Detection kit for prognosis of gastric cancer

Abstract

The invention discloses a detection kit for gastric cancer prognosis, which comprises the steps of establishing a gastric cancer prognosis model and obtaining a prognosis related target gene, the kit comprises a primer pair for detecting the expression level of the prognosis target gene, and the prognosis target gene at least comprises an OR51G2 protein gene and also comprises a CG beta8 protein gene and / or a UPK1B protein gene. The invention discovers that the expression level of the OR51G2 protein gene is related to poor prognosis of gastric adenocarcinoma patients for the first time, and RT-PCR (Reverse Transcription-Polymerase Chain Reaction) experiments discover that obvious expression difference exists between gastric cancer tissues and para-carcinoma tissues. CG [beta] 8 and UPK1B are proved to play an important role in malignant biological behaviors of gastric cancer in a plurality of researches, and are related to poor prognosis of gastric cancer patients. The primer pair designed aiming at three prognosis targets of OR51G2, CG beta8 and UPK1B is adopted, the gastric cancer prognosis can be rapidly and accurately carried out through RT-PCR (Reverse Transcription-Polymerase Chain Reaction), and the primer pair has a very good diagnosis application prospect.

Description

Technical Field

[0001] The present invention relates to a method for constructing a cancer prognosis model and a detection kit for cancer prognosis, and in particular to a method for constructing a gastric cancer prognosis model and a detection kit for gastric cancer prognosis. Background Art

[0002] Gastric cancer is a highly heterogeneous tumor with many prognostic factors. FA metabolism is significantly related to the occurrence, development and invasion of gastric cancer, affecting the prognosis of gastric cancer patients.

[0003] Fatty acid (FA) metabolism plays an important role in the occurrence and progression of various types of cancer in vitro and in vivo. Abnormal fatty acid metabolism affects the invasive ability of tumors. FA metabolism is significantly associated with biological behaviors such as cancer cell migration, angiogenesis, human immune surveillance of tumor cells, and tumor resistance to chemotherapeutic drugs. It interferes with the invasive ability of gastric cancer and affects the prognosis of gastric cancer patients. A variety of models for predicting gastric cancer prognosis have been established based on tumor immunology, histone lactylation, and tumor autophagy. However, there is a lack of relevant research on the heterogeneity of gastric cancer based on the subtypes of fatty acid metabolism-related genes (FMGs).

[0004] Clinical prediction models are statistical models that assess the likelihood of a disease or outcome based on the characteristics of the research population. They are mainly divided into diagnostic models and prognostic models. Bioinformatics methods gather knowledge from multiple disciplines such as statistics, biology, and computer science. They use biomacromolecules as research objects, mine large amounts of information data, and combine them with machine learning algorithms (ML) to perform comprehensive analysis such as feature screening and feature extraction of different omics, such as genomics, metabolomics, and proteomics data, obtain digital information, discover key gene combinations, train and build models, and perform well in predicting the progression of diseases.

[0005] For different omics, the characteristics of samples can be analyzed and classified into different subgroups through molecular classification methods. Applying it to oncology can help determine different types of tumor subtypes, analyze differences in survival time, treatment response, etc. among different group subtypes, and thus promote individualized treatment and prognosis assessment. Common classification algorithms include consensus clustering, non-negative matrix factorization (NMF), and self-organizing map (SOM Clustering).

[0006] NMF is a matrix decomposition method that performs dimensionality reduction analysis and feature extraction under the constraint that all elements in the matrix are non-negative numbers. The basic idea of ​​NMF can be simply understood as follows: for any given non-negative matrix V, the NMF algorithm can find a non-negative matrix W and a non-negative matrix H, so that V = W*H holds, thereby decomposing a non-negative matrix into the product of two non-negative matrices on the left and right, and first decomposing a large non-negative matrix into two small non-negative matrices. Since the decomposed matrix is ​​also non-negative, it can also be further decomposed. The NMF algorithm usually runs many times with different parameters, and the results are output and analyzed after the NMF rank metric diagram, including co-occurrence correlation coefficients, silhouette coefficients, dispersion coefficients, etc., which can reflect whether the instance is compact in various clusters, and can be given a certain degree of dominance to evaluate the clustering effect and find the best classification results. The commonly used standard for evaluating whether the clustering effect is compact is the front point with the largest decline range of the cophenetic curve. It can also be determined by the Silhouette curve and the dispersion curve. It can also be determined by the first r value of the inflection point of the residual sum of square (RSS) curve. The clustering effect can also be judged by directly visually observing the consensus matrix.

[0007] In the medical field, commonly used machine learning methods for risk prediction modeling include survival trees, Bayesian methods, neural networks, support vector machines, and other machine learning algorithms. According to the characteristics of survival time distribution, statistical methods used in survival analysis include parametric methods, semi-parametric methods, and non-parametric methods. When the survival time distribution is unknown and the proportional risk assumption does not hold, non-parametric methods are more applicable.

[0008] Random survival forest (RSF) is an extension of random forests (RF) to process survival data. As an integrated machine learning algorithm for processing right-censored data, RSF has the characteristics of a non-parametric method. RSF is a derivative of the binary decision tree. - A collection of binary survival trees. It trains a large number of survival trees. Each survival tree splits the node according to the criterion of maximizing the survival difference and outputs a prediction result. Then, the prediction results of the individual trees are comprehensively weighted and the final result is selected by voting. The general process is: (1) A subset of samples is randomly selected from a given data set with replacement through the "bootstrap method" to construct a survival tree. Because about 37% of the data in each sample is not extracted, these are called out-of-bag (OOB) data. (2) For each training sample set, features are randomly selected to construct its corresponding binary survival tree. The candidate features are used to split the root node of each survival tree to maximize the survival difference between subsets. (3) The Nelson-Aalen method is used to estimate the total cumulative risk of the random survival forest. (4) The value of Harrell's consistency index is generally used to evaluate the accuracy of the model constructed by RSF. The higher the consistency index value, the higher the accuracy of the model. C-index = 1-Error rate. Random survival forest can rank the importance of variables by using VIMP method or minimum depth method, and find out the important variables for outcome events. Compared with traditional survival statistical analysis methods such as Cox proportional hazards regression (CPH) model, random survival forest can be an alternative model to CPH model when the actual survival data is difficult to meet the equal proportional hazard assumption set for CPH model and the requirement that covariates are independent of each other. RSF has no requirements for data types and flexible structural design. Because Bootstrap sampling is performed on the data set, internal evaluation is performed during the generation process, and an unbiased estimate of the error is obtained, so there is no need to apply other independent test machines for verification.

[0009] CGβ8 is one of the alleles encoding the β subunit of human chorionic gonadotropin (HCG), and β-HCG is closely related to pregnancy. Similar to previous studies, CGβ8 is a prognostic factor for gastric cancer patients. Zhao R et al. found that the expression of HCG and its receptor encoded by CGβ8 in gastric cancer tissues was significantly higher than that in adjacent tissues, which may promote the proliferation of tumor cells by activating the expression of c-Met dependent on the PKA signaling pathway. Human urinary protein 1B (Uroplakin1B, UPK1B) is a structural protein of urothelial cells, which can promote the proliferation, invasion and metastasis of tumor cells. It is differentially expressed in a variety of tissue-specific tumors, such as urothelial carcinoma, renal cancer and other malignant tumors. The human olfactory receptor (OR) gene family is an important member of the G protein-coupled receptors in sensory neurons. It not only plays an important role in olfactory epithelial cells, but also plays a key role in the invasion of tumor cells. The OR gene family is significantly overexpressed in tumors and is considered to be a tumor biomarker, but the function of the OR family in gastric cancer and its potential molecular mechanism are still largely unknown. Summary of the invention

[0010] Purpose of the invention: The purpose of the present invention is to provide a method for constructing a gastric cancer prognosis model, to solve the problem of how to screen and use fatty acid metabolism-related genes to construct a gastric cancer prognosis model. Another purpose of the present invention is to provide a detection kit for gastric cancer prognosis, to solve the problem of how to use new target genes for gastric cancer prognosis.

[0011] Technical solution: The method for constructing a gastric cancer prognosis model described in the present invention comprises the following steps:

[0012] The gastric cancer transcriptome and clinical data were used to screen differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples;

[0013] Collect fatty acid metabolism genes, take the intersection of differentially expressed genes and fatty acid metabolism genes, and obtain differentially expressed fatty acid metabolism-related genes;

[0014] A non-negative matrix method was used to decompose a gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes to identify patient subtypes with different prognoses;

[0015] Enrichment analysis of differential functions, differential pathways, and tumor microenvironment of patient subtypes with different prognoses;

[0016] The differentially expressed fatty acid metabolism-related genes between patient subtypes with different prognoses were used as candidate genes to establish the SUM model, and the core genes were screened based on the gastric cancer patient dataset to construct a gastric cancer prognosis model.

[0017] Based on machine learning algorithms, this paper explores the relationship between differentially expressed FMGs in gastric cancer and the prognosis of gastric adenocarcinoma patients, and establishes a gastric cancer prognosis model based on FMGs classification. First, through the NMF cluster analysis algorithm, two molecular subtypes in the TCGA-STAD cohort were successfully identified: C1 subtype and C2 subtype. The KM method (Log-Rank test) was used to observe whether there was a survival difference between the subtypes. The results showed that the 5-year DSS, PFI and overall OS of C1 subtype patients had a more obvious survival advantage than those of C2 subtype patients (P<0.05).

[0018] Different functions, different pathways and tumor microenvironment are all intended to explain the classification and classify patients with two prognoses.

[0019] One of the key genes of the prognostic model constructed by the present invention, OR51G2, was first discovered to be associated with the prognosis of patients with gastric adenocarcinoma. Preferably, the use of gastric cancer transcriptome and clinical data to screen differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples is: using gastric cancer transcriptome and clinical data downloaded from the TCGA database, based on the LIMMA software package in R language, with the cutoff standard being the difference multiple FC, |log 2 FC|>1 and adjusted P value<0.05 were used to screen the differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples.

[0020] Preferably, the non-negative matrix method is used to decompose the gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes to identify patient subtypes with different prognoses: the rank is set from 2 to 10 to perform non-matrix decomposition on the gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes, the optimal rank is determined based on stability, and patient subtypes with different prognoses are generated based on the optimal rank.

[0021] Preferably, the screening of core genes based on a gastric cancer patient dataset to construct a gastric cancer prognosis model comprises:

[0022] Gastric cancer patients were randomly divided into training set and test set. Univariate Cox regression analysis was performed on candidate genes in the training set. Genes related to overall prognosis were screened out with P < 0.05 as the filtering condition. Then, multivariate Cox regression analysis was used to screen out core genes to construct a gastric cancer prognostic model.

[0023] The test set was repeatedly sampled in the gastric cancer prognostic model and the model was fitted. The random survival forest was used to further evaluate the importance of core genes to the patient's survival status. The Cox regression coefficient and gene variables were used to construct the model, and risk scores were given to patients. Gastric cancer patients were divided into high-risk patients and low-risk patients according to the median value of the risk score.

[0024] Preferably, the core genes are CGβ8 protein gene, UPK1B protein gene and OR51G2 protein gene.

[0025] The present invention further provides a detection kit for gastric cancer prognosis based on the core genes in the above-mentioned prognostic model, comprising a primer pair for detecting the expression level of the prognostic target gene, wherein the prognostic target gene comprises the OR51G2 protein gene, the CGβ8 protein gene and / or the UPK1B protein gene.

[0026] Preferably, the method for detecting the expression level of the prognostic target gene is fluorescent quantitative PCR. The relevant reagents of fluorescent quantitative PCR include dNTP, polymerase, etc. The existing commercial reagent SuperReal PreMix contains the necessary components required for fluorescent quantitative PCR. The reaction system of the fluorescent quantitative PCR is:

[0027]

[0028] Preferably, the primer pair for detecting the expression level of the OR51G2 protein gene is:

[0029] Upstream primer: 5′-TGGAGCACGAGAAATTAGCCA-3′;

[0030] Downstream primer: 5′-GCCACAATAGGGGAATCTTTTGA-3′.

[0031] The primer pair for detecting the expression level of the CGβ8 protein gene is:

[0032] Upstream primer: 5′-CTGGCTGTGGAGAAGGAG-3′;

[0033] Downstream primer: 5′-ACATCGCGGTAGTTGCAC-3′;

[0034] The primer pair for detecting the expression level of the UPK1B protein gene is:

[0035] Upstream primer: 5′-CCAAAGACAACTCAACTGTTCGT-3′;

[0036] Downstream primer: 5′-AATGCCGCAACAACCAATAATC-3′;

[0037] Preferably, the internal reference gene of the fluorescent quantitative PCR is GAPDH, and the above kit also includes a primer pair for detecting the expression level of GAPDH, the sequence of which is:

[0038] Upstream primer: 5′-ACAACTTTGGTATCGTGGAAGG-3′;

[0039] Downstream primer: 5′-GCCATCACGCCACAGTTTC-3′.

[0040] GAPDH can also be replaced by other internal reference genes such as β-actin and 18S rRNA according to detection needs.

[0041] Preferably, the above-mentioned kit also includes a reverse transcription PCR reagent and an RNA extraction reagent, and the reaction system of the reverse transcription PCR is:

[0042]

[0043] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0044] Based on FMGs subtypes, the present invention screened independent risk genes related to the prognosis of STAD patients by RSF combined with multifactor Cox regression model, and constructed a prognostic risk model composed of CGβ8, UPK1B and OR51G2. The accuracy of the model was about 64%, and the patients were divided into high-risk group and low-risk group according to the median value of the risk score. KM survival analysis showed that the DSS, OS and PFI of the high-risk and low-risk groups were significantly different (P<0.001), and the low-risk group had a more obvious survival advantage. The three FMGs (CGβ8, UPK1B, OR51G2) discovered by the present invention are closely related to the prognosis of gastric cancer patients, and are all adverse prognostic factors for gastric cancer.

[0045] The present invention discovered for the first time that the expression level of the OR51G2 protein gene is associated with a poor prognosis in patients with gastric adenocarcinoma, and RT-PCR experiments found significant expression differences between gastric cancer tissue and adjacent tissue. CGβ8 and UPK1B have been shown in multiple studies to play an important role in the malignant biological behavior of gastric cancer and are associated with a poor prognosis in patients with gastric cancer. The present invention uses primer pairs designed for three prognostic targets, OR51G2, CGβ8 and UPK1B, and can quickly and accurately predict the prognosis of gastric cancer by RT-PCR, which has a good prospect for diagnostic applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The Venn diagram of differentially expressed genes and genes related to fatty acid metabolism;

[0047] Figure 2 KEGG pathway enrichment analysis for FMGs subtypes of gastric cancer;

[0048] Figure 3 To determine the optimal decomposition level r value through the NMF algorithm;

[0049] Figure 4The differences in detection point-related genes between FMGs subtypes in the TCGA-STAD dataset;

[0050] Figure 5 The differences in the stromal cell scores, immune cell scores, and comprehensive estimated scores of the detection points between FMGs subtypes in the TCGA-STAD dataset;

[0051] Figure 6 The differences in immune cell scores between FMGs subtypes in the TCGA-STAD dataset;

[0052] Figure 7 To determine the core genes of the prognostic model for gastric cancer based on random forest and COX regression;

[0053] Figure 8 This is a statistical diagram of the differential expression of CGβ8, UPK1B and OR51G2 mRNA in gastric cancer / paracancerous tissues detected by RT-PCR. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0055] Example 1: A method for constructing a gastric cancer prognosis model is as follows:

[0056] The prognostic data of gastric cancer were obtained from the TCGA database and the GEO database, and a prediction model for gastric cancer was constructed based on a machine learning algorithm to assist clinicians in improving the prognosis and survival of gastric cancer patients.

[0057] The TCGA database used in the present invention is currently the largest cancer gene information data. It contains multi-omics data of 33 types of cancer, including gene expression data, DNA methylation data, miRNA expression data, copy number variation data, SNP data, etc., and also integrates clinical sample information of tumor patients: population data (Biospecimen) and clinical data (Clinical), including age, gender, TNM staging and follow-up results.

[0058] The non-negative matrix (NMF) method was used to decompose the gastric adenocarcinoma (STAD) patient cohort with differential expression of fatty acid metabolism genes from TCGA, and two subtypes with different prognoses were identified. Then, the differential functional enrichment, differential pathways, tumor microenvironment TM and other aspects of the two subtypes were compared and analyzed. Finally, a model based on CGβ8, OR51G2 and UPK1B genes was constructed using the random survival forest method to predict the prognosis of gastric cancer. The specific steps are:

[0059] 1.1 Download of gastric cancer patient data and related genes

[0060] Gastric cancer transcriptome and clinical data downloaded from the TCGA database were used to screen differentially expressed genes (DEGs) between gastric cancer tissue samples and adjacent tissue samples. Based on the LIMMA package in R language, the cutoff criteria were the absolute value of the fold change (FC) |log2FC|>1 and the adjusted P value <0.05. Finally, 2155 differentially expressed genes were screened, of which 975 were up-regulated genes and 1180 were down-regulated genes. The results are as follows Figure 1 shown.

[0061] 1.2 Extraction of differentially expressed fatty acid metabolism-related genes in gastric cancer

[0062] First, 309 fatty acid metabolism genes were obtained and collected from the fatty acid metabolism pathway of the molecular feature database KEGG, the fatty acid metabolism genes of Hallmark, and the specific genes related to fatty acid metabolism of Reactome. The intersection of differentially expressed genes in gastric cancer and fatty acid metabolism genes was identified, and 78 differentially expressed fatty acid metabolism-related genes were obtained.

[0063] 1.3 Identification of subtypes of differentially expressed fatty acid metabolism-related genes in gastric cancer

[0064] In this study, the "NMF" package was used to decompose the gastric adenocarcinoma patient cohort, and the rank was set from 2 to 10 for non-matrix decomposition to determine the optimal rank. Results: The "brunet" algorithm was used for 200 iterations. The cophenetic correlation coefficients and silhouette coefficients were obtained / determined by the cophenetic correlation coefficients, dispersion coefficients and mean value of silhouette width. Figure 3 As shown in the figure, Figure A is the NMF rank measurement diagram; Figure B is the subtype classification heat map with ranks from 2 to 10. It was determined that the optimal rank was 2 with the best stability, resulting in two subtypes, Cluster A and Cluster B. The Kaplan-Meier (KM) method was used to analyze the survival differences between subtypes. The analysis results showed that the OS and PFS of Cluster A patients were significantly better than those of Cluster B patients.

[0065] 1.4 Differences in biological functions between subtypes

[0066] In order to explore the molecular mechanism and potential regulatory pathways of how fatty acid metabolism genes regulate the malignant biological behavior of gastric cancer, GO function and KEGG pathway enrichment analysis were performed on differentially expressed genes. In this study, the "edgeR" package was first used to identify DEGs between C1 and C2 subtypes. The cutoff criteria were: absolute value of fold change (FC) |log2FC|>1 and adjusted P value <0.05. After the genes were arranged in descending order according to logF value, enrichment analysis was performed using the GSEA function in the "clusterProfiler" package. The reference gene set mainly used Gene Ontology (GO) function and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis to explore the various biological processes and signaling pathways that the selected genes may be involved in.

[0067] The strict threshold of pathway enrichment (FDR<0.05) was used to sort the significantly enriched gene sets according to the NES value / FDR value, and the top 50 KEGG pathways significantly enriched between Cluster A and Cluster B were listed. The previously screened DEGs were uploaded using the STRING online analysis website to obtain the protein-protein interaction (PPI) network. Then, the Cytoscape software was used to set the number of hubs to 10 and identify the top 10 key genes with significant expression in the protein interaction network (PPI). The results are shown in Figure 2. Figure 2 shown.

[0068] 1.5 Comparative analysis of immune-related characteristics among different subtypes

[0069] In order to further observe the potential differences in immune infiltration between gastric cancer patients in different model groups, the ESTIMATE algorithm was used for inference. The immune score (ImmuneScore) and stromal score (StromaScore) of gastric cancer patients were evaluated by the R package "ESTIMATE". The ESTIMATEScore is equal to the sum of the ImmuneScore and the StromaScore, which is used to evaluate the purity of the tumor. The higher the ESTIMATEScore, the lower the purity of the tumor. At the same time, the NMF algorithm was used to analyze the differences between the common immune detection genes of gastric cancer patients between the two subtypes, such as Figure 4 , Figure 5 and Figure 6As shown, in the STAD cohort, CTLA4, CXCL9, CXCL10, IDO1, and CD8A were significantly increased in Cluster B (p<0.05). The ESTIMATE evaluation results of the STAD cohort showed that the matrix score, immune score, and ESTIMATE score of subtype Cluster B were significantly higher than those of Cluster A (p<0.05). Neutrophils, type 17 helper cells, CD56 bright NK cells, and CD56 dim NK cells were highly expressed in the C1 subtype, and eosinophils, naive B cells, naive dendritic cells, myeloid suppressor cells, mast cells, natural killer cells, natural killer T cells, plasmacytoid dendritic cells, regulatory T cells, follicular helper T cells, type I helper cells, and type 2 helper cells were highly expressed in the C2 subtype.

[0070] 1.6 Construction of gastric cancer prognostic model

[0071] The construction of the model used the differentially expressed fatty acid metabolism-related genes between C1 and C2 subtypes as candidate genes to establish the SUM model. Gastric cancer patients were randomly divided into two groups, one for training and the other for testing, with 500 samples in each group. Univariate Cox regression analysis was used for differentially expressed genes in the training set, with P < 0.05 as the filtering condition, and a total of 4 genes related to overall prognosis were screened out. Then, multivariate Cox regression analysis was used to screen out 3 core genes (fatty acid metabolism-related genes CGβ8, UPK1B, and OR51G2) to construct a gastric cancer prognosis model, such as Figure 7 As shown, Figure 7 The left figure in the middle is the out-of-bag error rate curve of the random survival forest model; the right figure is the variable importance ranking. The test set was repeatedly sampled in the model and the model was fitted 1000 times. The random survival forest was used to further evaluate the importance of these three fatty acid metabolism-related genes to the patient's survival status. The Cox regression coefficient and the gene variables were used to construct the model, and the risk score was calculated for the patient. The specific calculation method was risk score = expression of CGβ8 × 0.004321797 + expression of UPK1B × 0.000139908 + expression of OR51G2 × 0.222133824. Patients were divided into high-risk patients and low-risk patients according to the median value of risk score. Patients with a value greater than the median value of risk score were defined as high-risk patients, and vice versa. The results of survival analysis showed that compared with patients in the high-risk group, patients in the low-risk group had better clinical prognosis in terms of overall survival (OS), disease-specific survival (DSS) and progression-free interval (PFI).

[0072] Example 2: Verify the differential expression of CGβ8, UPK1B and OR51G2 in gastric cancer and adjacent tissues, the method is as follows:

[0073] 1) Collection of clinical tissue samples

[0074] The experimental samples were all from patients with gastric adenocarcinoma pathologically confirmed by the Department of Gastrointestinal Surgery of Jining First People's Hospital. The collection time was July to August 2023, and a total of 4 patients' cancer and adjacent cancer tissues were collected. The ethics number was 2023 Lunshenyan No. (050).

[0075] 2) Real-time fluorescence quantitative PCR (RT-PCR) experiment

[0076] RNA extraction

[0077] (1) Tissue samples were collected on ice, and the centrifuge tubes, tweezers, scissors and other tools used for sampling were cooled on dry ice in advance.

[0078] (2) Take 50 mg of tissue and place it in an RNase-free EP tube. Add 1 ml of TRlpure Reagent and grind it on ice using an ultrasonic cell grinder until no obvious solid tissue is observed. Incubate at room temperature for 5 min.

[0079] (3) Add 0.2 ml of chloroform to 1 ml of TRlpure Reagent lysis buffer, mix for 30 seconds, and incubate at room temperature for 3 minutes.

[0080] (4) Centrifuge the EP tube at 4°C and 12,000 rpm / min for 5-10 minutes.

[0081] (5) Pipette the upper aqueous phase into an RNase-free centrifuge tube, add an equal volume of 70% ethanol, and vortex to mix.

[0082] (6) Add the above mixture to the RNA adsorption column and centrifuge at room temperature, 12000 rpm / min for 1 minute. Discard the waste liquid in the collection tube and repeat twice.

[0083] (7) Add 700 μL of deproteinized solution 1 from the TRlpure total RNA rapid extraction kit, centrifuge at room temperature, 12,000 rpm / min for 1 minute, and discard the waste liquid in the collection tube.

[0084] (8) Add 500 μL of rinse buffer and centrifuge at room temperature, 12,000 rpm / min for 1 min. Repeat twice.

[0085] (8) Place the RNA adsorption column in a new RNase-free centrifuge tube, add 60ul RNase-free H2O to the adsorption membrane, place at room temperature for 3-5min, and centrifuge at 12000rpm / min for 2 minutes. The solution in the centrifuge tube is the RNA sample.

[0086] (9) Detect the concentration of RNA samples.

[0087] Reverse transcription experiments

[0088] (1) Prepare the reverse transcription system by mixing pre-transcribed RNA, 5×FastKing-RT SuperMx and RNase-Free ddH2O in the following amounts.

[0089]

[0090] (2) After the system is configured, vortex the EP tube to mix it evenly, put it into the PCR instrument, and perform reverse transcription according to the following program: 42°C for 15 minutes (reverse transcription reaction) → 95°C for 3 minutes (enzyme inactivation).

[0091] (3) Place the reverse transcribed cDNA on ice before performing real-time quantitative PCR reaction; if long-term storage is required, store it in a -20°C refrigerator.

[0092] Real-Time PCR

[0093] (1) Real-time fluorescence quantitative PCR was used to detect the mRNA expression of CGβ8, UPK1B, OR51G2 and GAPDH in gastric cancer tissues and adjacent tissues. The primer pairs used for real-time fluorescence quantitative PCR detection are as follows:

[0094]

[0095]

[0096] (2) Prepare the RT-PCR reaction solution on ice. Perform three replicate wells for each sample and prepare the following reaction system on a 96-well plate:

[0097]

[0098] (3) Seal the membrane, centrifuge briefly, and set the program of the fluorescence quantitative PCR instrument: 95°C for 15 min, 40×[95°C for 10 sec → 65°C for 20 sec → 72°C for 20 sec], and collect the fluorescence signal at 65°C.

[0099] (4) Data processing: GAPDH was the internal reference gene, and the relative expression of each gene in each sample was analyzed by relative quantitative method, and the calculation method was 2^(-△△CT) method.

[0100] 3) Results

[0101] The expression of the core genes of the CGβ8, UPK1B and OR51G2 models in clinical gastric cancer samples was verified. The results of RT-PCR experiments showed that the expression of CGβ8 and UPK1B in gastric cancer tissues was significantly upregulated compared with that in adjacent tissues (P<0.05), while no expression difference was detected for OR51G2. Figure 8 As shown, Figure A is a comparison chart of the expression levels of CGβ8 protein gene, Figure B is a comparison chart of the expression levels of UPK1B protein gene, and Figure C is a comparison chart of the expression levels of OR51G2 protein gene. * represents P < 0.05.

Claims

1. A method for constructing a gastric cancer prognosis model, characterized in that: The steps include: The gastric cancer transcriptome and clinical data were used to screen differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples; Collect fatty acid metabolism genes, take the intersection of differentially expressed genes and fatty acid metabolism genes, and obtain differentially expressed fatty acid metabolism-related genes; A non-negative matrix method was used to decompose a gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes to identify patient subtypes with different prognoses; Enrichment analysis of differential functions, differential pathways, and tumor microenvironment of patient subtypes with different prognoses; The differentially expressed fatty acid metabolism-related genes between patient subtypes with different prognoses were used as candidate genes to establish the SUM model, and the core genes were screened based on the gastric cancer patient dataset to construct a gastric cancer prognosis model.

2. The method for constructing a gastric cancer prognosis model according to claim 1, characterized in that: The method of using gastric cancer transcriptome and clinical data to screen differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples is as follows: using gastric cancer transcriptome and clinical data downloaded from the TCGA database, based on the LIMMA software package of the R language, with a cutoff standard of differential fold FC, |log2FC|>1 and an adjusted P value <0.05 to screen differentially expressed genes between gastric cancer tissue samples and adjacent tissue samples.

3. The method for constructing a gastric cancer prognosis model according to claim 1, characterized in that: The non-negative matrix method is used to decompose the gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes, and identify patient subtypes with different prognoses: the rank is set from 2 to 10, the gastric adenocarcinoma patient cohort with differentially expressed fatty acid metabolism-related genes is subjected to non-matrix decomposition, the optimal rank is determined according to stability, and patient subtypes with different prognoses are generated according to the optimal rank.

4. The method for constructing a gastric cancer prognosis model according to claim 1, characterized in that: The method of screening core genes based on a gastric cancer patient data set to construct a gastric cancer prognosis model comprises: Gastric cancer patients were randomly divided into training set and test set. Univariate Cox regression analysis was performed on candidate genes in the training set. Genes related to overall prognosis were screened out with P < 0.05 as the filtering condition. Then, multivariate Cox regression analysis was used to screen out core genes to construct a gastric cancer prognostic model. The test set was repeatedly sampled in the gastric cancer prognostic model and the model was fitted. The random survival forest was used to further evaluate the importance of core genes to the patient's survival status. The Cox regression coefficient and gene variables were used to construct the model, and risk scores were given to patients. Gastric cancer patients were divided into high-risk patients and low-risk patients according to the median value of the risk score.

5. The method for constructing a gastric cancer prognosis model according to claim 1, characterized in that: The core genes are CGβ8 protein gene, UPK1B protein gene and OR51G2 protein gene.

6. A detection kit for gastric cancer prognosis, characterized in that: It comprises a primer pair for detecting the expression level of the prognostic target gene, wherein the prognostic target gene comprises the OR51G2 protein gene, the CGβ8 protein gene and / or the UPK1B protein gene.

7. The detection kit for gastric cancer prognosis according to claim 6, characterized in that: The method for detecting the expression level of the prognostic target gene is fluorescent quantitative PCR, and the reaction system of the fluorescent quantitative PCR is: 。 8. The detection kit for gastric cancer prognosis according to claim 6, characterized in that: The primer pair for detecting the expression level of the OR51G2 protein gene is: Upstream primer: 5′-TGGAGCACGAGAAATTAGCCA-3′; Downstream primer: 5′-GCCACAATAGGGGAATCTTTTGA-3′; The primer pair for detecting the expression level of the CGβ8 protein gene is: Upstream primer: 5′-CTGGCTGTGGAGAAGGAG-3′; Downstream primer: 5′-ACATCGCGGTAGTTGCAC-3′; The primer pair for detecting the expression level of the UPK1B protein gene is: Upstream primer: 5′-CCAAAGACAACTCAACTGTTCGT-3′; Downstream primer: 5′-AATGCCGCAACAACCAATAATC-3′.

9. The detection kit for gastric cancer prognosis according to claim 6, characterized in that: It also includes a primer pair for detecting GAPDH expression level, the sequence of which is: Upstream primer: 5′-ACAACTTTGGTATCGTGGAAGG-3′; Downstream primer: 5′-GCCATCACGCCACAGTTTC-3′; Also included are reverse transcription PCR reagents and RNA extraction reagents.

10. The detection kit for gastric cancer prognosis according to claim 9, characterized in that: The reaction system of the reverse transcription PCR is:

Images