Metabolic markers for gastric cancer diagnosis and their applications

Six metabolic markers for gastric cancer diagnosis were screened through ultra-high performance liquid chromatography-mass spectrometry technology, and combined with machine learning models, a gastric cancer diagnosis model was constructed, which solved the problems of low diagnostic accuracy and poor resource utilization efficiency in traditional methods, and achieved the Higher diagnostic accuracy and early recognition capabilities.

CN118671346BActive Publication Date: 2024-11-15ZHEJIANG CANCER HOSPITAL
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411157760.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-15
Estimated Expiration
2044-08-22

AI Technical Summary

Technical Problem

The existing gastric cancer diagnosis methods have problems with low accuracy and poor resource utilization efficiency, especially in the screening of high-risk populations. The traditional risk assessment model has poor effect and is difficult to effectively identify early gastric cancer.

Method used

Six metabolic markers of gastric cancer diagnosis (2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovaleryl carnitine, and glyceroldeoxycholic acid were screened through ultra-high performance liquid chromatography-mass spectrometry. , 2-hydroxy-3-methylbutyric acid and N-acetyl neuraminine), combined with lasso regression score and a variety of machine learning models, build gastric cancer diagnostic models to improve the sensitivity and specificity of diagnosis.

Benefits of technology

It significantly improves the diagnostic accuracy and early diagnosis ability of gastric cancer, has higher sensitivity and specificity compared with traditional methods, and can more effectively identify high-risk groups, reduce unnecessary interventions, and improve patient prognosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118671346B_ABST
    Figure CN118671346B_ABST
Patent Text Reader

Abstract

The present invention provides a metabolic marker for gastric cancer diagnosis and its application, wherein the metabolic marker is at least one of 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid. The metabolic marker has excellent diagnostic accuracy, and combined application can effectively diagnose gastric cancer and early gastric cancer, and its diagnostic efficiency is significantly better than traditional clinical tumor markers. Compared with traditional risk factors, it has higher diagnostic accuracy, better risk management, higher patient benefits and more effective resource utilization, and has excellent overall benefits in clinical applications, and ultimately improves the diagnosis and treatment effect and quality of life of patients.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and relates to metabolic markers for gastric cancer diagnosis and applications thereof, and in particular to a method for identifying key metabolic markers of gastric cancer and constructing a machine learning gastric cancer diagnostic model. Background Art

[0002] Gastric cancer (GC) is a common malignant tumor of the digestive tract. It has an insidious onset and more than 80% of patients are already in the middle or advanced stages when diagnosed. Therefore, the current status of gastric cancer prevention and treatment is very grim.

[0003] Endoscopic screening is the most effective means to reduce gastric cancer mortality. However, considering that gastroscopy is invasive and expensive, the population has poor compliance and low participation rate, and the diagnosis of gastric cancer is highly dependent on experienced physicians and sophisticated examination equipment. Limited by the large population base and limited medical resources, large-scale endoscopic screening does not meet the screening principles of effectiveness, ease, and economy. Identifying and concentrating high-risk populations is the key to increasing the benefits of endoscopic screening and is also the most important first step in gastric cancer screening.

[0004] The investigation of relevant epidemiological risk factors was once used as a preliminary screening enrichment method before gastroscopy. However, based on literature review and analysis of previous screening project practices, the investigation of relevant epidemiological risk factors based on traditional macro risk factors, Helicobacter pylori infection ( H.pylori The traditional gastric cancer risk assessment model based on 24-hour serum albumin (SAL), pepsinogen (PG) and gastrin-17 (G-17) has poor enrichment effect on high-risk populations. The gastric cancer detection rate based on the traditional risk assessment model is low (only 1%), the enrichment effect on high-risk populations is poor, and the efficiency is low. In view of the above bottleneck problems, there is an urgent need to develop non-invasive, accurate, reliable and suitable biomarkers for promotion for enrichment of high-risk populations for gastric cancer, which is an important breakthrough in reducing the disease burden of gastric cancer in the context of precise prevention and screening.

[0005] Blood metabolic markers have been proven to directly reveal cancer phenotypes and play an important role in the occurrence and development of gastric cancer. They are expected to serve as potential markers for identifying people at high risk of gastric cancer. Compared with upstream molecular biomarkers such as nucleic acids and proteins, metabolic molecules as downstream molecular biomarkers can directly reveal cancer phenotypes. Before the onset of the disease or during the accumulation of risk factors, endogenous substances will respond metabolically and more sensitively identify specific metabolic types caused by genes and environmental factors. In addition, the number of metabolites is far less than the number of genes and proteins they correspond to. Minor changes in gene and protein expression will be amplified at the metabolite level, which helps to detect minor changes in metabolic characteristics with higher sensitivity. At present, the widespread application of high-throughput sequencing and high-sensitivity mass spectrometry has greatly improved the detection level of blood metabolomes and the potential for discovering new biomarkers. Ultra-high performance liquid chromatography-tandem mass spectrometry (UPLC-MS / MS) is the current mainstream technology for metabolomics detection. It has the advantages of high sensitivity, high resolution, high throughput and deep coverage by accurately measuring the mass-to-charge ratio of metabolites and their fragments at a high resolution of about ppm. In summary, metabolomics, as the terminal of omics research, reflects the final effect of multiple factors on the organism, and has a strong comprehensive information advantage. Studying the metabolic reprogramming of tumor cells is becoming a new idea for tumor prevention, diagnosis and targeted treatment.

[0006] The above-mentioned record of background technical knowledge is intended to help ordinary technicians in this field understand the existing technologies that are relatively close to the present invention, and at the same time facilitate the understanding of the inventive concept and technical solution of the present invention. It should be clear that in the absence of clear evidence that the above-mentioned content has been disclosed before the filing date of this patent application, the above-mentioned background technology should not be used to evaluate the novelty of the technical solution of this application. Summary of the invention

[0007] Technical issues

[0008] In order to solve the above problems, the purpose of the present invention is to provide a metabolic marker for gastric cancer diagnosis and its application. The metabolic marker has excellent diagnostic accuracy, and can effectively diagnose gastric cancer and early gastric cancer when used alone or in combination. Its diagnostic efficacy is significantly better than traditional clinical tumor markers. Compared with traditional risk factors, it has higher diagnostic accuracy, better risk management, higher patient benefits and more effective resource utilization. It has excellent overall benefits in clinical applications, and ultimately improves the patient's diagnosis and treatment effects and quality of life.

[0009] Technical Solution

[0010] In order to achieve the above-mentioned purpose, the inventors of the present application conducted in-depth research and found that consistent differential metabolites were screened out through ultra-high performance liquid chromatography-mass spectrometry technology, optimized screening conditions, and external validation set verification. Finally, a lasso regression scoring method was adopted, and the AUC of the model was calculated under different numbers of selected features and different LRScore thresholds for optimization. Six gastric cancer diagnostic metabolic markers were obtained, which have higher AUC values ​​and significantly better sensitivity than traditional blood biomarkers. They can be used for gastric cancer diagnosis and early gastric cancer diagnosis, thus completing the present invention.

[0011] That is, the present invention includes the following aspects.

[0012] Scheme 1, a metabolic marker for gastric cancer detection, wherein the metabolic marker is at least one of 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid.

[0013] Furthermore, the metabolic marker is a combination of 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid.

[0014] Scheme 2: Use of a reagent for specifically detecting the aforementioned metabolite markers in the preparation of a product for diagnosing gastric cancer.

[0015] Furthermore, the diagnosis is to determine the disease condition of gastric cancer in the diagnosed subject based on the expression level of the marker in the biological sample of the diagnosed subject.

[0016] Scheme 3: Use of reagents for specifically detecting the enrichment of metabolites of ketone bodies, methylhistidine, β-alanine, fatty acids and histidine metabolic pathways in the preparation of products for diagnosing gastric cancer.

[0017] Furthermore, in Scheme 2-3, the diagnosis of gastric cancer includes diagnosing early gastric cancer and / or distinguishing between gastric cancer and non-malignant diseases of the stomach.

[0018] Furthermore, in Scheme 2-3, the product includes a chip, a test paper, a reagent or a kit.

[0019] Furthermore, in Scheme 2-3, the kit also includes a quality control product or a standard product.

[0020] Furthermore, in Schemes 2-3, the sample used in the diagnostic process is selected from serum, plasma, blood, dried blood spots, urine, dried urine spots or body fluids.

[0021] Scheme 4, the method for screening the aforementioned metabolite markers comprises the following steps:

[0022] (1) Blood samples from gastric cancer patients and healthy controls were collected, and initial plasma metabolites were identified by ultra-high performance liquid chromatography-mass spectrometry. The candidate plasma differential metabolites were screened out between gastric cancer patients and healthy controls by using P.adjusted and Fold Change screening conditions.

[0023] (2) consistent differential metabolites were verified in a multicenter independent external validation set;

[0024] (3) A lasso regression scoring method was used, and the AUC of the model was optimized by calculating different numbers of selected features and different LRScore thresholds. Finally, a panel of six differential metabolites was identified, namely 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid, and N-acetylneuraminic acid.

[0025] In the scheme of the present invention, firstly, candidate differential metabolites in plasma are screened between gastric cancer and healthy controls, then consistent differential metabolites are verified through the validation set, and then six differential metabolite panels are determined based on the lasso regression scoring method and AUC optimization, and a gastric cancer diagnosis machine learning model is constructed based on the six differential metabolite panels, and the model is used to trace the differential metabolites in tissues to screen out metabolic markers that can diagnose gastric cancer. The panel of six metabolites has excellent performance, including sensitivity, specificity, accuracy and area under the curve. The combination of six potential biomarkers has better net benefits than traditional risk factors, and the sensitivity is significantly better than traditional blood biomarker detection, and it also shows strong performance in the diagnosis of early gastric cancer.

[0026] Furthermore, the ultra-high performance liquid chromatography-tandem mass spectrometry technique uses four parallel methods:

[0027] Analysis was performed using a reverse phase column ultra-high performance liquid chromatography-positive electrospray ionization mode;

[0028] Analysis of lipid metabolites using reverse phase ultra-high performance liquid chromatography-positive ion electrospray ionization mode;

[0029] Reversed phase column ultra-high performance liquid chromatography-negative ionization electrospray ionization mode analysis;

[0030] Analysis by ultra-high performance liquid chromatography-negative ion electrospray ionization mode on a hydrophilic interaction chromatography column.

[0031] Scheme 5, a method for diagnosing gastric cancer and constructing an early gastric cancer diagnosis model based on the aforementioned metabolic markers, comprising the following steps:

[0032] (1) Blood samples from gastric cancer patients and healthy controls were collected, and initial plasma metabolites were identified by ultra-high performance liquid chromatography-mass spectrometry. The candidate plasma differential metabolites were screened out between gastric cancer patients and healthy controls by using P.adjusted and Fold Change screening conditions.

[0033] (2) consistent differential metabolites were verified in a multicenter independent external validation set;

[0034] (3) A lasso regression scoring method was used and the AUC of the model was calculated for optimization under different numbers of selected features and different LRScore thresholds, and the six metabolic markers were finally determined;

[0035] (4) Based on the six metabolic markers in step (3), a gastric cancer diagnosis model was constructed using machine learning models including neural network model (NN), support vector machine (SVM), ridge regression (RR), logistic regression (LR) and naive Bayes (NB) model.

[0036] Adopting advanced integrated learning mechanism, integrating multi-source heterogeneous model algorithms for collaborative learning, building a multi-level model fusion framework, including early feature engineering, mid-term model optimization and late decision fusion, through weighted fusion and stacking generalization technology, intelligent algorithm design adaptive weight dynamic adjustment allocation mechanism, reduce overfitting risk, optimize overall prediction performance. This study involves the strategic fusion of multiple machine learning algorithms such as neural network NN, naive Bayesian network BN, support vector machine SVM, ridge regression RR, logistic regression LR and random forest RF.

[0037] Scheme 6, the gastric cancer diagnosis and early gastric cancer diagnosis model obtained by the construction method described above.

[0038] Solution 7, a computer device, comprising a memory, a processor, a communication interface and a communication bus; wherein the memory, the processor and the communication interface communicate with each other via the communication bus; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory, and when the processor executes the computer program, the method described in Solution 5 is implemented.

[0039] Solution 8, a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method described in Solution 5 when executed by a processor.

[0040] Based on the common knowledge in this field, the above-mentioned preferred conditions can be combined with each other to obtain a specific implementation method.

[0041] Beneficial Effects

[0042] According to the present invention, initial plasma metabolites are centrally identified in blood samples of gastric cancer and healthy controls by ultra-high performance liquid chromatography-mass spectrometry technology, and then the screening conditions are optimized to screen out candidate plasma differential metabolites, and then the consistent differential metabolites are verified in a multi-center independent external validation set, and finally a method based on lasso regression scoring is adopted, and the AUC of the model is calculated under different numbers of selected features and different LRScore thresholds for optimization, and finally six differential metabolites are identified, and the panel performance of the six metabolites is superior, including sensitivity, specificity, accuracy and area under the curve. The combination of the six potential biomarkers has better net benefit than traditional risk factors, and the sensitivity is significantly better than traditional blood clinical tumor marker detection. It shows strong performance in the diagnosis of gastric cancer and early gastric cancer, and is significantly better than traditional risk factors and traditional clinical tumor markers in terms of diagnostic sensitivity, and has the potential to be a new diagnostic marker for gastric cancer. Its simple detection operation, minimally invasive sampling, easy interpretation and high acceptance are expected to achieve the identification and screening of high-risk populations for gastric cancer in large-scale applications. In addition, it helps to reduce unnecessary interventions, improve prognosis, and has excellent overall benefits in clinical applications, ultimately improving patients' diagnosis and treatment effects and quality of life.

[0043] The present invention adopts an advanced integrated learning mechanism, involving a variety of machine learning algorithms such as neural network NN, Bayesian network BN, support vector machine SVM, ridge regression RR, logistic regression LR and random forest RF for strategic fusion, integrating multi-source heterogeneous model algorithms for collaborative learning, and building a multi-level model fusion framework, including early feature engineering, mid-term model optimization and late decision fusion. Through weighted fusion and stacking generalization technology, the intelligent algorithm designs an adaptive weight dynamic adjustment allocation mechanism to reduce the risk of overfitting and optimize the overall prediction performance.

[0044] The 26 plasma differential metabolites screened through multi-center independent external validation were validated in tissues to verify whether these 26 plasma differential metabolites were differentially expressed in gastric cancer tissues and adjacent non-malignant tissues. Eight key functional metabolic factors that were also differentially expressed in gastric cancer tissues and adjacent non-malignant tissues were screened and validated, locating the upstream and downstream key regulatory gene targets, which can be used for mechanism exploration.

[0045] The present invention adopts the above technical solution to achieve the above purpose, which makes up for the shortcomings of the prior art and has reasonable design and convenient operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to make the above and / or other purposes, features, advantages and examples of the present invention more obvious and easy to understand, the drawings required for use in the specific implementation of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 Represents the ROC curve of the combined machine learning model of six blood metabolic markers in the training set and validation set for gastric cancer diagnosis;

[0048] Figure 2 Represents the ROC curves of six blood metabolic markers in the training set and validation set for gastric cancer diagnosis;

[0049] Figure 3 Represents the ROC curves of traditional risk factors, combination of six blood metabolic markers, and combination of traditional risk factors + six blood metabolic markers for gastric cancer diagnosis in the training set and validation set;

[0050] Figure 4 Clinical decision curves representing traditional risk factors, a combination of six blood metabolic markers, and a combination of traditional risk factors + six blood metabolic markers;

[0051] Figure 5 It shows the comparison results of the diagnostic sensitivity of six blood metabolic markers and traditional clinical tumor markers for gastric cancer in the training set and validation set;

[0052] Figure 6 The ROC curves of the machine learning model combining six blood metabolic markers in the training set and validation set for the diagnosis of early gastric cancer are shown;

[0053] Figure 7 Heat map showing the differential expression of eight gastric cancer tissue-tracing key metabolites in gastric cancer tissues and their paired adjacent non-malignant tissues;

[0054] Figure 8 It indicates the fold changes of the differential expression of the eight differential metabolites in the training set, validation set and tissues;

[0055] Fig. 9 It represents the OR values ​​and 95% confidence intervals of the eight differential metabolites in the training set, validation set and tissues;

[0056] Fig.10 Pathway enrichment analysis representing differential metabolites between gastric cancer tissues and paired adjacent normal tissues. DETAILED DESCRIPTION

[0057] Those skilled in the art can refer to the content of this article and appropriately replace and / or modify the process parameters to achieve the same. However, it should be particularly noted that all similar replacements and / or modifications are obvious to those skilled in the art and are considered to be included in the present invention. The products and preparation methods described in the present invention have been described through preferred examples. It is obvious that relevant personnel can modify or appropriately change and combine the products and preparation methods described herein without departing from the content, spirit and scope of the present invention to implement and apply the technology of the present invention.

[0058] Unless otherwise defined, technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. The present invention uses the methods and materials described herein; however, other suitable methods and materials known in the art may also be used. The materials, methods, and examples described herein are illustrative only and are not intended to be limiting. All publications, patent applications, patents, provisional applications, database entries, and other references mentioned herein are incorporated herein by reference in their entirety. In the event of a conflict, the present specification, including definitions, shall prevail.

[0059] Unless specifically stated, the materials, methods, and examples described herein are illustrative only and not limiting. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are still described herein.

[0060] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present invention. At the same time, the embodiments in this application and the features in the embodiments can be combined with each other without conflict.

[0061] It should be understood that any technical solution claimed for protection in the present invention does not involve the diagnosis and treatment of diseases.

[0062] In order to facilitate understanding of the embodiments of the present invention, the abbreviations and key terms that may be involved in the embodiments of the present invention are first explained or defined. For undefined abbreviations or key terms, they are commonly understood by those skilled in the art.

[0063] UPLS-MS: ultra-performance liquid chromatography-mass spectrometry;

[0064] LRScore: Lasso regression score;

[0065] NN: neural network model;

[0066] SVM: support vector machine;

[0067] RR: Ridge regression;

[0068] LR: logistic regression;

[0069] NB: Naive Bayes model;

[0070] GC: Gastric cancer.

[0071] It should be understood that the term "metabolomics" as used herein refers to the use of chromatography-mass spectrometry to detect metabolites with a molecular weight of less than 1000 Daltons. It should be understood that the term "metabolite" or "metabolite" as used herein refers to a substance produced during a chemical or physical process in the human body. It includes any chemical or biochemical product of a metabolic process, such as any compound produced by processing, cleavage or consumption of a biomolecule. Examples of such molecules include, but are not limited to: acids and related compounds; mono-, di- and tri-carboxylic acids (saturated, unsaturated and aliphatic and cyclic, aromatic, alkaryl); aldehydes, keto acids; lactone forms; gibberellins; abscisic acid; alcohols, polyols, derivatives and related compounds; ethanol, benzyl alcohol, methanol; propylene glycol, glycerol, phytol; inositol, furfuryl alcohol, menthol; aldehydes, ketones, quinones, derivatives and related compounds; acetaldehyde, butyraldehyde, benzaldehyde, acrolein, furfural, glyoxal; acetone, butanone; anthraquinone; carbohydrates; monosaccharides, disaccharides, trisaccharides; alkaloids, amines and other bases; pyridines (including nicotinic acid, nicotinamide); pyrimidines (including cytosine, thymine ); purines (including guanine, adenine, xanthine / hypoxanthine, kinetin); pyrroles; quinolines (including isoquinolines); morphinans, tropanes, cinchonans, nucleotides, oligonucleotides, derivatives and related compounds; guanosine, cytosine, adenosine, thymidine, inosine; amino acids, oligopeptides, derivatives and related compounds; esters; phenols and related compounds; heterocyclic compounds and derivatives; pyrroles, tetrapyrroles; flavonoids; indoles; lipids (including fatty acids and triglycerides), derivatives and related compounds; carotenoids, phytoene and sterols, isoprenoids, including terpenes; and modified forms of the above molecules.

[0072] Unless otherwise indicated, the experiments and procedures described in the examples were performed essentially according to conventional methods well known in the art and described in various references. For example, conventional techniques of immunology, biochemistry, chemistry, molecular biology, microbiology, cell biology, genomics and recombinant DNA used in the present invention can be found in Sambrook, Fritsch and Maniatis, MOLECULAR CLONING: A LABORATORY MANUAL, 2nd edition (1989); CURRENT PROTOCOLS IN MOLECULAR BIOLOGY (F.M. Ausubel et al., eds., (1987)); METHODS IN ENZYMOLOGY series (Academic Press): PCR 2: A PRACTICAL APPROACH (M.J. MacPherson, B.D. Heims, et al., eds., (1987)); .Hames and G.R.Taylor (eds. (1995)), and ANIMAL CELL CULTURE (R.I.Freshney (ed. (1987)).

[0073] In addition, the test methods used in the examples are conventional methods unless otherwise specified; the materials, reagents, etc. used are all commercially available unless otherwise specified. The reagents or instruments used without indicating the manufacturer are all conventional products that can be obtained commercially. All the publications and other references mentioned in this article are incorporated herein by reference in their entirety.

[0074] Embodiment 1:

[0075] Obtain blood samples from patients with gastric cancer and healthy controls

[0076] Based on high-throughput non-targeted and ultra-efficient metabolomics detection, it is divided into a training set and an independent external validation set, and a gastric cancer diagnostic model is constructed through multiple machine learning models.

[0077] The present invention collects a total of 277 samples from gastric cancer patients and 320 samples from healthy controls, which are randomly divided into a training set and a validation set. The training set has a total of 144 cases, including 50 cases of gastric cancer and 94 healthy controls, and all clinical blood samples used are from Zhejiang Cancer Hospital; the validation set has a total of 453 cases, including 227 cases of gastric cancer and 226 healthy controls, and the clinical blood samples used are from seven multicenters, including Zhejiang Cancer Hospital, Sichuan Cancer Hospital, Zhejiang Provincial Hospital of Traditional Chinese Medicine, Fenghua People's Hospital, Xinchang People's Hospital, Daishan First People's Hospital, and Tiantai Hospital of Traditional Chinese Medicine.

[0078] Embodiment 2:

[0079] Screening for metabolic markers in gastric cancer

[0080] Through ultra-high performance liquid chromatography-mass spectrometry (UPLS-MS) technology, a total of 1008 plasma metabolites were identified in the discovery set; a total of 93 plasma candidate differential metabolites were screened between gastric cancer and healthy people; the candidate differential metabolites were verified in the validation set, and finally 26 blood differential metabolites were verified in the multi-center independent external validation set; among the 26 blood differential metabolites, 10 were upregulated and 16 were downregulated, mainly lipids and amino acids, and were mainly enriched in ketone bodies, methylhistidine, β-alanine, fatty acids and histidine metabolic pathways; a method based on The lasso regression score (LRScore) method was used, and the AUC of the model was optimized by calculating different numbers of selected features and different LRScore thresholds (TLRScore) to further screen the differential metabolites. Based on the optimized TLRScore (0.87), six differential metabolite panels and metabolic markers for gastric cancer detection were finally determined, namely 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid.

[0081] 1. Sample pretreatment and metabolite extraction: Plasma samples need to be mixed with methanol to precipitate proteins, and tissue samples need to be minced and ground first. Methanol is added to the sample at a ratio of 1:4, and after oscillation for 3 minutes to mix, insoluble matter is precipitated by centrifugation at 20°C 4000×g for 10 minutes. After centrifugation, 4 portions of 100 μL of supernatant are taken from each sample and placed in 4 sample plates, dried with nitrogen, and added to the reconstitution solution for subsequent UPLC-MS / MS detection.

[0082] 2. Ultra-high performance liquid chromatography-tandem mass spectrometry detection: Methanol was added to the sample at a ratio of 1:4, and after oscillation for 3 minutes to mix, the insoluble matter was precipitated by centrifugation at 4000×g at 20°C for 10 minutes. After centrifugation, 4 portions of 100 μL of supernatant were taken from each sample and placed in 4 sample plates, dried with nitrogen, and added to the reconstitution solution for subsequent UPLC-MS / MS detection. The four different UPLC-MS / MS methods were all carried out using ACQUITY 2D UPLC (Waters, Milford, MA, USA) ultra-high performance liquid chromatography combined with QExactive (QE) high-resolution mass spectrometry (Thermo Fisher Scientific, San Jose, USA). The QE mass spectrometry parameters were a scanning resolution of 35,000 and a scanning range of 70-1000 m / z.

[0083] The specific UPLC-MS / MS method parameters of the four parallel methods are as follows: Method 1: QE was detected in positive ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLC BEH C18, 2.1x100mm, 1.7um; Waters), and the mobile phase was water (A) and methanol (B) containing 0.05% PFPA and 0.1% FA; Method 2: QE was detected in negative ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLC BEH C18, 2.1x100mm, 1.7um; Waters), and the mobile phase was water (A) and methanol (B) containing 6.5mM ammonium bicarbonate; Method 3: QE was detected in positive ion electrospray ionization (ESI) mode, and the liquid phase was separated using a C18 column (UPLC BEH C18, 2.1x100mm, Method 4: QE was detected in negative ion electrospray ionization (ESI) mode, and the liquid phase was separated using a HILIC column (UPLC BEH Amide, 2.1x150mm, 1.7um; Waters), and the mobile phase was water containing 10mM ammonium formate (A) and acetonitrile (B).

[0084] 3. Identification and quantification of searched substances: First, pre-process the test data, filter the data, and confirm that it meets the quality control requirements. Then use the self-developed data processing system to extract the chromatographic peaks, and then compare the peaks based on the retention time, the first-level accurate mass number and the second-level fragment spectrum. Figure 3 The metabolites were identified by strict matching of the indicators, and the area under the chromatographic peak curve was calculated for metabolite quantification.

[0085] 4. Data standardization and normalization: First, the original peak area of ​​each metabolite was transformed Log2 to reduce the skewed distribution of the overall value and make the data close to normal distribution. Then the median was used for normalization, and finally the minimum value of all samples was used to fill the missing values.

[0086] 5. Metabolic detection quality control: A strict quality control system is in place during sample preparation and instrument analysis. An internal standard mixture (IS) is added to each sample to help chromatographic peak calibration and instrument stability monitoring. The average relative standard deviation (RSD) of the IS in each sample is calculated to determine the instrument variability. The average RSD in this study was ≤5%, which met the quality control standards. In addition, a small portion of each experimental sample was combined as a mixed quality control sample (Pooled QC), and 4 pooled quality control samples were evenly inserted before, during, and after each analytical sequence for quality control. To avoid elution carryover, a blank sample was inserted every 10 samples for monitoring quality control (Blank QC).

[0087] 5. Statistical analysis: The screening of significantly different metabolites uses parametric tests (student's t-Test, ANOVA, etc.) or non-parametric tests (Wilcox's rank test, Kruskal-Wallis, etc.). Principal component analysis (PCA) was used to explore the law of the overall distribution of high-dimensional data and compare the differences in metabolic patterns between groups. The discriminative ability of metabolic profiles was explored by supervised learning method orthogonal partial least squares discriminant analysis (OPLS-DA). Multivariate logistic regression analysis was used to explore the association between metabolic molecule expression and the risk of gastric cancer. Differential expression analysis was performed in combination with Benjamini-Hochberg FDR multiple correction P value, fold change (FC) value, and variable projection importance (VIP) to screen differential metabolites in gastric cancer, and differential metabolites were visualized through volcano plots, box plots, heat maps, etc. Representative metabolic features were selected by Lasso regression threshold (LRScore) to construct a gastric cancer metabolic feature panel, and a total of six blood metabolic markers for gastric cancer detection were screened, including 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid.

[0088] 6. Construct a gastric cancer diagnostic model: Comprehensively and systematically integrate macro risk factors such as H. pylori infection, environmental lifestyle, etc., and use neural network NN, Bayesian network BN, support vector machine SVM, ridge regression RR, Lasso regression LR and random forest RF machine learning model algorithms to construct a precise gastric cancer model. Model validation combines 5-fold crossover internal validation and independent external validation. Sensitivity, specificity, accuracy, area under the curve, positive predictive value, negative predictive value, and the number of gastroscopes required for each gastric cancer detection are used as evaluation indicators to evaluate the modeling effects of various machine learning models and determine the optimal model. Functional pathway enrichment analysis is performed based on the KEGG database, focusing on key metabolic pathways and functional metabolic molecular regulatory genes.

[0089] Figure 1 The ROC curves of the six blood metabolic markers combined with the machine learning model for gastric cancer diagnosis in the training set and the validation set are shown. It can be seen that in the training set, the AUC of the six blood metabolic markers combined with the machine learning model for gastric cancer diagnosis is 0.947-0.982, especially the LR, RR, SVM and NN models can all reach above 0.95, and in the validation set, the AUC is 0.937-0.951, which has a relatively excellent diagnostic accuracy and can effectively diagnose gastric cancer.

[0090] Figure 2 The ROC curves of six blood metabolic markers used alone and in combination for gastric cancer diagnosis in the training set and validation set are shown. It can be seen that plasma metabolic markers can be used for gastric cancer diagnosis in both the training set and the validation set, and the combination of six blood metabolic markers is more effective in diagnosing gastric cancer than using them alone.

[0091] Figure 3 The ROC curves of traditional risk factors, a combination of six blood metabolic markers, and a combination of traditional risk factors + six blood metabolic markers for gastric cancer diagnosis in the training set and validation set are shown. It can be seen that the AUC values ​​of traditional risk factors are 0.692 (training set) and 0.595 (validation set), which means that they can only exert the lowest diagnostic efficacy. The application of the combination of six blood metabolic markers can exert a diagnostic efficacy of AUC values ​​as high as 0.982 (training set) and 0.951 (validation set), which is much better than the traditional risk factors. In addition, it was unexpectedly found that the combination of traditional risk factors + six blood metabolic markers reduced the diagnostic effect of the metabolic marker combination. The possible reason is that the traditional risk factors disrupt the diagnostic effect of the metabolic markers.

[0092] Figure 4The clinical decision curves of traditional risk factors, a combination of six blood metabolic markers, and a combination of traditional risk factors + six blood metabolic markers are shown. It can be seen that the net benefit of using a combination of six blood metabolic markers combined with a neural network NN to diagnose gastric cancer is the highest, which is better than the traditional risk factors + six blood metabolic markers combination, and significantly better than traditional risk factors. Therefore, it can be seen that the application of a combination of six blood metabolic markers combined with a neural network model to diagnose gastric cancer has higher diagnostic accuracy, better risk control, higher patient benefits, and more efficient resource utilization, which helps to reduce unnecessary interventions, improve prognosis, and has excellent overall benefits in clinical applications, ultimately improving patients' diagnosis and treatment effects and quality of life.

[0093] Figure 5 The comparison results of the diagnostic sensitivity of six blood metabolite markers and traditional clinical tumor markers (CA724, CA199, CA242, CEA, CA125, AFP) for gastric cancer in the training set and validation set are shown. It can be seen that the diagnostic sensitivity of the six blood metabolite combination for gastric cancer is significantly better than that of traditional clinical tumor markers, and it also shows strong performance in the diagnosis of early gastric cancer.

[0094] Figure 6 The ROC curves of the machine learning model combined with six blood metabolic markers for the diagnosis of early gastric cancer in the training set and validation set are shown. It can be seen that in the training set, the AUC of the machine learning model combined with six blood metabolic markers for the diagnosis of early gastric cancer is 0.914-0.961, especially the SVM model can reach above 0.95, while in the validation set, the AUC is 0.894-0.940, which has a relatively excellent diagnostic accuracy and can effectively diagnose early gastric cancer.

[0095] Embodiment 3:

[0096] Screening key metabolic factors in gastric cancer

[0097] Obtain paired samples of gastric cancer tissue and its adjacent non-malignant tissue, trace the expression differences of blood differential metabolites in gastric cancer tissue, and screen key metabolic molecules of gastric cancer. The specific steps include: verifying the 26 blood differential metabolites screened by multi-center independent external validation in Example 2 in tissues, verifying whether these 26 blood differential metabolites are differentially expressed in gastric cancer tissue and adjacent non-malignant tissue, and verifying and screening to obtain eight key metabolic factors that are also differentially expressed in gastric cancer tissue and adjacent non-malignant tissue, that is, eight differential metabolites: N-acetyl Neuraminidase, acisoga, 3-hydroxybutyric acid (BHBA), butyroylcarnitine (C5:1-DC), linoleoylarachidonylglyceride (18:2 / 20:4), oleoylarachidonylglyceride (18:1 / 20:4), C-glycosyltryptophan and N6-succinyladenosine; the enrichment of eight key metabolic factors in gastric cancer pathways showed that glycine and serine metabolism, methionine metabolism and amino sugar metabolism pathways were significantly enriched in gastric cancer patients, providing new clues for the etiology of gastric cancer and new evidence for the development of new therapeutic targets for gastric cancer. Figure 7 The heat map of the key metabolites of the above eight gastric cancer tissues is shown. Figure 8 The fold changes of the differential expression of eight differential metabolites in the training set, validation set and tissues are shown. Fig. 9 The OR values ​​and 95% confidence intervals of eight differential metabolites in the training set, validation set, and tissues are shown. Fig.10 The schematic diagram of the metabolite enrichment pathway of gastric cancer tissue traceability is shown. Figure 7-10 It can be seen that the above eight key metabolites traceable to gastric cancer tissues are differentially expressed in gastric cancer tissues and adjacent normal tissues.

[0098] Based on the above content, the above eight differential metabolites can also be used as metabolic markers for diagnosing gastric cancer. Furthermore, reagents for specifically detecting the eight differential metabolites can also be used in the preparation of products for diagnosing gastric cancer. Furthermore, reagents for specifically detecting the enrichment level of metabolites in the above metabolic pathways can also be used in the preparation of products for diagnosing gastric cancer.

[0099] Machine learning and artificial intelligence play a key role in the processing and interpretation of high-dimensional data. Traditional analysis methods are difficult to handle, while machine learning algorithms efficiently process large-scale data, simplify the data analysis process, help reduce data dimensionality and select the most representative features, dig out metabolic markers related to disease states, automatically discover potential patterns and laws in data, and help understand the complexity of biological systems. At the same time, by training classification models to distinguish different physiological states and disease types, the diagnostic efficacy, generalization ability and interpretability of diseases are improved. The present invention adopts an advanced integrated learning mechanism, integrates multi-source heterogeneous model algorithms for collaborative learning, and constructs a multi-level model fusion framework, including early feature engineering, mid-term model optimization and late decision fusion. Through weighted fusion and stacked generalization technology, intelligent algorithms design adaptive weight dynamic adjustment allocation mechanisms to reduce the risk of overfitting and optimize overall prediction performance. This study involves a variety of machine learning algorithms such as neural network NN, Bayesian network BN, support vector machine SVM, ridge regression RR, logistic regression LR and random forest RF for strategic fusion. By adopting these independent and efficient algorithms in the composite model, implementing hierarchical nonlinear feature abstraction, and using distributed computing of fusion algorithms to improve the robustness and accuracy of the prediction model, as well as accelerate the model training and parameter optimization process. The predicted output of each algorithm instance will be used as the input of the meta-learning process, and the meta-model with hyperparameters will be optimized through heuristic search and adaptive algorithms to ensure the maximum capture of potential data correlations. This method will open up new paths and possibilities in future data processing and analysis.

[0100] Embodiment 4:

[0101] Based on the aforementioned gastric cancer diagnostic metabolic markers: 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid, an early gastric cancer diagnostic model is constructed, and the construction method includes the following steps:

[0102] (1) Blood samples from gastric cancer patients and healthy controls were collected, and initial plasma metabolites were identified by ultra-high performance liquid chromatography-mass spectrometry. The candidate plasma differential metabolites were screened out between gastric cancer patients and healthy controls by using P.adjusted and Fold Change screening conditions.

[0103] (2) consistent differential metabolites were verified in a multicenter independent external validation set;

[0104] (3) A lasso regression scoring method was used and the AUC of the model was calculated for optimization under different numbers of selected features and different LRScore thresholds, and the six metabolic markers were finally determined;

[0105] (4) Based on the six metabolic markers in step (3), a gastric cancer diagnosis model was constructed using machine learning models including neural network model (NN), support vector machine (SVM), ridge regression (RR), logistic regression (LR) and naive Bayes (NB) model.

[0106] Embodiment 5:

[0107] Also provided is a method for calculating a gastric cancer survival assessment score of a subject using the aforementioned gastric cancer detection metabolite marker for non-diagnostic or therapeutic purposes, wherein the gastric cancer detection metabolite marker includes 2,6-dihydroxybenzoic acid, cysteine ​​s-sulfate, isovalerylcarnitine (C5), glycoursodeoxycholate, 2-hydroxy-3-methylvalerate and N-acetylneuraminate, which are respectively denoted as: Di, Cy, Is, Gl, Hy, Ac, and the method includes the following steps:

[0108] 1) Obtain the concentrations of 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid in the sample, respectively, and record them as C Di , C Cy , C Is , C Gl , C Hy , C ac ;

[0109] 2) Based on the coxph function in the survival package of R language, multivariate COX proportional hazard regression analysis was performed on 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid, and a survival assessment score calculation model was established according to the partial regression coefficient corresponding to each metabolite marker:

[0110]

[0111] Among them, SC represents the survival assessment score; C Di Indicates the concentration of 2,6-dihydroxybenzoic acid; C Cy represents the concentration of cysteine-S-sulfate; C Is represents the concentration of isovalerylcarnitine (C5); C Gl represents the concentration of ursodeoxycholate; CHy Indicates the concentration of 2-hydroxy-3-methylbutyric acid; C Ac Indicates the concentration of N-acetylneuraminic acid.

[0112] The median of the survival assessment score SC is used as the evaluation tendency cutoff point. The higher the survival assessment score SC of the subject is above the evaluation tendency cutoff point, the higher the possibility that the subject will be evaluated as having a shorter survival period after suffering from gastric cancer. The corresponding results of Example 1 and Example 2 are substituted into the above-mentioned survival assessment score calculation model, and the area under the ROC curve AUC for evaluating the one-year survival period after suffering from gastric cancer is not less than 0.85, indicating that the above-mentioned survival assessment score calculation model has a good discrimination degree for gastric cancer survival period, which helps to verify the predictive efficacy of the above-mentioned markers for survival period from another dimension, thereby achieving accurate diagnosis, and then early detection and early treatment, and maximizing the diagnosis and treatment effect and quality of life of patients.

[0113] Based on the above analysis, the present invention can also provide a reagent for specifically detecting metabolic markers including 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine (C5), ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid for use in preparing a product for evaluating the survival period after suffering from gastric cancer, wherein the survival period includes a one-year survival period, a three-year survival period and / or a five-year survival period.

[0114] Embodiment 6:

[0115] A computer-readable storage medium is also provided, in which a computer program executable by a processor is stored. When the computer program is executed by the processor, at least one step of the method for constructing an early gastric cancer diagnostic model described in the aforementioned embodiment or the method for calculating a subject's gastric cancer assessment score using the gastric cancer detection metabolic marker for non-diagnostic or therapeutic purposes in the aforementioned embodiment is executed, and the same technical effect can be achieved. To avoid repetition, this embodiment will not be described in detail.

[0116] Embodiment 7:

[0117] A computer device, comprising a memory, a processor, a communication interface and a communication bus; wherein the memory, the processor and the communication interface communicate with each other via the communication bus; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory, and when the processor executes the computer program, it implements at least one step of the method for constructing an early gastric cancer diagnosis model described in the aforementioned embodiment or the method for calculating a subject's gastric cancer assessment score using the gastric cancer detection metabolic marker for non-diagnostic or therapeutic purposes in the aforementioned embodiment, and can achieve the same technical effect. To avoid repetition, this embodiment will not be repeated.

[0118] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PR AM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0119] The conventional techniques in the above embodiments are prior arts known to those skilled in the art, and thus will not be described in detail here.

[0120] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

[0121] Although the present invention has been described in detail and some specific embodiments have been cited, it is obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit and scope of the present invention.

[0122] Although the above-mentioned specific embodiments have been shown, described and pointed out the novel features applied to various embodiments, it should be understood that various omissions, substitutions and changes can be made to the form and details of the described device or method without departing from the spirit of the present disclosure. In addition, the above-mentioned various features and methods can be used independently of each other, or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of the present disclosure. Many of the above-mentioned embodiments include similar components, and therefore, these similar components are interchangeable in different embodiments. Although the present invention has been disclosed in the context of certain embodiments and examples, it should be understood by those skilled in the art that the present invention can extend beyond the specifically disclosed embodiments to other alternative embodiments and / or applications and obvious modifications and equivalents thereof. Therefore, the present invention is not intended to be limited by the specific disclosure of the preferred embodiments herein.

[0123] Matters not covered in the present invention are all known technologies.

Claims

1. A metabolic marker for gastric cancer diagnosis, characterized in that: The metabolic marker is a combination of 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine C5, ursodeoxycholate, 2-hydroxy-3-methylbutyric acid and N-acetylneuraminic acid.

2. Use of a reagent for specifically detecting the metabolic marker described in claim 1 in the preparation of a product for diagnosing gastric cancer.

3. The use according to claim 2, characterized in that: The diagnosis is to determine the disease condition of gastric cancer in the diagnosed subject according to the expression level of the marker in the biological sample of the diagnosed subject.

4. The method for screening metabolic markers according to claim 1, characterized in that The following steps are involved: (1) Blood samples from gastric cancer patients and healthy controls were collected, and initial plasma metabolites were identified by ultra-high performance liquid chromatography-mass spectrometry. The candidate plasma differential metabolites were screened out between gastric cancer patients and healthy controls by using P.adjusted and Fold Change screening conditions. (2) consistent differential metabolites were verified in a multicenter independent external validation set; (3) A lasso regression scoring method was used, and the AUC of the model was optimized by calculating different numbers of selected features and different LRScore thresholds. Finally, a panel of six differential metabolites was identified, namely 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine C5, ursodeoxycholate, 2-hydroxy-3-methylbutyric acid, and N-acetylneuraminic acid.

5. A method for diagnosing gastric cancer and constructing an early gastric cancer diagnosis model based on the metabolic markers described in claim 1, characterized in that The following steps are involved: (1) Blood samples from gastric cancer patients and healthy controls were collected, and initial plasma metabolites were identified by ultra-high performance liquid chromatography-mass spectrometry. The candidate plasma differential metabolites were screened out between gastric cancer patients and healthy controls by using P.adjusted and Fold Change screening conditions. (2) consistent differential metabolites were verified in a multicenter independent external validation set; (3) A lasso regression scoring method was used and the AUC of the model was calculated for optimization under different numbers of selected features and different LRScore thresholds. Finally, six metabolic markers were identified: 2,6-dihydroxybenzoic acid, cysteine-S-sulfate, isovalerylcarnitine C5, ursodeoxycholate, 2-hydroxy-3-methylbutyric acid, and N-acetylneuraminic acid. (4) Based on the six metabolic markers in step (3), a gastric cancer diagnosis model was constructed using machine learning models including a neural network model, a support vector machine, a ridge regression, a logistic regression, and a naive Bayes model.

6. The gastric cancer diagnosis and early gastric cancer diagnosis model obtained by the construction method of claim 5.

7. A computer device, comprising a memory, a processor, a communication interface and a communication bus; wherein: The memory, processor, and communication interface communicate with each other via the communication bus; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory, and is characterized in that: when the processor executes the computer program, it implements the construction method described in claim 5.

Citation Information

Patent Citations

  • Biomarkers for theranostics

    CN103237901A

  • Biomarkers for gastric cancer and uses thereof

    CN104204807A

  • Salivary biomarker for cancer, method and device for assaying same, and method for determining salivary biomarker for cancer

    CN105765383A

  • Process and its application for improving reproducibility in maldi-TOF glycan profiling of human serum: experimental procedure and application to the screening for ovarian tumors

    US20150017669A1

  • Method for the diagnosis and prognosis of cancer

    US20160169899A1