CN105044361B - A kind of diagnostic marker and its screening technique for being suitable for esophageal squamous cell carcinoma early diagnosis - Google Patents

Abstract

The invention discloses a diagnostic marker suitable for early diagnosis of esophageal squamous cell carcinoma and a screening method thereof. The invention discovers 25 serum metabolic markers and 10 metabolic pathways related thereto. A diagnostic marker for the diagnosis of esophageal cancer can be obtained through the combination of these 25 serum metabolic markers. The diagnostic marker screening method of the present invention has strong operability, and the diagnostic marker can be used to construct a diagnostic model. The diagnostic model has good effect, high sensitivity and good specificity, and is not only suitable for the diagnosis of advanced esophageal cancer, but also suitable for the diagnosis of early esophageal cancer. . The diagnostic model constructed by the diagnostic markers of the present invention can be diagnosed only by taking blood, which is non-invasive and low-cost, and can well replace the current endoinvasive diagnostic model, greatly reducing the suffering of patients, and the present invention is fast and easy to diagnose. The method is convenient, takes a short time, improves work efficiency, is beneficial to early detection and early treatment of esophageal cancer, and has good clinical application and promotion value.

Description

A diagnostic marker suitable for early diagnosis of esophageal squamous cell carcinoma and its screening method

technical field

The present invention relates to the diagnosis of esophageal squamous cell carcinoma, in particular to a diagnostic marker for esophageal squamous cell carcinoma, a screening method for the diagnostic marker, a diagnostic model based on the diagnostic marker, and a method for constructing the diagnostic model. The diagnostic marker and diagnostic model have good diagnostic effects on esophageal carcinoma in situ, early and advanced esophageal cancer, are particularly suitable for early diagnosis of esophageal cancer, and belong to the technical field of diagnosis of esophageal squamous cell carcinoma.

Background technique

Esophageal cancer is a malignant lesion formed by abnormal proliferation of esophageal squamous or glandular epithelium. According to the latest data from the World Health Organization, about 400,000 people die from esophageal cancer every year in the world. my country is the country with the highest incidence and mortality rate of esophageal cancer, and 90% of patients have squamous cell carcinoma (ESCC). The onset of esophageal cancer is occult, with no symptoms or very atypical symptoms in the early stage. It is already in the late clinical stage when it is discovered, and the prognosis is generally poor (the 5-year survival rate is about 13%). Therefore, early diagnosis and early treatment are the key to improve the prognosis and reduce the mortality of esophageal cancer. At present, the platform for early diagnosis and early treatment in areas with a high incidence of esophageal cancer and commonly used clinical early screening and diagnostic methods include esophageal mesh cytology, X-ray barium meal contrast, esophageal ultrasonography, and esophageal endoscopy. However, these methods are invasive, complicated and expensive, which limits their wide application in screening and early diagnosis of esophageal cancer.

The occurrence of esophageal cancer involves a complex process of multi-factors, multi-stages, multi-gene mutation accumulation and interaction with environmental factors, including many proto-oncogenes, tumor suppressor genes and protein changes at the molecular level, as well as long-term poor life or diet The influence of habits (eating foods containing more nitrosamines, such as liking pickled sauerkraut or moldy food, long-term preference for hot food, smoking, bad habits of drinking, etc.). Metabolomics is the qualitative and quantitative detection of all small molecular metabolites (such as fatty acids, amino acids, nucleosides and steroids) in biological samples (such as serum, urine, saliva, etc.) The metabolic response of endogenous substances after the body is disturbed by diseases or accumulation of risk factors. Biological information in the body is transferred from genes to proteins through transcription, and finally manifested as small molecule metabolites. Unlike the internal differences of organisms reflected by genomics and proteomics, the research field of metabolomics extends to the interaction and interaction between the organism and the environment. Small molecule metabolites are not only the material basis of the body's life activities and biochemical metabolism, but also reflect the changes of some external factors to the metabolic environment in the body. Therefore, the differences in the concentration of some unique metabolites between different individuals actually reflect the inherent disease. manifestations and external causes. In recent years, studies have found that during the development of diseases such as metabolic diseases and malignant tumors (ovarian cancer), the basic biochemical metabolism of the body has undergone significant changes, which will play an important role in understanding the metabolic mechanisms of complex diseases. Screening and early diagnosis provide new technical methods.

The occurrence and development of esophageal cancer is caused by the interaction of multiple genes and environmental factors. First, the expression of related functional genes is changed or mutated, followed by a series of changes in cell signal transduction and protein synthesis. Metabolites change. The occurrence of esophageal cancer is the result of gradual accumulation of environmental risk factors and continuous damage to the homeostasis of various metabolic pathways in the body. At present, some people have used metabolomics to study esophageal cancer, such as Wu et al (Wu H, Xue R, Lu C, et al. Life Sci,2009,877(27):3111-7.), Zhang et al. (Zhang J, Bowers J, LiuL, et al.Esoph ageal cancer metabolite biomarkers detected by LC-MS and NMRmethods.PLoS O ne,2012,7 (1):e30181.), Xu et al. (Xu J, Chen Y, Zhang R, et al.Globaland targeted metabolomics of esophageal squamous cell carcinoma discovers potential diagnostic and therapeutic biomarkers.Mol Cell Proteomics,2013,12(5):1306- 18.) Both used metabolomics technology to study esophageal cancer.

The occurrence and development of esophageal cancer is accompanied by changes in various metabolites in the body, and it usually takes several years or even ten years. If cancer can be detected in the early stage and treated early, it can effectively improve the prognosis. In fact, studies have shown that before the onset of disease or the accumulation of risk factors, endogenous substances will make corresponding metabolic responses. For example, Zhao et al. revealed the metabolic characteristics of pre-diabetic patients through the study of metabolic fingerprints, and confirmed that the metabolic changes of fatty acids, tryptophan, uric acid, and bile acids occurred in a long period of time before the clinical symptoms of the disease appeared. Screening, early diagnosis and intervention of diseases provide new possibilities. However, the metabolomics studies of esophageal cancer mentioned above mainly included patients with advanced esophageal cancer, and most of them did not include or rarely included early-stage esophageal cancer cases. Lymph node and distant metastases have occurred in advanced esophageal cancer, and even a state of tumor cachexia has occurred. At this time, the body's metabolism has undergone great changes. Therefore, these studies can only find differences in the metabolic profile of advanced esophageal cancer compared with healthy controls. According to These differences in metabolic profiles can only better diagnose advanced esophageal cancer, but cannot diagnose early esophageal cancer, that is, the early diagnosis of esophageal cancer cannot be achieved. Secondly, the metabolomics study of esophageal cancer mentioned above only obtained a small part of metabolites related to the pathogenesis of esophageal cancer (from the research level, it is only the analysis of metabolic targets, not the analysis of metabolic profiling or metabolomics). In addition, most of the above studies did not evaluate the translational medicine potential and application effect of metabolomics from the perspective of objective molecular screening/early diagnosis standard model of esophageal cancer, and most of them did not report the screening/diagnosis of metabolites for esophageal cancer. Sensitivity, specificity, and area under the ROC curve (AUC).

The inventor has conducted a series of studies on esophageal cancer in the early stage, using metabolomics technology to study a metabolomics analysis model that can be used for early screening of esophageal cancer. This technology has application in the discovery of high-risk groups in high-incidence areas of esophageal cancer in my country and promotion value, it has been piloted in Feicheng City, Shandong Province. However, there are still many differences between the screening of high-incidence groups of esophageal cancer and early clinical diagnosis. Esophageal cancer screening in high-incidence areas is to observe healthy or apparently healthy people in high-incidence areas. The purpose is to find those who appear to be healthy but suspected of having esophageal lesions (high-risk individuals) in the healthy population, and the screening test is positive. Patients need to make further diagnosis or intervention; and diagnosis is to observe patients or suspicious patients in a clinical environment, the purpose is to distinguish whether patients have corresponding symptoms, to make timely and correct judgments on patients' conditions, so as to take corresponding and effective measures. Treatment measures, clinically positive patients should be given treatment (such as surgery, chemotherapy or radiotherapy). At present, hospitals generally use traumatic, complicated, and expensive imaging examinations to clinically diagnose esophageal cancer cases, and most of the patients who actively seek treatment are already in the advanced stage, so there is still a lack of simple and effective clinical diagnosis of esophageal cancer (especially Serum biomarkers for early esophageal cancer diagnosis. Therefore, finding specific, sensitive, economical and non-invasive serum metabolic markers for early diagnosis of esophageal cancer and establishing a safe and effective early molecular diagnosis model of esophageal cancer have important clinical application value.

Contents of the invention

Aiming at the shortcomings of esophageal squamous cell carcinoma (referred to as esophageal cancer) in the prior art, which are complex, expensive, and traumatic, and the current markers are only highly sensitive to advanced esophageal cancer, and cannot achieve early diagnosis of esophageal cancer, this paper The invention provides a diagnostic marker suitable for the early diagnosis of esophageal squamous cell carcinoma. The diagnostic marker has good sensitivity and specificity for esophageal carcinoma in situ, early esophageal cancer, and advanced esophageal cancer. It can not only be used for The diagnosis of advanced esophageal cancer can also be better used for the early diagnosis of esophageal squamous cell carcinoma, which is of great significance for improving the prognosis and reducing mortality of esophageal squamous cell carcinoma.

The present invention also provides a screening method for the diagnostic markers suitable for early diagnosis of esophageal squamous cell carcinoma. The markers obtained by this method have good sensitivity and specificity for early and advanced esophageal cancer, and are especially suitable for Early diagnosis of cancer has important clinical significance for the treatment of esophageal cancer.

The present invention also provides a diagnostic model of esophageal squamous cell carcinoma and a method for constructing the diagnostic model. The method for constructing the model is simple and can replace current invasive diagnostic methods. Carcinoma of the esophagus, early esophagus, and advanced esophagus all have good sensitivity and specificity, providing effective technical support for the early diagnosis and treatment of esophageal squamous cell carcinoma.

The present invention also provides a method for diagnosing esophageal squamous cell carcinoma using the diagnostic model. The model of the present invention can be used for diagnosis only by taking blood, which is convenient and quick without internal trauma, especially for early esophageal cancer with high sensitivity and specificity. It has good clinical application value.

At present, most of the research and screening of diagnostic markers for esophageal squamous cell carcinoma are carried out from the aspects of genes and macromolecular proteins. Based on this idea, markers that are especially suitable for early diagnosis of esophageal squamous cell carcinoma have been discovered, which makes it a good diagnostic method for early esophageal squamous cell carcinoma that is not easy to detect. The present invention relies on the esophageal cancer screening and follow-up population cohort of the "National Esophageal Cancer Early Diagnosis and Early Treatment Demonstration Base (Feicheng City, Shandong Province)" to obtain esophageal carcinoma in situ (abbreviated as carcinoma in situ, 0 stage 39 cases), early esophageal cancer (referred to as early cancer, 17 cases of stage I, 11 cases of stage II) and advanced esophageal cancer (referred to as advanced cancer, 30 cases of stage III), and randomly selected patients without any esophageal lesions and other metabolic diseases (such as hyperthyroidism) , hypothyroidism, hypertension, diabetes, kidney disease, etc.) were healthy controls, using UPLC-QTOF/MS to obtain metabolic fingerprints of 1466 small molecule metabolites, after small molecule metabolism in patients with esophageal cancer and healthy subjects The comparison and analysis of the metabolic fingerprints of esophageal squamous cell carcinoma obtained diagnostic markers suitable for the early diagnosis of esophageal squamous cell carcinoma. These diagnostic markers were used to construct a model to obtain a diagnostic model of esophageal cancer, which can be used to quickly diagnose whether it is a squamous cell carcinoma of the esophagus Esophageal cancer, especially early esophageal cancer, can be diagnosed with high sensitivity and specificity, and has clinical application and promotion value.

In the present invention, the esophageal carcinoma in situ refers to stage 0 in the TNM staging standard, which refers to atypical hyperplasia (severe) in the epithelial layer of the mucosa or in the epidermis of the skin involving the whole layer of the epithelium, but has not yet penetrated the basement membrane and descended Invasive and growing cancer; early esophageal cancer refers to stage I and II in the TNM staging standard, and refers to cancer confined to the retromucosa or submucosa without lymph node involvement and distant metastasis; advanced esophageal cancer refers to stage III in the TNM staging standard And stage IV, refers to the cancer that has involved the muscle layer or reached the adventitia or beyond the adventitia, with local or distant lymph node metastasis. The TNM staging standard is based on the American joint Committee on Cancer (AJCC) TNM Classfication of Carcinoma of the Esophagus and Esophagogastric Junction (7th ed, 2010).

The diagnostic marker and diagnostic model of the present invention can diagnose asymptomatic or insignificant early esophageal cancer without internal invasiveness, which reduces the pain of the tester, and the diagnosis process is simple and fast, which improves work efficiency. For esophageal cancer Early diagnosis and early treatment of cancer, improvement of prognosis, and reduction of mortality are all of great significance. Realize the concrete technical scheme of the present invention as follows:

A diagnostic marker suitable for the early diagnosis of esophageal squamous cell carcinoma, which is any one or a combination of more than one of the following 25 serum metabolic markers: beta-alanine-lysine (beta-Ala -Lys), L-Carnosine, cis-9-Palmitoleic acid, Palmitic acid, Oleic Acid, Lysophosphatidic acid LPA (18: 1(9Z)/0:0), Lysolecithin LysoPC(14:0/0:0), Lysolecithin LysoPC(18:2(9Z,12Z)), Lysolecithin LysoPC(24:0), Phospholipids PC(14:1(9Z)/P-18:1(11Z)), phospholipid PC(16:0/18:2(9Z,12Z)), phospholipid PC(24:1(15Z)/22:6( 4Z, 7Z, 10Z, 13Z, 16Z, 19Z)), linoleic acid (Linoleic acid), nicotinamide adenine dinucleotide (NADH), cortisol (Cortisol), L-tyrosine (L-Tyrosine) , L-Tryptophan, Glycocholic Acid, Taurocholate, Hypoxanthine, Allantoic acid, Inosine, Sphingosine 1-phosphate (Sphingosine 1-phosphate), galactosylceramide sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0), lactosylceramide (d18:1/22:0).

Among the above-mentioned diagnostic markers, it may be any one of the above-mentioned 25 serum metabolic markers, or any combination of two or more of them. When a combination of two or more serum metabolic markers is used as a diagnostic marker, the diagnostic effect will be better than that of a single serum metabolic marker as a diagnostic marker.

Further, the above-mentioned diagnostic markers can be any combination of serum metabolic markers in the following (a)-(h): (a) can be beta-alanine-lysine (beta-Ala-Lys ) and L-Carnosine; (b) or a combination of cis-9-Palmitoleic acid, Palmitic acid and Oleic Acid ; (c) or lysophosphatidic acid LPA (18:1 (9Z)/0:0), lyso-lecithin LysoPC (14:0/0:0), lyso-lecithin LysoPC (18:2 (9Z,12Z) ) and LysoPC (24:0); 9Z,12Z)), phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)) and linoleic acid (Linoleic acid); (e) or smoke A combination of NADH, L-Tyrosine, and L-Tryptophan; (f) or Cortisol, Glycocholic Acid (Glycocholic Acid) and taurocholate (Taurocholate); (g) or a combination of hypoxanthine (Hypoxanthine), allantoic acid (Allantoic acid) and inosine (Inosine); (h) or 1 -Sphingosine 1-ph osphate, galactosylceramide sulfate 3-O-Sulfogalactosylceramide(d18:1/20:0) and lactosylceramide(d18:1/22:0) combination.

Further, the above-mentioned diagnostic markers can be two or more combinations of the following 15 serum metabolic markers: lysophosphatidic acid LPA (18:1(9Z)/0:0), lysophosphatidic acid LysoPC ( 14:0/0:0), LysoPC (18:2(9Z,12Z)), LysoPC (24:0), Phospholipid PC (14:1(9Z)/P-18:1( 11Z)), phospholipid PC (16:0/18:2 (9Z, 12Z)), phospholipid PC (24:1 (15Z)/22:6 (4Z, 7Z, 10Z, 13Z, 16Z, 19Z)), smoke Amide adenine dinucleotide (NA DH), cortisol (Cortisol), L-tryptophan (L-Tryptophan), taurocholate (Taurocholate), hypoxanthine (Hypo xanthine), inosine (Inosine) ), galactosylceramide sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0), lactosylceramide (d18:1/22:0).

Further, the above-mentioned diagnostic markers can be two or more combinations of the following seven serum metabolic markers: lysophosphatidic acid LPA (18:1(9Z)/0:0), lysophosphatidic acid LysoPC ( 14:0/0:0), LysoPC (18:2(9Z,12Z)), LysoPC (24:0), Phospholipid PC (14:1(9Z)/P-18:1( 11Z)), phospholipid PC (16:0/18:2(9Z,12Z)) and phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)).

Further, the above-mentioned diagnostic markers can be two or more combinations of the following five serum metabolic markers: L-Tyrosine, L-Tryptophan, Glycocholic Acid, Taurocholate and Cortisol.

Preferably, the above-mentioned diagnostic marker is a combination of the following 25 serum metabolic markers (referred to as diagnostic marker A, the same below): beta-alanine-lysine (beta-Ala-Lys), L-carnosine ( L-Carnosine), cis-9-hexadecenoic acid (cis-9-Palmitoleic acid), palmitic acid (Palmitic acid), oleic acid (Oleic Acid), lysophosphatidic acid LPA (18:1(9Z)/0 :0), LysoPC(14:0/0:0), LysoPC(18:2(9Z,12Z)), LysoPC(24:0), LysoPC(14:1( 9Z)/P-18:1(11Z)), Phospholipid PC(16:0/18:2(9Z,12Z)), Phospholipid PC(24:1(15Z)/22:6(4Z,7Z,10Z, 13Z, 16Z, 19Z)), Linoleic acid (Linoleic acid), Nicotinamide adenine dinucleotide (NADH), Cortisol (Cortisol), L-Tyrosine (L-Tyrosine), L-Tryptophan (L-Tryptophan), Glycocholic Acid, Taurocholate, Hypoxanthine, Allantoic acid, Inosine e, Sheath 1-phosphate Amino alcohol (Sphingosine1-phosphate), galactosylceramide sulfate 3-O-Sulfogalactos ylceramide (d18:1/20:0) and lactosylceramide (d18:1/22:0).

Preferably, the above-mentioned diagnostic markers are a combination of the following seven serum metabolic markers (referred to as diagnostic marker B, the same below): lysophosphatidic acid LPA (18:1(9Z)/0:0), lysolecithin LysoPC(14:0/0:0), LysoPC(18:2(9Z,12Z)), LysoPC(24:0), LysoPC(14:1(9Z)/P-18: 1(11Z)), phospholipid PC (16:0/18:2(9Z,12Z)) and phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)) .

Preferably, the above-mentioned diagnostic marker is a combination of the following five serum metabolic markers (referred to as diagnostic marker C, the same below): L-tyrosine (L-Tyrosine), L-tryptophan (L-Tryptophan ), Glycocholic Acid, Taurocholate and Cortisol.

The present invention provides diagnostic markers composed of various serum metabolites or combinations of serum metabolites. The above diagnostic markers involve a total of 25 serum metabolic markers, and these 25 serum metabolic markers are closely related to 10 metabolic pathways. Among them, the two serum metabolic markers, beta-Ala-Lys and L-Carnosine, are closely related to the metabolic pathway of beta-Alanine metabolism; Three serum metabolic markers, cis-9-Palmitoleic acid, palmitic acid and oleic acid, are closely related to the metabolic pathway of fatty acid biosynthesis; Lysophosphatidic acid LPA (18:1(9Z)/0:0), lysolecithin LysoPC (14:0/0:0), lysolecithin LysoPC (18:2(9Z,12Z)) and lysolecithin LysoPC (24:0) These four serum metabolic markers are closely related to the metabolic pathway of glycerophospholipid metabolism; phospholipid PC(14:1(9Z)/P-18:1(11Z)), phospholipid PC(16: 0/18:2(9Z,12Z)), phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)) and linoleic acid (Linoleic acid) Serum metabolic markers are closely related to two metabolic pathways: glycerophospholipid metabolism and linoleic acid metabolism; nicotinamide adenine dinucleotide (NADH) and oxidative phosphorylation (Oxidative phosphorylation) metabolic pathway Closely related; L-Tyrosine (L-Tyrosine) and L-Tryptophan (L-Tryptophan), two serum metabolic markers, are closely related to phenylalanine/tyrosine and tryptophan and tryptophan biosynthesis) metabolic pathway; two serum metabolic markers, Glycocholic Acid and Taurocholate, are closely related to the primary bile acid biosynthesis metabolic pathway; cortisol ( Cortisol) is closely related to cancer pathway and bile secretion (Pathways in cancer, and Bile secretion) metabolic pathway; the three serum metabolic markers of hypoxanthine (Hypoxanthine), allantoic acid (Allantoic acid) and inosine (Inosine) are closely related to Purine metabolism lism) metabolic pathways are closely related; 1-phosphate sphingosine (Sphingosine 1-phosphate), sulfate galactosylceramide 3-O-Sulfogalactosylceramide (d18:1/20:0) and lactosylceramide (d18: 1/22:0) These three serum metabolic markers are closely related to the metabolic pathway of sphingolipid metabolism.

The present invention also provides a screening method for various diagnostic markers suitable for the early diagnosis of esophageal squamous cell carcinoma, comprising the following steps:

(1) Serum samples from patients with esophageal squamous cell carcinoma and healthy people were collected as analysis samples, where esophageal squamous cell carcinoma serum samples included esophageal carcinoma in situ serum samples, early esophageal cancer serum samples and advanced esophageal cancer serum samples;

(2) Analyze each analysis sample using UPLC-QTOF/MS serum metabolomics technology to obtain the original metabolic fingerprint of each serum sample;

(3) Use the R language XCMS software package to preprocess the original metabolic fingerprints of esophageal squamous cell carcinoma serum samples and healthy serum samples respectively, and obtain a two-dimensional matrix of metabolite information for each row analysis sample and each column, and Use the CAMERA software package of R language to identify metabolite peaks in the two-dimensional matrix for further statistical analysis;

(4) The two-dimensional matrix in step (3) is subjected to principal component analysis and partial least squares discriminant analysis in turn to obtain the PLS-DA model, which shows that patients with esophageal squamous cell carcinoma and healthy people have metabolic patterns differences and clear classification trends;

(5) According to the PLS-DA model obtained above, the differential metabolites were screened with the help of the variable importance score of the PLS-DA modeling and the univariate non-parametric test. The screening criteria were: VIP≥1, and the false discovery rate FDR The q value after multiple testing correction is less than 0.05;

(6) Determine the quasi-molecular ions, adducts and isotope information of the differential metabolites obtained through the above screening according to the CAMERA package of the R language, and obtain potential metabolic markers;

(7) On the basis of the above potential metabolic markers, combined with the primary and secondary mass spectrum information, quasi-molecular ion information, adduct information and isotope information of the potential metabolic markers, the molecular mass and molecular formula of the diagnostic markers are estimated, And compared and matched with existing standard compounds to obtain serum metabolic markers. A single serum metabolic marker or a combination of serum metabolic markers can be used as a diagnostic marker suitable for early diagnosis of esophageal squamous cell carcinoma.

In the above screening method, the healthy population is the population without upper gastrointestinal lesions and other metabolic diseases (such as hyperthyroidism, hypothyroidism, hypertension, diabetes, kidney disease, etc.).

In the above screening method, when LC-MS serum metabolomics analysis is performed, a quality control sample is added to every 10 analysis samples for real-time monitoring of the quality control of the analysis samples from the pre-injection treatment to the analysis process. The quality control sample was a mixture of 5 esophageal cancer serum samples and 5 healthy serum samples.

In the above-mentioned screening method, the following pretreatments are carried out before the analysis sample and the quality control sample are injected:

(1) Use a pipette to extract 50 μl of analytical samples or quality control samples and place them on a 96-well plate of the Bravo automatic specimen processing system;

(2) Add 150 μl of methanol for extraction, vortex for 30 s, and incubate at -20°C to precipitate protein;

(3) Centrifuge at 4000 rpm for 20 minutes at 4°C in a high-speed centrifuge;

(4) Pour the supernatant from step (3) into an LC-MS sampling vial and store at -80°C for LC-MS detection.

In the above screening method, the spectral preprocessing of the original metabolic fingerprint refers to converting the obtained original metabolic fingerprint into an MZdata data file with Masshunter software, and then using the XCMS software package to perform retention time correction and peak identification on the MZdata data file. , peak matching and peak alignment preprocessing operations to obtain a two-dimensional matrix.

In the above screening method, the R software package CAMERA is used to identify metabolite peaks on the two-dimensional matrix, including isotope peaks, adducts and fragment ions.

In the above screening method, when each analysis sample is analyzed by LC-MS serum metabolomics technology, the chromatographic column used in liquid chromatography is WatersACQUITYUPLC HSS T3 chromatographic column, the specification is 100mm×2.1mm, 1.8μm; the injection volume is 6 μL, the injection temperature is 4°C, the flow rate is 0.5ml/min; the chromatographic mobile phase contains two solvents A and B: A in positive ion ESI+ mode is 0.1wt% formic acid aqueous solution, and A in negative ion ESI- mode is 0.5 Mmol/L ammonium fluoride aqueous solution, B in positive ion ESI+ mode is 0.1wt% formic acid in acetonitrile solution, B in negative ion ESI- mode is pure acetonitrile; chromatographic gradient elution condition: 0-1min is 1% B, 1-8min is 1% B-100% B gradually increasing, 10-10.1min is 100% B rapidly reduced to 1% B, and then 1% B lasts for 1.9min.

In the above screening method, when each analysis sample is analyzed by LC-MS serum metabolomics technology, the mass spectrometry detection uses the quadrupole time-of-flight mass spectrometer Q-TOF, and uses the positive ion mode ESI+ and negative ion of the electrospray ion source Mode ESI-, the ion source temperature is 400°C, the cone gas flow rate is 12L/min, the desolvation temperature is 250°C, the desolvation gas flow rate is 16L/min; the capillary voltage is +3kV and +3kV and -3kV, the cone voltage is 0V; the cone pressure in positive ion mode is 20psi, and the cone pressure in negative ion mode is 40psi; s.

In the preferred solution of the present invention, 39 patients with esophageal carcinoma in situ, 28 patients with early esophageal cancer, 30 patients with advanced esophageal cancer, and 105 healthy people were used for screening.

In the preferred solution of the present invention, the R2X=0.167, R2Y=0.569, and Q2Y=0.523 of the PLS-DA model obtained in the screening process.

The present invention also provides a method for constructing a diagnostic model of esophageal squamous cell carcinoma, comprising the following steps:

(1) Serum samples from patients with esophageal squamous cell carcinoma and healthy people were collected as analysis samples, where esophageal squamous cell carcinoma serum samples included esophageal carcinoma in situ serum samples, early esophageal cancer serum samples and advanced esophageal cancer serum samples;

(2) Analyze each analysis sample using LC-MS serum metabolomics technology to obtain the original metabolic fingerprint of each serum sample;

(3) Use the R language XCMS software package to preprocess the original metabolic fingerprints of each serum sample to obtain a two-dimensional matrix of metabolite information for each row analysis sample, and use the R software package CAMERA to analyze the two-dimensional Matrix for metabolite peak identification for further statistical analysis;

(4) Screen out the information of the diagnostic markers suitable for the early diagnosis of esophageal squamous cell carcinoma from the two-dimensional matrix according to the mass-to-charge ratio and retention time, and obtain the two-dimensional matrix of diagnostic markers;

(5) According to the two-dimensional matrix of diagnostic markers, a random forest model was constructed using the randomForest software package in R language to obtain a diagnostic model for esophageal squamous cell carcinoma.

In the above construction method, the esophageal carcinoma in situ refers to stage 0 in the TNM staging standard, which refers to the atypical hyperplasia (severe) in the epithelial layer of the mucosa or in the epidermis of the skin, involving the whole layer of the epithelium, but has not yet penetrated the basement membrane. Cancer that infiltrates and grows below; early esophageal cancer refers to stage I and II in the TNM staging standard, and refers to cancer confined to the retromucosa or submucosa without lymph node involvement and distant metastasis; advanced esophageal cancer refers to stage III in the TNM staging standard Stage and stage IV refer to cancers that have involved the muscular layer or reached the adventitia or beyond the adventitia, with local or distant lymph node metastasis. The TNM staging standard is based on the American joint Committee on Cancer (AJCC) TNM Classfication of Carcinoma of the Esophagus and Esophagogastric Junction (7th ed, 2010).

In the preferred solution of the present invention, when constructing the random forest model, the modeling parameter ntree=5000.

In the preferred solution of the present invention, the model is constructed based on the following number of samples: 39 patients with esophageal carcinoma in situ, 28 patients with early esophageal cancer, 30 patients with advanced esophageal cancer, and 105 healthy people.

In a preferred version of the present invention, when the diagnostic marker suitable for the early diagnosis of esophageal squamous cell carcinoma is a combination of 25 serum metabolic markers (diagnostic marker A), the diagnostic threshold (Threshold) of the resulting diagnostic model is 0.3552; when the diagnostic marker suitable for the early diagnosis of esophageal squamous cell carcinoma is a combination of 7 serum metabolic markers (diagnostic marker B), the diagnostic threshold (Threshold) of the resulting diagnostic model is 0.7431; when suitable for When the diagnostic marker for early diagnosis of esophageal squamous cell carcinoma is a combination of 5 serum metabolic markers (diagnostic marker C), the diagnostic threshold (Threshold) of the obtained diagnostic model is 0.4943. When the prediction value given by the diagnostic model is greater than or equal to the diagnostic cut-off value, it means that you have esophageal squamous cell carcinoma, and when it is less than the diagnostic cut-off value, it means that you do not have esophageal squamous cell carcinoma.

The present invention also provides a diagnostic model of esophageal squamous cell carcinoma, which is constructed according to the method for constructing the diagnostic model of esophageal squamous cell carcinoma. As above, in the preferred solution of the present invention, when the diagnostic marker used in the diagnostic model is diagnostic marker A, the diagnostic cutoff value of the diagnostic model is 0.3552; when it is diagnostic marker B, the diagnostic cutoff value of the diagnostic model is 0.7431; When it is the diagnostic marker C, the diagnostic cut-off value of the diagnostic model is 0.4943.

The present invention also provides a method for using a diagnostic model of esophageal squamous cell carcinoma, that is, a method for diagnosing esophageal squamous cell carcinoma using the diagnostic model of esophageal squamous cell carcinoma, comprising the following steps:

(1) Take the serum sample to be tested, meet the sample injection requirements through pretreatment, analyze the pretreated serum sample to be tested by LC-MS serum metabolomics technology, and obtain the original metabolic fingerprint of the serum sample to be tested;

(2) Use the R language XCMS software package to preprocess the original metabolic fingerprint and identify metabolite peaks to obtain a two-dimensional matrix that can be used for statistical analysis;

(3) Screen out the information of the diagnostic markers suitable for the early diagnosis of esophageal squamous cell carcinoma according to the mass-to-charge ratio and retention time from the two-dimensional matrix, and obtain the two-dimensional matrix of diagnostic markers;

(4) Bring the two-dimensional matrix of diagnostic markers into the diagnostic model of esophageal squamous cell carcinoma, and judge whether it is early esophageal squamous cell carcinoma according to the values given by the model and the diagnostic threshold (Threshold) of the model.

In the preferred version of the present invention, when the combination of 25 serum metabolic markers (diagnostic marker A) is suitable for the early diagnosis of esophageal squamous cell carcinoma, the value given by the diagnostic model is greater than or equal to 0.3552 When the diagnosis is esophageal cancer, otherwise it is not; when the combination of 7 serum metabolic markers (diagnostic marker B) is suitable for the early diagnosis of esophageal squamous cell carcinoma, the value given by the diagnostic model is greater than or equal to 0.7431, it is diagnosed as esophageal cancer, otherwise it is not; when the combination of 5 serum metabolic markers (diagnostic marker C) is suitable for the early diagnosis of esophageal squamous cell carcinoma, the diagnostic model gives When the value of is greater than or equal to 0.4943, it is diagnosed as esophageal cancer, otherwise it is not.

The advantages of the present invention are: the present invention uses serum metabolomics technology and data statistical analysis technology to obtain diagnostic markers and esophageal squamous cell carcinoma diagnostic models suitable for early diagnosis of esophageal squamous cell carcinoma, and has found a close relationship with the diagnostic markers. Related 10 metabolic pathways. The diagnostic marker screening method of the present invention has strong operability, simple model construction method, good diagnostic effect, high sensitivity and good specificity, and is not only suitable for the diagnosis of advanced esophageal cancer, but also suitable for the diagnosis of early esophageal cancer, especially suitable for Early diagnosis of esophageal cancer. The present invention can realize diagnosis only by taking blood, is non-invasive and low-cost, and can well replace the current internal invasive diagnosis mode, greatly reducing the pain of patients, and the present invention is fast and convenient in diagnosis, requires a short time, and improves the The work efficiency is conducive to the early detection and early treatment of esophageal cancer, and has good clinical use and promotion value.

Description of drawings

Figure 1. The total ion chromatogram of the original metabolic fingerprint (LC-QTOF/MS, +ESI is the positive ion mode, -ESI is the negative ion mode), the horizontal axis is the retention time (Retention Time, min), and the vertical axis is the metabolite Relative concentration, Normal is a healthy serum sample, ESCC is an esophageal cancer serum sample.

Figure 2. PCA score plot of metabolic profile pre-analysis, where Normal is a healthy serum sample, ESCC is an esophageal cancer serum sample, and QC is a quality control sample.

Figure 3A is the PLS-DA three-dimensional score map comparing the metabolic profiles of esophageal cancer and healthy controls, the modeled R2X=0.167, R2Y=0.569, Q2Y=0.523; Figure 3B is the corresponding PLS-DA model based on the random permutation method Validation diagram. Among them, Normal is a healthy serum sample, and ESCC is a serum sample of esophageal cancer.

Figure 4. The flow chart of the identification of L-Tryptophan (L-Tryptophan), where Figure (a): m/z is 205.0974 retention time characteristics in the chromatogram; Figure (b): retention time is 181.38s first-order Mass Spectrum; Figure (c): Secondary ion MS/MS fragmentation with a retention time of 181.38s and m/z of 205.0974; Figure (d): Fragmentation of metabolites (RT: 181.38s, m/z: 205.0974) cracking mechanism.

Figure 5. The ROC curve of the external test sample of the random forest model constructed by 25 metabolic markers.

Figure 6. The external test sample ROC curve of the random forest early esophageal cancer diagnosis model constructed by 7 metabolic markers.

Figure 7. The external test sample ROC curve of the random forest early esophageal cancer diagnosis model constructed by 5 metabolic markers.

detailed description

Below, the present invention is further explained through the following specific embodiments, and the advantages of the present invention are further proved.

The screening of the diagnostic markers of the present invention, the construction method of the diagnostic model and the effect verification are as follows:

1. Research object

Relying on the esophageal cancer screening platform of "National Demonstration Base for Early Diagnosis and Treatment of Esophageal Cancer (Feicheng City, Shandong Province)", this study targeted iodine-stained biopsy subjects aged 40-69 under gastroscope in Feicheng City, Shandong Province (confirmed as the gold standard) Collect esophageal carcinoma in situ (39 cases at stage 0), early esophageal cancer (17 cases at stage I, 11 cases at stage II) and advanced esophageal cancer (30 cases at stage III); The test subjects were 105 healthy subjects without upper gastrointestinal lesions as healthy samples.

2. Serum metabolomics detection by LC-MS

All collected serum samples were centrifuged and stored in a -80°C refrigerator, using an ultra-high performance liquid chromatography-mass spectrometer (UPLC-QTOF/MS 6550, Agilent) and a Bravo automatic specimen pretreatment system (Agilent, USA). Metabolomics detection (in 3 large batches of detection, and quality control), to obtain the original metabolic fingerprint of the sample including chromatographic and mass spectral information. The specific operation is as follows:

2.1 Instruments and equipment

Experimental equipment includes: UPLC-QTOF/MS 6550 system (Agilent, USA), Bravo system (Agilent, USA), high-speed low-temperature centrifuge, vibrating vortex machine, nitrogen drying device, 4 ℃ refrigerator (Haier), pure water instrument (Siemens).

Experimental consumables include: Waters ACQUITY HSS T3 (particle size, 1.8μm; 100mm (length) × 2.1mm) chromatographic column, liquid nitrogen, high-purity nitrogen; conical bottom sample bottle, 2ml centrifugal rotor, 2ml centrifuge tube (round bottom), pipette, 1000μl Pipette tip, 200μl pipette tip, marker pen, latex gloves, mask.

Experimental reagents include: methanol (Dima, HPLC grade), acetonitrile (Dima, HPLC grade), formic acid (Guangfu Institute of Fine Chemistry, Tianjin), pure water (TOC<10ppb).

2.2 Serum sample pretreatment

Before the pretreatment of serum samples, 21 quality control samples (QC) were prepared, and all early esophageal cancer serum samples, esophageal carcinoma in situ serum samples, advanced esophageal cancer serum samples, healthy serum samples and quality control samples were randomly numbered, and early Esophageal cancer serum samples, esophageal carcinoma in situ serum samples, advanced esophageal cancer serum samples, and healthy serum samples were used as analysis samples, and a quality control sample was added every 10 analysis samples. Early esophageal cancer serum samples, esophageal carcinoma in situ serum samples, and advanced esophageal cancer serum samples were collectively referred to as esophageal cancer serum samples, and the quality control sample was a mixed sample of 5 esophageal cancer serum samples and 5 healthy serum samples. Esophageal cancer serum samples, healthy serum samples and quality control samples were all pretreated, and the pretreatment included the following 4 steps:

(1) Use a pipette to extract 50 μl of analytical samples or quality control samples, and place them on a 96-well plate of the Bravo automatic specimen processing system (Agilent, USA);

(2) Add 150 μl of methanol for extraction, vortex for 30 s, and incubate at -20°C to precipitate protein.

(3) Centrifuge at 4000 rpm for 20 minutes at 4°C in a high-speed centrifuge;

(4) Pour the supernatant of step (3) into an LC-MS injection bottle, and store it at -80°C for LC-MS detection;

2.3 Serum UPLC-QTOF/MS detection

UPLC system (1290series, Agilent) injected 6 μL aliquots of pretreated samples into ACQUITY UPLCHSS T3 (particle size, 1.8 μm; 100 mm (length) × 2.1 mm) chromatographic column (Waters, Milford, USA). The injection sequence was completely randomized to eliminate the bias caused by the injection sequence. The chromatographic mobile phase contains two solvents: A is 0.1wt% formic acid (diluted in water, positive ion ESI+) or 0.5mM ammonium fluoride (diluted in water, negative ion ESI-), B is 0.1wt% formic acid (diluted in acetonitrile, positive ion ESI+ ) or 100% acetonitrile (negative ion ESI-). The chromatographic gradient is: 1% B for 0-1 min, 1% B-100% B for 1-8 min, gradually increasing, then 100% B for 10-10.1 min, rapidly reduced to 1% B, and then 1% B for 1.9 min. The flow rate is 0.5ml/min. The entire sample detection process was maintained at 4°C. Wherein, the percentages of A and B refer to volume percentages.

Agilent quadrupole time-of-flight mass spectrometer Q-TOF (6550, Agilent) was used for mass spectrometry detection, and the positive ion mode (ESI+) and negative ion mode (ESI-) of the electrospray ion source were used. The ion source temperature was set at 400 °C, and the cone gas flow was 12 L/min. At the same time, the desolvation temperature was set at 250° C., and the desolvation gas flow rate was 16 L/min. The capillary voltage was +3kV and -3kV in positive and negative ion modes, respectively, and the cone voltage was 0V. Cone pressure is 20psi (positive ions) and 40psi (negative ions). The range of mass-to-charge ratio for spectral data collection is 50-1200m/z, and the scanning frequency of collection is 0.25s. In MS/MS MS analysis, high-purity nitrogen is used as a collision gas to generate target ion fragments, and the collision energy is set to 10, 20, or 40eV.

3. XCMS Spectrum Preprocessing

UPLC-QTOF/MS serum positive ion ESI+ and negative ion ESI- detection to obtain the original metabolic fingerprint data (see Figure 1), through Agilent's Masshunter software into Mzdata data files, and then use the R language XCMS software package for XCMS spectrum pre-processing Processing, preprocessing includes retention time correction, peak identification, peak matching, peak alignment, noise filtering, overlapping peak analysis, threshold selection, standardization, etc. The relevant parameters of XCMS pretreatment are: peak width at half waist is 10 (fwhm=10), retention time window is set to 10 (bw=10), and other parameters are default values. After XCMS spectrum preprocessing, a two-dimensional matrix that can be used for statistical analysis is obtained, in which each row is a sample (observation), each column is a metabolite (variable), and the matrix median value is the corresponding metabolite concentration. And each metabolite peak is characterized by retention time (retention time, RT) and mass-to-charge ratio (mass-to-chargeratio, m/z). This two-dimensional matrix was then used for metabolite peak identification (including isotopic peaks, adducts and fragment ions) using the R package CAMERA. The samples were standardized before statistical analysis, and the retention time range to be analyzed was set at 0.5-10 min. After XCMS spectrum preprocessing, the data matrix generated by UPLC-QTOF/MS spectrum in positive ion detection mode contains 981 metabolite peaks, and 485 metabolite peaks in negative ion detection mode, with a total of 1466 metabolite peaks.

4. LC-MS experiment quality control

When serum samples are tested for metabolomics, the prepared QC samples are evenly inserted into the analyzed samples in the order of 1 QC for every 10 analyzed samples, so as to monitor the quality control from sample pretreatment to sample detection in real time . After the obtained original metabolic fingerprint was preprocessed by XCMS spectrum, the %RSD value (coefficient of variation) of each metabolite in the QC sample was calculated, and the %RSD value of most metabolites was controlled below 30%, indicating that the sample pretreatment to The quality control of the sample product testing process is good, and the obtained metabolomics data is authentic and credible.

5. Metabolic profile pre-analysis based on PCA

The unsupervised analysis method, principal component analysis (PCA), was used to initially observe the classification trends and outliers between groups, as shown in Figure 2. The repeatability of the QC samples in the figure can indicate that the quality control of the LC-MS experiment is good. It can also be seen from the figure that there is a certain classification trend between esophageal cancer and healthy controls, but there are still some overlaps, and a supervised learning method is needed to achieve further classification.

6. Metabolic profile analysis based on PLS-DA

Randomly assign 4/5 of the obtained two-dimensional matrix data as the training sample training data, and the other 1/5 as the external test sample test data (see Table 1). In order to eliminate the deviation caused by the difference within the group as much as possible and obtain a more obvious grouping trend, a supervised analysis method, that is, partial least squares-discriminant analysis (PLS-DA), was further used for the training samples to show the difference between esophageal cancer and healthy controls. Differences and classification trends in metabolic profiles among As shown in Figure 3, there are differences in metabolic patterns and obvious classification trends between esophageal cancer and healthy controls, and the modeled R2X=0.167, R2Y=0.569, and Q2Y=0.523.

Table 1. Baseline and clinicopathological features of metabolomic studies for early diagnosis of esophageal cancer

7. Screening of differential metabolites and identification of chemical substances for early diagnosis of esophageal cancer

In order to screen out differential metabolites for the diagnosis of early esophageal cancer, we screened with the help of variable importance score (VIP) in PLS-DA modeling and univariate nonparametric test (nonparametric Kruskal-Wallis rank sum test). The variable screening criteria were: VIP≥1; and the q value after multiple test correction of the false discovery rate FDR was less than 0.05. According to this standard, a total of 551 differentially expressed serum metabolome markers between esophageal cancer and healthy controls were screened out, and the quasi-molecular ions, adducts and isotope information of the differential metabolites were further determined according to the CAMERA package of the R language, excluding chemical metabolites. Signals and those not found in the human body, 242 potential metabolic markers were obtained.

For the above 242 potential metabolic markers, follow the identification steps of chemical substances (such as the identification process of the metabolic marker L-tryptophan RT 181.38s, m/z 205.0974, see Figure 4) to identify the metabolic markers:

(1) According to the first-order mass spectrometry fragmentation distribution characteristics of potential metabolic markers, combined with the R language CAMERA software package (CAMERA is the R language package software package Collection of annotation related methods for massspectrometry data, http://bioconductor.org/packages/ release/bioc/html/CAMERA.html) determine the quasi-molecular ions, adducts and isotope information of potential metabolic markers, and infer the molecular mass and molecular formula of potential metabolic markers;

(2) Search the online human metabolite database HMDB (http://www.hmdb.ca/) and METLIN (http://metlin.scripps.edu/) according to the molecular mass, and determine several candidate compounds;

(3) Conduct RRLC-QTOF/MS/MS secondary mass spectrometry experiments on 242 potential metabolic markers, further obtain the mass spectrometry ion fragment information corresponding to metabolites, and perform secondary mass spectrometry fragment ion matching with candidate compounds in the database; The final substance identification was performed by comparing the chromatographic and mass spectral characteristics of the compound standard sample library.

According to the above identification method, a total of 25 serum metabolic markers were successfully identified by secondary mass spectrometry or standard products. The common names (Common Name) of these 25 serum size markers are: beta-Ala-Lys; L-Carnosine; cis-9-Palmitoleic acid; Palmitic acid; Oleic Acid; LPA(18:1(9Z)/0:0); LysoPC(14:0/0:0); LysoPC(18:2(9Z,12Z )); LysoPC(24:0); PC(14:1(9Z)/P-18:1(11Z)); PC(16:0/18:2(9Z,12Z)); PC(24:1 (15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)); Linoleicacid; NADH; Cortisol; L-Tyrosine; L-Tryptophan; Glycocholic Acid; Taurocholate; Hypoxanthine; Allantoic acid; Inosine; phosphate; 3-O-Sulfogalactosylceramide (d18:1/20:0); Lactosylceramide (d18:1/22:0).

Compared with the published literature, the 25 serum metabolic markers of the present invention are all found in early esophageal squamous cell carcinoma for the first time, which has very important significance for the diagnosis and treatment of early esophageal cancer. Among them, the Chinese translation of beta-Ala-Lys is beta-alanine-lysine, the Chinese translation of L-Carnosine is L-carnosine, and the Chinese translation of cis-9-Palmitoleic acid is cis-9-hexadecenoic acid The Chinese translation of Palmitic acid is palmitic acid, the Chinese translation of Oleic Acid is oleic acid, the Chinese translation of LP A is lysophosphatidic acid, the Chinese translation of LysoPC is lysolecithin, the Chinese translation of PC is phospholipid, and the Chinese translation of Li noleic acid The Chinese translation of NADH is nicotinamide adenine dinucleotide, the Chinese translation of Cortisol is cortisol, the Chinese translation of L-Tyrosine is L-tyrosine, and the Chinese translation of L-Tryptophan is L- Tryptophan, the Chinese translation of Glycocholic Acid is glycocholic acid, the Chinese translation of Taurocholate is taurocholate, the Chinese translation of Hy poxanthine is hypoxanthine, the Chinese translation of Allantoic acid is allantoic acid, and the Chinese translation of Inosine Inosine, the Chinese translation of Sphingosine 1-phosphate is 1-phosphate sphingosine, the Chinese translation of 3-O-Sulfogalactosylceramide is sulfate galactosylceramide, and the Chinese translation of Lactosylceramide is lactose ceramide. Due to possible deviations in translation, the English standard name shall prevail.

The database retrieval information of the above 25 serum metabolic markers in HMDB and METLIN is shown in Table 2 below. Those skilled in the art can obtain the detailed information of these 25 serum metabolic markers according to the HMDB ID number and METLIN ID number in the following table , such as the chemical formula:

Table 2 The database retrieval information of 25 serum metabolic markers in HMDB and METLIN

In addition, through KEGG enrichment and metabolic pathway analysis, it was found that the above 25 serum metabolic markers were closely related to the following 10 metabolic pathways: Glycerophospholipid metabolism; Linoleicacid metabolism; beta-Alanine metabolism; Fatty acid biosynthesis; Oxidativephosphorylation ; Phenylalanine, tyrosine and tryptophan biosynthesis; Primary bile acid biosynthesis; Pathways in cancer, and Bile secretion; Purine metabolism; Sphingolipid metabolism. The above 10 metabolic pathways are the standard names in the metabolic pathway KEGG (http://www.genome.jp/kegg/), and their corresponding Chinese translation names are: glycerophospholipid metabolism, linoleic acid metabolism metabolism), beta-Alanine metabolism (beta-Alanine metabolism), fatty acid synthesis (Fatty acid biosynthesis), oxidative phosphorylation (Oxidative phosphorylation), phenylalanine/tyrosine and tryptophan metabolism (Phenylalanine, tyrosine and Tryptophan biosynthesis), Primary bile acid biosynthesis, Pathways in cancer, and Bile secretion, Purine metabolism, Sphingolipid metabolism. This proves that these 10 metabolic pathways are disturbed in the early stage of esophageal cancer, and this discovery of the present invention has a very good guiding effect on the prevention of esophageal cancer and the development of drugs.

Table 3 below shows the difference information of the 25 serum metabolic markers screened between patients with esophageal cancer and healthy people, in which L-Tryptophan was found in both positive and negative ion modes. FC is the fold change between esophageal cancer and healthy controls. According to the FC information, it can be seen that: beta-alanine-lysine (beta-Ala-Lys), lysophosphatidic acid LPA (18:1( 9Z)/0:0), LysoPC (14:0/0:0), Phospholipid PC (16:0/18:2(9Z,12Z)), Nicotinamide Adenine Dinucleotide (NADH) , L-Tyrosine (L-Tyrosine), L-Tryptophan (L-Tryptophan), Glycocholic Acid (Glycocholic Acid), Allantoic acid (Allantoic acid), Inosine (Inosine) and 1-phosphate sphingosine The expression of alcohol (Sphingosine1-phosphate) in the esophageal cancer group was significantly higher than that in the healthy group, while the expression of other metabolic markers was significantly lower in the esophageal cancer group than in the healthy group.

FDR is the false discovery rate based on non-parametric multiple comparison correction, and its value is less than 0.05; AUC is the area under the ROC curve AUC value of the diagnostic test evaluation of a single metabolome marker, from which it can be seen that the 25 serum metabolites When markers are used alone to diagnose esophageal cancer and non-esophageal cancer, the lowest AUC value is 0.61, and the highest AUC value is 0.85. It can be seen that, as a single-component diagnostic marker, the diagnostic effect of the 25 serum metabolic markers screened by the present invention is more significant, and diagnosis with a single serum metabolic marker has certain clinical research value .

In order to make the diagnosis better, serum metabolic markers can be used in combination, for example, the following 8 diagnostic markers can be formed according to the relationship between serum metabolic markers and metabolic pathways: (a) beta-alanine - A combination of lysine (beta-Ala-Lys) and L-carnosine (L-Carnosine); (b) cis-9-hexadecenoic acid (cis-9-Palmitoleic acid), palmitic acid (Palmitic acid) and Oleic acid (OleicAcid) combination; (c) lysophosphatidic acid LPA (18:1 (9Z)/0:0), lysolecithin LysoPC (14:0/0:0), lysolecithin LysoPC (18:2 (9Z,12Z)) and LysoPC (24:0); (d) Phospholipid PC (14:1(9Z)/P-18:1(11Z)), Phospholipid PC (16:0/18 :2(9Z,12Z)), the combination of phospholipid PC(24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)) and linoleic acid (Linoleic acid); (e) smoke A combination of NADH, L-Tyrosine and L-Tryptophan; (f) Cortisol, GlycocholicAcid ) and Taurocholate; (g) Hypoxanthine, Allantoic acid and Inosine; (h) Sphingosine 1-phosphate -phosphate), a combination of 3-O-Sulfogalactosylceramide (d18:1/20:0) and lactosylceramide (d18:1/22:0).

Several metabolic markers with good AUC effects can also be selected to be combined to form diagnostic markers. For example, the diagnostic markers can be a combination of two or more of the following serum metabolic markers: lysophosphatidic acid LPA (18: 1(9Z)/0:0), Lysolecithin LysoPC(14:0/0:0), Lysolecithin LysoPC(18:2(9Z,12Z)), Lysolecithin LysoPC(24:0), Phospholipids PC(14:1(9Z)/P-18:1(11Z)), phospholipid PC(16:0/18:2(9Z,12Z)), phospholipid PC(24:1(15Z)/22:6( 4Z, 7Z, 10Z, 13Z, 16Z, 19Z)), Nicotinamide Adenine Dinucleotide (NADH), Cortisol, L-Tryptophan, Taurocholate ), hypoxanthine (Hypoxanthine), inosine (Inosine), galactosylceramide sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0), lactosylceramide (d18:1/22:0 ).

A combination of two or more of the following serum metabolic markers with good AUC effect can also be selected: lysophosphatidic acid LPA (18:1(9Z)/0:0), lysophosphatidic acid LysoPC (14:0/ 0:0), LysoPC(18:2(9Z,12Z)), LysoPC(24:0), LysoPC(14:1(9Z)/P-18:1(11Z)), Phospholipid PC (16:0/18:2(9Z,12Z)) and Phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)).

You can also choose a combination of two or more of the following serum metabolic markers with good AUC effects: L-Tyrosine, L-Tryptophan, Glycocholic Acid (Glycocholic Acid), Taurocholate and Cortisol.

Table 3. Twenty-five serum metabolic markers differentially expressed between esophageal cancer and healthy controls

FDR is the false discovery rate (multiple comparisons); AUC is the area under the ROC curve; FC is the fold change; RSD% is the coefficient of variation calculated based on quality control samples.

Below, the further application effects of the three preferred diagnostic markers of the present invention will be listed in detail, and the application of other diagnostic markers will not be listed here.

8. Esophageal cancer early diagnosis model and external verification

8.1 Using the combination of the 25 serum metabolic markers identified above as diagnostic markers, an early diagnosis model for esophageal cancer was constructed based on random forest in the training samples. The random forest is implemented using the randomForest software package in the R language, and the modeling parameter ntree=5000 (equivalent to b below).

The random forest modeling steps are as follows:

(1) The sample content of the original training set is N, apply the bootstrap method to randomly select b new self-service sample sets with replacement, and construct b classification trees from this, and the undrawn samples form b Out-of-bag data (out-of-bag, OOB);

(2) If there are m _all variables, m _try variables are randomly selected at each node of each tree (m _try << m _all ), and then a variable with the most classification ability is selected in m _try , The threshold for variable classification is determined by examining each classification point;

(3) Each classification tree in the random forest is a binary tree, and its generation follows the top-down recursive splitting principle, that is, the training set is divided sequentially from the root node. Each tree grows to its maximum without any pruning.

(4) Combine the generated multiple classification trees into a random forest, use the random forest classifier to discriminate and classify the new data, and the classification result depends on the number of votes of the tree classifier.

(5) Then the ROC curve analysis can be performed with the voting score and the actual classification situation to obtain the diagnostic cut-off value (Threshold) of diagnosis. The diagnostic threshold (Threshold) of this model is 0.3552.

The random forest model constructed above can be used as a diagnostic model for esophageal cancer. When the constructed random forest model is used for diagnosis, the data information of 25 serum metabolic markers in the serum to be tested is imported into the random forest model. If the model classifier If the voting result is greater than or equal to the diagnostic cutoff value, it is judged as a positive diagnosis (with esophageal squamous cell carcinoma), and if it is lower than the diagnostic cutoff value, it is judged as a negative diagnosis (without esophageal squamous cell carcinoma).

Substitute the two-dimensional matrix data of 25 serum metabolic markers of the external test sample into the random forest model established above to obtain the predicted value of the probability of esophageal cancer of the test sample, and compare it with the actual pathological results (esophageal cancer or healthy) Perform ROC curve analysis (see Figure 5) to obtain the sensitivity, specificity and AUC value of the area under the ROC curve of the random forest model, and the results are shown in Table 4. As can be seen from Fig. 5 and Table 4, the esophageal cancer diagnostic model constructed above in the present invention has a good effect, and its area under the ROC curve AUC for esophageal cancer diagnosis is 0.895 (0.784～1), the sensitivity is 85.00%, and the specificity is 90.48%.

Furthermore, ROC curve analysis was performed on the predicted values of the esophageal cancer prevalence probability of different stages of the test samples compared with the actual pathological results (esophageal cancer or healthy), to evaluate the diagnostic effect of the diagnostic model on different stages of esophageal cancer. The sensitivity, specificity and AUC values of the area under the ROC curve of the random forest model for different stages of esophageal cancer are shown in Table 4 below. It can be seen from the table that: with the further deterioration of esophageal cancer, the AUC value and specificity tend to increase. Sensitivity is better in carcinoma in situ and advanced carcinoma, and decreases in early carcinoma. Overall, the model has a better diagnostic effect on advanced esophageal cancer, but the diagnostic effect (AUC) of carcinoma in situ and early esophageal carcinoma is also low. It can reach an acceptable value of 0.85 or more, which also has the value of early diagnosis. It also shows that the serum metabolic markers screened by the present invention have metabolic changes in the early stage of esophageal cancer or even carcinoma in situ.

Carcinoma in situ is an earlier stage than early stage (I and II) esophageal cancer, which is more difficult to diagnose in the early stage and relatively easier in the late stage. From the data in the table, the diagnostic model of the present invention can well diagnose whether to suffer from esophageal cancer, and not only has a good diagnostic effect on advanced esophageal cancer, but also has high accuracy, sensitivity and specificity for early esophageal cancer and carcinoma in situ. The accuracy is also good, and it can effectively diagnose carcinoma in situ and early esophageal cancer with no obvious symptoms, which reduces the rate of missed diagnosis of cancer, and is very conducive to the early detection and early treatment of esophageal cancer. The mortality rate is very helpful, and has good clinical use and promotion value.

Table 4. ROC analysis results of external generalization of esophageal cancer diagnostic model

8.2 A combination of 7 serum metabolic markers was used as a diagnostic marker for modeling and used to diagnose esophageal cancer, as follows:

Randomly assign 4/5 of the obtained two-dimensional matrix data as the training sample training data, and the other 1/5 as the external test sample test data (see Table 1). Only use lysophosphatidic acid LPA(18:1(9Z)/0:0), lyso-lecithin Lyso PC(14:0/0:0), lyso-lecithin LysoPC(18:2(9Z,12Z)), hemolys Lecithin LysoPC(24:0), phospholipid PC(14:1(9Z)/P-18:1(11Z)), phospholipid PC(16:0/18:2(9Z,12Z)) and phospholipid PC(24 :1(15Z)/22:6(4Z, 7Z, 10Z, 13Z, 16Z, 19Z)) Seven metabolic markers were used as diagnostic markers, and the early diagnosis of esophageal cancer was constructed based on random forest (random forest t) in the training samples Model. The random forest is implemented using the randomForest software package in the R language, the modeling parameter ntree=5000, and the random forest modeling steps are the same as above.

When using the constructed random model for diagnosis, the data information of the 7 serum metabolic markers in the serum to be tested is imported into the random forest model, and if the voting result of the model classifier is greater than or equal to the diagnostic cut-off value, it is judged as a positive diagnosis ( Esophageal squamous cell carcinoma), if it is lower than the diagnostic cut-off value, it is judged as negative diagnosis (no esophageal squamous cell carcinoma). The diagnostic threshold (Threshold) of this model was 0.7431.

Substitute the two-dimensional matrix data of 7 serum metabolic markers of the external test sample into the random forest model established above to obtain the predicted value of the probability of esophageal cancer of the test sample, and compare it with the actual pathological results (esophageal cancer or healthy) Perform ROC curve analysis (see Figure 6) to obtain the sensitivity, specificity and AUC value of the area under the ROC curve of the random forest model, and the results are shown in Table 5. It can be seen from Fig. 6 and Table 5 that the esophageal cancer diagnostic model constructed above in the present invention has good effect, and its AUC for esophageal cancer diagnosis is 0.876 (0.752-1), the sensitivity is 90%, and the specificity is 85.71%.

Furthermore, ROC curve analysis was performed on the predicted values of the esophageal cancer prevalence probability of different stages of the test samples compared with the actual pathological results (esophageal cancer or healthy), to evaluate the diagnostic effect of the diagnostic model on different stages of esophageal cancer. The sensitivity, specificity and AUC values of the area under the ROC curve of the random forest model for different stages of esophageal cancer are shown in Table 5 below. The degree is better in carcinoma in situ and advanced carcinoma, and decreased in early carcinoma. Generally speaking, the model has a better diagnostic effect on advanced esophageal cancer, but the diagnostic effect (AUC) of carcinoma in situ and early esophageal carcinoma is also lower. It can reach an acceptable value of 0.83 or more, which also has the value of early diagnosis. It also shows that the serum metabolic markers screened by the present invention have metabolic changes in the early stage of esophageal cancer or even carcinoma in situ.

As can be seen from the data in the table, the diagnostic model constructed by the information of 7 serum metabolic markers of the present invention is less effective than the diagnostic model constructed by using the information of 25 serum metabolic markers, but the diagnostic model can also be very good. Diagnose whether you have esophageal cancer, and not only has a good diagnostic effect on advanced esophageal cancer, but also has good accuracy, sensitivity and specificity on early esophageal cancer and carcinoma in situ, and can effectively diagnose in situ cancer with no obvious symptoms. Cancer and early esophageal cancer reduce the rate of missed diagnosis of cancer, which is very conducive to the early detection and early treatment of esophageal cancer, and is very helpful for improving the prognosis of esophageal cancer and reducing the mortality rate of esophageal cancer. value.

Table 5 ROC analysis results of external extension of esophageal cancer diagnostic model

8.3. The combination of 5 serum metabolic markers is used as a diagnostic marker for modeling and used to diagnose esophageal cancer, as follows:

Randomly assign 4/5 of the obtained two-dimensional matrix data as the training sample training data, and the other 1/5 as the external test sample test data (see Table 1). Five serum metabolic markers are used: L-Tyrosine, L-Tryptophan, Glycocholic Acid, Taurocholate and Cortisol As diagnostic markers, an early diagnosis model of esophageal cancer was constructed based on random Forest in the training samples. The random forest is implemented using the randomForest software package in the R language, the modeling parameter ntree=5000, and the random forest modeling steps are the same as above.

When the constructed random model is used for diagnosis, the data information of the five serum metabolic markers in the serum to be tested is imported into the random forest model, and if the voting result of the model classifier is greater than or equal to the diagnostic cut-off value, it is judged as a positive diagnosis ( Esophageal squamous cell carcinoma), if it is lower than the diagnostic cut-off value, it is judged as negative diagnosis (no esophageal squamous cell carcinoma). The diagnostic threshold (Threshold) of this model was 0.4943.

Substitute the two-dimensional matrix data of the five serum metabolic markers of the external test sample into the random forest model established above to obtain the predicted value of the probability of esophageal cancer of the test sample, and compare it with the actual pathological results (esophageal cancer or healthy) Perform ROC curve analysis (see Figure 7), and obtain the sensitivity, specificity and AUC value of the area under the ROC curve of the random forest model, and the results are shown in Table 6. It can be seen from Figure 7 and Table 6 that the esophageal cancer diagnostic model constructed above in the present invention has good effect, and its AUC for esophageal cancer diagnosis is 0.84 (0.703-0.978), the sensitivity is 95%, and the specificity is 76.19%.

Furthermore, ROC curve analysis was performed on the predicted values of the esophageal cancer prevalence probability of different stages of the test samples compared with the actual pathological results (esophageal cancer or healthy), to evaluate the diagnostic effect of the diagnostic model on different stages of esophageal cancer. The sensitivity, specificity and AUC values of the area under the ROC curve of the random forest model for different stages of esophageal cancer are shown in Table 6 below. show different trends.

As can be seen from the data in the table, the diagnostic model constructed by the information of 5 serum metabolic markers of the present invention is less effective than the diagnostic model constructed by using the information of 25 and 7 serum metabolic markers, but the diagnostic model can also It is very good at diagnosing whether you have esophageal cancer, and it is not only good for the diagnosis of advanced esophageal cancer, but also has good accuracy, sensitivity and specificity for early esophageal cancer and carcinoma in situ, and can effectively diagnose symptoms without obvious symptoms. Carcinoma in situ and early esophageal cancer reduce the rate of missed diagnosis of cancer, which is very conducive to early detection and early treatment of esophageal cancer, and is very helpful for improving the prognosis of esophageal cancer and reducing the mortality rate of esophageal cancer. Use and promote value.

Table 6 ROC analysis results of the external extension of the early diagnosis model of esophageal cancer

9. Conclusion

9.1 Any one of the 25 serum metabolic markers obtained in the present invention has a good diagnostic effect as a diagnostic marker for esophageal cancer, but the combined application of multiple serum metabolic markers has a better effect.

9.2 The three preferred diagnostic markers of the present invention (diagnostic markers A, B, C) and the constructed diagnostic model have a good diagnostic effect on esophageal cancer and have clinical application value.

After verification, the diagnostic marker and diagnostic model obtained by the present invention have good application value, and the diagnostic marker and diagnostic model of the present invention can be used clinically to diagnose esophageal cancer. The steps are as follows:

(1) Collect the serum to be tested, and after centrifugation, use the steps (1)-(4) in the above 2.2 to pretreat the serum for sample injection;

(2) Perform LC-MS detection on the pretreated serum sample to be tested according to the above-mentioned 2.3 steps to obtain the original metabolic fingerprint;

(3) Preprocessing the original metabolic fingerprint according to the method of step 3 above, and carrying out metabolite peak identification to obtain a two-dimensional matrix of the serum to be tested;

(4) Screen out the corresponding diagnostic marker (diagnostic marker A, B or C) information from the two-dimensional matrix according to the mass-to-charge ratio and retention time to obtain a two-dimensional matrix of diagnostic markers;

(5) Bring the two-dimensional matrix of diagnostic markers into the corresponding diagnostic model, and judge whether it is esophageal squamous cell carcinoma according to the value given by the model and the diagnostic threshold (Threshold) of the model. When the value given by the model is greater than or equal to the diagnostic cut-off value, it is judged as a positive diagnosis (with esophageal squamous cell carcinoma), and if it is lower than the diagnostic cut-off value, it is judged as a negative diagnosis (without esophageal squamous cell carcinoma).

In addition, in order to speed up the efficiency, serum samples of multiple people can be collected at the same time and numbered, and multiple samples can be subjected to LC-MS detection, spectrum preprocessing, metabolic peak identification, two-dimensional matrix screening of diagnostic markers and data import.

In practical application, more samples can be selected for modeling according to the modeling method of the present invention, so as to increase the accuracy of the model. The above is a description of the patent of the present invention without limitation, and other implementations based on the idea of the patent of the present invention are within the protection scope of the present invention.

Claims (14)

1. A diagnostic marker suitable for early diagnosis of esophageal squamous cell carcinoma, characterized in that it comprises: Serum metabolic marker A: Phospholipid PC (14:1(9Z)/P-18:1(11Z)), and Serum metabolic marker B: Phospholipid PC (24:1(15Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z)). 2. The diagnostic marker according to claim 1, characterized in that: it also includes serum metabolic marker C, which is selected from one or more of the following serum metabolic markers: beta-propanol Amino acid-lysine (beta-Ala-Lys), L-carnosine (L-Carnosine), cis-9-hexadecenoic acid (cis-9-Palmitoleic acid), palmitic acid (Palmitic acid), oleic acid ( Oleic Acid), lysophosphatidic acid LPA(18:1(9Z)/0:0), lysolecithin LysoPC(14:0/0:0), lysolecithin LysoPC(18:2(9Z,12Z)), LysoPC (24:0), phospholipid PC (16:0/18:2 (9Z, 12Z)), linoleic acid (Linoleic acid), nicotinamide adenine dinucleotide (NADH), cortisol ( Cortisol), L-Tyrosine, L-Tryptophan, Glycocholic Acid, Taurocholate, Hypoxanthine, Urine Allantoic acid, Inosine, Sphingosine 1-phosphate, 3-O-Sulfogalactosylceramide (d18:1/20:0), lactose Ceramide Lactosylceramide (d18:1/22:0). 3. The diagnostic marker according to claim 2, characterized in that: it also includes a serum metabolic marker C, which is selected from one or more of the following serum metabolic markers: lysophosphatidic acid LPA(18:1(9Z)/0:0), LysoPC(14:0/0:0), LysoPC(18:2(9Z,12Z)), LysoPC(24: 0), phospholipid PC (16:0/18:2 (9Z,12Z)), nicotinamide adenine dinucleotide (NADH), cortisol (Cortisol), L-tryptophan (L-Tryptophan), bovine Taurocholate, Hypoxanthine, Inosine, Galactosylceramide Sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0), Lactosylceramide (d18 :1/22:0). 4. The diagnostic marker according to claim 3, characterized in that: it also includes a serum metabolic marker C, which is selected from one or more of the following serum metabolic markers: lysolecithin LysoPC(18:2(9Z,12Z)), LysoPC(24:0), Nicotinamide Adenine Dinucleotide(NADH), Cortisol, Taurocholate, Sulfate Galactosylceramide 3-O-Sulfogalactosylceramide (d18:1/20:0), Lactosylceramide (d18:1/22:0). 5. The diagnostic marker according to claim 4, characterized in that: it also includes a serum metabolic marker C, which is selected from one or more of the following serum metabolic markers: lysolecithin LysoPC(18:2(9Z,12Z)), LysoPC(24:0), Taurocholate, Lactosylceramide(d18:1/22:0). 6. The diagnostic marker according to claim 1, characterized in that: it also includes a serum metabolic marker C, which is selected from one or more of the following serum metabolic markers: lysophosphatidic acid LPA(18:1(9Z)/0:0), LysoPC(14:0/0:0), LysoPC(18:2(9Z,12Z)), LysoPC(24: 0), phospholipid PC (16:0/18:2(9Z,12Z)). 7. The diagnostic marker according to claim 1, characterized in that: it also includes a serum metabolic marker C, which is selected from one or more of the following combinations: Combination 1: the combination of beta-Ala-Lysine (beta-Ala-Lys) and L-Carnosine (L-Carnosine); Combination 2: a combination of cis-9-Palmitoleic acid, Palmitic acid and Oleic acid; Combination 3: lysophosphatidic acid LPA (18:1(9Z)/0:0), lyso-lecithin LysoPC(14:0/0:0), lyso-lecithin LysoPC(18:2(9Z,12Z)) and hemolysate Combination of lecithin LysoPC (24:0); Combination 4: Combination of phospholipid PC (16:0/18:2(9Z,12Z)) and linoleic acid (Linoleic acid); Combination five: a combination of nicotinamide adenine dinucleotide (NADH), L-tyrosine (L-Tyrosine) and L-tryptophan (L-Tryptophan); Combination 6: Combination of Cortisol, Glycocholic Acid and Taurocholate; Combination 7: a combination of Hypoxanthine, Allantoic acid and Inosine; Combination eight: Sphingosine 1-phosphate, galactosylceramide sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0) and lactosylceramide (d18:1/22: 0) combination. 8. The diagnostic marker according to claim 1, characterized in that: it also includes serum metabolic marker C, which is a combination of the following 23 serum metabolic markers: beta-alanine- Lysine (beta-Ala-Lys), L-Carnosine (L-Carnosine), cis-9-Hexadecenoic acid (cis-9-Palmitoleic acid), Palmitic acid (Palmitic acid), Oleic acid (Oleic Acid), Lysophosphatidic acid LPA(18:1(9Z)/0:0), lysolecithin LysoPC(14:0/0:0), lysolecithin LysoPC(18:2(9Z,12Z)), lysolecithin LysoPC (24:0), phospholipid PC (16:0/18:2 (9Z,12Z)), linoleic acid (Linoleic acid), nicotinamide adenine dinucleotide (NADH), cortisol (Cortisol), L -L-Tyrosine, L-Tryptophan, Glycocholic Acid, Taurocholate, Hypoxanthine, Allantoic Acid ( Allantoic acid), Inosine, Sphingosine 1-phosphate, 3-O-Sulfogalactosylceramide (d18:1/20:0) and Lactosylceramide (d18:1/22:0). 9. The diagnostic marker according to claim 7, characterized in that: beta-alanine-lysine (beta-Ala-Lys) and L-carnosine (L-Carnosine) and beta alanine metabolism (beta-Ala-Lys) Alanine metabolism) metabolic pathways are closely related; cis-9-Palmitoleic acid, Palmitic acid and Oleic acid are closely related to the metabolic pathway of fatty acid biosynthesis; Lysophosphatidic acid LPA (18:1(9Z)/0:0), lysolecithin LysoPC (14:0/0:0), lysolecithin LysoPC (18:2(9Z,12Z)) and lysolecithin LysoPC (24:0) is closely related to the metabolic pathway of glycerophospholipid metabolism; Phospholipid PC(14:1(9Z)/P-18:1(11Z)), Phospholipid PC(16:0/18:2(9Z,12Z)), Phospholipid PC(24:1(15Z)/22:6 (4Z, 7Z, 10Z, 13Z, 16Z, 19Z)) and linoleic acid are closely related to two metabolic pathways, Glycerophospholipid metabolism and Linoleic acid metabolism; Nicotinamide adenine dinucleotide (NADH) is closely related to oxidative phosphorylation (Oxidative phosphorylation) metabolic pathway; L-Tyrosine (L-Tyrosine) and L-Tryptophan (L-Tryptophan) are closely related to the metabolic pathway of Phenylalanine/tyrosine and tryptophan biosynthesis; Glycocholic Acid and Taurocholate are closely related to the metabolic pathway of primary bile acid biosynthesis; Cortisol is closely related to cancer pathways and bile secretion (Pathways in cancer, and Bile secretion) metabolic pathways; Hypoxanthine, Allantoic acid and Inosine are closely related to the metabolic pathway of Purine metabolism; 1-phosphate sphingosine (Sphingosine 1-phosphate), galactosylceramide sulfate 3-O-Sulfogalactosylceramide (d18:1/20:0) and lactosylceramide (d18:1/22:0) with Sphingolipid metabolism (Sphingolipid metabolism) metabolic pathways are closely related. 10. A method for screening diagnostic markers suitable for early diagnosis of esophageal squamous cell carcinoma according to any one of claims 1-9, characterized in that it comprises the following steps: (1) Serum samples from patients with esophageal squamous cell carcinoma and healthy people were collected as analysis samples, where esophageal squamous cell carcinoma serum samples included esophageal carcinoma in situ, early esophageal cancer and advanced esophageal cancer; (2) Analyze each analysis sample using LC-MS serum metabolomics technology to obtain the original metabolic fingerprint of each serum sample; (3) Use the R language XCMS software package to preprocess the original metabolic fingerprints of esophageal squamous cell carcinoma serum samples and healthy serum samples respectively, and obtain a two-dimensional matrix of metabolite information for each row analysis sample and each column, and Use the R package CAMERA to identify metabolite peaks in the two-dimensional matrix for further statistical analysis; (4) Perform principal component analysis and partial least squares discriminant analysis on the two-dimensional matrix in step (3) in turn to obtain a PLS-DA model, which shows that patients with esophageal squamous cell carcinoma have metabolic patterns different from healthy people and clear classification trends; (5) According to the PLS-DA model obtained above, the differential metabolites were screened with the help of the variable importance score of the PLS-DA modeling and the univariate non-parametric test. The screening criteria were: VIP≥1, and the false discovery rate FDR The q value after multiple testing correction is less than 0.05; (6) Determine the quasi-molecular ions, adducts and isotope information of the differential metabolites obtained from the above screening according to the CAMERA package of the R language, and obtain potential metabolic markers; (7) On the basis of the above potential metabolic markers, combined with the primary and secondary mass spectrometry information, quasi-molecular ion information, adduct information and isotope information of the potential metabolic markers, the molecular mass and molecular formula of the diagnostic markers were estimated, And compared and matched with the existing standard compounds to obtain a diagnostic marker suitable for early diagnosis of esophageal squamous cell carcinoma. 11. The screening method according to claim 10, characterized in that: when performing LC-MS serum metabolomics analysis, every 10 analysis samples are added with a quality control sample for real-time monitoring of the analysis samples from pre-injection processing As to the quality control situation in the analysis process, the quality control sample is a mixed sample of 5 esophageal cancer serum samples and 5 healthy serum samples. 12. The screening method according to claim 10, characterized in that: the analysis sample and the quality control sample are subjected to the following pretreatment before sample introduction: (1) Use a pipette to draw 50 μl of analytical samples or quality control samples, and place them on a 96-well plate of the Bravo automatic specimen processing system; (2) Add 150μl methanol for extraction, vortex for 30s, and incubate at -20°C to precipitate protein; (3) Then centrifuge in a high-speed centrifuge at 4°C at 4000 rpm for 20 minutes; (4) Pour the supernatant from step (3) into an LC-MS sample bottle and store at -80°C for LC-MS detection. 13. The screening method according to claim 10, characterized in that: carrying out spectrum preprocessing to the original metabolic fingerprints refers to: converting the obtained original metabolic fingerprints into MZdata data files with Masshunter software, and then using the Mzdata data files The XCMS software package performs preprocessing operations including retention time correction, peak identification, peak matching and peak alignment to obtain a two-dimensional matrix; use the R software package CAMERA to identify metabolite peaks in the two-dimensional matrix, including isotope peaks, adducts and fragments Metabolite peak identification for the ion. 14. The screening method according to claim 10, characterized in that: when each analysis sample is analyzed by LC-MS serum metabolomics technology, the chromatographic column used in liquid chromatography is Waters ACQUITY UPLC HSS T3 chromatographic column, specification It is 100 mm×2.1 mm, 1.8 μm; the injection volume is 6 µL, the injection temperature is 4°C, and the flow rate is 0.5 ml/min; the chromatographic mobile phase contains two solvents A and B: A in positive ion ESI+ mode is 0.1 wt% formic acid aqueous solution, A under negative ion ESI-model is 0.5mmol/L ammonium fluoride aqueous solution, B under positive ion ESI+ mode is 0.1wt% formic acid acetonitrile solution, B under negative ion ESI-model is pure acetonitrile; Gradient elution conditions are: 0-1min is 1%B, 1-8min is 1%B-100%B gradually increasing, 10-10.1min is 100%B rapidly reduces to 1%B, and then 1%B lasts for 1.9min ; The mass spectrometry detection uses the quadrupole time-of-flight mass spectrometer Q-TOF, and adopts the positive ion mode ESI+ and the negative ion mode ESI- of the electrospray ion source. The temperature is 250℃, the flow rate of desolvation gas is 16L/min; the capillary voltage is +3kV and -3kV respectively in positive ion mode and negative ion mode, and the cone voltage is 0V; the cone pressure in positive ion mode is 20psi, and in negative ion mode The cone pressure is 40psi; the mass-to-charge ratio range of the spectral data collection is 50-1200 m/z, and the scanning frequency of collection is 0.25s.