19-9.

[0015] Preferably, the biomarker is used to distinguish between healthy people, pancreatitis patients and pancreatic cancer patients.

[0016] Preferably, the microRNA is derived from small extracellular vesicles.

[0017] Preferably, the small extracellular vesicles are derived from any one or a combination of at least two of blood, serum, serosal fluid, plasma, lymph, urine, cerebrospinal fluid, saliva, vaginal secretions, milk, tears or ascites.

[0018] Preferably, the microRNA is derived from small extracellular vesicles in blood.

[0019] Preferably, the CA19-9 is derived from serum.

[0020] Preferably, the nucleic acid sequence of the hsa-miR-486-5p includes the sequence shown in SEQ ID NO.1.

[0021] Preferably, the nucleic acid sequence of the hsa-miR-423-5p includes the sequence shown in SEQ ID NO.2.

[0022] Preferably, the nucleic acid sequence of hsa-miR-26b-5p includes the sequence shown in SEQ ID NO.3.

[0023] Preferably, the nucleic acid sequence of the hsa-miR-95-3p includes the sequence shown in SEQ ID NO.4.

[0024] Preferably, the nucleic acid sequence of the hsa-miR-122-5p includes the sequence shown in SEQ ID NO.5.

[0025] SEQ ID NO. 1:UCCUGUACUGAGCUGCCCCGAG.

[0026] SEQ ID NO. 2: UGAGGGGCAGAGAGCGAGACUUU.

[0027] SEQ ID NO. 3: UUCAAGUAAUUCAGGAUAGGU.

[0028] SEQ ID NO. 4: UUCAACGGGUAUUUAUUGAGCA.

[0029] SEQ ID NO. 5: UGGAGUGUGACAAUGGUGUUUG.

[0030] In a second aspect, the present invention provides use of the biomarker described in the first aspect in the diagnosis of pancreatic cancer.

[0031] In a third aspect, the present invention provides a composition or a kit for evaluating, diagnosing or monitoring pancreatic cancer, wherein the composition or the kit comprises a detection reagent for the biomarker described in the first aspect.

[0032] Preferably, the composition or kit comprises a reagent for detecting the content of microRNA and a reagent for detecting the content of CA19-9.

[0033] Preferably, the reagents for detecting the content of microRNA include RNA reverse transcription reagents, DNA amplification primers and amplification reagents.

[0034] Preferably, the reagent for detecting the CA19-9 content includes an antibody that recognizes CA19-9, preferably a monoclonal antibody.

[0035] In a fourth aspect, the present invention provides a diagnostic model for the risk of pancreatic cancer, using the biomarkers hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p and CA19-9 described in the first aspect, and adopting five machine learning algorithms including K nearest neighbor, random forest, extreme random tree, neural network and gradient boosting machine, presetting different hyperparameters for each algorithm, and training the machine learning algorithm with training set sample data to obtain the diagnostic model.

[0036] The present invention combines microRNA markers with CA19-9 and uses a variety of machine learning algorithms to construct diagnostic models for healthy people, pancreatitis and pancreatic cancer with high sensitivity and specificity.

[0037] Preferably, the risk scoring model of the diagnostic model is as follows:

[0038]

[0039] Where: x i represents the model's prediction result for sample i, P(x i ) is the probability of the corresponding classification of sample i predicted by the diagnostic model, where 0 represents a healthy person, 1 represents pancreatitis, and 2 represents pancreatic cancer. The classification with the highest probability is taken as the final prediction result of sample i.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention discovers and verifies highly effective miRNA biomarkers, preferably miRNAs derived from small extracellular vesicles in the blood, and simultaneously establishes a highly effective diagnostic model in combination with serum CA19-9 levels, and applies the detection reagent of the biomarker to the preparation of a composition or kit for evaluating, diagnosing or monitoring pancreatic cancer, thereby improving the accuracy of early diagnosis of pancreatic cancer, determining whether a sample has pancreatic cancer, pancreatitis or is not sick, and reducing the false positive rate of pancreatic cancer diagnosis, reducing the phenomenon of over-medicalization, alleviating the patient's additional suffering, and prolonging the survival time of pancreatic cancer patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is the result of transmission electron microscopy identification of small extracellular vesicles;

[0043] Figure 2 is the particle size distribution diagram of small extracellular vesicles;

[0044] Figure 3 This is the ROC curve diagram for the training cohort;

[0045] Figure 4 ROC curve for the validation cohort. DETAILED DESCRIPTION

[0046] To further illustrate the technical means and effects of the present invention, the present invention is further described below in conjunction with the embodiments and drawings. It should be understood that the specific implementation methods described herein are only used to explain the present invention, rather than to limit the present invention.

[0047] If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the field or the product instructions are used. If no manufacturer is specified for the reagents or instruments used, they are all conventional products that can be purchased through regular channels.

[0048] Reagents and instruments used in the following examples:

[0049] 0.45 μm centrifugal filter tube: purchased from Corning, USA, product number is CLS8163-100EA;

[0050] 0.22 μm centrifugal filter tube: purchased from Corning, USA, product number is CLS8161-100EA;

[0051] Exosome precipitant: purchased from Shanghai Siludi Biomedical Technology Co., Ltd., product number L3525;

[0052] Transmission electron microscope: purchased from Hitachi High-Technologies, Japan, product number H-7650;

[0053] Nanoparticle tracking analyzer: purchased from Malvern, UK, product model: NanoSight NS300;

[0054] miRNeasy Serum / Plasma Kit: purchased from QIAGEN, China, product number 217184;

[0055] Agilent 2100 analyzer: purchased from Agilent, USA, product number G2939B;

[0056] NEBNext Multiplex Small RNA Library Prep Set for Illumina: purchased from NEB, USA, product number is E7300L;

[0057] NucleoSpin Gel and PCR Clean-up: purchased from MACHEREY-NAGEL, Germany, product number 740609.50;

[0058] GX Touch TM HT nucleic acid analyzer: purchased from PerkinElmer, USA, product number is CLS137031.

[0059] In the following embodiments and test examples, the present invention extracts microRNA from small extracellular vesicles in blood and applies it to the determination of biomarkers. It can be understood that microRNA from other sources, such as microRNA from blood, serum, serous fluid, plasma, lymph, urine, cerebrospinal fluid, saliva, vaginal secretions, milk, tears or ascites, can be applied to the present invention.

[0060] Example 1

[0061] This embodiment provides a method for discovering biomarkers for pancreatic cancer diagnosis and establishing a diagnostic model.

[0062] (1) Establishment of training cohort and validation cohort

[0063] Blood samples were collected, and samples with pathological diagnosis of non-pancreatic ductal adenocarcinoma, no surgical operation, and severe hemolysis (hemolysis grade ≥ 5) were removed, and patients with chronic pancreatitis and pancreatic ductal adenocarcinoma were retained. In addition, blood samples from healthy people were collected, and 30% of the samples were randomly selected as the validation cohort, and the rest were used as the training cohort.

[0064] (2) Extraction and characterization of blood small extracellular vesicles

[0065] (a) Blood collection and extraction of small extracellular vesicles

[0066] Collect 10 mL of blood sample into an anticoagulant vacuum blood collection tube, place the tube in a precooled centrifuge and centrifuge at 1600 g, 4 °C for 10 min, transfer the supernatant to a new 1.5 mL centrifuge tube, centrifuge again at 16000 g, 4 °C for 15 min, and finally divide the supernatant into 1 mL per tube and store in a -80 °C refrigerator.

[0067] Take out the plasma sample, incubate it in a 37℃ water bath until it is completely thawed, centrifuge at 12000g, 4℃ for 10min, transfer 500μL of supernatant to a 0.45μm centrifugal filter tube, centrifuge at 12000g, 4℃ for 5min to filter, transfer the filtrate to a 0.22μm centrifugal filter tube, centrifuge at 12000g, 4℃ for 5min to filter again. Pipette the filtrate into a 1.5mL centrifuge tube, add 1 / 4 volume of exosome precipitant, mix well, incubate at 4℃ for 30min, then centrifuge at 4700g, 4℃ for 30min, discard the supernatant after centrifugation, and finally add 200μL PBS to suspend small extracellular vesicles.

[0068] (b) Characteristics of blood small extracellular vesicles

[0069] In order to detect the characteristics of small extracellular vesicles in the blood of pancreatic cancer, chronic pancreatitis and healthy people, the present invention uses transmission electron microscopy to detect the morphology of small extracellular vesicles and uses nanoparticle tracking analysis (NTA) to detect the particle size distribution of small extracellular vesicles.

[0070] Identification of morphological characteristics of small extracellular vesicles: First, the isolated small extracellular vesicles were resuspended in PBS, and 4% paraformaldehyde was added to fix the small extracellular vesicles, and then the small extracellular vesicles were transferred to a carbon-coated 200-mesh electron microscope copper grid. The copper grid was first washed twice with PBS, and then washed with fresh PBS containing glycine (50mM) for 3 minutes, and then washed again with fresh PBS containing 0.5% BSA for 10 minutes, and finally stained with 2% uranyl acetate. After staining, the morphology of small extracellular vesicles was observed using a transmission electron microscope.

[0071] Detection of the particle size distribution of small extracellular vesicles: First, take the separated small extracellular vesicles and dilute them with PBS to 1×10 7 -1×10 9 / mL. Subsequently, NTA was turned on and 1mL of the diluted sample was injected into the sample chamber. The 488nm excitation module was used, and the lens parameters of the camera were set. The shutter value was selected as 890, the gain value was selected as 146, and the detection threshold was set to 7. At least 200 complete tracks were analyzed and obtained for each video. Finally, the nanoparticle tracking data of small extracellular vesicles were analyzed using the NTA analysis software.

[0072] (3) Extraction and expression of blood small extracellular vesicle miRNA

[0073] (a) Blood EV miRNA was extracted using miRNeasy Serum / Plasma Kit. The specific experimental steps were in accordance with the product instructions, and then miRNA quantification and fragment distribution were detected using Agilent 2100 Bioanalyzer.

[0074] (b) Expression detection of small extracellular vesicles in blood

[0075] The library was constructed using the NEBNext Multiplex Small RNA Library Prep Set for Illumina kit. The specific experimental steps were in accordance with the product manual. The amount of miRNA loaded in each plasma sample was 100 ng, and the total volume did not exceed 6 μL. The library was constructed by connecting the 3' adapter, hybridizing the reverse transcription primer, connecting the 5' adapter, reverse transcription, and PCR amplification. The PCR product was then purified using the NucleoSpin Gel and PCR Clean-up kit, and the DNA was finally eluted with 30 μl NE buffer. DNA library quality control was performed using GX Touch TM HT nucleic acid analyzer and its supporting chips and reagents were used for sequencing on the Illumina Nova 6000 platform with a sequencing strategy of PE150 to obtain small RNA sequencing data.

[0076] (4) Sequencing data analysis process

[0077] (a) Sequencing data alignment: After removing the sequencing adapters of the small RNA sequencing data, the sequencing data were aligned to the human reference genome hg19 using BWA software, and the number of reads aligned to the miRNA was counted.

[0078] (b) miRNA annotation: miRNAs were annotated using the Gencode v25 and miRBase v21 databases, and those annotated as known mature miRNAs were retained for subsequent analysis.

[0079] (c) miRNA filtering: For the training cohort, mature miRNAs with a length of less than or equal to 30 nt and covered by at least g samples in the training cohort data with no less than 10 reads were retained for subsequent analysis, where g is the minimum number of samples in the group after the cohort was grouped; for the validation cohort, miRNAs screened by the training cohort were retained for subsequent analysis.

[0080] (d) Normalization of miRNA expression: The trimmed mean of M-values ​​(TMM) method in the limma analysis package in R language was used to normalize the miRNA expression levels of the training cohort samples and the validation cohort samples, respectively.

[0081] (5) Discovery of biomarkers

[0082] Based on the expression of miRNAs in the training cohort, the samples were grouped according to the pathological test results, and statistical and machine learning methods were used to discover miRNAs as biomarkers that can be used to distinguish healthy people, pancreatitis, and pancreatic ductal adenocarcinoma patients. The process is as follows:

[0083] (a) Training cohort grouping: Based on the results of sample pathology tests, the patient samples in the training cohort were divided into three groups: healthy subjects, chronic pancreatitis, and pancreatic ductal adenocarcinoma.

[0084] (b) Statistical methods for biomarker screening: First, the Kruskal-Wallis H test and Conover test were used to analyze the expression differences of all miRNAs among the three groups, and the mutual information (MI) between all miRNA expressions and sample groups was calculated. The miRNAs with inter-group statistical differences p value ≤ 0.1 and mutual information scores in the top 10% of all markers were screened. Then, the correlation coefficients between these miRNAs were calculated, and only the miRNAs with the largest variance were retained as candidate biomarkers among the miRNAs with correlation coefficients > 0.8 for subsequent analysis.

[0085] (c) Screening of biomarkers using machine learning methods: The miRNAs screened in step (c) were used as initial markers and were first screened using three machine learning feature screening methods.

[0086] Method 1: Evaluate the classification performance of each marker on the sample separately, and mark the markers with a performance score <0.8 as candidate discarded markers; Method 2: First, take all the initial markers as a whole to evaluate the classification performance of the sample, then randomly shuffle the expression of the markers in the sample in turn, calculate the degree of reduction in the classification performance score after the expression of each marker is shuffled, and mark the markers with a performance score reduction of <0.01 as candidate discarded markers; Method 3: First, take all the initial markers as a whole to evaluate the classification performance of the sample, and sort the markers according to their importance, then add one marker at a time starting from the most important marker and evaluate the improvement in the model score, and mark the markers with a score improvement of <0.01 as candidate discarded markers.

[0087] The intersection markers of the candidate discarded markers obtained by the above three methods were discarded to obtain the remaining 15 markers. Subsequently, five machine learning algorithms, including K-nearest neighbor (KNN), random forest, extremely randomized trees, neural network, and gradient boosting machine (GBM), were used to train and evaluate the classification effect and feature weights of the remaining markers as a whole on the training set data. Markers with feature weights <5% were marked as candidate discarded markers. If a marker was marked as discarded by all algorithms, it was not retained. Combining the above methods, 7 candidate markers were obtained for subsequent analysis.

[0088] (d) Evaluation of candidate marker combinations: The candidate markers were combined with CA19-9 for permutation and combination, and five machine learning algorithms, including K-nearest neighbor, random forest, extreme random tree, neural network, and gradient boosting machine, were used. The classification effect and feature weight of all marker combinations were trained and evaluated on the training set data using the 5-fold cross-validation method. Based on the training results, markers with low feature weights and no improvement in model classification performance were removed, and finally the five miRNA markers and CA19-9 combinations were confirmed for risk scoring model construction.

[0089] (6) Construction of risk scoring model for pancreatic cancer, healthy subjects and chronic pancreatitis

[0090] Healthy people, chronic pancreatitis and pancreatic ductal adenocarcinoma were used as classification prediction targets. The marker combination discovered in step (5) was used. Five machine learning algorithms, namely K nearest neighbor, random forest, extreme random tree, neural network and gradient boosting machine, were used. Different hyperparameters were preset for each algorithm. The machine learning algorithm was trained using the training set sample data to obtain a diagnostic model. A risk scoring model of the diagnostic model was constructed. The model formula is as follows:

[0091]

[0092] Where: x i represents the model's prediction result for sample i, P(x i ) is the probability of the corresponding classification of sample i predicted by the diagnostic model, where 0 represents a healthy person, 1 represents pancreatitis, and 2 represents pancreatic cancer. The classification with the highest probability is taken as the final prediction result of sample i. The trained model is saved in the form of a file on the hard disk. When calling the model, the sample marker expression value is input to obtain the model prediction result.

[0093] (7) Evaluation and verification of risk scoring model performance

[0094] In the training set, the 5-fold cross-validation method was used to evaluate the classification effect and feature weight of the model. In the validation set, the self-sampling method (bootstrap, 100 times) was used to evaluate the classification effect and feature weight of the model. Model evaluation and validation indicators include the area under the receiver operating characteristic curve (AUC, ranging from 0 to 1), balanced accuracy (ranging from 0 to 1), specificity (ranging from 0 to 1), and sensitivity (ranging from 0 to 1). The higher the value, the better the model classification performance.

[0095] (8) Application of risk scoring model

[0096] (a) Peripheral blood was collected from patients with suspected pancreatic cancer in clinical diagnosis, small extracellular vesicles were obtained, and the expression of biomarkers was obtained using small RNA sequencing;

[0097] (b) using the risk scoring model, obtaining the expression values ​​of blood small extracellular vesicles hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p, and serum CA19-9 expression levels, and entering the values ​​into the trained model formula and program to calculate the probability of each patient being predicted to be healthy, pancreatitis or pancreatic cancer;

[0098] (c) Select the output result with the highest probability and give the predicted results of pancreatic cancer risk for each patient.

[0099] Test Example 1

[0100] This test case obtained biomarkers for pancreatic cancer diagnosis and characterized their specificity and sensitivity.

[0101] (1) Establishment of training cohort and validation cohort

[0102] Blood samples from 107 patients were collected. Samples with pathological diagnosis of non-pancreatic ductal adenocarcinoma, no surgical operation and severe hemolysis (hemolysis grade ≥ 5) were removed. 32 patients with chronic pancreatitis and 56 patients with pancreatic ductal adenocarcinoma were retained and blood samples were collected from patients before surgery. In addition, blood samples from 32 healthy people were collected. Finally, a total of 120 samples were included in the group, 30% of which were randomly selected as the validation cohort, and the rest were used as the training cohort.

[0103] The training cohort consisted of 84 samples, including 23 samples from healthy people, 22 samples from chronic pancreatitis, and 39 samples from pancreatic ductal adenocarcinoma. The validation cohort consisted of 36 samples, including 9 samples from healthy people, 10 samples from chronic pancreatitis, and 17 samples from pancreatic ductal adenocarcinoma. Table 1 shows the clinical information of the patients, including the gender and age information of chronic pancreatitis and pancreatic ductal adenocarcinoma samples and the serum CA19-9 expression results of all samples. The analysis results showed that there was no significant difference in gender, age, serum CA19-9 expression, and the proportion of samples of different classifications between the two groups of patients.

[0104] Table 1

[0105]

[0106] (2) Extraction and characteristics of blood small extracellular vesicles

[0107] Small extracellular vesicles were extracted from the plasma of pancreatic cancer, chronic pancreatitis and healthy people, and the morphology of small extracellular vesicles was detected by transmission electron microscopy, such as Figure 1 As shown, small extracellular vesicles present a typical "horseshoe-shaped" morphology, and the NTA detection results are as follows Figure 2 As shown, the average particle size of small extracellular vesicles in representative samples is 118.4 nm, which is consistent with the particle size distribution of small extracellular vesicles.

[0108] (3) Discovery of biomarkers

[0109] The present invention uses small RNA sequencing to detect the expression level of small extracellular vesicle miRNA in the blood of pancreatic cancer patients, chronic pancreatitis and healthy people. The statistical methods Kruskal-Wallis H test and Conover test are used to analyze the expression differences of all miRNAs between pancreatic cancer patients, healthy people and chronic pancreatitis people and the correlation coefficients between miRNAs. At the same time, three machine learning feature screening methods are used to screen miRNAs that can be used to distinguish pancreatic cancer patients from healthy people and chronic pancreatitis people as biomarker candidate molecular markers, including hsa-let-7b-5p, hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-140- 3p, hsa-miR-142-5p, hsa-miR-144-3p, hsa-miR-150-5p, hsa-miR-200a-3p, hsa-miR-223-3p, hsa-miR-24-2-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p and hsa-miR-95-3p, a total of 15, were used for subsequent analysis. Subsequently, five machine learning algorithms, including K nearest neighbor, random forest, extreme random tree, neural network, and gradient boosting machine, were used to mark markers with feature weights <5% as candidate discarded markers. If a marker was marked as discarded by all algorithms, it was not retained. A total of seven markers were obtained as candidate markers, including hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p, and hsa-miR-95-3p , the last 7 candidate markers were combined with CA19-9 for permutations and combinations (a total of 255 combinations), and the markers hsa-let-7e-5p and hsa-miR-30d-5p, which had low feature weights and did not improve the model classification performance, were removed. The final marker combination was CA19-9 and 5 miRNAs (hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p, and hsa-miR-122-5p), which was used to subsequently construct a risk score model for pancreatic cancer, healthy subjects, and chronic pancreatitis.

[0110] (4) Risk scoring model for pancreatic cancer, healthy subjects, and chronic pancreatitis

[0111] In order to construct a risk scoring model for pancreatic cancer, healthy people and chronic pancreatitis, the miRNA expression data of the training cohort were used. In the training set, the classification effect and feature weight of the model were evaluated using the 5-fold cross-validation method. Combined with the results of pathological examination, a risk scoring model for pancreatic cancer, healthy people and chronic pancreatitis was constructed. The predicted results of the risk of pancreatic cancer for each patient were given according to the model formula and procedure.

[0112] The model evaluation indicators include the area under the receiver operating characteristic curve (ROC curve) (AUC, value range 0-1), balanced accuracy (value range 0-1), recall (value range 0-1), precision (value range 0-1), specificity (value range 0-1) and sensitivity (value range 0-1). The results are shown in Table 2, which are 0.96, 0.87, 0.87, 0.85, 0.86 and 0.90 respectively. The ROC curve is shown in Figure 3 The results showed that in the training cohort, this risk prediction model had a high AUC, balanced accuracy, recall, precision, specificity and sensitivity, and the model had better prediction performance.

[0113] Table 2

[0114] Algorithm Model AUC Balanced Accuracy Accuracy Recall Specificity Sensitivity Gradient Boosting Machine (LightGBM) 0.9629 0.8721 0.8514 0.8706 0.8622 0.8964 Neural Network (NeuralNetFastAI) 0.9581 0.8643 0.8677 0.8588 0.8778 0.8429 Random Forest (Gini) 0.9662 0.8617 0.8703 0.8699 0.8329 0.9250 Random Forest (Entr) 0.9649 0.8617 0.8703 0.8699 0.8329 0.9250 Extreme Random Trees (Entr) 0.9610 0.8271 0.8128 0.8346 0.8093 0.8714 Extreme Random Trees (Gini) 0.9589 0.8238 0.8089 0.8346 0.8044 0.8714 Neural Network (NeuralNetTorch) 0.9558 0.8133 0.8738 0.8449 0.7473 0.9500 K nearest neighbor (Dist) 0.8780 0.8083 0.8437 0.8338 0.7547 0.9250 K nearest neighbors (Unif) 0.8568 0.7467 0.7777 0.7868 0.6511 0.9500

[0115] (5) Application and effectiveness verification of risk scoring model

[0116] The expression values ​​of blood small extracellular vesicles hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p and serum CA19-9 expression values ​​of the validation cohort were brought into the trained model formula and program to calculate the probability of each patient being predicted to be healthy, chronic pancreatitis and pancreatic cancer. The output result with the highest probability was selected to give the predicted result of each patient's risk of pancreatic cancer.

[0117] In order to verify the effectiveness of the risk score model in predicting pancreatic cancer, the self-sampling method (bootstrap, 100 times) was used to evaluate the classification effect and feature weights of the model for the validation set. Taking the pathological test results as the true value, the model evaluation indicators include the area under the receiver operating characteristic curve (ROC curve) (AUC, value range 0-1), balanced accuracy (value range 0-1), recall rate (value range 0-1), precision rate (value range 0-1), specificity (value range 0-1) and sensitivity (value range 0-1). The results are shown in Table 3, which are 0.98, 0.89, 0.89, 0.90, 0.89 and 0.88 respectively. The ROC curve is shown in Figure 4The results showed that in the validation cohort, this risk prediction model had a high AUC, balanced accuracy, recall, precision, specificity, and sensitivity, and the model had better predictive performance.

[0118] Table 3

[0119] Algorithm Model AUC Balanced Accuracy Accuracy Recall Specificity Sensitivity Extreme Random Trees (Entr) 0.9778 0.9137 0.9082 0.8889 0.9474 0.8235 Extreme Random Trees (Gini) 0.9761 0.9137 0.9082 0.8889 0.9474 0.8235 Random Forest (Entr) 0.9741 0.9063 0.9082 0.8889 0.9474 0.8235 Gradient Boosting Machine (LightGBM) 0.9805 0.8941 0.8953 0.8889 0.8947 0.8824 Neural Network (NeuralNetFastAI) 0.9673 0.8941 0.8953 0.8889 0.8947 0.8824 Random Forest (Gini) 0.9800 0.8693 0.8769 0.8611 0.8947 0.8235 Neural Network (NeuralNetTorch) 0.9672 0.8490 0.8477 0.8056 0.9474 0.6471 K nearest neighbors (Unif) 0.8985 0.7649 0.7698 0.7500 0.7895 0.7059 K nearest neighbor (Dist) 0.9082 0.7257 0.7305 0.6944 0.7895 0.5882

[0120] In summary, the present invention has developed a biomarker for distinguishing pancreatic cancer, healthy individuals and chronic pancreatitis, the biomarker comprising blood small extracellular vesicle miRNA and CA19-9, and the detection reagent of the biomarker is applied to the preparation of a composition or kit for evaluating, diagnosing or monitoring pancreatic cancer, with high sensitivity and high specificity. The present invention also establishes a high-performance diagnostic model for the risk of pancreatic cancer, reducing the false positive rate of pancreatic cancer diagnosis.

[0121] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.

Claims

1. Use of a biomarker detection reagent in the preparation of a composition or kit for evaluating, diagnosing or monitoring pancreatic cancer, characterized in that: The biomarker includes microRNA; the microRNA includes any one of hsa-let-7b-5p, hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-140-3p, hsa-miR-142-5p, hsa-miR-144-3p, hsa-miR-150-5p, hsa-miR-200a-3p, hsa-miR-223-3p, hsa-miR-24-2-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p or hsa-miR-95-3p, or a combination of at least two thereof.

2. The use according to claim 1, characterized in that: The microRNA includes any one of hsa-let-7e-5p, hsa-miR-122-5p, hsa-miR-26b-5p, hsa-miR-30d-5p, hsa-miR-423-5p, hsa-miR-486-5p and hsa-miR-95-3p, or a combination of at least two thereof.

3. The use according to claim 1, characterized in that: The microRNAs include hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p; Preferably, the biomarker further includes CA19-9; Preferably, the biomarker is used to distinguish between healthy people, pancreatitis patients and pancreatic cancer patients.

4. The use according to claim 1, characterized in that: The microRNA is derived from small extracellular vesicles; Preferably, the small extracellular vesicles are derived from any one or a combination of at least two of blood, serum, serosal fluid, plasma, lymph, urine, cerebrospinal fluid, saliva, vaginal secretions, milk, tears or ascites; Preferably, the microRNA is derived from small extracellular vesicles in blood; Preferably, the CA19-9 is derived from serum.

5. The use according to claim 1, characterized in that: The nucleic acid sequence of hsa-miR-486-5p includes the sequence shown in SEQ ID NO.1; Preferably, the nucleic acid sequence of hsa-miR-423-5p includes the sequence shown in SEQ ID NO.2; Preferably, the nucleic acid sequence of hsa-miR-26b-5p includes the sequence shown in SEQ ID NO.3; Preferably, the nucleic acid sequence of hsa-miR-95-3p includes the sequence shown in SEQ ID NO.4; Preferably, the nucleic acid sequence of the hsa-miR-122-5p includes the sequence shown in SEQ ID NO.

5.

6. Use of the biomarker as claimed in any one of claims 1 to 5 in the diagnosis of pancreatic cancer.

7. A composition or kit for evaluating, diagnosing or monitoring pancreatic cancer, characterized in that: The composition or kit comprises a detection reagent for the biomarker described in any one of claims 1-5.

8. The composition or kit according to claim 7, characterized in that The composition or kit includes a reagent for detecting the content of microRNA and a reagent for detecting the content of CA19-9; Preferably, the reagents for detecting the content of microRNA include RNA reverse transcription reagents, DNA amplification primers and amplification reagents; Preferably, the reagent for detecting the CA19-9 content includes an antibody that recognizes CA19-9, preferably a monoclonal antibody.

9. A diagnostic model for the risk of pancreatic cancer, characterized in that: Using the biomarkers hsa-miR-486-5p, hsa-miR-423-5p, hsa-miR-26b-5p, hsa-miR-95-3p and hsa-miR-122-5p and CA19-9 described in any one of claims 1 to 5, five machine learning algorithms including K nearest neighbor, random forest, extreme random tree, neural network and gradient boosting machine are adopted, different hyperparameters are preset for each algorithm, and the machine learning algorithm is trained using the training set sample data to obtain the diagnostic model.

10. The diagnostic model according to claim 9, characterized in that The risk scoring model of the diagnostic model is as follows: Where: x i represents the model's prediction result for sample i, P(x i ) is the probability of the corresponding classification of sample i predicted by the diagnostic model, where 0 represents a healthy person, 1 represents pancreatitis, and 2 represents pancreatic cancer. The classification with the highest probability is taken as the final prediction result of sample i.

Images