Application of biomarker in diagnosis / prediction of peritoneal relapse after local progression stage gastric cancer operation

Abstract

The invention discloses application of a biomarker in diagnosis / prediction of peritoneal relapse after a local progression stage gastric cancer operation, and particularly relates to a biomarker which is BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2 or a combination of the BUB1, the CKS2, the PCNA, the CHEK1, the NEK2 and the NCAPG2. Through the systematic and comprehensive biomarker discovery and multi-sample verification method, a six-genome and risk stratification model is determined to detect the risk of peritoneal relapse after a high-risk gastric cancer operation, and if the risk is discovered as early as possible, information decision can be provided for clinic, and the treatment result of a patient can be improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of biotechnology, and relates to the application of biomarkers in the diagnosis / prediction of postoperative peritoneal recurrence of locally advanced gastric cancer, and in particular to the biomarkers BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2 or a combination thereof. BACKGROUND

[0002] Gastric cancer is the second leading cause of cancer-related mortality worldwide. It is predominantly manifested in the advanced stage, leading to a considerable proportion of patients being initially diagnosed with metastatic disease or relapsing after adjuvant therapy. Peritoneal dissemination is the main mechanism of recurrence and distant metastasis in locally advanced gastric cancer (LAGC), ultimately leading to poor prognosis with a median survival of less than 12 months. The ineffectiveness of traditional systemic chemotherapy and limited treatment options are the main reasons for the poor prognosis of patients with peritoneal metastasis (PM) of gastric cancer. In addition, the lack of a robust diagnostic tool for early detection of PM is a major obstacle to improving patient outcomes. Compared with computed tomography (CT) or positron emission tomography-CT (PET-CT), staging laparoscopy has higher diagnostic accuracy in identifying PM in patients with gastric cancer, and is therefore increasingly being applied. Despite the advantages of staging laparoscopy, its invasiveness, requirement for general anesthesia and relatively high cost hinder its widespread adoption as a standard procedure. Although many retrospective studies emphasize the importance of staging laparoscopy in the diagnosis of PM, there is still no consensus on its routine application in patients with LAGC.

[0003] An increasing number of studies have confirmed that the presence of intraperitoneal cancer cells is associated with an increased incidence of peritoneal recurrence or metastasis in patients with gastric cancer. The 8th edition of the American Joint Committee on Cancer (AJCC) / International Union Against Cancer (UICC) classification employs cytological examination of washed cells to detect malignant cells, thereby contributing to more accurate tumor staging. Tumors with positive cytological examination, even without visible metastatic lesions (P0CY1), are classified as advanced cancer with distant metastasis. Early identification of intraperitoneal cancer cells is crucial, as it requires a treatment method that significantly improves patient prognosis. Detection of PM before surgery facilitates the implementation of aggressive treatments such as neoadjuvant intraperitoneal and systemic chemotherapy (NlPS), hyperthermic intraperitoneal chemotherapy (HlPEC) and cytoreductive surgery (CRS). Although these treatments are not yet standardized, emerging evidence emphasizes their potential to significantly improve prognosis. This highlights the importance of early detection of metastatic tumors and the key role of early detection of peritoneal dissemination in improving treatment effectiveness. Therefore, the development of accurate molecular biomarkers for early detection of PM, through timely intervention, has the potential to significantly reduce morbidity and mortality associated with this disease. SUMMARY

[0004] In order to solve the technical problems in the prior art, the present application provides the following technical solutions:

[0005] The present application provides the use of a substance for detecting a biomarker in a sample in any one of the following:

[0006] A1. Use in the preparation of a product for diagnosing recurrence of gastric cancer after surgery;

[0007] A2. Use in the preparation of a product for predicting recurrence of gastric cancer after surgery;

[0008] The biomarker is BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof.

[0009] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, and NCAPG2.

[0010] Further, the biomarker further includes depth of invasion, tumor size, and vascular cancer embolus invasion.

[0011] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, depth of invasion, tumor size, and vascular cancer embolus invasion.

[0012] Further, the product includes reagents, kits, test papers, primers, probes, and chips.

[0013] Further, the kit can further include one or more substances selected from the group consisting of containers, instructions for use, positive controls, negative controls, buffers, adjuvants, and solvents.

[0014] Further, the chip includes gene chips and protein chips.

[0015] In some embodiments, the gene chip includes a solid support and probes immobilized on the solid support.

[0016] In some embodiments, the probes include probes against the biomarkers for detecting the transcription level of the biomarkers.

[0017] In some embodiments, the protein chip includes a solid support and antibodies that specifically bind to proteins encoded by the biomarkers immobilized on the solid support.

[0018] Further, the kit further includes an mRNA expression amount auxiliary detection reagent.

[0019] Further, the mRNA expression assisting detection reagent includes a reaction reagent for visualizing an amplicon corresponding to the primer, an RNA extraction reagent, a reverse transcription reagent, a cDNA amplification reagent, and / or a standard used for preparing a standard curve.

[0020] Further, the reaction reagent for visualizing the amplicon corresponding to the primer includes a reagent for visualizing the amplicon in agarose gel electrophoresis, enzyme-linked gel method, chemiluminescence method, in situ hybridization method, and fluorescence detection method.

[0021] Further, the kit further includes a protein expression assisting detection reagent.

[0022] Further, the protein expression assisting detection reagent includes a blocking solution, an antibody diluent, a washing buffer, and a color developing termination solution.

[0023] Further, the sample includes blood, serum, plasma, peripheral blood, tissue, blood cells, bone marrow, ascites, fine needle biopsy sample, cell-containing body fluid, free floating nucleic acid, sputum, saliva, urine, cerebrospinal fluid peritoneal fluid, pleural fluid, feces, lymph, skin swab, oral swab, nasal swab or lavage.

[0024] Further, the sample includes blood, serum, plasma, peripheral blood, tissue, or fine needle biopsy sample.

[0025] In some embodiments, the sample is taken directly from a subject without further processing, e.g., blood can be obtained from the peripheral circulatory system of a subject (e.g., by a blood collection needle). The sample can include, for example, blood, urine, feces, saliva, cerebrospinal fluid, and sweat, among others. Non-limiting examples of the sample include blood (or components of blood, e.g., white blood cells, red blood cells, platelets) obtained from any anatomical location of a subject (e.g., tissue, circulatory system, bone marrow), cells obtained from any anatomical location of a subject, skin, heart, lung, kidney, exhaled breath, bone marrow, feces, semen, vaginal fluid, tissue fluid derived from tumor tissue, breast, pancreas, cerebrospinal fluid, tissue, throat swab, biopsy, placental fluid, amniotic fluid, liver, muscle, smooth muscle, bladder, gall bladder, colon, intestine, brain, cavity fluid, sputum, pus, microbiota, meconium, milk, prostate, esophagus, thyroid, serum, saliva, urine, gastric and digestive fluids, tears, ocular fluid, sweat, mucus, earwax, oil, glandular secretions, spinal fluid, hair, nails, skin cells, blood plasma, nasal swab or nasopharyngeal wash, spinal fluid, cord blood, lymphatic fluid, and / or other excretions or body tissues.

[0026] Further, the gastric cancer includes locally advanced gastric cancer.

[0027] Further, the gastric cancer postoperative recurrence includes gastric cancer postoperative peritoneal recurrence, gastric cancer postoperative lymph node recurrence, gastric cancer postoperative lung recurrence, gastric cancer postoperative liver recurrence, gastric cancer postoperative bone recurrence.

[0028] Further, the sample is from a subject to be detected.

[0029] Further, the subject to be detected includes a human or a non-human mammal.

[0030] Further, the subject to be detected is a human.

[0031] The term "agent" as used herein can be used to refer to a compound or entity of any chemical class, including, for example, a polypeptide, a nucleic acid, a carbohydrate, a lipid, a small molecule, a metal, or a combination or composite thereof. Where appropriate, as will be clear to those skilled in the art, the term can be used to refer to a cell or organism or a fraction, extract or component thereof or an entity comprising a cell or organism or a fraction, extract or component thereof. Alternatively or additionally, as will be clear from the context, the term can be used to refer to a natural product, as it is found in and / or obtained from nature. In some cases, again as will be clear from the context, the term can be used to refer to one or more artificial entities, as it is designed, engineered and / or produced by the hand of man and / or not found in nature. In some embodiments, an agent can be used in isolated or pure form; in some embodiments, an agent can be used in crude form. In some embodiments, a collection or library of potential agents can be provided, for example, which can be screened to identify or characterize active agents therein. In some cases, the term "agent" can refer to a compound or entity that is or comprises a polymer; in some cases, the term can refer to a compound or entity that comprises one or more polymer moieties. In some embodiments, the term "agent" can refer to a compound or entity that is not and / or is substantially free of any polymer and / or one or more particular polymer moieties. In some embodiments, the term can refer to a compound or entity that lacks or is substantially free of any polymer moieties. In some embodiments, an agent is a compound. In some embodiments, the enhancer can refer to a food grade compound that binds with salt. The enhancer binds with salt through forming electrostatic interaction, hydrogen bond, ionic interaction, coordination interaction, non-covalent interaction, van der Waals interaction, or any combination thereof.

[0032] The present application provides a product comprising a detection reagent for detecting the content of a biomarker, the biomarker being BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof.

[0033] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2.

[0034] Further, the biomarker further comprises depth of invasion, tumor size, vascular tumor embolus invasion.

[0035] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, depth of invasion, tumor size, vascular tumor embolus invasion.

[0036] Further, the product comprises reagents, kits, test papers, primers, probes.

[0037] Further, the kit comprises an immunohistochemical detection kit, an immunoblotting detection kit, an immunochromatographic detection kit, a flow cytometry kit, a qPCR kit, an ELISA kit.

[0038] Further, the product further comprises reagents commonly used in PCR reactions, RT-PCR derivative reactions, 3SR amplification, LCR, SDA, NASBA, TMA, SYBR Green, TaqMan probes, molecular beacons, dual hybridization probes, complex probes, ISH, microarrays, Southern blotting, Northern blotting, multi-analyte profiling tests, enzyme-linked immunosorbent assays (ELISA), radioimmunoassays, immunofluorescence assays, enzyme immunoassays, immunoprecipitation assays, chemiluminescence assays, immunohistochemical assays, dot blot assays, or slot blot assays.

[0039] The present application provides a system for diagnosing whether gastric cancer has recurred after surgery or predicting the risk of recurrence of gastric cancer after surgery, comprising:

[0040] An input unit for obtaining expression level detection data of biomarkers in a sample;

[0041] A result determination unit for inputting the data of the detection unit into a computer program for processing and determination, and obtaining a comparison result after comparing with a set critical value in the computer;

[0042] An output unit for outputting the result of the result determination unit to obtain whether the corresponding object of the detection sample has recurred after gastric cancer surgery or the probability of recurrence of gastric cancer after surgery;

[0043] All units in the system are operated using a computer program or artificial intelligence.

[0044] The biomarker is BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof.

[0045] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2.

[0046] Further, the biomarker further comprises infiltration depth, tumor size, and vascular cancer embolus infiltration.

[0047] Further, the biomarker is a combination of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, infiltration depth, tumor size, and vascular cancer embolus infiltration.

[0048] The device, electronic equipment, non-volatile computer storage medium and method provided by the embodiments of the present specification are corresponding, and therefore the device, electronic equipment, non-volatile computer storage medium also have similar beneficial technical effects to the corresponding method, since the beneficial technical effects of the method have been described in detail above, therefore, the beneficial technical effects of the corresponding device, electronic equipment, non-volatile computer storage medium will not be described here.

[0049] The present application provides a gastric cancer postoperative recurrence risk prediction model, the gastric cancer postoperative recurrence risk prediction model includes using a machine learning method to train an algorithm model for diagnosing whether gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk prediction.

[0050] Further, the algorithm model processes data to obtain a gastric cancer postoperative recurrence evaluation result, the gastric cancer postoperative recurrence evaluation result includes a gastric cancer postoperative recurrence probability and an important characteristic factor, the important characteristic factor reflects the correlation with gastric cancer postoperative recurrence; the important characteristic factor is the expression level data of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2 or a combination thereof extracted from the information parameters of the to-be-detected object.

[0051] Further, the important characteristic factor further comprises infiltration depth, tumor size, and vascular cancer embolus infiltration data extracted from the information parameters of the to-be-detected object.

[0052] Further, one RSA model in the algorithm model = [(3.665 x 6-mRNA marker combination) + (5.009 x infiltration depth) + (2.009 x tumor size) + (1.417 x vascular cancer embolus infiltration) + (-12.340)], the RSA model is used to evaluate the risk of gastric cancer postoperative recurrence.

[0053] Further, the machine learning method comprises a decision tree learning method, a random forest learning method, a K-nearest neighbor algorithm learning method, a naive Bayes learning method, a support vector machine learning method, a neural network learning method, and an Adeboost learning method.

[0054] In some embodiments, the training set data is input into an AdaBoost algorithm model, a series of weak classifiers are trained using the training set data, and the weak classifiers are linearly combined into a strong classifier. Each time a weak classifier is trained, the AdaBoost algorithm changes the probability distribution of the training set data, so that each weak classifier obtains different weights of the training set data. Each sample that is misclassified will obtain a greater weight in the next weak classifier. Then, the AdaBoost adopts a weighted majority voting method, increases the weight of the weak classifier with a small classification error rate, reduces the weight of the weak classifier with a large classification error rate, and linearly combines the classifiers to obtain the final classifier. The obtained classifier is tested according to test set data in the to-be-processed data, and the AdaBoost algorithm is adjusted according to the test result, so as to obtain an evaluation model that can be used for risk analysis of postoperative recurrence of gastric cancer, and realize risk prediction of postoperative recurrence of gastric cancer.

[0055] The present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is stored in the computer readable storage medium and is executed by a processor to implement the gastric cancer postoperative recurrence risk prediction model described above.

[0056] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment mainly describes the difference from other embodiments. In particular, for the device, electronic equipment, and non-volatile computer storage medium embodiments, because they are basically similar to the method embodiments, the description is relatively simple, and the related parts can be referred to the part of the method embodiment.

[0057] The system, device, module or unit described in the above embodiments can be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer. Specifically, the computer may, for example, be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0058] In some embodiments, the modules or steps described above can be implemented by a computing device. In some embodiments, the modules or steps can be centralized on a single computing device. In some embodiments, the modules or steps can be distributed across multiple computing devices, which are linked by a network or other linkage. In some embodiments, the computing devices store program code for performing the modules or steps described above, which is stored in a memory device. In some embodiments, the modules or steps can be prepared as individual integrated circuits, or combined into a single integrated circuit. The present application is not limited to any particular implementation of the modules or steps, whether in hardware or software.

[0059] The terms "subject," "patient," "subject," "test subject," or "individual" as used herein refer to any subject for which treatment or prevention is desired, particularly a vertebrate subject, and even more particularly a mammalian subject. Suitable vertebrates falling within the scope of the present application include, but are not limited to, any member of the subphylum chordata, including a primate (e.g., humans, monkeys, and apes, including monkey species (e.g., macaques (genus Macaca) (e.g., cynomolgus monkeys (Macaca fascicularis), and / or rhesus monkeys (Macaca mulatta)) and baboons (Papio ursinus), as well as marmosets (species from the genus Callithrix), squirrel monkeys (species from the genus Saimiri), and tamarins (species from the genus Saguinus), and species of chimpanzees, such as Pan troglodytes), a rodent (e.g., a mouse, rat, guinea pig), a rabbit (e.g., a rabbit, hare), a bovine (e.g., a cow), an ovine (e.g., a sheep), a caprine (e.g., a goat), a porcine (e.g., a pig), an equine (e.g., a horse), a canine (e.g., a dog), a feline (e.g., a cat), an avian (e.g., a chicken, a turkey, a duck, a goose, an avian companion bird such as a canary, a parrot, etc.), a marine mammal (e.g., a dolphin, a whale), a reptile (a snake, a frog, a lizard, etc.), and a fish.

[0060] The present application provides a computer evaluation method for whether gastric cancer recurs after surgery or the risk of recurrence of gastric cancer after surgery, the method comprising:

[0061] Obtaining to-be-processed data, wherein the to-be-processed data comprises detection data of a BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof biomarker in a sample;

[0062] inputting the to-be-processed data into the gastric cancer postoperative recurrence risk prediction model described above, obtaining the evaluation result of whether the to-be-processed data has gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk;

[0063] outputting the evaluation result of whether the to-be-processed data has gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk.

[0064] Further, the to-be-processed data further includes detection data of the subject's depth of invasion, tumor size, and vascular plug invasion.

[0065] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the present application. In some cases, the actions or steps described in the present application can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are possible or can be advantageous.

[0066] The term "machine learning" used in the present application means an algorithm that gives a computer learning ability without explicit programming, including an algorithm that learns from data and makes predictions on data. The machine learning algorithm used in the embodiments disclosed herein can include, but is not limited to, random forest (RF), least absolute shrinkage and selection operator (LASSO) logistic regression, regularized logistic regression, XGBoost, decision tree learning, artificial neural network (ANN), deep neural network (DNN), support vector machine, rule-based machine learning, etc. Algorithms such as linear regression or logistic regression can be used as part of the machine learning process.

[0067] The term "comprising" used in the present application means that the composition or method of one or more named elements or steps is open, meaning that the named elements or steps are necessary, but other elements or steps can be added within the scope of the composition and method. To avoid tediousness, it should also be understood that any composition or method described as "comprising" one or more named elements or steps also describes a corresponding more limited "consisting essentially of the same named elements or steps, meaning that the composition or method includes the named essential elements or steps, and can also include other elements or steps that do not materially affect the basic and novel characteristics of the composition or method.

[0068] Advantages and beneficial effects of the present application:

[0069] Using these genetic markers, a risk prediction model was developed to successfully predict surgical specimens, endoscopic specimens, and blood specimens from a multi-center cohort of patients with peritoneal recurrence of gastric cancer, transitioning from invasive tissue specimens to non-invasive liquid biopsy methods. Importantly, given the association between peritoneal recurrence and metastasis, we evaluated the diagnostic genome to assess the performance of multiple independent tests for peritoneal metastasis in a prospective cohort of patients with primary metastasis or P0 CY1 tumors. Through this systematic, comprehensive biomarker discovery and multi-sample validation approach, we identified a six-gene panel and risk stratification model to detect high-risk gastric cancer postoperative peritoneal recurrence, which, if detected early, can provide information for clinical decision-making and improve patient outcomes. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 is a flow chart of the study design for discovery and validation of 6-mRNA markers in different LAGC patient populations;

[0071] Figure 2 is a diagram of the discovery process and preliminary validation results of peritoneal recurrence candidate markers in LAGC patients based on transcriptomics; wherein (A) 6 candidate mRNAs were discovered using appropriate transcriptome data from TCGA database and GEO database, as well as 3 pairs of peritoneal recurrence and 3 pairs of no peritoneal recurrence mRNA sequencing Venn diagram; (B) the relevant genes of the 6 genes in the panel were intersected, and 39 common relevant genes were listed; (C) 6 pairs of mRNA drug sensitivity analysis of 198 anti-cancer drugs in GDSC database; (D) comparison of the expression of 6 mRNAs (BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2) in cancer tissues and normal tissues in TCGA database; (E) comparison of the expression of 6 mRNAs in cancer tissues and normal tissues in 29 cases of fresh frozen tissues; (F) comparison of the expression of 6 mRNAs in 22 cases of gastric cancer patient and healthy person peripheral blood samples; (G) construction of PPI network of the 6 mRNAs using online STRING database (https: / / string-db.org); (H) correlation heat map analysis of 6 mRNAs and common metastasis genes based on TCGA database;

[0072] Figure 3FIG. 1 is a graph of the results of the transcriptome training and validation phase for identifying peritoneal recurrence of surgical specimens of LAGC patients; wherein (A) a nomogram for peritoneal recurrence of LAGC patients was constructed based on 6-mRNA combined with clinical characteristics; (B) ROC curves of different predictive variables in the training set; (C) ROC curves of different predictive factors in the validation set; (D-I) prediction confusion matrix of RSA models constructed using clinical characteristics, 6-mRNA in the training set and the validation set; (J) calibration curve of RSA models in the training set; (K) calibration curve of RSA models in the validation set; (L-Q) clinical benefit graph of RSA models constructed using clinical characteristics, 6-mRNA and their combination in the training set and the validation set; (R) Log-rank test survival curve graph of patients in the training set divided into low-risk group and high-risk group according to the critical value obtained from the nomogram Youden index; (S) Log-rank test survival curve was drawn, and patients in the validation set were divided into low-risk group and high-risk group according to the critical value obtained from the nomogram Youden index;

[0073] Figure 4 FIG. 2 is a graph of the results of the transcriptome validation phase for identifying peritoneal recurrence of gastroscope biopsy specimens of LAGC patients; wherein (A-F) correlation analysis of 6-mRNAs in gastroscope biopsy specimens and paired surgical resection specimens; (G-L) comparison of the expression of 6-mRNAs in gastroscope biopsy specimens and paired surgical resection specimens; (M) ROC curve of different predictive variables in gastroscope biopsy specimens; (N) calibration curve of RSA models in gastroscope biopsy specimens; (O) Log-rank test survival curve was drawn, and patients were divided into low-risk group and high-risk group according to the critical value obtained from the nomogram Youden index; (P-R) prediction confusion matrix graph constructed by clinical characteristics, 6-mRNA and RSA models; (S-U) clinical benefit graph of RSA models constructed using clinical characteristics, 6-mRNA and their combination in the gastroscope biopsy specimen validation set;

[0074] Figure 5FIG. 1 is a result map for identifying peritoneal recurrence in peripheral blood samples of LAGC patients in the transcriptome verification stage; wherein (A) is a nomogram for peritoneal recurrence of LAGC patients based on 6-mRNA combined with clinical characteristics; (B) is an ROC curve of different predictive variables in the training set; (C) is an ROC curve of different predictive variables in the verification set; (D) is an ROC curve of different predictive variables in the tumor marker negative set; (E-M) are prediction confusion matrix maps of RSA models constructed using clinical characteristics, 6-mRNA, and both in the training set, the verification set, and the tumor marker negative set; (N-V) are clinical benefit maps of RSA models constructed using clinical characteristics, 6-mRNA, and their combination in the training set, the verification set, and the tumor marker negative set; (W1) is a calibration curve of the RSA model in the training set; (W2) is a calibration curve of the RSA model in the verification set; (W3) is a calibration curve of the RSA model in the tumor marker negative set; (X) is a Log-rank test survival curve drawing, in which the patients in the training set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (Y) is a Log-rank test survival curve drawing, in which the patients in the verification set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (Z) is a Log-rank test survival curve drawing, in which the patients in the tumor marker negative set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram;

[0075] Figure 6 FIG. 2 is a result map for identifying peritoneal metastasis and micrometastasis in peripheral blood samples of LAGC patients in the transcriptome verification stage; wherein (A) is an ROC curve of different predictive variables for predicting the occurrence of peritoneal micrometastasis (P0CY1); (B) is a HE staining of free cancer cells shed from the abdominal cavity; (C) is an ROC curve of different predictive variables for predicting the occurrence of PM; (D) is a diagnostic laparoscopy showing peritoneal metastatic nodules; (E-G) are confusion matrix maps of RSA models constructed using clinical characteristics, 6-mRNA, and both for predicting the occurrence of peritoneal micrometastasis (P0CY1); (H-J) are confusion matrix maps of RSA models constructed using clinical characteristics, 6-mRNA, and their combination for predicting the occurrence of PM; (K-M) are clinical benefit maps of RSA models constructed using clinical characteristics, 6-mRNA, and their combination for predicting the occurrence of peritoneal micrometastasis (P0CY1); (N-P) are clinical benefit maps of RSA models constructed using clinical characteristics, 6-mRNA, and their combination for predicting the occurrence of PM; (Q) is a calibration curve of the RSA model for predicting peritoneal micrometastasis (P0CY1) of patients; (R) is a calibration curve of the RSA model for predicting the PM cohort; (S) is a Log-rank test survival curve drawing, in which the patients in the peritoneal micrometastasis set (P0CY1) are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (T) is a Log-rank test survival curve drawing, in which the patients in the PM set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (U) is a Log-rank test survival curve drawing, in which the patients in the peritoneal micrometastasis set (P0CY1) are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (V) is a Log-rank test survival curve drawing, in which the patients in the PM set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (W) is a Log-rank test survival curve drawing, in which the patients in the peritoneal micrometastasis set (P0CY1) are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (X) is a Log-rank test survival curve drawing, in which the patients in the PM set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (Y) is a Log-rank test survival curve drawing, in which the patients in the peritoneal micrometastasis set (P0CY1) are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram; (Z) is a Log-rank test survival curve drawing, in which the patients in the PM set are divided into a low-risk group and a high-risk group according to the critical value obtained from the Youden index of the nomogram.

[0076] ROC curves for 6-mRNA prediction of peritoneal metastasis in colorectal cancer; (V) ROC curves for 6-mRNA prediction of peritoneal metastasis in hepatocellular carcinoma; (W) ROC curves for 6-mRNA prediction of peritoneal metastasis in pancreatic ductal carcinoma. Detailed Implementation

[0077] Example

[0078] 1. Experimental Methods

[0079] 1.1 Discovering biomarkers in whole-genome expression profiling datasets

[0080] The biomarker discovery and validation process used in this study is as follows: Figure 1 As shown, firstly, mRNA sequencing data from multiple sources, including the Gene Expression Omnibus (GEO) database, the Cancer Genome Atlas (TCGA) database, and samples matched for PM recurrence, were used for biomarker identification. For validation, we used a cohort of fresh frozen postoperative specimens, endoscopic biopsy specimens, and peripheral blood samples from multiple medical clinics to ensure thorough and reproducible validation.

[0081] During the discovery phase, stringent screening criteria for the application of the GEO database (https: / / www.ncbi.nlm.nih.gov / geo / ) led to the selection of the GSE15081 dataset for label development. mRNA sequencing data were collected from three pairs of cancer tissues that underwent radical resection of LAGC followed by adjuvant therapy and subsequently developed peritoneal metastasis, and from three pairs of non-recurrent cancer tissues at the Fourth Hospital of Hebei Medical University.

[0082] We first assessed the expression of selected mRNAs in a pilot cohort of 29 matched GC and adjacent non-malignant tissue (ANM) samples using real-time quantitative polymerase chain reaction (RT-qPCR). Peripheral blood samples were collected from 22 gastric cancer patients and 22 healthy individuals undergoing routine physical examinations during the same period to detect the expression of candidate mRNAs in the bloodstream. Samples were obtained from the Fourth Hospital of Hebei Medical University (FHHMU) between January and March 2023.

[0083] 1.2 Clinical cohorts for biomarker validation

[0084] In this study, 329 fresh-frozen specimens from LAGC patients in two independent cohorts were collected. The biomarkers for predicting PM recurrence were developed and validated. The specimens were collected from August 2016 to March 2019, and patients who received neoadjuvant therapy and gastric remnant cancer after partial gastrectomy were excluded. 196 patients were from the Fourth Hospital of Hebei Medical University (FHHMU) and Shijiazhuang People's Hospital (S1ZPH), and 133 patients were from Baoding Central Hospital (BDCH), Jinling Hospital of Nanjing University (INNI), and Renmin Hospital of Wuhan University (WHPH).

[0085] Further analysis included 103 matched gastroscopic biopsy specimens from 5 institutions to validate the transition of biomarkers from surgical specimens to gastroscopic biopsy specimens.

[0086] In addition, we retrospectively analyzed serum samples of LAGC patients, including patients with postoperative PM recurrence and without PM recurrence, to determine the tissue-based biomarkers for liquid biopsy. 120 LAGC patients admitted to Zhongshan Hospital Affiliated to Fudan University from February 2017 to December 2019 were selected. The validation cohort included serum samples from 123 LAGC patients from 4 other institutions from January to December 2016.

[0087] In addition, we conducted an analysis to evaluate the effectiveness of biomarkers in detecting PM and micrometastasis in patients with gastric cancer. Peripheral blood samples from 66 LAGC patients (including 12 patients with P0CY1 status confirmed by peritoneal cytology [FHHMU. registration. nct03718624, ChiCTR1800014817]) and 95 LAGC patients (including 15 patients with PM confirmed by laparoscopic exploration [FHHMU, registration: NCT02555358, NCT01516944]) were used to predict the occurrence of P0CY1 and the incidence of PM, respectively. In addition, we evaluated the specificity of our mRNA combination as a biomarker for PM in LAGC and other gastrointestinal malignancies (including colorectal cancer, pancreatic ductal carcinoma, and hepatocellular carcinoma) using RT-qPCR on serum samples collected from FHHMU patients from 2019 to 2021.

[0088] According to the treatment guidelines for gastric cancer, patients were followed up for recurrence or disease progression through laboratory tests, endoscopy, and abdominal and pelvic CT. Tissue specimens were immediately frozen in liquid nitrogen and stored at -80°C. Surgical specimens were processed according to the guidelines of the Chinese Society of Clinical Oncology. Tumor and lymph node staging was performed according to the 8th edition of the AJCC staging criteria. All procedures were performed in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants, and approval was obtained from the institutional review board of all participating institutions.

[0089] 1.3 RNA extraction and gene expression analysis

[0090] Total RNA was isolated from fresh frozen surgical tissues using TRlzol reagent according to the manufacturer's protocol. For serum samples, 200 μΐ aliquots were thawed on ice, centrifuged at 3000 rpm for 5 min to remove cell debris and then mixed with 5 volumes of TRlzol. Total RNA was extracted using Qiagen miRNeasy kit. One fifth volume of chloroform was added to the mixture, mixed by inversion and the sample was placed on ice for 5 min. The centrifuge was spun at high speed for 15 min at 12.000 rpm at 4°C. After 15 min, the supernatant was aspirated and an equal volume of isopropanol was added. The sample was mixed and placed on ice for 10 min. It was spun again at high speed for 10 min at 12.000 rpm at 4°C. After discarding the supernatant, 400 μΐ of 75% ethanol was added and the sample was spun at 9000 rpm at 4°C for 5 min. The supernatant was discarded and the pellet was allowed to dry before dissolving the pellet in an appropriate volume of DEPC water. Reverse transcription was performed using the High Capacity cDNA Reverse Transcription Kit from Roche, Germany, followed by qRT-PCR on an Applied Biosystems Real-Time PCR system. The pre-denaturation program was set at 95°C for 10 min. Then, 40 cycles of reactions were performed, each cycle ending with 15 sec denaturation at 95°C, 30 sec annealing at 58°C and 30 sec elongation at 72°C. Finally, the melting curve program was set at 95°C for 15 sec, ending elongation, maintaining the system temperature at 60°C for 1 min, and finally increasing the temperature by 0.3°C per min to 95°C for 15 sec.

[0091] Several measures were taken to ensure the reproducibility of the detection, including proper controls, exclusion of low quality RNA samples and repeated detection over time points. Gene expression was quantified using the Applied Biosystems QuantStudio 6 Flex Real-Time PCR System software. The relative gene abundance was determined using the 2^-ΔΔct method, with ΔCT being the difference in CT values between the target gene and GAPDH, with GAPDH as the internal control.

[0092] 1.4 PPI protein-protein interaction network analysis, pathway analysis and chemosensitivity analysis

[0093] The protein-protein interaction (PPI) network of target genes was constructed using the STRlNG database (https: / / string-db.org). Genes with a combined score of ≥0.4 were retained. The protein-protein interaction network in the STRING database was imported into Cytoscape v3.9.1 for topological analysis. Cytoscape provides 11 topological methods. In this paper, the degree method was used to identify key genes based on the connectivity within the network. Cytoscape 3.9.1 also supports visual PPI network representation. In addition, Enrichr (https: / / maayanlab.cloud / Enrichr / ) software was used to perform gene set and pathway analysis on the four candidate genes.

[0094] Using Oncopredict, we determined the potential therapeutic drugs for each gene using data from the Genomics of Drug Sensitivity in Cancer (GDSC) database (https: / / www.cancerrxgene.org / ), which divides genes into high and low expression groups based on their median expression levels. Wilcoxon rank-sum test was used to compare the differences in drug sensitivity between the two groups.

[0095] 1.5 Statistical analysis

[0096] IBM SPSS 23 software was used for statistical analysis. R version 3.6.3 and GraphPad Prism version 8.0. Single and multiple regression analyses were used to identify significant clinical pathological variables and mRNA categorical variables as covariates. Variables that were significant in the single factor analysis were included in the multiple regression analysis. In the discovery phase, Wilcoxon rank-sum and Bonferroni tests were used to compare the expression of different genes between the PM recurrence group and the non-PM recurrence group. In the clinical validation phase, a gene-based risk score model was established using a backward elimination logistic regression analysis, and the model performance was evaluated using the receiver operating characteristic (ROC) curve and the area under the curve (AUC) value.

[0097] ROC curves were plotted using the pROC package in R, and the area under the curve (AUC) was plotted. The DeLong test was used to compare the ROC curves. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), precision, and accuracy of the 6-mRNA biomarker set in all cohorts were determined using the reportROC package shown in the confusion matrix. The best cutoff value of the ROC curve was established using the Youden index in the pROC program. According to the Youden index and the median risk score, patients were divided into a high-risk group and a low-risk group, which helped to predict recurrence and detect metastasis. We analyzed the recurrence-free survival (RFS) using the Kaplan-Meier method, which was defined as the time interval from surgery to the confirmation of peritoneal recurrence or all-cause death, and the recurrence-free patients who survived at this milestone were followed up for 5 years. For the lost patients who had no evidence of recurrence 5 years ago, they were evaluated at the last visit. Statistical significance was set at p<0.05.

[0098] 2. Results of the experiment

[0099] 2.1 Identification of overexpressed candidate mRNAs in gastric cancer patients with peritoneal recurrence by whole genome gene expression profiling

[0100] In this study, we performed preliminary biomarker discovery by analyzing the transcriptome data of two publicly available gastric cancer datasets (TCGA and GSE15081), in addition, we also performed mRNA sequencing on 3 pairs of gastric cancer tissues with peritoneal metastatic recurrence and 3 pairs without peritoneal metastasis after radical surgery. The GSE15081 dataset included 33 patients with peritoneal recurrence and 75 patients without recurrence. Through differential gene expression analysis (Wilcoxon rank-sum test for GSE15081 and EdgeR for TCGA, both P<0.05, and Bonferroni correction was applied) and correlation analysis (r<0.5), we identified 6 genes (BUB1, CKS2, PCNA, CHEK1, NEK2, and NCAPG2) that were differentially expressed between patients with peritoneal recurrence and those without peritoneal recurrence. Figure 2 A).

[0101] Further analysis of TCGA data showed that these genes were significantly overexpressed in cancer tissues compared to ANM (P<0.05) Figure 2 D). To verify these results, we established a pilot cohort in FHHMU, including gastric cancer and healthy screening populations from January to December 2023. Analysis of 29 matched GC and ANM tissue samples by RT-qPCR confirmed the high expression levels of these mRNAs in GC tissues (P<0.05) Figure 2 E). Analysis of peripheral blood from 22 gastric cancer patients and 22 healthy individuals at the same time also showed that the candidate genes were upregulated in cancer patientsFigure 2 F)。

[0102] To investigate potential associations between gene expression and drug sensitivity in our panel, we used the "oncoppredict" R package. This software analyzes expression matrices and drug response data from the Genomicsofdrug Sensitivity in Cancer (GDSC) database, which includes information for 198 anticancer drugs. Analysis of the GSE15081 dataset showed that the high expression group of 6 identified genes had increased sensitivity to BMS-754807 and AZD8186 Figure 2 C)。

[0103] Protein interaction networks were drawn using the STRING database and visualized with Cytoscape 3.9.1 Figure 2 B, Figure 2 G) in gastric cancer. Moreover, correlation analysis with metastasis-related genes (MMP1, MMP3 and VEGFA) by Timer 2.0 (http: / / timer.cistrome.org / ) showed positive correlation, and other correlations are detailed in Figure 2 H。

[0104] 2.2 Validation of the 6-mRNA combination to predict peritoneal recurrence in LAGC patients from surgical resection specimens

[0105] We evaluated the 6-mRNA combination by RT-qPCR in 164 LAGC patients without peritoneal recurrence after radical resection and 32 LAGC patients with peritoneal recurrence after radical resection. Receiver operating characteristic curve analysis was used to determine the accuracy of individual and combined biomarkers to distinguish peritoneal recurrence cases. Results are shown in Figure 3 B, although individual mRNA markers were effective, the 6-mRNA combination showed better diagnostic performance (AUC = 0.902, 95% CI: 0.851-0.953, P < 0.001). Further clinical evaluation of the 6-mRNA classification included analysis of its ability to detect peritoneal recurrence together with clinical variables.

[0106] We used a formula based on the coefficients of the logistic regression and a constant to estimate the probability of peritoneal recurrence [(3.665 x 6-mRNA marker combination) + (5.009 x depth of invasion) + (2.009 x tumor size) + (1.417 x vascular cancer embolus invasion) + (-12.340)], which was depicted as a nomogram to visualize the prediction of peritoneal recurrence Figure 3A). After calibration of the model using data from the training cohort, the same statistical parameters were applied to the validation cohort. This process allowed us to stratify patients into low- and high-risk categories based on the cut-off determined by the Youden index. Combining the 6-mRNA panel with clinical variables, we developed a risk stratification assessment (RSA) model that showed superior predictive ability for peritoneal recurrence with an AUC of 0.966 (95% CI: 0.944-0.988, P < 0.001), as shown in Figure 3 D-F. The DeLong test showed that the RSA model had a higher AUC value in the training set than the clinical model (0.966 vs 0.869; P = 0.001) and the 6-mRNA panel model (0.966 vs 0.902; P = 0.038). The calibration curve of the model further highlighted its excellent predictive accuracy Figure 3 J).

[0107] Subsequently, the model was applied to an independent external validation cohort of 133 LAGC patients, including 20 patients with peritoneal recurrence and 113 patients without peritoneal recurrence. This application highlighted the robust predictive ability of the RSA model, with an AUC of 0.961 (95% CI: 0.931-0.991, P < 0.001) Figure 3 C). In this cohort, the RSA model showed unparalleled sensitivity (65.0%) and specificity (95.6%), outperforming the clinical model (sensitivity 20.0%, specificity 100.0%) and the 6-mRNA panel model (sensitivity 45.0%, specificity 97.3%) in terms of predictive performance Figure 3 G-I). Calibration curve analysis further confirmed the improvement in predictive accuracy Figure 3 K).

[0108] We further evaluated the potential of the RSA model in improving the cost-effectiveness of clinical decision-making. In our training cohort, using existing clinical parameters, 36.7% of patients were considered high-risk for peritoneal recurrence, while the remaining 63.3% were classified as low-risk. Follow-up results showed that 13.8% (27 / 196) and 2.6% (5 / 196) of patients in the high-risk and low-risk groups, respectively, developed peritoneal recurrence. This outcome suggests that the initial risk classification based on clinical features resulted in 22.9% of patients receiving unnecessary intensive follow-up, while 2.6% of patients did not receive necessary close monitoring Figure 3 L). In contrast, when the 6-mRNA classification was applied to the same cohort, a more accurate risk stratification occurred, with 29.6% classified as high-risk and 70.4% classified as low-risk. Peritoneal recurrence occurred in 14.7% (29 patients) of the high-risk group and 1.5% (3 patients) of the low-risk group Figure 3 M).

[0109] In the high-risk category of the RSA model, the incidence rate during follow-up was significantly reduced to 12.3% (24 patients) out of 342 follow-up cases. Figure 3 N). This trend is consistent in the external validation cohort; compared with other models, the RSA model significantly reduces unnecessary follow-up for high-risk patients and minimizes missed diagnoses in the low-risk group. Figure 3 This indicates a substantial improvement in clinical decision-making, reducing the likelihood of high-risk patients receiving unnecessary interventions and increasing the detection rate of low-risk patients.

[0110] Further follow-up of the enrolled patients based on the high-risk and low-risk groups of the nomogram revealed that, in the training group, the 5-year RFS was significantly worse in the high-risk group than in the low-risk group (25.0% vs. 59.3%, P<0.0001). A similar difference was observed in the validation set (28.2% vs. 57.4%, P=0.00012). Figure 4 RS).

[0111] 2.3 Validating the predictive value of 6-mRNA combination for peritoneal recurrence in patients with laminar dysplasia of the vagina (LAGC) using endoscopic biopsy specimens.

[0112] In this study, in addition to surgical resection specimens from the training and validation cohorts, we obtained 103 matched endoscopic biopsy specimens, of which 15 showed peritoneal recurrence and 88 did not. Significant high correlations were found in the expression profiles of six genes between the biopsy and surgical specimens. Figure 4 Comparison of gene expression in these matched samples showed no significant difference in the expression of any gene. Figure 4 GL). The AUC for peritoneal recurrence detected by the 6-mRNA combination was 0.899 (95% CI: 0.830-0.967, P<0.001), while the AUC of the RSA model was 0.956 (95% CI: 0.910-1.000, P<0.001), highlighting its effectiveness for preoperative biopsy specimens. Figure 4 M). Calibration curve analysis further confirmed that the RSA model has higher predictive accuracy. Figure 4 N).

[0113] In the biopsy cohort, the RSA model exhibited the highest sensitivity (73.3%) and specificity (97.7%), both exceeding those of the clinical model (sensitivity 33.3%; specificity 100.0%) and the 6-mRNA combination model (sensitivity 40.0%; specificity 97.7%). Figure 4P-R). Comparative analysis of the clinical utility of different models showed that, compared with the clinical characteristics and 6-mRNA combination model, the RSA model increased the diagnostic rate of high-risk group peritoneal recurrence from 10.7% to 13.6%, and decreased the diagnostic rate of low-risk group from 3.9% to 1.0% Figure 4 S-U). These findings highlight the ability of the RSA model to refine clinical decision-making and reduce unnecessary interventions.

[0114] As in the follow-up of the previous cohort, based on the high-risk group and low-risk group of nomogram, we performed log-rank test on patients with endoscopic biopsy specimens, and found that the 5-year RFS of high-risk group was significantly lower than that of low-risk group (37.2% vs 56.7%, P=0.0056) Figure 5 O).

[0115] 2.4 Validation of 6-mRNA combination in predicting peritoneal recurrence in LAGC patients in peripheral blood specimens

[0116] The main purpose of this study is to develop a liquid biopsy-based detection method for predicting peritoneal recurrence in LAGC patients. We expanded the research scope from tissue samples to peripheral blood samples, and used the 6-mRNA combination to detect mRNA expression in serum samples of 120 LAGC patients (including 102 without peritoneal recurrence and 18 with recurrence). RT-qPCR was used for preliminary analysis to evaluate the expression levels of these mRNAs and assess their diagnostic utility. Based on multivariate Logistic regression, we constructed a prediction nomogram for peritoneal recurrence Figure 5 A), the AUC of the 6-mRNA biomarker combination was 0.903 (95% CI: 0.843-0.963, P<0.001; Figure 5 B), showing its predictive accuracy. The risk probability formula for peritoneal recurrence was calculated using logistic regression coefficients as [(3.873 x 6-mRNA combination) + (5.226 x invasion depth) + (2.761 x tumor size) + (1.848 x vascular cancer embolus invasion) + (-15.657)]. It is worth noting that the performance of the RSA model based on liquid biopsy is superior to that of the clinical model and the 6-mRNA model in predicting lymph node metastasis Figure 5 B). After risk stratification according to the Youden index, the confusion matrix analysis showed its predictive ability Figure 5 E-G), while the calibration curve analysis confirmed its excellent performance Figure 5 W1). Clinical benefit comparative analysis showed that the RSA model increased the detection rate of peritoneal recurrence in high-risk patients from 11.7% to 14.2%, and decreased the detection rate of peritoneal recurrence in low-risk patients from 3.3% to 0.8% Figure 5N-P). Patients were divided into low- and high-risk groups according to the Youden index, and follow-up showed that the 5-year RFS of the high-risk group was significantly lower than that of the low-risk group (36.7% vs. 54.9%, P = 0.016) Figure 5 X). This highlights the potential of the RSA model to improve clinical decision-making and optimize patient management.

[0117] Using the same statistical model and coefficients initially applied in the training cohort, we evaluated an independent external validation cohort comprising 104 LAGC patients without peritoneal recurrence and 19 patients with recurrence. The AUC was 0.960 (95% CI: 0.928-0.993, P < 0.001), confirming the ability of the RSA model to predict peritoneal recurrence Figure 5 C). Further analysis using a confusion matrix plot confirmed that the RSA model outperformed the 6-mRNA combination and the clinical characteristics model in predicting peritoneal recurrence Figure 5 H-J). Calibration curve analysis also confirmed the improved predictive performance of the RSA model Figure 5 W2). These results indicate that the tissue-based 6-mRNA combination was successfully applied to liquid biopsy using peripheral blood specimens and effectively predicted peritoneal recurrence in patients through the RSA model combining serum 6-mRNA data with clinical characteristics. Analysis of the clinical benefit rate for each model again confirmed the ability of the RSA model to improve the detection rate of peritoneal recurrence in the high-risk group (from 11.4% to 14.6%) while reducing the detection rate of peritoneal recurrence in the low-risk group (from 4.1% to 0.8%) Figure 5 Q-S). Using the nomogram risk grouping, a time-to-event test was performed on the validation cohort, which showed that the 5-year RFS of the high-risk group was significantly lower than that of the low-risk group (28.9% vs. 57.6%, P < 0.0001) Figure 5 Y), highlighting the potential of the RSA model to improve clinical outcomes through more accurate risk stratification.

[0118] In current clinical practice, peripheral blood tumor markers such as CA19-9, CA72-4, and CEA are commonly used indicators for monitoring peritoneal recurrence in LAGC patients. To evaluate whether our RSA model can identify tumor marker-negative patients who are likely to develop peritoneal recurrence, we analyzed peripheral blood samples from the combined training and validation cohorts. Among them, 46 patients had no peritoneal recurrence, and 10 patients had recurrence, and all the above tumor markers were negative. In the training set, the RSA model (AUC = 0.939, 95% CI: 0.870-1.000) outperformed the clinical characteristics model (AUC = 0.798; 95% CI: 0.637-0.959) and the 6-mRNA combination model (AUC = 0.911, 95% CI: 0.832-0.990) in predicting peritoneal recurrence Figure 5D). Other analyses using the confusion matrix plot and calibration curve confirmed that the RSA model was superior to other models in predicting peritoneal recurrence Figure 6 K-M, Figure 6 W3). We analyzed the clinical benefit rates between different models and again found that the RSA model was significantly superior to clinical features and 6-mRNA combination in detecting peritoneal recurrence Figure 6 T-V). Furthermore, survival analysis based on the high and low risk classification of the nomogram showed that in patients with tumor marker negative, the 5-year RFS of the high risk group was significantly lower than that of the low risk group (35.3% vs. 59.0%, P = 0.043) Figure 6 Z), highlighting the potential of the RSA model in improving the clinical decision and prognosis of LAGC patients.

[0119] 2.5 6-mRNA combination identified the presence of peritoneal metastasis and micrometastasis in the disease diagnosis of LAGC patients

[0120] A nomogram model was established using the training set data of peripheral blood to evaluate the early detection ability of 6-mRNA combination for peritoneal metastasis and micrometasis (P0CY1) in newly diagnosed LAGC patients. Unfortunately, current diagnostic techniques often ignore P0CY1 tumors, making them the main candidate for new therapies that can significantly improve the treatment outcome of this subgroup of gastric cancer. To verify the predictive ability of the RSA model combined with 6-mRNA and clinical features for P0CY1, we referred to two prospective studies (NCT03718624 and ChiCTR1800014817) in which P0CY1 patients received conversion therapy. No distant metastasis was found by preoperative abdominal CT, and the clinical consideration was local advanced stage. Through diagnostic laparoscopy and peritoneal free cancer cell detection, P0CY1 was detected in 12 of 66 LAGC patients. By analyzing the levels of 6 mRNAs in preoperative peripheral blood serum by RT-qPCR, we found that the RSA model (AUC = 0.941, 95% CI: 0.874-1.000) was significantly superior to clinical features (AUC = 0.852, 95% CI: 0.722-0.982) and 6-mRNA combination (AUC = 0.904, 95% CI: 0.832-0.976) in predicting P0CY1 Figure 6 A-B). In addition, confusion matrix and calibration curve analysis confirmed the superior prediction accuracy Figure 6 E-G,6Q). Comparative analysis of clinical benefit showed that the RSA model increased the detection rate of intraperitoneal micrometastasis in high-risk patients from 13.6% to 16.7%, and reduced the misdiagnosis rate of peritoneal recurrence in low-risk patients from 4.5% to 1.5% Figure 6K-M). According to the Youden index, patients were divided into low-risk and high-risk groups, and our follow-up results showed that the 5-year RFS of the high-risk group was significantly lower than that of the low-risk group (37.5% vs. 61.8%, P = 0.029) Figure 6 S), indicating that the RSA model has the potential to improve the prognosis of LAGC patients and guide clinical decision-making.

[0121] Next, we used serum samples from 95 LAGC patients in two previously registered prospective cohort studies (NCT02555358 and NCT01516944), and observed that 15 patients (15.8%) developed PM after diagnostic laparoscopy before treatment. Using the previously constructed nomogram model, the prediction accuracy of the RSA model (AUC = 0.970, 95% CI: 0.936-1.000) for PM was superior to that of clinical characteristics (AUC = 0.884, 95% CI: 0.736-0.964) and the 6-mRNA panel (AUC = 0.863, 95% CI: 0.701-0.941) Figure 6 C-D). Further analysis using a confusion matrix chart and calibration curve confirmed the excellent prediction performance ​ H-J,6R). Notably, our analysis of the clinical benefit rate of different models showed that the RSA model increased the detection rate of PM in high-risk patients from 11.6% to 14.7%, while reducing the false identification rate of peritoneal recurrence in the overall risk group from 4.2% to 1.1% ​ N-P). Follow-up studies showed that the 5-year recurrence-free survival rate (RFS) of the high-risk group was significantly lower than that of the low-risk group (29.0% vs. 58.3%, P = 0.0045) ​ T). Overall, our systematic and thorough biomarker discovery and validation approach led to the identification and validation of a new gene prediction panel, which significantly strengthened the management of gastric cancer by improving the detection rate of peritoneal metastasis, thus potentially improving overall survival.

[0122] In addition, we also evaluated the diagnostic performance of the 6-mRNA combination in serum samples of LAGC patients and evaluated its efficacy in other gastrointestinal malignancies, including colorectal cancer (n = 35), hepatocellular carcinoma (n = 38), and pancreatic cancer (n = 29). Interestingly, the diagnostic accuracy of our 6-mRNA combination in predicting PM (AUC = 0.879) was higher than that of all other gastrointestinal cancers evaluated (colorectal cancer: AUC = 0.756; hepatocellular carcinoma: AUC = 0.642; pancreatic cancer: AUC = 0.682; ​u-w). DeLong's test further confirmed that the 6-mRNA combination diagnosis of PM in LAGC has higher specificity compared to other gastrointestinal cancers (P < 0.001 for gastric cancer, colorectal cancer, hepatocellular carcinoma, and pancreatic ductal carcinoma). In summary, these findings suggest that our 6-mRNA biomarker combination has high specificity in differentiating LAGC patients from other patients with gastrointestinal malignancies, even as a non-invasive blood test.

Claims

1. Use of a substance for detecting a biomarker in a sample in any one of the following, comprising: A1. use in the manufacture of a product for diagnosing recurrence of gastric cancer after surgery; A2. use in the manufacture of a product for predicting recurrence of gastric cancer after surgery; the biomarker is BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof.

2. The use of claim 1, wherein the biomarker further comprises depth of invasion, tumor size, vascular cancer embolus invasion; preferably, the product comprises reagent, kit, test paper, primer, probe, chip; preferably, the kit further comprises one or more substances selected from the group consisting of container, instruction manual, positive control, negative control, buffer, adjuvant or solvent; preferably, the chip comprises gene chip, protein chip; preferably, the kit further comprises mRNA expression level auxiliary detection reagent; preferably, the mRNA expression level auxiliary detection reagent comprises primer corresponding amplicon visualization reaction reagent, RNA extraction reagent, reverse transcription reagent, cDNA amplification reagent, and / or standard used for preparing standard curve; preferably, the primer corresponding amplicon visualization reaction reagent comprises reagent for making amplicon visualizable in agarose gel electrophoresis, enzyme-linked gel method, chemiluminescence method, in situ hybridization method, fluorescence detection method; preferably, the kit further comprises protein expression level auxiliary detection reagent; preferably, the protein expression level auxiliary detection reagent comprises blocking solution, antibody diluent, washing buffer, color developing termination solution.

3. The use of claim 1, wherein the sample comprises blood, serum, plasma, peripheral blood, tissue, blood cells, bone marrow, ascites, fine needle biopsy sample, cell-containing body fluid, free floating nucleic acid, sputum, saliva, urine, cerebrospinal fluid peritoneal fluid, pleural fluid, feces, lymph, skin swab, oral swab, nasal swab or lavage; preferably, the sample comprises blood, serum, plasma, peripheral blood, tissue, or fine needle biopsy sample; preferably, the gastric cancer comprises locally advanced gastric cancer; preferably, the recurrence of gastric cancer after surgery comprises peritoneal recurrence of gastric cancer after surgery, lymph node recurrence of gastric cancer after surgery, lung recurrence of gastric cancer after surgery, liver recurrence of gastric cancer after surgery, bone recurrence of gastric cancer after surgery; preferably, the sample is from a subject to be detected; preferably, the subject to be detected comprises human or non-human mammal; preferably, the subject to be detected is human.

4. A product comprising detection reagent for detecting the content of a biomarker, the biomarker is BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof.

5. The product of claim 4, wherein the biomarker further comprises depth of invasion, tumor size, vascular cancer embolus invasion; preferably, the product comprises reagent, kit, test paper, primer, probe; preferably, the kit comprises immunohistochemical detection kit, immunoblotting detection kit, immunochromatographic detection kit, flow cytometry kit, qPCR kit, ELISA kit; Preferably, the product further comprises reagents conventionally used in PCR reactions, RT-PCR derivative reactions, 3SR amplification, LCR, SDA, NASBA, TMA, SYBR Green, TaqMan probes, molecular beacons, dual hybridization probes, complex probes, ISH, microarrays, Southern blots, Northern blots, multi-analyte profiling tests, enzyme-linked immunosorbent assays (ELISA), radioimmunoassays, immunofluorescence assays, enzyme immunoassays, immunoprecipitation assays, chemiluminescence assays, immunohistochemical assays, dot blot assays or slot blot assays.

6. A system for diagnosing whether a gastric cancer has recurred after surgery or predicting the risk of recurrence of a gastric cancer after surgery, the system comprising: an input unit configured to acquire expression level detection data of biomarkers in a sample; a result determination unit configured to input data from the detection unit into a computer program for processing and determination, and obtain a comparison result after comparing the data with a set threshold value in the computer; an output unit configured to output the result from the result determination unit to obtain whether the subject corresponding to the sample has a recurrence of a gastric cancer after surgery or the probability of recurrence of a gastric cancer after surgery; all units in the system are operated using a computer program or artificial intelligence; the biomarkers are BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof; Preferably, the biomarkers further comprise depth of invasion, tumor size, and vascular cancer embolus invasion.

7. A gastric cancer recurrence risk prediction model, the gastric cancer recurrence risk prediction model comprising an algorithm model trained using a machine learning method for diagnosing whether a gastric cancer has recurred after surgery or predicting the risk of recurrence of a gastric cancer after surgery.

8. The gastric cancer recurrence risk prediction model of claim 7, wherein the algorithm model processes data to obtain a gastric cancer recurrence evaluation result, the gastric cancer recurrence evaluation result comprising a gastric cancer recurrence probability and an important feature factor reflecting the correlation with the recurrence of a gastric cancer after surgery; and the important feature factor is expression level data of BUB1, CKS2, PCNA, CHEK1, NEK2, NCAPG2, or a combination thereof extracted from information parameters of the subject to be detected. Preferably, the important feature factor further comprises depth of invasion, tumor size, and vascular cancer embolus invasion data extracted from information parameters of the subject to be detected. Preferably, one RSA model in the algorithm model = [(3.665 x 6-mRNA marker combination) + (5.009 x depth of invasion) + (2.009 x tumor size) + (1.417 x vascular cancer embolus invasion) + (-12.340)], and the RSA model is used to evaluate the risk of recurrence of a gastric cancer after surgery. Preferably, the machine learning method comprises a decision tree learning method, a random forest learning method, a K-nearest neighbor algorithm learning method, a naive Bayes learning method, a support vector machine learning method, a neural network learning method, and an Adeboost learning method. 9.A computer readable storage medium having a computer program stored thereon, wherein the computer program is stored in the computer readable storage medium and, when executed by a processor, implements the gastric cancer postoperative recurrence risk prediction model of claim 7 or 8. 10.A computer evaluation method for gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk, the method comprising: obtaining to-be-processed data, wherein the to-be-processed data comprises detection data of a BUB1, a CKS2, a PCNA, a CHEK1, a NEK2, a NCAPG2, or a combination thereof biomarker in a sample; inputting the to-be-processed data into the gastric cancer postoperative recurrence risk prediction model of claim 7 or 8 to obtain an evaluation result of the to-be-processed data for gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk; and outputting the evaluation result of the to-be-processed data for gastric cancer postoperative recurrence or gastric cancer postoperative recurrence risk. Preferably, the to-be-processed data further comprises detection data of a subject's depth of invasion, tumor size, and vascular cancer embolus infiltration. ​ ​ ​

Images