Application of phenylbutyric acid, valeramide or salicylic acid as biomarker for early screening and diagnosis of lung cancer
By detecting the differential expression of phenbutyric acid, valeramide or salicylic acid in exhaled condensate and establishing a machine learning model, the problem of false positive and invasive risks of existing lung cancer diagnosis methods is solved, and efficient and non-invasive early-stage lung cancer screening and diagnosis is achieved.
Patent Information
- Application Number
- CN202510500629.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-21
- Publication Date
- 2025-06-27
AI Technical Summary
Existing lung cancer diagnosis methods have the potential risks of false positives, overdiagnosis and invasiveness, and it is difficult to effectively screen and diagnose early.
The differential expression of phenylbutyric acid, valeramide or salicylic acid was found in the exhaled condensate by using high-performance liquid-mass spectrometry combination detection method, and early lung cancer screening model was established through machine learning methods to improve the accuracy of diagnosis.
By detecting specific metabolites in exhaled condensate, an early stage lung cancer screening model with high prediction accuracy was established, which reduced the false positive rate and the risk of overdiagnosis and treatment, and provided a non-invasive auxiliary diagnostic method.
Smart Images
Figure CN120214167A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lung cancer biomarkers, and particularly to the application of phenylbutyric acid, valeramide or salicylic acid as biomarkers for early screening and diagnosis of lung cancer. Background Art
[0002] Lung cancer is the leading cause of morbidity and mortality of malignant tumors. Early screening is the primary strategy to reduce the mortality rate of lung cancer. Early detection and treatment can significantly improve the prognosis of lung cancer patients. At present, the diagnostic methods for lung nodules adopted clinically generally have potential risks of false positives, overdiagnosis and invasiveness. Currently, clinical diagnosis of lung cancer mostly relies on repeated low-dose computed tomography (LDCT) or invasive examinations, which have difficulties in differentiating benign / malignant nodules, overdiagnosis and potential risks of invasiveness.
[0003] Exhaled breath condensate (EBC) refers to the liquid substance formed by the condensation of exhaled gas during quiet breathing, which contains volatile and non-volatile compounds captured and diluted by water vapor condensation. EBC detection is a non-invasive method for detecting biomarkers in the lower respiratory tract, with the advantages of being simple and easy to operate, non-invasive, subject-friendly, and repeatable. EBC contains various substances such as volatile organic compounds, small molecule metabolites, polypeptides and nucleic acids, providing a non-invasive detection sample for the screening of early lung cancer biomarkers, and having great research value and clinical application potential. Summary of the Invention
[0004] The purpose of the present invention is to provide the application of phenylbutyric acid, valeramide or salicylic acid as biomarkers for early screening and diagnosis of lung cancer. The present invention first discovered phenylbutyric acid, valeramide and salicylic acid in exhaled breath condensate by high performance liquid chromatography-mass spectrometry (HPLC-MS) detection method, and they showed differential expression in early lung cancer and were correlated with the pathological characteristics of lung cancer. And an early lung cancer screening model based on machine learning method was established, and the area under the ROC curve (AUC) was 0.96, showing high prediction accuracy.
[0005] In order to achieve the above invention purpose, the present invention provides the following technical solutions:
[0006] The present invention provides a biomarker for early screening and diagnosis of lung cancer, and the biomarker combination is one of phenylbutyric acid, valeramide and salicylic acid.
[0007] Preferably, the expression of valeramide is up-regulated in the exhaled breath condensate of lung cancer patients; the expressions of phenylbutyric acid and salicylic acid are down-regulated in the exhaled breath condensate of lung cancer patients.
[0008] Preferably, the valeramide is positively correlated with the stage, type and diameter of lung cancer tumors; phenylbutyric acid is negatively correlated with the stage, type and diameter of lung cancer tumors; salicylic acid is negatively correlated with the stage and type of lung cancer tumors and positively correlated with the diameter of lung cancer tumors.
[0009] The present invention also provides the use of the biomarker combination in the preparation of a detection reagent for early screening and diagnosis of lung cancer.
[0010] Preferably, the detection reagent is a kit.
[0011] The present invention also provides a kit for early screening and diagnosis of lung cancer, comprising the biomarker combination.
[0012] Compared with the prior art, the present invention has the following beneficial effects:
[0013] The present invention uses liquid chromatography-mass spectrometry technology to carry out metabolomic detection and analysis on EBC samples, discovers metabolic pathways related to early lung cancer, and determines the expression differences of valeramide, phenylbutyric acid and salicylic acid in EBC of lung cancer patients and healthy people.
[0014] In addition, the present invention also determines through pathological correlation analysis that phenylbutyric acid, valeramide, and salicylic acid are correlated with tumorigenesis and development, and establishes an early lung cancer screening model based on logistic regression, support vector machine, Ridge Regression, Lasso regression, Elasticnet Regression, Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and LightGradientBoosting Machine (Lightboost). After comparison with the ROC curve and confusion matrix, the optimal detection model is established and verified. The results show that the AUC is above 0.91. Among them, the recall rate of the Ridge and Lasso regression models is 100%, the positive predictive value is 82.4%, and the AUC score is 0.96, showing good prediction accuracy.
[0015] The above metabolites are correlated with tumor stage, tumor type and tumor diameter. Among them, valeramide is positively correlated with tumor stage, type and diameter; phenylbutyric acid is negatively correlated with tumor stage, type and diameter; salicylic acid is negatively correlated with tumor stage and type and positively correlated with tumor diameter, and can be used to assist in the clinical diagnosis of lung cancer. Description of the Drawings
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on the provided drawings.
[0017] Figure 1 OPLS-DA analysis of metabolic ions for the tumor group (red) and the healthy control group (green).
[0018] Figure 2 Volcano plot analysis of metabolic ions for the tumor group and the healthy control group.
[0019] Figure 3 Pathway enrichment map of differential metabolites involved; A is KEGG enrichment analysis; B is RaMP-DB enrichment analysis.
[0020] Figure 4 Expression of five differential metabolites; A is benzoic acid; B is valeramide; C is betaine; D is phenylbutyric acid; E is salicylic acid.
[0021] Figure 5 ROC curve of the single-factor regression model for a single differential metabolite; A is benzoic acid; B is phenylbutyric acid; C is valeramide; D is betaine; E is salicylic acid.
[0022] Figure 6 ROC curve and confusion matrix of the SVM model.
[0023] Figure 7 ROC curve and confusion matrix of the linear model; A is the Ridge model; B is Lasso regression; C is the Elasticnet model.
[0024] Figure 8 ROC curve and confusion matrix of the integrated model; A is the RF model; B is the XGBoost model; C is the Lightboost model. Specific embodiments
[0025] The following will elaborate on the technical solutions provided by the present invention in combination with the embodiments, but they cannot be construed as limiting the protection scope of the present invention.
[0026] Embodiment 1: Screening of biomarkers
[0027] Sample collection: A total of 75 patients were recruited who visited Peking University First Hospital from February 2022 to December 2023, were diagnosed with solid pulmonary nodules, highly suspected of early lung cancer clinically, and were scheduled for surgical resection or non-surgical biopsy (bronchoscopic biopsy, transthoracic needle biopsy). The control group consisted of 84 healthy individuals who underwent physical examinations at the hospital's physical examination center from September 2022 to December 2023. Patient information was completed through hospital medical records, etc. (Table 1). There were no statistically significant differences in age, gender, and smoking history between the tumor group and the healthy group (Table 2).
[0028] Table 1 Summary of pathological information of lung cancer patients and healthy individuals
[0029]
[0030] Table 2 Baseline characteristics
[0031]
[0032] Detection sample: Exhaled breath condensate (EBC) of the subjects. The condensate was collected after the patients and healthy individuals breathed calmly for 10 minutes. The supernatant was aspirated after centrifugation at 3,000 rpm for 10 minutes. The samples were aliquoted and stored in a -80°C refrigerator.
[0033] Sample preparation: 400 μL of extraction solution (methanol:acetonitrile = 1:1 (v / v)) was added to the sample, and the sample was extracted by shaking at a speed of 1500 rpm at 10°C for 15 minutes. The sample was placed at -20°C for 20 minutes, and then centrifuged at a centrifugal force of 18000g at 4°C for 20 minutes. The supernatant was taken, dried under nitrogen, and then re-dissolved in 100 μL of acetonitrile aqueous solution (acetonitrile:water = 1:1). The mixture was shaken and mixed evenly at a speed of 1500 rpm at 10°C for 5 minutes, and the supernatant was transferred to an injection vial after centrifugation for 10 minutes.
[0034] Sample detection: An LC-MS / MS system combining ultra-high performance liquid chromatography and mass spectrometry was used. The chromatographic column was an ACQUITY UPLC HSS T3 column (2.1×100 mm, 1.8 μm); mobile phase A was 0.1% formic acid water, and mobile phase B was 95% acetonitrile + 5% water (containing 0.1% formic acid). The flow rate was 0.4 mL / min, the injection volume was 3 μL, and the elution was performed in a gradient manner.
[0035] Mass spectrometry data acquisition parameters: The first-stage resolution was 60000, the second-stage resolution was 30000, the electrospray voltage was 3.5 kV, the ion transfer tube temperature was 300°C, and the m / z scanning range was 7-1050.
[0036] Data analysis: In this invention, Thermo Fisher Xcalibur software was used to extract metabolite-related ions with relative ion signal intensities ranging from 0.5% to 100%, and XCMS software was used to complete the preprocessing of the original data. MetaboAnalyst 4.0 software was used to perform model discriminant analysis, model goodness-of-fit analysis, and differential metabolite ion analysis on the samples. Orthogonal partial least squares discriminant analysis (OPLS-DA) was used to perform discriminant analysis on the dataset. Characteristic metabolite ions with |Log2 FC| > 2 (P < 0.05) compared to the control group were screened out.
[0037] Online databases mzCloud (https: / / www.mzcloud.org) and HMDB database (http: / / www.hmdb.ca) were used to qualitatively analyze the characteristic metabolite ions. First, the differential metabolites were preliminarily determined by the mass-to-charge ratio of the first-order mass spectrometry (error < 10 ppm). The 243 screened differential ions were subjected to second-order mass spectrometry detection. MetaboAnalyst was used to perform t-test analysis, heat map analysis, visualization analysis, and metabolic pathway enrichment analysis on the differential metabolites.
[0038] Research results: As Figure 1 and Figure 2 shown, the OPLS-DA analysis showed that the tumor group and the healthy control group were separately distributed without overlap, and there were obvious differences between the two clusters (Q2: 0.78; R2: 0.787). Differential ions with |Log2FC| > 1 and p < 0.01 were screened out, and it was found that the tumor group had 142 upregulated differential metabolite ions (red dots) and 101 downregulated differential metabolite ions (blue dots). The 243 screened differential ions were subjected to second-order mass spectrometry detection, and more than 170 metabolites were obtained through data comparison.
[0039] KEGG and RaMP-DB pathway enrichment analysis of the differential metabolites found that the main KEGG pathways involved in the differential metabolites were starch and sucrose metabolism, galactose metabolism, glycine, serine, and threonine metabolism, phenylalanine metabolism pathway, etc. Figure 3 A).
[0040] The RaMP-DB signaling pathways involved in the differential metabolites included metabolic activities related to small molecule transport such as transmembrane transport mediated by solute carriers (SLC), transmembrane transporter disorders, bile salts and organic acids, metal ion and amine compound transport, etc., and also included molecular pathways related to tumors such as base excision repair, DNA methylation, heat shock response regulation mediated by heat shock transcription factor 1 (HSF1), and DNA repair. Figure 3B). Finally, five metabolites with significantly differential expression in lung cancer were found by comparing with standard samples, namely benzoic acid and valeramide with up-regulated expression in the lung cancer group, and phenylbutyric acid, betaine and salicylic acid with down-regulated expression in the lung cancer group( Figure 4 ).
[0041] Implementation Case 2 Biomarkers for Establishing an Early Screening Model for Lung Cancer
[0042] Pathological correlation analysis: The univariate logistic regression method was used to analyze the correlation between the above differential metabolites and the patients' clinical data (including clinical and pathological staging, typing) and radiomics features (including nodule size in CT).
[0043] Establishment of a single metabolite prediction model: Univariate logistic regression analysis was performed on the above 159 samples, the ROC curve was plotted, the area under the curve (AUC) was calculated, and its sensitivity and specificity were analyzed. Results with P<0.05 were considered statistically significant.
[0044] Establishment of a metabolite combination prediction model: Multivariate logistic regression analysis was performed on 159 samples to establish an early lung cancer prediction model based on a combination of 2 to 4 biomarkers. The main evaluation indicators were accuracy, sensitivity, positive predictive value and AUC value.
[0045] Establishment of a 5-metabolite combination model for early screening of lung cancer: The patients in Implementation Case 1 were randomly divided into a training set (123 cases) and a validation set (36 cases) at a ratio of 8:2. The leave-one-out cross-validation method (LOOCV) was used to construct a lung cancer differential diagnosis model in combination with 7 machine learning algorithms, including SVM, Ridge, Lasso, Elasticnet, RF, XGBoost, and Lightboost. 5-fold cross-validation was used to screen the optimal parameters of the model. The ROC curve and confusion matrix were plotted, and the main evaluation indicators were accuracy, sensitivity, positive predictive value and AUC.
[0046] Statistical methods: Continuous variables with normal distribution were expressed as mean ± standard deviation (SD), and non-normal distribution data were expressed as median [Q25, Q75]; non-categorical data were summarized as counts. Student's t-test, Kruskal-Wallis analysis of variance (ANOVA) or Wilcoxon signed-rank test were used for continuous variables, and χ2 test was used for comparison of categorical variables of clinical and imaging features. In all analyses, a two-tailed test with P<0.05 was considered statistically significant.
[0047] Research results: Correlation analysis was performed on the clinicopathological information including AJCC stage (I, II), tumor type (lung squamous cell carcinoma, lung adenocarcinoma, others), and tumor diameter with the above 5 differential metabolites. Univariate analysis found that the potential risk factors for tumor stage included benzoic acid (OR = 1.000002, 95% CI: 1.000001 - 1.000003, P = 0.0016), valeramide (OR = 1.000003, 95% CI: 1.000001 - 1.000004, P < 0.001), phenylbutyric acid (OR = 0.999967, 95% CI: 0.999945 - 0.999986, P = 0.0012), betaine (OR = 0.999956, 95% CI: 0.999905 - 0.999983, P = 0.0175), and salicylic acid (OR = 0.999971, 95% CI: 0.999946 - 0.99999, P = 0.0101).
[0048] The potential risk factors for tumor type included benzoic acid (OR = 1.000002, 95% CI: 1.000001 - 1.000003, P < 0.001), valeramide (OR = 1.000003, 95% CI: 1.000002 - 1.000004, P < 0.001), phenylbutyric acid (OR = 0.999966, 95% CI: 0.99995 - 0.99998, P < 0.001), betaine (OR = 0.999959, 95% CI: 0.999927 - 0.999979, P = 0.00147), and salicylic acid (OR = 0.999976, 95% CI: 0.99996 - 0.99999, P = 0.00218).
[0049] The potential risk factors for tumor diameter included benzoic acid (OR = 1.000002, 95% CI: 1.000001 - 1.000003, P < 0.001), valeramide (OR = 1.000002, 95% CI: 1.000001 - 1.000003, P < 0.001), phenylbutyric acid (OR = 0.999953, 95% CI: 0.999935 - 0.99997, P < 0.001), betaine (OR = 0.999943, 95% CI: 0.99989 - 0.999972, P = 0.00408), and salicylic acid (OR = 1.000004, 95% CI: 1.000002 - 1.000005, P < 0.001).
[0050] The results suggest that benzoic acid and valeramide are positively correlated with the stage, type, and diameter of tumors; phenylbutyric acid and betaine are negatively correlated with the stage, type, and diameter of tumors; salicylic acid is negatively correlated with the stage and type of tumors and positively correlated with the tumor diameter.
[0051] Single-factor logistic regression models were constructed for each of the five metabolites, and ROC curve analysis was performed ( Figure 5 ). Among them, betaine (AUC = 0.996) and valeramide (AUC = 0.904) had better accuracy, the AUC of benzoic acid was 0.738, the AUC of phenylbutyric acid was 0.779, and the AUC of salicylic acid was 0.655.
[0052] Furthermore, the present invention performed a multi-factor logistic regression analysis on 159 samples to establish an early lung cancer prediction model based on combinations of 2 to 4 markers. The main evaluation indicators were accuracy, sensitivity, positive predictive value, and AUC. As shown in Table 3, it can be seen that the various markers provided by the present invention can be used alone or in combination, and all have good sensitivity, accuracy, positive predictive value, and AUC scores.
[0053] Table 3 Prediction model scores of multi-factor logistic regression for marker combinations
[0054]
[0055]
[0056] Furthermore, seven machine learning methods were used to construct an early lung cancer prediction model based on combinations of five metabolomes, and internal validation was performed. The validation set scores are shown in Table 4. The accuracy of the SVM model (65%) was the lowest ( Figure 6 ), and the accuracies of the three models of linear learning (Ridge, Lasso, Elasticnet) were the highest (80%, 91.7%, 91.7%) ( Figure 7 ). The accuracies of the three models of ensemble learning (RF, XGBoost, Lightboost) were the next highest (88.9, 80.6, 86.1) ( Figure 8 ). From the perspective of the AUC scores, the three models of linear learning had the highest AUC scores, all of which were 0.96. The results suggest that Lasso and Elasticnet performed the best and there was no overfitting phenomenon, making them the optimal models.
[0057] Table 4 Model scores of machine learning models for validation set validation
[0058]
[0059]
[0060] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.
Claims
1. A biomarker for early screening and diagnosis of lung cancer, characterized in that: The biomarker combination is one of phenylbutyric acid, valeramide, and salicylic acid.
2. The marker combination according to claim 1, characterized in that: The expression of valeramide is up-regulated in the exhaled breath condensate of lung cancer patients; the expression of phenylbutyric acid and salicylic acid is down-regulated in the exhaled breath condensate of lung cancer patients.
3. The marker combination according to claim 1 or 2, characterized in that: The valeramide is positively correlated with the stage, type and diameter of lung cancer tumors; phenylbutyric acid is negatively correlated with the stage, type and diameter of lung cancer tumors; salicylic acid is negatively correlated with the stage and type of lung cancer tumors, and positively correlated with the diameter of lung cancer tumors.
4. Use of the biomarker combination according to any one of claims 1 to 3 in the preparation of a detection reagent for early screening and diagnosis of lung cancer.
5. The use according to claim 4, characterized in that: The detection reagent is a kit.
6. A kit for early screening and diagnosis of lung cancer, characterized in that: The method comprises the biomarker combination according to any one of claims 1 to 3.