MicroRNA biomarkers for gastric cancer diagnosis

Abstract

Methods for detecting or diagnosing gastric cancer in a subject or determining the likelihood that a subject will develop gastric cancer are provided. [Solution] The method includes a step of determining the expression level of at least one microRNA (miRNA) in a non-cellular biological fluid sample obtained from a subject, wherein the miRNA is either an "upregulated" miRNA selected from a specific group of miRNAs; or a "downregulated" miRNA selected from a specific group of miRNAs, and downregulation of the "downregulated" miRNA or upregulation of the "upregulated" miRNA compared to a control indicates the subject has cancer, and the control is a subject without gastric cancer.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of priority to Singapore Patent Application No. 10201404742W filed on August 7, 2014, the contents of which are incorporated herein by reference in their entirety for all purposes.

[0002] The present invention relates to biochemistry, and in particular to biomarkers. In particular, the present invention relates to cancer-associated biomarkers and methods of using the biomarkers to determine the likelihood that a patient will suffer from cancer, e.g., gastric cancer. [Background technology]

[0003] Gastric cancer is the fourth most common cancer (>1 million cases per year) and the second leading cause of cancer deaths worldwide. The prognosis of gastric cancer is poor, with a 5-year survival rate of less than 10%. This is mainly because symptoms usually present at an advanced stage of the disease. Therefore, early detection and accurate monitoring of the disease are necessary to improve clinical outcomes. Currently, endoscopy is the only reliable method for early diagnosis, but its high cost and risks limit its use as a screening test. A less invasive gastric cancer screening test is highly desirable and is expected to eliminate unnecessary endoscopy. Summary of the Invention [Problem to be solved by the invention]

[0004] Thus, the present invention provides an alternative method for detecting gastric cancer. [Means for solving the problem]

[0005] In one aspect, provided is a method of detecting or diagnosing gastric cancer in a subject, or determining the likelihood that a subject will develop gastric cancer, comprising measuring the expression level of at least one miRNA having at least 90% sequence identity to a miRNA in a non-cellular biological fluid sample obtained from the subject, wherein differential expression of the miRNA as compared to a control is indicative of the subject having gastric cancer, and wherein the miRNA is selected from the group consisting of hsa-miR-142-5p, hsa-miR-29c-3p, hsa-miR-93-5p, hsa-miR-142-6p, hsa-miR-142-7p, hsa-miR-29c-3p, hsa-miR-93-7p, hsa-miR-142-8p, hsa-miR-142-9p, hsa-miR-142-10p, hsa-miR-142-12p, hsa-miR-142-26p, hsa-miR-142-36p, hsa-miR-142-46p, hsa-miR-142-5p, hsa-miR-142-6p, hsa-miR-142-16p, hsa-miR-142-18p, hsa-miR-142-28p, hsa-miR-142-36p, hsa-miR-142-16p, hsa-miR-142-18p, hsa-miR-142-16 ... miR-140-5p, hsa-miR-148a-3p, hsa-miR-183-5p, hsa-miR-29b-3p, hsa-miR-424-5p, hsa-miR-101-3p, hsa-miR-106b-3p, hsa-miR-128, hsa-miR-12 80, hsa-miR-140-3p, hsa-miR-15b-3p, hsa-miR-186-5p, hsa-miR-18b-5p, hsa-miR-197-3p, hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-20b-5p, hs a-miR-21-3p, hsa-miR-23a-5p, hsa-miR-25-3p, hsa-miR-27a-5p, hsa-miR-29a-3p, hsa-miR-29b-2-5p, hsa-miR-29c-5p, hsa-miR-338-5p, hsa-miR -425-3p, hsa-miR-4306, hsa-miR-450a-5p, hsa-miR-486-5p, hsa-miR-500a-3p, hsa-miR-501-5p, hsa-miR-532-3p, hsa-miR-550a-5p, hsa-miR-579 , hsa-miR-589-5p, hsa-miR-590-5p, hsa-miR-598, hsa-miR-616-5p, hsa-miR-627, hsa-miR-629-3p, hsa-miR-629-5p, hsa-miR-93-3p, hsa-miR-195 -5p, hsa-miR-18a-3p, hsa-miR-363-3p, hsa-miR-181a-2-3p, hsa-miR-16-5p, hsa-miR-501-3p, hsa-miR-23a-3p, hsa-miR-339-3p, hsa-miR-15a-5p,hsa-miR-320b, hsa-miR-374b-5p, hsa-miR-650, hsa-miR-1290, hsa-miR-22-3p, hsa-miR-320c, hsa-miR-130a-3p, hsa-miR-320 e, hsa-miR-378a-3p, hsa-miR-9-5p, hsa-miR-200b-3p, hsa-miR-141-3p, hsa-miR-191-5p, hsa-miR-628-5p, hsa-miR-484, hsa-m or an "upregulated" miRNA selected from the group consisting of hsa-miR-103a-3p, hsa-miR-30a-5p, hsa-miR-181a-5p, hsa-miR-107, hsa-miR-26a-5p, hsa-miR-126-3p, hsa-miR-99b-5p, hsa-miR-339-5p, hsa-miR-122-5p, hsa-miR-136-5p, hsa-miR-139-5p, hsa-miR-425-5p; miR-146a-5p, hsa-miR-154-5p, hsa-miR-193b-3p, hsa-miR-23c, hsa-miR-30b-5p, hsa-miR-337-5p, hsa-miR-382-5p, hsa-miR- 409-3p, hsa-miR-411-5p, hsa-miR-485-3p, hsa-miR-487b, hsa-miR-495, hsa-miR-885-5p, hsa-miR-99a-5p, hsa-miR-362-5p, hs a-miR-671-3p, hsa-miR-454-3p, hsa-miR-328, hsa-miR-320a, hsa-miR-126-5p, hsa-miR-27a-3p, hsa-miR-30d-5p, hsa-miR-10a-5p, hsa-miR-10b-5p, hsa-miR-497-5p, hsa-miR-134, and hsa-miR-150-5p.

[0006] In another aspect, provided is a method of detecting or diagnosing a stage of gastric cancer in a subject, or determining the likelihood that a subject has a stage of gastric cancer, comprising measuring in a non-cellular biological fluid sample obtained from the subject an expression level of at least one miRNA having at least 90% sequence identity to a microRNA (miRNA) listed in at least one table selected from the group consisting of Table 8, Table 15, Table 16, and Table 17, wherein differential expression of miRNA in the sample compared to a control diagnoses the subject as having one of stage 1, stage 2, stage 3, or stage 4 gastric cancer.

[0007] In yet another aspect, provided is a method for detecting or diagnosing gastric cancer in a subject, comprising: (a) measuring an expression level of at least one miRNA in a non-cellular biological fluid sample, the miRNA being selected from the group consisting of hsa-miR-103a-3p, hsa-miR-142-5p, hsa-miR-29c-3p, hsa-miR-30a-5p, hsa-miR-107, hsa-miR-26a-5p, hsa-miR-140-5p, hsa-miR-148a-3p, hsa-miR-29b-3p, hsa-miR-1 26-3p, hsa-miR-181a-5p, hsa-miR-27a-5p, hsa-miR-616-5p, hsa-miR-484, hsa-miR-4306, hsa-miR-590-5p, hsa-miR-362-5p, hsa-miR-106b -3p, hsa-miR-497-5p, hsa-miR-18b-5p, hsa-miR-122-5p, hsa-miR-200b-3p, hsa-miR-197-3p, hsa-miR-486-5p, hsa-miR-99a-5p, hsa-miR-88 (b) determining whether the miR-16-5p has at least 90% sequence identity with a miRNA selected from the group consisting of miR-130a-3p, miR-150-5p, miR-30d-5p, miR-10b-5p, miR-532-3p, miR-23a-5p, miR-21-3p, miR-136-5p, miR-1280, and miR-16-5p; and generating a score based on said expression level of RNA; and (c) predicting the likelihood that said subject has gastric cancer using said score, wherein said score is calculated by a classification algorithm that compares said expression level of said subject to that of a positive control (a sample from a subject with gastric cancer) or a negative control (a sample from a subject without gastric cancer), and wherein said score indicates whether said subject has: i. gastric cancer, where said score falls within the range of the score of the positive control; or ii.The method identifies the likelihood that the score falls within the range of the score of a negative control and that the subject does not have gastric cancer.

[0008] The invention will be better understood with reference to the detailed description and when considered in conjunction with the following non-limiting examples and the accompanying drawings. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 shows a schematic diagram categorizing miRNAs detected in patients with gastric cancer. [Diagram 2]Figure 2 shows an example of a workflow diagram for measuring miRNAs detected in clinical samples by high-throughput miRNA RT-qPCT. Figure 2 shows the steps that occur during the process, from isolation of total miRNA from clinical samples to final data processing and statistical analysis. As illustrated in Figure 2, clinical samples are first subjected to an "isolation" step, which refers to isolating and purifying miRNAs from a sample (e.g., patient serum). Next, spike-in miRNAs (artificial synthetic miRNA mimics (i.e., small single-stranded RNAs with lengths ranging from 22 to 24 bases)) are added to the sample. These spike-in miRNAs are added to the sample to monitor the efficiency of various steps (including isolation, reverse transcription, amplification, and qPCR). Next, the miRNAs are divided into a number of groups. This step is called "multiplex design" (design to engineer miRNA assays to be divided in silico into multiple multiplex groups (45-65 miRNAs per group) to minimize non-specific amplification and primer-primer interactions during the RT and amplification processes). The samples then undergo a "multiplex reverse transcription" process (a process of combining various reverse transcription primer pools and adding them to different multiplex groups to synthesize cDNA). During cDNA synthesis, a PCR primer pool is combined with each synthesized cDNA pool (from a certain multiplex group) and an optimized touchdown PCR is performed to simultaneously increase the total cDNA amount of the group. The step of increasing the total cDNA amount is called the "amplification" step. After amplification and before data processing and statistical analysis, the amplified cDNA undergoes singleplex qPCR (referring to the process of distributing the amplified cDNA pools to different wells in a 384-well plate and then performing a singleplex qPCR reaction). Simultaneously with the measurement of miRNAs in clinical samples, a standard curve of synthetic miRNAs was measured to obtain absolute copy numbers for all measurements by interpolation. Thus, Figure 2 provides a detailed step-by-step workflow of the miRNA sample processing that occurs during the experimental section below. [Diagram 3]Figure 3 shows a heatmap of miRNAs consistently detected in samples from subjects. Figure 3 shows a heatmap representation of all robustly detected miRNAs (listed in Table 4). miRNA expression levels (copies / ml) are presented on a log2 scale and normalized to a mean of zero. The color of each point corresponds to its concentration. Hierarchical clustering is performed on both dimensions (miRNA and sample) based on Euclidean distance. For the horizontal lines: black represents gastric cancer subjects and white represents normal / control subjects. [Figure 4] Figure 4 shows a heat map of regulated miRNAs. In particular, the heat map shown in Figure 4 shows all the regulated miRNAs in gastric cancer, which are listed in Table 6. The expression levels of miRNAs (copies / ml) are presented in log2 scale and normalized to have a mean of zero. The grey scale corresponds to the concentration of miRNAs. Hierarchical clustering is performed on both dimensions (miRNA and sample) based on Euclidean distance. For the horizontal lines: black represents gastric cancer subjects and white represents normal / control subjects. [Diagram 5] Figure 5 shows box plots depicting the expression levels of the top-ranked upregulated miRNA (hsa-miR-142-5p) and downregulated miRNA (hsa-miR-99b-5p) in all gastric cancer subjects (regardless of subtype or stage) compared to normal / control subjects. Expression levels (copies / ml) are shown on a log2 scale. Box plots show the miRNA expression levels of subjects falling in the 25th, 50th, and 75th percentiles of the expression level distribution. AUC refers to the area under the receiver operating characteristic curve. [Figure 6]Figure 6 shows dot plots showing the results of correlation analysis between all reliably detected miRNAs. Based on the log2 scale expression levels (copies / mL), Pearson's linear correlation coefficients were calculated between all 191 reliably detected miRNA targets (listed in Table 4). Each dot represents a miRNA pair with a p-value (relative to the correlation difference from zero) less than 0.01 and a correlation coefficient higher than 0.5 (left panel, positively correlated) or lower than -0.5 (right panel, negatively correlated). [Figure 7] Figure 7 shows dot plots showing the results of correlation analysis of gastric cancer biomarkers. Based on the log2 scale expression levels (copies / mL), Pearson's linear correlation coefficients were calculated between all 75 miRNA targets that were altered in gastric cancer subjects (listed in Table 6). Each dot represents a miRNA pair with a p-value (correlation difference from zero) less than 0.01 and a correlation coefficient higher than 0.5 (left panel, positively correlated) or lower than -0.5 (right panel, negatively correlated). Thus, in summary, Figures 6 and 7 show that many cancer-related and non-related miRNAs are positively correlated, which makes it difficult to select the optimal miRNA combination for gastric cancer diagnosis. [Figure 8] Figure 8 shows a graph plotting the sensitivity and specificity of various top-ranked miRNAs in various stages of gastric cancer compared to normal / control. More specifically, Figure 8 shows the receiver operating characteristic curves of the top upregulated (based on AUC (area under the receiver operating characteristic curve)) miRNAs (row 1) and downregulated (row 2) in various stages of gastric cancer subjects compared to miRNAs in normal / control subjects. [Figure 9]Figure 9 shows Venn diagrams illustrating the overlap between biomarkers for various stages of gastric cancer. These Venn diagrams are generated based on Tables 8, 9, 10, and 11. Thus, Figure 9 shows that the miRNAs used to determine various stages of gastric cancer appear to be specific to the stage they correlate with (except for the large overlap observed in both stage 3 and stage 4, two stages considered to be clinically closely related). [Figure 10] Figure 10 shows box plots of the diagnostic power (AUC) of multivariate biomarker panels (number of miRNAs ranged from 3 to 10) in the discovery and validation phases during in silico four-fold cross validation. Biomarker panels with 3 to 10 miRNAs were identified using sequence forward floating search (SFFS), which uses a linear support vector machine as the model based on a discovery set of samples and validated on a separate independent sample set. Four-fold cross validation was performed multiple times. Box plots show the 25th, 50th, and 75th percentiles of the AUC of the miRNA sets for classification of normal / control and gastric cancer subjects. Figure 10 shows the robustness of the data obtained from the various multivariate biomarker panels. [Figure 11] Figure 11 shows a bar graph depicting the average AUC of different multivariate biomarker panels in the discovery set (black bars) and validation set (grey bars) during the cross-validation process. Error bars indicate the standard deviation of the AUC. To test the significance of the improvement in AUC in the validation set when more miRNAs are included in the panel, a right-tailed t-test was performed to compare all adjacent grey bars. *: p-value < 0.05; **: p-value < 0.01; ***: p-value < 0.001. Thus, Figure 11 shows the difference in AUC between different numbers of miRNAs in the multivariate biomarker panel set. [Figure 12] FIG. 12 shows a schematic diagram summarizing the steps of the methods described herein that are used to determine the likelihood that a subject has gastric cancer. [Figure 13]FIG. 13 shows a schematic diagram summarizing the steps of the methods described herein that are used to determine the likelihood that a subject has one of the stages of gastric cancer. [Figure 14] FIG. 14 shows a schematic summarizing the steps of the multivariate analysis methods described herein in determining the likelihood that a subject has gastric cancer. [Figure 15] Figure 15 shows (A) box plot representation of cancer risk scores for control and gastric cancer subjects in this study; (B) probability distribution of the cancer risk scores fitted to a logistic distribution for the control and gastric cancer subjects in this study; (C) the probability of an unknown subject (belonging to the high-risk population) having gastric cancer depends on the value of the cancer risk score (the prevalence of gastric cancer in the high-risk population was 0.0067 based on data from the National University Hospital); (D) the fold increase in the probability (risk) of an unknown subject having gastric cancer at various cancer risk score levels compared to the prevalence of gastric cancer in the high-risk population. The dotted lines demarcate the threshold score boundaries that divide subjects into high (H), medium (M), or low (L) cancer risk groups. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Brief explanation of the table The invention will be better understood by reference to the detailed description and when considered in conjunction with the following non-limiting examples and the accompanying tables.

[0011] Table 1 shows a summary of serum / plasma microRNA biomarker studies for gastric cancer. Table 1 summarizes the different studies that measure miRNA levels in cell-free serum / plasma samples. Only results validated by RT-qPCR are included in this table. "GC" refers to gastric cancer subjects and "C" refers to control subjects.

[0012] Table 2 shows the clinical information of the gastric cancer subjects analyzed in the experimental section of this disclosure. All sera were collected from 236 gastric cancer subjects before any treatment and stored at -80°C until use. H. Pylori infection status is shown in the column labeled "H Pylori" (for Helicobacter pylori).

[0013] Table 3 shows the clinical information of the normal subjects (controls). All normal / control subjects (237 subjects) were confirmed to be free of gastric cancer by endoscopy at the time the samples were collected. Subjects were followed for 3-5 years with endoscopic screening every 2 years to determine if they had developed gastric cancer within this time period. All sera were stored at -80°C until use. H. Pylori status is shown in the column labeled "H Pylori" (for Helicobacter pylori).

[0014] Table 4 shows the sequences of 191 miRNAs that were reliably detected in serum samples from both gastric cancer and normal subjects (controls). A miRNA was considered "reliably detected" if at least 90% of the serum samples had a concentration higher than 500 copies / ml. The miRNAs were named according to the miRBase V18 release.

[0015] Table 5 shows a summary of the characteristics of both gastric cancer subjects and healthy controls (normal subjects / controls).

[0016] Table 6 lists the miRNAs that are differentially expressed between normal subjects (control) and gastric cancer subjects. When comparing between normal / control subjects and all gastric cancer subjects (regardless of subtype and stage), 75 miRNAs had p-values ​​less than 0.01 after false discovery rate correction (Bonferroni method). AUC-area under the receiver operating characteristic curve. Fold change-mean expression level (copies / ml) of miRNA in cancer population divided by that of normal / control population.

[0017] Table 7 shows a comparison between this study and other literature reports. The miRNAs not listed in Table 4 (i.e., miRNAs with expression levels of ≧500 copies / ml) are considered below the detection limit (ND) of this study. Therefore, Table 7 shows that the majority of differentially regulated miRNAs described in the art are different from the exemplary miRNAs of this disclosure.

[0018] Table 8 lists the miRNAs that are differentially expressed between normal subjects (control) and stage 1 gastric cancer subjects. When comparing between normal / control and stage 1 gastric cancer subjects, 20 miRNAs had p-values ​​less than 0.01 after false discovery rate correction (Bonferroni method). AUC - area under the receiver operating characteristic curve. Fold change - the mean expression level of the miRNA in the cancer population divided by that of the normal / control population.

[0019] Table 9 lists the miRNAs that are differentially expressed between normal subjects (control) and stage 2 gastric cancer subjects. When comparing between normal / control and stage 2 gastric cancer subjects, 62 miRNAs had p-values ​​less than 0.01 after false discovery rate correction (Bonferroni method). AUC - area under the receiver operating characteristic curve. Fold change - the mean expression level (copies / ml) of the miRNA in the cancer population divided by that of the normal / control population.

[0020] Table 10 lists the miRNAs that are differentially expressed between normal subjects (control) and stage 3 gastric cancer subjects. When comparing between normal / control subjects and stage 3 gastric cancer subjects, 43 miRNAs had p-values ​​less than 0.01 after false discovery rate correction (Bonferroni method). AUC - area under the receiver operating characteristic curve. Fold change - the mean expression level of the miRNA in the cancer population divided by that of the normal / control population.

[0021] Table 11 lists the miRNAs that are differentially expressed between normal subjects (control) and stage 4 gastric cancer subjects. When comparing between normal / control and stage 4 gastric cancer subjects, 74 miRNAs had p-values ​​less than 0.01 after false discovery rate correction (Bonferroni method). AUC - area under the receiver operating characteristic curve. Fold change - the mean expression level of the miRNA in the cancer population divided by that of the normal / control population.

[0022] Table 12 lists the miRNAs identified in the multivariate biomarker panel identification process. The features of the miRNAs selected for biomarker panels assembled with 6, 7, 8, 9, and 10 miRNAs are summarized. The occurrence rate is defined as the number of occurrences of a miRNA in all panels divided by the total number of panels. Panels with the top and bottom 10% AUC were excluded to avoid counting false discovery biomarkers resulting from applying inaccurate data from subpopulations generated by the randomization process of the cross-validation analysis. Among them, only miRNAs with an occurrence rate of 4 or more are listed. The miRNA changes in various stages of gastric cancer were defined based on Tables 6 and 8-11.

[0023] Table 13 shows the statistics of the prevalence of miRNAs frequently selected in the multivariate biomarker panel selection process. The percentage of miRNA panels using various numbers of miRNAs from two miRNA groups (hsa-miR-134, hsa-miR-1280, hsa-miR-27a-5p group and hsa-miR-425-5p, hsa-660-5p group). The top and bottom 10% of biomarker panels of 6, 7, 8, 9, and 10 miRNAs, as defined by the AUC of the validation set (Figure 10), were excluded from the count.

[0024] Table 14 lists the miRNAs that are differentially expressed between normal / control and gastric cancer (regardless of stage) that have not been previously reported in the art (listed in Table 1).

[0025] Table 15 lists the miRNAs that are differentially expressed between normal / control and stage 2 gastric cancer, which have not been previously reported in the art.

[0026] Table 16 lists the miRNAs that are differentially expressed between normal / control and stage 3 gastric cancer, which have not been previously reported in the art.

[0027] Table 17 lists the miRNAs that are differentially expressed between normal / control and stage 4 gastric cancer, which have not been previously reported in the art.

[0028] Table 18 lists miRNAs (not previously reported in the art) that are differentially expressed between normal / control and stage 1, stage 2, stage 3, stage 4, or all stages of gastric cancer. Table 18 is a combination of Tables 14, 8, 15, 16, and 17.

[0029] Table 19 lists miRNAs whose expression levels were altered in gastric cancer subjects and that are frequently selected for use in multivariate biomarker panels (i.e., the significant groups in Table 12) (not previously reported in the art).

[0030] Table 20 lists miRNAs whose expression levels were not altered in gastric cancer subjects (i.e., the non-significant group in Table 12) that are frequently selected for use in multivariate biomarker panels (not previously reported in the art).

[0031] Table 21 lists 12 miRNAs (along with their Ki coefficient values) that are frequently selected in the multivariate biomarker panel identification process with an occurrence rate of >20%, and are combined to calculate the cancer risk score (Equation 1) in Figure 15.

[0032] Detailed Description of the Invention MicroRNAs (miRNAs) are small non-coding RNAs that regulate gene expression, and their aberrant expression is involved in the pathogenesis of various cancers (e.g., gastric cancer). It is known that microRNAs can be used to determine the likelihood that a subject will develop cancer.

[0033] Thus, in one aspect, provided is a method for determining the likelihood that a subject will develop or have gastric cancer. In one example, FIG. 12 is a summary of steps of a method for determining the likelihood that a subject will develop or have gastric cancer. In one example, the method may also be for detecting or diagnosing gastric cancer in a subject. In one example, the method is an in vitro method. In one example, the method described herein may include measuring or determining the expression level of at least one miRNA having at least 90% sequence identity with a microRNA (miRNA) listed in Table 18 in a non-cellular biological fluid sample obtained from a subject, wherein differential expression of miRNA expression compared to a control is indicative of the subject having gastric cancer, and the miRNA is either a miRNA listed in Table 18 as "upregulated" or a miRNA listed in Table 18 as "downregulated".

[0034] Table 18 lists miRNAs that can be used to determine the likelihood that a patient will develop or have gastric cancer, which are:

[0035] [Table 1-1]

[0036] [Table 1-2]

[0037] As used herein, the term "miRNA", "microRNA", or "miR" refers to an RNA (RNA analog) that includes endogenous non-coding gene products, whose precursor RNA transcripts can form small stem-loops where the mature "miRNA" is cleaved by the endonuclease Dicer. miRNAs are encoded by genes, but unlike mRNAs, miRNAs control the expression of mRNAs. In one example, the term "miRNA" or "microRNA" refers to a short RNA molecule with at least 10 nucleotides and no more than 35 nucleotides covalently linked. In one example, the polynucleotide is a molecule of 10-33 nucleotides, 15-30 nucleotides, 17-27 nucleotides, 18-26 nucleotides in length, i.e., 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, or 35 nucleotides in length, optionally excluding labels or sequence extensions (e.g., biotin stretches). Sequences of the miRNAs described herein are provided in Table 4 provided herein, which includes the miRNAs of SEQ ID NO:1 through SEQ ID NO:191. It will be understood by one of skill in the art that miRNAs are a type of polynucleotide having a sequence that includes letters such as "ATGC." It will be understood that, unless otherwise noted, nucleotides are in 5'->3' order, from left to right, with "A" representing deoxyadenoside, "C" representing deoxycytidine, "G" representing deoxyguanoside, and "T" representing deoxythymidine. The letters A, C, G, and T are used to refer to the base itself, the nucleoside, or the nucleotide containing the base, as is standard in the art. The term includes oligonucleotides composed of naturally occurring nucleobases, sugars, and internucleoside (backbone) linkages, as well as oligonucleotides having non-natural portions that function similarly or with specific improved function. In naturally occurring polynucleotides, the internucleoside linkages are typically phosphodiester linkages, and the components are referred to as "nucleotides."The term "oligonucleotide" may also include oligonucleotides that are fully or partially modified or substituted (eg, of the bases or sugars).

[0038] It will be understood in the art that in any of the methods described herein, the miRNA may not have 100% sequence identity with the sequence of the miRNA listed in Table 4. Thus, in one example, the miRNA evaluated may have at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 97.5%, at least 98%, at least 99%, or at least 99.9% sequence identity with the miRNA listed in any one of Tables 8, 15, 16, 17, 18, 19, or 20 (see Table 4). In one example, the miRNA evaluated may have one, two, three, or four nucleotide substitutions. Thus, the miRNA used in the methods described herein is detectable using a reagent that may be capable of specifically hybridizing or binding to a sequence listed in Table 8, 15, 16, 17, 18, 19, or 20 (see Table 4). As used herein, the terms "hybridize to" and "hybridization" are used interchangeably with the terms "specific to" and "specifically bind to" and refer to a sequence-specific non-covalent interaction with a complementary nucleic acid, for example, the interaction between a target nucleic acid sequence and a target-specific nucleic acid sequence primer or probe. In one example, a hybridizing nucleic acid probe is one that hybridizes with a selectivity of greater than 70%, 80%, 90%, or 100% (i.e., cross-hybridization with one of the miRNAs described herein may occur at less than 30%, less than 20%, or less than 10%). As will be understood by those skilled in the art, a nucleic acid probe that "hybridizes" to a miRNA described herein can be determined in consideration of its length and composition. In one example, a nucleic acid probe that hybridizes to a miRNA described herein may have one, two, or three base pairs that do not match. In one example, the term "miRNA" as used herein includes miRNAs in which the 3' and / or 5' ends of the miRNA may be modified or digested during their production.In one example, the assays of the invention pick up miRNAs whose 3' and / or 5' ends have been modified or digested during their production. In one example, the miRNAs disclosed herein may include miRNAs that may differ from the sequences listed in Table 4 by 3, 4, or 5 nucleotides.

[0039] The term "cancer" as used herein refers to the presence of cells that have typical properties of cells from which cancer arises (e.g., uncontrolled proliferation, immortalization, metastatic potential, rapid growth and proliferation rate, and certain characteristic morphological features known in the art). In one example, "cancer" may be gastric cancer or stomach cancer. In one example, "cancer" may include pre-cancer and malignant cancer. Thus, the term "gastric cancer" covers all stages of gastric cancer as described by the National Cancer Institute of the National Health.

[0040] In one example, it will be understood by those skilled in the art that the methods described herein do not include steps performed by a doctor / physician. Thus, the results obtained from the methods described herein may need to be combined with clinical data and other clinical symptoms, after which a final diagnosis made by a physician can be presented to the subject. The final diagnosis of whether a subject has gastric cancer is within the domain of a physician and is not considered part of this disclosure. Thus, the terms "determine", "detect", and "diagnose" as used herein refer to the identification of the probability or likelihood that a subject has a disease (e.g., gastric cancer) at any stage of its development, or the determination of the predisposition of a subject to develop the disease. In one example, "diagnose", "determine", and "detect" occur prior to the presentation of symptoms. In one example, "diagnose", "determine", and "detect" (in conjunction with other clinical symptoms) allow a clinician / physician to confirm gastric cancer in a subject suspected of having gastric cancer.

[0041] As used herein, the term "differential expression" refers to the difference in expression level of one or more miRNAs, measured as miRNA abundance, in one sample compared to the expression level of the same one or more miRNAs in a second sample, or compared to a predetermined standard or control. Differential expression can be measured by methods known in the art, such as array hybridization, next generation sequencing, RT-PCR, and other methods known to those of skill in the art.

[0042] The control group used to determine whether differential expression is observed may vary depending on the purpose of the method described herein. However, as a general rule, for any given method described herein, the control group can be easily determined by one of skill in the art. Unless otherwise specifically stated, in one example, the control group may be subjects without any gastric cancer-related disease. In one example, the control group may be healthy individuals with no pre-existing conditions. In one example, the control group may be subjects known to have a non-gastric cancer-related disease (e.g., but not limited to, gastritis, intestinal metaplasia, intestinal atrophy, etc.). In one example, the controls may vary in terms of sample characteristics and other factors (e.g., the sex, age, and race of the subject based on the subjects from which the training data set was obtained). In one example, the control subjects may be Asian or of Asian descent (e.g., Chinese, Chinese descent).

[0043] In one example, differential expression is observed when a miRNA listed as "upregulated" in Table 18 is found to be upregulated in a sample obtained from a subject suspected of having gastric cancer, as compared to a control. Thus, in one example, upregulation of a miRNA listed as "upregulated" in Table 18, as compared to a control, may be indicative of, or may diagnose, that the subject has gastric cancer.

[0044] In any of the examples or methods described herein, the term "upregulated" as used herein refers to an increase in miRNA expression levels or detection of more copies of miRNA compared to a control. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.01 or greater. In one example, the p-value of a statistical test for the change may be less than 0.05. In one example, a p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.02 to about 2.5. In one example, a miRNA can be considered "upregulated" when the fold change in expression level relative to a control is at least about 1.05 to about 1.53, as illustrated in Table 6. In one example, a miRNA can be considered "upregulated" when the fold change in expression level relative to a control is at least about 1.10-fold, about 1.15-fold, about 1.20-fold, about 1.25-fold, about 1.30-fold, about 1.35-fold, about 1.40-fold, about 1.45-fold, about 1.50-fold. A control as used herein may be a subject without gastric cancer.

[0045] In one example, differential expression is observed when a miRNA listed as "downregulated" in Table 18 is found to be downregulated in a sample obtained from a subject suspected of having gastric cancer compared to a control. Thus, in one example, downregulation of a miRNA listed as "downregulated" in Table 18 compared to a control may be indicative of a subject having gastric cancer, or may thereby diagnose a subject as having gastric cancer.

[0046] In any of the examples or methods described herein, the term "downregulated" as used herein refers to a decrease in miRNA expression levels or detection of fewer copies of miRNA compared to a control. In one example, a miRNA can be considered "downregulated" if it shows a fold change in expression level relative to a control of at least about 0.99 or less. In one example, the p-value of a statistical test for the change can be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "downregulated" if the fold change in expression level relative to a control is at least about 0.98 to about 0.3. In one example, a miRNA can be considered "downregulated" when it exhibits a fold change in expression level relative to a control of at least about 0.5 to about 0.92, as illustrated in Table 6. In one example, a miRNA expression can be considered "downregulated" when it exhibits a fold change in expression level relative to a control of at least about 0.95, about 0.90, about 0.85, about 0.80, about 0.75, about 0.70, about 0.65, about 0.60, about 0.55. The control used herein may be a subject without gastric cancer.

[0047] The term "about" is used herein to indicate that a value includes the inherent variation of error known in the art that may result from the use of the equipment or method employed to determine the value, or the variation that exists between test subjects. Thus, in one example, the term "about" in the context of fold changes in expression as used throughout this disclosure herein means ±5%, ±4%, ±3%, ±2%, ±1%, or ±0.5% of the referenced value.

[0048] In one example, the methods described herein can measure the differential expression of one or more miRNAs listed in Table 18. Thus, in one example, the methods can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 85, at least 90, at least 95, at least 100, at least 105, at least 107, or all of the miRNAs listed in Table 18. In one example, the method can measure 1-108 miRNAs listed in Table 18, or 2-10, 11-30, 31-50, 50-70, 71-90, or 91-108 miRNAs listed in Table 18. In one example, the method can measure differential expression of all miRNAs listed in Table 18.

[0049] In one example, the methods described herein are capable of measuring differential expression of at least one miRNA listed in Table 18 as "upregulated" and at least one miRNA listed in Table 18 as "downregulated." In one example, the methods are capable of measuring differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60 miRNAs listed in Table 18 as "upregulated." In one embodiment, the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or 38 of the miRNAs listed as "downregulated" in Table 18 can be measured.

[0050] In one example, the methods described herein may include measuring the expression levels of microRNAs (miRNAs) listed in Table 14 in a non-cellular biological fluid sample obtained from a subject.

[0051] Table 14 lists miRNAs that can be used to determine the likelihood that a patient will develop or have gastric cancer, which are:

[0052] [Table 2]

[0053] As known in the art, when a proliferative disease progresses (with or without treatment), the clinical symptoms of the disease vary between stages of progression.Therefore, to help those skilled in the art determine how gastric cancer / stomach cancer has progressed, the National Cancer Institute of the National Institutes of Health has published the general clinical symptoms of stages observed in gastric cancer / stomach cancer.The stages of gastric cancer include, but are not limited to, the following stages as described below:

[0054] Stage I: In stage I, cancer has already formed in the lining of the mucosa (innermost layer) of the stomach wall. Stage I is divided into stage IA and stage IB, depending on where the cancer has spread.

[0055] Stage IA: Cancer may have already spread into the submucosa (layer of tissue next to the mucosa) of the stomach wall.

[0056] Stage IB: Cancer may have already spread into the submucosa (the layer of tissue next to the mucosa) of the stomach wall and is found in one or two lymph nodes near the tumor; or cancer has already spread into the muscle layer of the stomach wall.

[0057] Stage II: Stage II stomach cancer is divided into stage IIA and stage IIB, depending on where the cancer has spread.

[0058] Stage IIA: Cancer has spread to the subserosa (layer of tissue next to the serosa) of the stomach wall; has spread to the muscle layer of the stomach wall and is found in 1 or 2 lymph nodes near the tumor; or has spread to the submucosa (layer of tissue next to the mucosa) of the stomach wall and is found in 3 to 6 lymph nodes near the tumor.

[0059] Stage IIB: Cancer has spread to the serosa (outermost layer) of the stomach wall; has spread to the subserosa (layer of tissue next to the serosa) of the stomach wall and is found in 1 or 2 lymph nodes near the tumor; has spread to the muscle layer of the stomach wall and is found in 3 to 6 lymph nodes near the tumor; or has spread to the submucosa (layer of tissue next to the mucosa) of the stomach wall and is found in 7 or more lymph nodes near the tumor.

[0060] Stage III: Stage III gastric cancer is divided into stage IIIA, stage IIIB, and stage IIIC, depending on where the cancer has spread.

[0061] Stage IIIA: Cancer has spread to the serosal (outermost) layer of the stomach wall and is found in 1 or 2 lymph nodes near the tumor; has spread to the subserous layer (the tissue layer next to the serosa) of the stomach wall and is found in 3 to 6 lymph nodes near the tumor; or has spread to the muscle layer of the stomach wall and is found in 7 or more lymph nodes near the tumor.

[0062] Stage IIIB: Cancer has spread to nearby organs (e.g., spleen, transverse colon, liver, diaphragm, pancreas, kidneys, adrenal glands, or small intestine) and is found in 1 or 2 lymph nodes near the tumor; it has spread to the serosa (outermost layer) of the stomach wall and is found in 3 to 6 lymph nodes near the tumor; or it has spread to the subserosa (layer of tissue next to the serosa) of the stomach wall and is found in 7 or more lymph nodes near the tumor.

[0063] Stage IIIC: Cancer has already spread to nearby organs (e.g., the spleen, transverse colon, liver, diaphragm, pancreas, kidneys, adrenal glands, or small intestine) and may be found in 3 or more lymph nodes near the tumor; or it has already spread to the serosa (outermost layer) of the stomach wall and is found in 7 or more lymph nodes near the tumor.

[0064] Stage IV: In stage IV, the cancer has already spread to distant parts of the body.

[0065] As will be appreciated by those skilled in the art, determining whether a subject is at a particular stage may be useful to provide clinicians with clear guidelines on how to best approach any treatment or palliative care. Thus, in one aspect, the present disclosure provides a method for detecting or diagnosing a stage of gastric cancer in a subject. In one example, the method determines the likelihood that the subject is at a stage of gastric cancer. In this aspect, the method includes measuring an expression level of a microRNA (miRNA) having at least 90% sequence identity to a miRNA listed in at least one table selected from the group consisting of Table 8, Table 9, Table 15, Table 10, Table 16, Table 11, or Table 17 in a non-cellular biological fluid sample obtained from the subject, and a differential expression of miRNA expression in the sample obtained from the subject compared to a control may be indicative of the likelihood that the subject is at any one of stage 1, stage 2, stage 3, or stage 4 of gastric cancer, or the differential expression may diagnose the subject as being at any one of stage 1, stage 2, stage 3, or stage 4 of gastric cancer. As used herein, a control may be a subject without gastric cancer. In one example, Figure 13 is a summary of the steps of a method for determining the likelihood that a subject will develop or have one of the stages of gastric cancer.

[0066] In one example, differential expression is observed when a miRNA listed as "upregulated" in Table 8, when compared to a control, may be indicative of or may diagnose a subject as having stage 1 gastric cancer, and the miRNA may comprise at least one of the miRNAs listed as "upregulated" in Table 8. Table 8 is described in more detail in the Experimental Section. In one example, a miRNA may be considered "upregulated" when the fold change in expression level relative to the control is at least about 1.01 to about 2.5 or more. In one example, the statistical test for the change may have a p-value of less than 0.05. In one example, the p-value may be calculated using a statistical test known in the art, such as a t-test (p-value<0.01). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.10-fold, about 1.15-fold, about 1.20-fold, about 1.25-fold, about 1.30-fold, about 1.35-fold, about 1.40-fold, about 1.45-fold, about 1.50-fold. In one example, as illustrated in Table 8, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.11 to about 1.57.

[0067] In one example, the methods described herein can measure the differential expression of one or more miRNAs listed as "upregulated" in Table 8. Thus, in one example, the methods can measure the differential expression of at least two, at least three, at least four, at least five, at least six, at least seven, or eight miRNAs listed as "upregulated" in Table 8.

[0068] In one example, differential expression is observed when a miRNA listed as "downregulated" in Table 8, when compared to a control, may be indicative of or may diagnose a subject as having stage 1 gastric cancer. In one example, the miRNA may include at least one of the miRNAs listed as "downregulated" in Table 8. In one example, a miRNA may be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.9 or less. In one example, the p-value of the statistical test for the change may be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach was illustratively used in the following examples. In one example, miRNA expression can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.95, about 0.90, about 0.85, about 0.80, about 0.75, about 0.70, about 0.65, about 0.60, about 0.55. In one example, as exemplified in Table 8, a miRNA can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.68 to about 0.84.

[0069] In one example, the methods described herein can measure the differential expression of one or more miRNAs listed as "downregulated" in Table 8. Thus, in one example, the methods can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or 12 miRNAs listed as "downregulated" in Table 8.

[0070] In one example, the methods described herein are capable of measuring the differential expression of at least one miRNA listed in Table 8 as "upregulated" and at least one miRNA listed in Table 8 as "downregulated." Thus, the methods can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, or 8 miRNAs listed in Table 8 as "upregulated," and can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or 12 miRNAs listed in Table 8 as "downregulated."

[0071] In one example, differential expression is observed when a miRNA listed in Table 9 or Table 15 as "upregulated" when compared to a control may be indicative of or may be diagnostic of a subject having stage 2 gastric cancer. Table 9 is described in further detail in the Experimental Section. In one example, the miRNA can include at least one of the miRNAs listed in Table 15 as "upregulated."

[0072] Table 15 lists the miRNAs used to determine the likelihood that a patient will develop or have stage 2 gastric cancer, which are:

[0073] [Table 3-1]

[0074] [Table 3-2]

[0075] In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.01 to about 2.5 or more. In one example, the p-value of a statistical test for the change can be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art, such as a t-test (p-value<0.01). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.10-fold, about 1.15-fold, about 1.20-fold, about 1.25-fold, about 1.30-fold, about 1.35-fold, about 1.40-fold, about 1.45-fold, about 1.50-fold. In one example, as exemplified in Table 9, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.13 to about 2.13.

[0076] In one example, the methods described herein can measure differential expression of one or more miRNAs listed as "upregulated" in Table 9 or Table 15. Thus, in one example, the methods can measure differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, or 54 miRNAs listed as "upregulated" in Table 9 or 15.

[0077] In one example, differential expression is observed when a miRNA listed as "downregulated" in Table 9 or Table 15, when compared to a control, may be indicative of or may diagnose a subject as having stage 2 gastric cancer. In one example, the miRNA may include at least one of the miRNAs listed as "downregulated" in Table 15. In one example, a miRNA may be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.9 or less. In one example, the p-value of the statistical test for the change may be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach was illustratively used in the following examples. In one example, miRNA expression can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.95, about 0.90, about 0.85, about 0.80, about 0.75, about 0.70, about 0.65, about 0.60, about 0.55. In one example, as exemplified in Table 9, miRNA can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.56 to about 0.87.

[0078] In one example, the methods described herein can measure the differential expression of one or more miRNAs listed as "downregulated" in Table 9 or Table 15. Thus, in one example, the methods can measure the differential expression of at least two, at least three, at least four, at least five, at least six, at least seven, or eight miRNAs listed as "downregulated" in Table 9 or Table 15.

[0079] In one example, the methods described herein are capable of measuring the differential expression of at least one miRNA listed in Table 9 or Table 15 as "upregulated" and at least one miRNA listed in Table 9 or Table 15 as "downregulated." In one example, the method can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, or 54 miRNAs listed as "upregulated" in Table 9 or Table 15; and can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, or 8 miRNAs listed as "downregulated" in Table 9 or Table 15.

[0080] In one example, when a method described herein that can provide a diagnosis or indication as to whether a subject is likely to have stage 2 gastric cancer uses Table 15, the method may further include comparing the expression levels of any one or more miRNAs, including, but not limited to, miR-20a-5p, miR-223-3p, miR-17-5p, miR-106b-5p, miR-423-5p, and miR-21-5p.

[0081] In one example, differential expression is observed when a miRNA listed in Table 10 or Table 16 as "upregulated" when compared to a control may be indicative of or may be diagnostic of a subject having stage 3 gastric cancer. Table 10 is described in more detail in the Experimental Section. In one example, the miRNA can include at least one of the miRNAs listed in Table 16 as "upregulated."

[0082] Table 16 lists the miRNAs used to determine the likelihood that a patient will develop or have stage 3 gastric cancer, which are:

[0083] [Table 4]

[0084] In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.01 to about 2.5 or more. In one example, the p-value of a statistical test for the change can be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art, such as a t-test (p-value<0.01). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.10-fold, about 1.15-fold, about 1.20-fold, about 1.25-fold, about 1.30-fold, about 1.35-fold, about 1.40-fold, about 1.45-fold, about 1.50-fold. In one example, as exemplified in Table 10, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.20 to about 1.93.

[0085] In one example, the methods described herein can measure differential expression of one or more miRNAs listed as "upregulated" in Table 10 or Table 16. Thus, in one example, the methods can measure differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or 37 miRNAs listed as "upregulated" in Table 10 or Table 16.

[0086] In one example, when differential expression is observed, a miRNA listed as "downregulated" in Table 10 or Table 16 may be indicative of a subject having stage 3 gastric cancer when compared to a control, or the miRNA can diagnose the subject as having stage 3 gastric cancer. In one example, the miRNA may include at least one of the miRNAs listed as "downregulated" in Table 16. In one example, a miRNA may be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.9 or less. In one example, the p-value of the statistical test for the change may be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach was used illustratively in the following examples. In one example, miRNA expression can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.95, about 0.90, about 0.85, about 0.80, about 0.75, about 0.70, about 0.65, about 0.60, about 0.55. In one example, as exemplified in Table 10, miRNA can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.72 to about 0.88.

[0087] In one example, the methods described herein can measure the differential expression of one or more miRNAs listed as "downregulated" in Table 10 or Table 16. Thus, in one example, the methods can measure the differential expression of at least two, at least three, at least four, at least five, or six miRNAs listed as "downregulated" in Table 10 or Table 16.

[0088] In one example, the methods described herein can measure the differential expression of at least one miRNA listed in Table 10 or Table 16 as "upregulated" and at least one miRNA listed in Table 10 or Table 16 as "downregulated." In one example, the methods can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or 37 miRNAs listed in Table 10 or Table 16 as "upregulated," and can measure the differential expression of at least 2, at least 3, at least 4, at least 5, or 6 miRNAs listed in Table 10 or Table 16 as "downregulated."

[0089] In one example, a method described herein for providing a diagnosis or indication as to whether a subject is likely to have stage 3 gastric cancer uses Table 16. The method may further include comparing the expression levels of any one or more miRNAs, including but not limited to miR-21-5pm, miR-223-3p, and miR-423-5p.

[0090] In one example, differential expression is observed when a miRNA listed in Table 11 or Table 17 as "upregulated" when compared to a control may be indicative of or may be diagnostic of a subject having stage 4 gastric cancer. Table 10 is described in more detail in the Experimental Section. In one example, the miRNA can include at least one of the miRNAs listed in Table 17 as "upregulated."

[0091] Table 17 lists the miRNAs used to determine the likelihood that a patient will develop stage 4 gastric cancer, which are:

[0092] [Table 5]

[0093] In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.01 to about 2.5 or more. In one example, the p-value of a statistical test for the change can be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art, such as a t-test (p-value<0.01). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach is illustratively used in the following examples. In one example, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.10-fold, about 1.15-fold, about 1.20-fold, about 1.25-fold, about 1.30-fold, about 1.35-fold, about 1.40-fold, about 1.45-fold, about 1.50-fold. In one example, as exemplified in Table 11, a miRNA can be considered "upregulated" if the fold change in expression level relative to a control is at least about 1.12 to about 1.86.

[0094] In one example, the methods described herein can measure differential expression of one or more miRNAs listed as "upregulated" in Table 11 or Table 17. Thus, in one example, the methods can measure differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 17, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, or 46 miRNAs listed as "upregulated" in Table 11 or 15.

[0095] In one example, differential expression is observed when a miRNA listed as "downregulated" in Table 11 or Table 17, when compared to a control, may be indicative of or may diagnose a subject as having stage 4 gastric cancer. In one example, the miRNA may include at least one of the miRNAs listed as "downregulated" in Table 17. In one example, a miRNA may be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.9 or less. In one example, the p-value of the statistical test for the change may be less than 0.05. In one example, the p-value can be calculated using a statistical test known in the art (t-test (p-value<0.01)). In another example, the statistical test is a t-test (p-value<0.01) corrected for false discovery rate (FDR) using a Bonferroni-type multiple comparison method known in the art. This approach was illustratively used in the following examples. In one example, miRNA expression can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.95, about 0.90, about 0.85, about 0.80, about 0.75, about 0.70, about 0.65, about 0.60, about 0.55. In one example, as exemplified in Table 11, miRNA can be considered "downregulated" when the fold change in expression level relative to a control is at least about 0.48 to about 0.90.

[0096] In one example, the methods described herein can measure differential expression of one or more miRNAs listed as "downregulated" in Table 11 or Table 17. Thus, in one example, the methods can measure differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, or 28 miRNAs listed as "downregulated" in Table 11 or 17.

[0097] In one example, the methods described herein are capable of measuring the differential expression of at least one miRNA listed in Table 11 or Table 17 as "upregulated" and at least one miRNA listed in Table 11 or Table 17 as "downregulated." In one example, the method can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, or 46 miRNAs listed as "upregulated" in Table 11 or 17, and can measure the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, or 28 miRNAs listed as "downregulated" in Table 11 or Table 17.

[0098] In one example, a method described herein for providing a diagnosis or indication of whether a subject is likely to have stage 4 gastric cancer uses Table 17. The method may further include comparing the expression levels of any one or more miRNAs, including but not limited to miR-21-5pm, miR-223-3pm, miR-20a-5pm, miR-106b-5p, and miR-17-5p.

[0099] In yet another aspect, provided is a method of detecting or diagnosing gastric cancer in a subject or determining the likelihood that a subject has gastric cancer, comprising: (a) measuring in a non-cellular biological fluid sample an expression level of at least one miRNA having at least 90% sequence identity to a miRNA listed in Table 12 or Table 19; (b) generating a score based on the expression level of the miRNA measured in step (a); and (c) predicting the likelihood that the subject has gastric cancer using the score, wherein the score is calculated by a classification algorithm that compares the expression level of the subject to a positive control (a sample from a subject with gastric cancer) or a negative control (a sample from a subject without gastric cancer), and the score identifies the likelihood that the subject either (i) has gastric cancer, where the score falls within the range of the score of the positive control; or (ii) does not have gastric cancer (i.e., is free of gastric cancer), where the score falls within the range of the score of the negative control. In one example, the miRNA can include at least one of the miRNAs listed in Table 19. In one example, a method for detecting or diagnosing gastric cancer in a subject or determining the likelihood that a subject has gastric cancer may be referred to as a multivariate method. In one example, a multivariate method may use miRNAs that are representative of subpopulations or groups as determined by clinical data. In one example, a multivariate biomarker panel may include the biomarkers listed in Table 12, which includes both miRNAs known to be associated with gastric cancer and miRNAs known not to be associated with gastric cancer. In one example, miRNAs associated with gastric cancer that can be used in a multivariate method may be included in Table 19. Table 12 is described in more detail in the experimental section. In one example, FIG. 14 is a summary of the steps of a multivariate method for determining the likelihood that a subject will develop (have) gastric cancer.

[0100] Table 19 lists miRNAs whose expression levels were altered in gastric cancer subjects and which are frequently selected for use in multivariate biomarker panels, including:

[0101] [Table 6]

[0102] As discussed above, the multivariate method may include determining the biomarkers listed in Table 12, which includes both miRNAs known to be associated with gastric cancer and miRNAs known not to be associated with gastric cancer. Thus, in one example, where the multivariate method uses miRNAs listed in Table 19, the method may further include measuring the expression level of at least one miRNA whose expression level is unchanged in the subject when compared to a negative control. In one example, the miRNA whose expression level is unchanged in the subject when compared to a negative control may be any one of the miRNAs listed in Table 20.

[0103] Table 20 lists miRNAs frequently selected for use in multivariate biomarker panels whose expression levels were not associated with gastric cancer subjects, and are as follows:

[0104] [Table 7]

[0105] In one example, in the multivariate method known in the art and described herein, the score can be calculated using a classification algorithm.In one example, the classification algorithm includes, but is not limited to, a support vector machine algorithm, a logistic regression algorithm, a multinomial logistic regression algorithm, Fisher's linear discriminant algorithm, a quadratic classification algorithm, a perceptron algorithm, a k-nearest neighbor algorithm, an artificial neural network algorithm, a random forest algorithm, a decision tree algorithm, a naive Bayes algorithm, an adaptive Bayes network algorithm, and an ensemble learning method that combines multiple learning algorithms.As will be understood by those skilled in the art, the classification algorithm can be pre-trained using the expression levels of positive and negative controls.

[0106] In certain aspects of the multivariate methods described herein, the positive control may be a sample obtained from a subject with gastric cancer, and the negative control may be a sample obtained from a control without gastric cancer. In one example, the control without gastric cancer may be as described above. That is, in one example, the control without gastric cancer may be a subject without any gastric cancer-related disease. In one example, the control without gastric cancer may be a healthy person with no pre-existing medical conditions. In one example, the control without gastric cancer may be a subject known to have a non-gastric cancer-related disease (e.g., but not limited to, gastritis, intestinal metaplasia, intestinal atrophy, etc.). In one example, the controls may vary with respect to sample characteristics and other factors (e.g., gender, age, race of the subject based on the subjects from which the training data set was obtained). In one example, the control subject may be of Chinese or Chinese descent.

[0107] As will be appreciated by one of skill in the art, the expression level of a miRNA may be any one of the following: concentration, log(concentration), Ct / Cq value, 2 to the power of Ct / Cq value, and the like.

[0108] In one example, the methods described herein may use a data-trained formula that may be able to determine the likelihood that a subject has gastric cancer. In one example, the algorithm combines measurements of multiple miRNAs to provide a score for predicting the risk that a subject has gastric cancer. The algorithm combines information of multiple gene targets to provide better prediction accuracy than using a single miRNA biomarker. In one example, the minimum number of miRNA biomarkers to build a useful panel is 4. For example, FIG. 11 shows that a biomarker panel with 4 miRNAs has a mean AUC value of >0.8 in the validation set to be considered a useful biomarker panel. The improvement in the performance of the biomarker panel (in terms of AUC value) does not increase significantly when the number of miRNA types is greater than 7. Thus, in one example, the minimum number of miRNAs to build an optimal panel with maximum performance is 7 miRNAs. In one example, a formula that can be used to calculate a score for determining the likelihood that a subject has gastric cancer is as follows:

[0109] A × miRNA1 + B × miRNA2 + C × miRNA3 + D.

[0110] In the formula, miRNA1, miRNA2, miRNA3 are values ​​associated with the expression level of miRNA, which may be copy number, log(copy number), Ct / Cq value, or other values. And A, B, C, D are coefficients, which may be positive or negative. For example, a score for determining the likelihood that a subject has gastric cancer (i.e., a cancer risk score) can be calculated by combining 12 miRNAs (Table 21) that are often selected in the multivariate biomarker panel identification process with an occurrence rate of >20%, as described in the following examples and FIG. 15. In one example, when the score is obtained from the formula, the higher / lower the score, the higher the probability that the subject has gastric cancer. In one example, the range may not be clearly defined because the coefficients may be adjusted. For example, all the coefficients (i.e., A, B, or C) may be half-quantitative, or the value of D may change depending on the training with the training dataset. In one example, the value of the score may be positively or negatively correlated with the risk of having gastric cancer. The actual correlation may depend on the training data set, and those skilled in the art are familiar with the necessary adjustments required in establishing the training data set.

[0111] In one example, the range may not be clearly defined because the coefficients may be adjusted. For example, all coefficients (i.e., A, B, or C) may be multiplied or the value of D may change depending on training with a training data set. In one example, the value of the score may be positively or negatively correlated with the risk of having gastric cancer. The actual correlation may depend on the training data set, and one of skill in the art is familiar with the necessary adjustments required in establishing a training data set.

[0112] In one example, all coefficients (i.e., A, B, or C) can be intentionally adjusted halfway and / or the value of D can also be intentionally changed to adjust the range of values ​​calculated based on the formula for most subjects to 0-100. In one example, the value of the score can be positively or negatively correlated with the risk of having gastric cancer. The actual correlation may depend on the training data set, and one of skill in the art is familiar with the necessary adjustments required in establishing the training data set.

[0113] In one example, the methods described herein can measure the expression level of at least one miRNA listed in Table 12 or Table 19 as "significant" and at least one miRNA listed in Table 12 or Table 20 as "non-significant." In one example, the methods can measure differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or 42 miRNAs listed in Table 12 as "significant," or at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or at least 42 miRNAs listed in Table 19. or the differential expression of at least 12, at least 13, at least 14, at least 15, at least 20, at least 25, at least 30, at least 35, or 38 miRNAs can be measured; and the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, or 12 miRNAs listed in Table 12 as "non-significant" can be measured, or the differential expression of at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, or 10 miRNAs listed in Table 20 can be measured.

[0114] In one example, when the multivariate methods described herein use gastric cancer-associated miRNAs listed in Table 19, the method may further comprise measuring the expression level of any one or more miRNAs selected from the group consisting of miR-21-5p, miR-20a-5pm, miR-17-5p, miR-423-5p, and miR-223-3p.

[0115] In one example where the multivariate methods described herein use unrelated miRNAs listed in Table 20, the methods described herein may further comprise measuring the expression level of any one or more miRNAs selected from the group consisting of miR-34a-5p and miR-27b-3p.

[0116] An exemplary approach to arrive at a set of miRNAs correlated with gastric cancer is summarized below, and the general work flow is shown in Figures 1, 2, and 14.

[0117] Step 1: Total RNA (or a subfraction thereof) is extracted from a sample (e.g., plasma, serum, or other blood fraction) of a subject(s) with gastric cancer using an appropriate kit and / or purification method.

[0118] Step 2: From each sample, the amount (expression level) of one miRNA or a set of at least one miRNA selected from the group consisting of SEQ ID NO:1 to SEQ ID NO:191 is measured using experimental techniques. In one example, the at least one miRNA can be selected from the group listed in Tables 12, 19, and 20. These techniques are known in the art and include, but are not limited to, array-based approaches, amplification methods (PCR, RT-PCR, or qPCR), sequencing, next generation sequencing, and / or mass spectrometry.

[0119] Step 3: Apply mathematical methods to gather information about the diagnostic value and overlap of each single miRNA biomarker. These methods are known in the art and include, but are not limited to, basic mathematical approaches (e.g., fold quotients, signal-to-noise ratio, correlation), statistical methods (such as hypothesis testing (e.g., t-test, Wilcoxon-Mann-Whitney test)), area under the receiver operating characteristic curve, information theory approaches (e.g., mutual information, cross-entropy), probability theory (e.g., joint probability and conditional probability), or combinations and modifications of the methods mentioned above.

[0120] Step 4: Using the information obtained in step 3), a diagnostic content or diagnostic value is inferred for each miRNA biomarker. However, this diagnostic value is usually too small to obtain a highly accurate diagnosis with an accuracy rate, specificity, and sensitivity exceeding the 80% barrier. The diagnostic content of miRNAs suitable for the diagnosis of gastric cancer is listed in Figure 1. This figure includes the miRNAs listed in Tables 12, 19, and 20.

[0121] Step 5: More than one miRNA biomarker may be employed to improve the diagnostic performance of individuals suffering from gastric cancer. Thus, to select / define a miRNA biomarker set (which may include miRNAs listed in Table 12 or Tables 19 and 20) adapted for the detection of gastric cancer, statistical learning / machine learning / bioinformatics / computational approaches are applied for set selection. These methods include, but are not limited to, wrapper subset selection methods (e.g., forward stepwise, backward stepwise, combinatorial, optimization methods), filter subset selection methods (e.g., methods described in step 3), principal component analysis, or combinations and modifications of such methods (e.g., hybrid methods).

[0122] Step 6: The subset selected / defined in step 5 (which may range from only a few (at least one, at least two, or more sets) to all measured miRNAs) is then used to perform a diagnosis of gastric cancer. For this purpose, statistical learning / machine learning / bioinformatics / computational approaches are applied, including but not limited to any type of objective or non-objective analysis (classification methods (e.g., naive Bayes methods, linear discriminant analysis, neural networks for quadratic discriminant analysis, tree-based approaches, support vector machines, nearest neighbor methods), regression methods (e.g., linear regression, multiple regression, logistic regression, probit regression, ordered logistic regression, ordered probit regression, Poisson regression, negative binomial regression, multinomial logistic regression, truncated regression), clustering methods (e.g., k-means, hierarchical clustering, PCA), adaptations, extensions, and combinations of the approaches mentioned above).

[0123] Step 7: By combining the subset selection (step 5) and the machine learning approach (step 6), an algorithm or mathematical function for diagnosing gastric cancer is obtained. This algorithm or mathematical function is applied to the miRNA profile (miRNA expression profile data) of an individual (subject) to perform the diagnosis of gastric cancer.

[0124] In any of the methods described herein, the non-cellular biological fluid may include any bodily fluid that does not contain any cellular components of the bodily fluid. Thus, in one example, when extracted from a subject, the biological fluid or fluid may contain cellular components typically found in the subject's biological fluid or fluid. However, the biological fluid or fluid may also be processed using methods known in the art to remove any cellular components contained within the biological fluid or fluid. For example, a centrifugation or purification process may be performed to remove the cellular components of the biological fluid or fluid, after which the methods of the present disclosure may be performed. Thus, in one example, the biological fluid or non-cellular biological fluid includes, but is not limited to, amniotic fluid, breast milk, bronchial lavage fluid, cerebrospinal fluid, colostrum, interstitial fluid, peritoneal fluid, pleural fluid, saliva, semen, urine, tears, and non-cellular components of whole blood (including plasma, red blood cells, white blood cells, and serum). In one example, the non-cellular biological fluid may be serum or plasma.

[0125] In any of the methods described herein, the terms "subject" and "patient" as used herein may be used interchangeably to refer to animals (e.g., mammals, fish, amphibians, reptiles, birds, and insects). In one example, the subject may be a mammal (e.g., a non-human mammal (e.g., a dog, a cat, or a primate) and a human). In one example, the subject may be a primate (e.g., a chimpanzee and a human). In one example, the subject may be a human. In one example, the subject may be a human with or without gastric cancer (stomach cancer). In the experimental section of the present disclosure, the training dataset was obtained from subjects who are Chinese or of Chinese descent. Thus, in one example, the subject may be of Asian descent. In one example, the subject may be of Chinese or of Chinese descent.

[0126] In one example, the methods described herein can be provided as a kit. The kit may include components necessary or components for detecting the miRNAs described herein. The components may include primers / probes hybridizable with the miRNAs described herein, reagents, containers, and / or instruments configured to detect the miRNAs and primers / probes described herein. In one example, containers in the kit may include a primer / probe set capable of detecting or measuring the level of the miRNAs described herein, which is radioactively labeled prior to use. The kit may include containers of materials for reverse transcribing and amplifying one or more miRNAs (e.g., multiple containers containing dNTPs (each of the four deoxynucleic acids dATP, dCTP, dGTP, and dTTP), buffers, reverse transcriptase, and DNA polymerase). The kit may also include nucleic acid template(s) for positive and / or negative control reactions. The kit may further include any reaction components or buffers required for isolation of miRNA from a sample obtained from a subject.

[0127] The invention illustratively described herein can be suitably practiced in the absence of any element(s), limitation(s) not specifically disclosed herein. Thus, for example, the terms "comprising," "including," "containing," etc., should be read broadly and without limitation. Moreover, the terms and expressions employed herein are used as terms of description and not of limitation. And there is no intention to use terms and expressions that exclude any equivalents of the features shown and described or portions thereof. However, it is recognized that various modifications are possible within the scope of the invention as claimed. Thus, it should be understood that although the invention has been specifically disclosed by preferred embodiments and optional features, modifications and variations of the invention embodied in this specification and other disclosures can be restored by those skilled in the art, and such modifications and variations are considered to be within the scope of the invention.

[0128] The invention has been described broadly and generically herein. Narrower species and subgeneric groupings falling within this generic disclosure also form part of the invention. This includes the generic description of the invention with hypothetical or negative limitations that remove any subject matter from the genus, regardless of whether the featured material is specifically described herein.

[0129] Other embodiments are within the following claims and non-limiting examples. Also, when features or aspects of the invention are described in terms of a Markush group, those of skill in the art will recognize that the invention is also described thereby with respect to any individual members or subgroups of members of the Markush group. EXAMPLES

[0130] Experimental Section method Pre-analytical processing (sample collection and microRNA extraction): Serum samples from normal subjects (i.e., gastric cancer-free controls) and gastric cancer subjects were collected by centrifugation, processed, and stored frozen at -80°C by members of the Singapore Gastric Cancer Society. Total RNA from 200 μl of serum was isolated using the well-established column-based miRNeasy kit (Qiagen, Hilden, Germany) on a semi-automated QiaCube system. Serum contains small amounts of RNA. To reduce RNA loss and monitor extraction efficiency, a rationally designed isolation enhancer (MS2) and spike-in control RNA (MiRXES) were added to the samples prior to isolation.

[0131] RT-qPCR: Isolated total RNA and synthetic RNA standards were converted to cDNA in an optimized multiplex reverse transcription reaction with the addition of a second set of spike-in control RNA to detect the presence of inhibitors and monitor RT-qPCR efficiency. Reverse transcription reactions were performed using Impron II (Promega®) reverse transcriptase according to the manufacturer's instructions. The synthetic cDNA was then subjected to a multiplex amplification step and quantified using a Sybr Green-based single-plex qPCR assay (MIQE compatible) (MiRXES®). An ABI® ViiA-7 384 instrument was used for the qPCR reactions. An overview and details of the work flow for RT-qPCR measurements of miRNAs are summarized in Figure 2.

[0132] Data processing: Raw cycles to threshold (Ct) were processed and the absolute copy number of the target miRNA in each sample was determined by interpolating the synthetic miRNA standard curve. Technical variation introduced during RNA isolation and RT-qPCR processing was normalized by spike-in control RNA. For single miRNA analysis, biological variation was further normalized by a set of validated endogenous reference miRNAs stably expressed across all control and disease samples.

[0133] Test Design A well-designed clinical trial (case-control study) was conducted to ensure accurate identification of diagnostic biomarkers for gastric cancer. A total of 237 Chinese patients with gastric cancer (stages 1-4) with a mean age of 68.0 ± 10.9 years were used in this study and compared with another 236 age- and sex-matched normal (healthy / no gastric cancer) Chinese subjects to serve as the control group. All cancer subjects were confirmed by endoscopy and biopsy, and serum samples were collected before any treatment. All normal (healthy) subjects were confirmed to be free of gastric cancer by endoscopy at the time of sample collection, and were followed up for 3-5 years with biennial endoscopy screening to ensure that subjects did not develop gastric cancer within this period. Detailed clinical information of the subjects is tabulated in Tables 2 and 3. All serum samples were stored at -80°C until use. Characteristics of the subjects studied are summarized in Table 5. Many control subjects (normal / healthy / non-gastric cancer subjects) had gastric diseases (e.g., gastritis, intestinal metaplasia, and intestinal atrophy). Thus, the control group represented a high-risk population for gastric cancer. Although the percentage of gastric cancer patients with H. pylori (78.8%) was higher than that of the control group (55.7%), the difference was too small to support the use of H. pylori as a biomarker for gastric cancer.

[0134] [Table 8-1]

[0135] [Table 8-2]

[0136] [Table 8-3]

[0137] [Table 8-4]

[0138]

Table 8-5

[0139]

Table 8-6

[0140]

Table 9-1

[0141]

Table 9-2

[0142]

Table 9-3

[0143]

Table 9-4

[0144]

Table 9-5

[0145]

Table 9-6

[0146]

Table 10

[0147] It is known that cell-released miRNAs circulating in blood may originate from various tissue sources. As a result, changes in miRNA levels due to the presence of solid gastric tumors may be complicated by the presence of the same miRNAs from other sources. Thus, determining the difference in miRNA expression levels found in cancer and control groups is difficult and, as expected, the differences are unclear. It is also known that most cell-released miRNAs are very low in blood due to the dilution effect of the large blood volume (5 L in adults). Thus, accurate measurement of multiple miRNA targets from limited amounts of serum / plasma samples is important and presents a great challenge. Instead of using low sensitivity or semi-quantitative screening methods (microarray, sequencing), to best facilitate the discovery of significantly altered miRNA expression and the identification of a multivariate miRNA biomarker panel for the diagnosis of gastric cancer, we chose to perform a qPCR-based assay in this experimental study with the work flow summarized in Figure 2.

[0148] In the workflow used in this experimental section, all reactions were performed at least twice in a single-plex manner for miRNA targets and at least four times for synthetic RNA "spike-in" controls. To ensure the accuracy of the results of such a high-throughput qPCR study, after multiple iterations, the study was designed / established into a powerful workflow for circulating biomarker discovery (see the "Methods" section and Figure 2). In this workflow, various artificially designed "spike-in" controls were used to monitor and correct technical variations in the isolation, reverse transcription, amplification, and qPCR processes. All spike-in controls were non-natural synthetic miRNA mimics (small single-stranded RNAs ranging from 22-24 bases in length) that were designed in silico to have very low sequence homology to all known human miRNAs, thus minimizing cross-hybridization to the primers used in the assay. The miRNA assay was also engineered to be divided in silico into multiple multiplex groups to minimize non-specific amplification and primer-primer interactions. Synthetic miRNAs were used to generate standard curves for the interpolation of absolute copy numbers in all measurements, thus further correcting for technical variability.Using this workflow with multiple levels of control, the expectation is that even circulating miRNAs with low expression levels can be reliably and reproducibly detected.

[0149] miRNA biomarkers The step towards biomarker identification involves comparing the expression levels of each miRNA in normal (control) and disease states.Expression levels of 578 human miRNAs (miRBase version 10 release) in all 472 serum samples (gastric cancer and normal / control) were quantitatively measured using a robust workflow and highly sensitive qPCR assay (MiRXES, Singapore).

[0150] In this experimental design, 200 μl of serum was extracted, total RNA was reverse transcribed, and amplified by touchdown amplification to increase the amount of cDNA but not to change the ratio of miRNA expression levels (Figure 2). The amplified cDNA was then diluted for qPCR measurement. A simple calculation based on the dilution effect revealed that miRNAs (expressed at levels ≦500 copies / ml in serum) were quantified at levels close to the detection limit of the singleplex qPCR assay (≦10 copies / well). At such concentrations, the measurement becomes very difficult due to technical limitations (pipetting and qPCR errors). Therefore, miRNAs expressed at concentrations ≦500 copies / ml were excluded from the analysis and considered as undetectable in this study.

[0151] Approximately 33% of the total miRNAs assayed (a total of 578 miRNAs were assayed) were found to be highly expressed. These 191 miRNAs (expression levels ≥ 500 copies / ml; Table 4) were reliably detected in more than 90% of serum samples, and these were much more abundant than previously reported using other techniques. This highlights the importance of using a strong experimental design and a well-controlled workflow.

[0152] [Table 11-1]

[0153] [Table 11-2]

[0154] [Table 11-3]

[0155] [Table 11-4]

[0156] [Table 11-5]

[0157] A heatmap was then generated to represent the expression levels of all detected serum miRNAs (191 detected miRNAs, Figure 3). Although the various subpopulations of miRNAs were clustered together, there appeared to be no obvious differences between the cancer and control groups by unsupervised hierarchical clustering based on the total number of detected miRNAs. Thus, without any feature selection, the overall miRNA profiles for both groups appeared similar.

[0158] The expression of 191 serum miRNAs was then compared.

[0159] · A] Between normal / control and gastric cancer (regardless of stage or subtype). B] Between various stages of gastric cancer (stages 1 to 4), and ·C] Between normal / control and various stages of gastric cancer.

[0160] The significance of differential expression between two groups was calculated based on t-tests (p-value < 0.01) corrected for false discovery rate (FDR) estimates using Bonferroni-type multiple comparison procedures known in the art.

[0161] A] Identification of differentially expressed miRNAs between normal / control and gastric cancer (regardless of stage or subtype). Samples from patients clinically confirmed to have either diffuse, intestinal, or mixed type of gastric cancer were grouped together and compared with samples from normal / control donors. A pool of 75 miRNAs was identified that showed significant differential expression between the control and cancer groups (p-value < 0.01, Table 6). Unlike other reports (Table 1) in which most of the expression of the putative biomarkers was found to be higher in gastric cancer subjects, this study demonstrated that 51 miRNAs were upregulated, whereas 24 were downregulated in cancer subjects (Table 6). Thus, this experimental design identified a larger number of regulated biomarkers.

[0162] [Table 12]

[0163] [Table 13-1]

[0164] [Table 13-2]

[0165] Comparing the results of this study (Table 6) with the 30 miRNAs identified as differentially expressed in blood in other reports (Table 1), only six miRNAs (hsa-miR-106b-5p, hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-21-5p, hsa-miR-223-3p, and hsa-miR-423-5p) were found to be commonly upregulated (Table 7). Thus, the majority of miRNAs reported to be differentially regulated in the literature were not confirmed in this study. Interestingly, we identified 69 novel miRNAs here that are potential biomarkers for gastric cancer.

[0166] [Table 14]

[0167] Using this set of 75 biomarkers, more differential clustering between gastric cancer subjects and normal / control subjects was observed in the heat map of miRNA profiles (Figure 4). Looking at the horizontal lines, the majority of gastric cancer subjects (black) were clustered in a concentrated area, while the majority of normal / control subjects were left in the remaining space in the map. The AUC values ​​for the top-ranked upregulated miRNA (hsa-miR-142-5p) and downregulated miRNA (hsa-miR-99b-5p) were only 0.71 and 0.67, respectively (Figure 5). Each of the 75 differential miRNAs or a combination of several of them can serve as a biomarker or biomarker panel for the diagnosis of gastric cancer.

[0168] The miRNA expression was found to be clustered into subgroups, as illustrated in the heatmaps (Figs. 3, 4). Each cluster had about 10–20 miRNAs (Figs. 3, 4; between dotted lines). Analysis of all detectable miRNAs revealed that many of them were positively correlated (Pearson correlation coefficient >0.5) (Fig. 6). The same observation was found for gastric cancer-specific miRNAs (Fig. 7). These results indicated that a group of miRNAs was similarly regulated among all subjects. As a result, a panel of miRNAs can be assembled by replacing one or more differentially altered miRNAs with another to improve diagnostic performance. And all 75 significantly altered miRNAs were important for the development of a multivariate index diagnostic assay for gastric cancer.

[0169] B] Identification of differentially expressed miRNAs between normal / control and various stages of gastric cancer. The expression of miRNAs among different stages of gastric cancer was then compared.Using a two-way ANOVA taking into account stage and subtype, 36 miRNAs (from a pool of 191 serum miRNAs) were found to be differentially expressed among different stages of gastric cancer subjects (p-value of ANOVA test after correction for false discovery rate was <0.01; data not shown).

[0170] Knowing the differences between the various stages, we next addressed the identification of miRNAs that were differentially expressed between normal / control and each cancer stage. The total number of differential miRNAs found in this novel search process was 113 (Tables 8-11; profiles in these tables were overlapping). Among these, a pool of 20 miRNAs (8 upregulated and 12 downregulated), 62 miRNAs (54 upregulated and 8 downregulated), 43 miRNAs (37 upregulated and 6 downregulated), or 74 miRNAs (28 upregulated and 46 downregulated) showed differential expression between normal / control and the corresponding stages 1, 2, 3, and 4, respectively (FDR-corrected p-values ​​<0.01; Tables 8-11).

[0171] [Table 15]

[0172] [Table 16-1]

[0173] [Table 16-2]

[0174] [Table 17-1]

[0175] [Table 17-2]

[0176] [Table 18-1]

[0177] [Table 18-2]

[0178] With the exception of three miRNAs (hsa-miR-486-5p, hsa-miR-19b-3p, and hsa-miR-411-5p), the rest of the 75 miRNAs (Table 6) found in the analysis using all cancer stages together were found to be a subset of these 113 miRNAs identified when each cancer stage was analyzed independently. Most of the 36 miRNAs (except one) found to be differentially regulated between various stages of gastric cancer were found to be differentially expressed when comparing normal / control (N) with each stage of gastric cancer (S1-S4) (Tables 8-11). The differences between each stage resulted in different miRNA pools that were useful for detecting gastric cancer. Thus, the results of comparing the expression levels in different stages relative to normal / control resulted in a larger pool of informative miRNAs.

[0179] Apparently, fewer miRNAs were identified in early stage 1 than in later stages. The miRNA features ranked highest by AUC value were not always the same across all stages (Fig. 8). A larger number of miRNAs were regulated (Tables 8-11), and similar features were repeatedly found across stages (Fig. 10). Only one miRNA (hsa-miR-27a-5p) was found to be common among all four stages. These differential miRNAs (Tables 8-11) can be used individually, and collecting the results of multiple miRNAs may be useful for diagnosing gastric cancer in clinical settings where the cancer stage is unknown at presentation.

[0180] From the above studies, a biomarker panel consisting of at least one miRNA from each stage-specific miRNA list can provide an alternative method for determining the likelihood that a subject has gastric cancer.

[0181] III. Discovery of multivariate biomarker panels An important criterion for assembling a multivariate panel is to include at least one miRNA from the specific list for each cancer stage to ensure coverage of all cancer subgroups. However, the miRNAs defining the four cancer stages were not completely distinct, since the same miRNAs were similarly found among the stages (Fig. 9). At the same time, many cancer-related and non-related miRNAs were found to be positively correlated (Figs. 6, 7), which makes it difficult to select the optimal miRNA combination for gastric cancer diagnosis.

[0182] Seeing the complexity of this task, we decided that the study should use the SFFS (sequence forward floating search) algorithm to identify the miRNA panel with the highest AUC. It was also known in the art to use a state-of-the-art linear support vector machine, a well-used and recognized modeling tool for constructing variable panels, to assist in the selection of miRNA combinations. This model generates a score based on a linear equation that describes the expression level of each member and its weighting. This linear model is easily accepted and applied in clinical practice.

[0183] A key requirement for the success of such a process is the availability of high-quality data: quantitative data of all detected miRNAs in a large number of well-defined clinical samples not only improved the accuracy and precision of the results, but also ensured the consistency of the identified biomarker panel for clinical application using qPCR.

[0184] In the situation of a large number of clinical samples (500+), the problem of overfitting of data in modeling was minimized because only 191 feature candidates were to be selected. Also, to ensure the accuracy of the results, four-fold cross-validation was performed multiple times to test the performance of the identified biomarker panel on an independent set of validation samples (the remaining 1 / 4 of samples in each fold) based on the discovery set (3 / 4 of samples in each fold). During the cross-validation process, samples were matched for sex, subtype, and stage.

[0185] Box plots of the results (AUC of biomarker panels in both discovery and validation phases) are shown in Figure 10. As expected, there was a decrease in the AUC values ​​in the validation set for each discovery (0.04-0.06). Although the box size was larger for the validation phase data (indicating a wider range of values), the difference in AUC values ​​was ≦0.1. This range of values ​​was even narrower (<0.05) when the number of biomarkers included in the analysis was greater than six. This observation indicated that more robust results were obtained with a number of miRNAs greater than six.

[0186] A more quantitative representation of the results is shown in Figure 11, where no improvement in AUC value was observed when the number of biomarkers included in the analysis was greater than seven. Although the difference between the six-miRNA biomarker panel and the seven-miRNA biomarker panel was statistically significant, the improvement in AUC value was only 0.02. Therefore, the AUC value of 0.855 obtained using the six-miRNA panel is sufficient for clinical use.

[0187] Composition of IV multivariate miRNA biomarker panels To investigate the composition of multivariate biomarker panels, we counted the prevalence of miRNAs in all panels containing 6–10 miRNAs, but excluded panels with AUCs in the top and bottom 10%. We excluded those with AUCs in the top and bottom 10% to avoid counting false discovery biomarkers resulting from applying inaccurate data from subpopulations generated by the randomization process of the cross-validation analysis. If we excluded those miRNAs with counts less than 4, a total of 55 miRNAs were selected in the discovery process (Table 12). We also found that the expression of 43 of these was significantly altered in cancer. We found that the inclusion of the other 12 (which were unchanged in cancer) significantly improved the AUC values. Without direct and quantitative measurement of all miRNA targets, these miRNAs would never have been selected in high-throughput screening methods (microarrays, sequencing) and would have been excluded for qPCR validation. These miRNAs were not randomly selected. This is because the top two miRNAs (hsa-miR-532-5p, hsa-miR-30e-5p) were consistently found in 57.8% and 48.4% of these panels.

[0188] [Table 19-1]

[0189] [Table 19-2]

[0190] The miRNAs selected to form the multivariate panel (Table 12) showed differences in detecting various stages of cancer (Tables 8-11). Regarding the top 10 frequently selected miRNAs listed as having an abundance higher than 20%, only three of them (hsa-miR-21-5p, hsa-miR-27a-5p, hsa-miR-142-5p) were found to be commonly regulated in more than two different stages of cancer. Meanwhile, for the remaining ones, hsa-miR-103a-3p was significant only for stage 1, hsa-miR-181a-5p was significant only for stage 4, hsa-miR-26a-5p was significant for stages 1&4, hsa-miR-20a-5 and hsa-miR-17-5p were significant for stages 2&4, and hsa-miR-616-5p and hsa-miR-30a-5p were significant for stages 3&4. Two miRNAs specific to each of stages 1-3 were frequently selected, except for stage 4 (Figure 9), where a large number of miRNAs overlapped with the list of the other three stages.

[0191] When comparing the identity of the multiple miRNAs selected for multivariate panels and single miRNAs as diagnostic markers, they were not necessarily identical. For example, both the top upregulated miRNA (hsa-miR-532-5p) and downregulated miRNA (hsa-miR-30e-5p) for all cancers (Figure 5) were not included in the list. Therefore, the best biomarkers identified for various stages could not be simply combined to form an optimal biomarker panel; rather, the marker panels provided complementary information that gave the best results.

[0192] Interestingly, five miRNAs (those with >48% occurrence, see Table 12) were frequently selected for multivariate panels. Three of these (hsa-miR-21-5p, hsa-miR-103a-3p, and hsa-miR-20a-5p) were cancer-specific, and two (non-significant group; hsa-miR-532-5p and hsa-miR-30e-5p) were from other groups. The co-existence rate of these five miRNAs in all multivariate panels was analyzed (Table 13). Then, 85.9% of the panels had either hsa-miR-532-5p or hsa-miR-30e-5p. Therefore, the use of either one of these two miRNAs may be useful enough to increase the AUC value in the multivariate index panel.

[0193] Overall, 93.8% of the miRNA biomarker panels had at least one miRNA from each group. Thus, in one example, a multivariate method for determining the likelihood that a subject has gastric cancer may include at least one miRNA from each of these two frequently selected miRNA groups.

[0194] [Table 20]

[0195] Calculating your cancer risk score The miRNAs can be combined to form a biomarker panel to calculate a cancer risk score (Equation 1). For example, 12 miRNAs (Table 21) that have an occurrence rate of >20% and are often selected in the multivariate biomarker panel identification process can be combined to form a biomarker panel to calculate a cancer risk score. The equation here shows the use of a linear model for gastric cancer risk prediction, where the cancer risk score (unique for each subject) indicates the likelihood that the subject has gastric cancer. This calculation is performed by summing the weighted measurements of the 12 miRNAs with a constant of 50. Formula 1-

[0196]

number

[0197] [Table 21]

[0198] The control subjects and cancer subjects in this study have different cancer risk score values ​​calculated based on the formula (Figure 15A). The fit of the cancer risk scores of the control subjects and the cancer subjects to the probability distribution can be seen in Figure 15B, and a clear separation can be found between the two groups. In this study, the control subjects were non-cancer subjects selected from the high-risk population, and the probability of having gastric cancer is 0.0067. Based on this prior probability and the fit to the probability distribution in Figure 15B, the probability (risk) of the unknown subjects having cancer can be calculated based on their cancer risk score values ​​(Figure 15C). The higher the score, the higher the risk of the subject having gastric cancer. Furthermore, the cancer risk score can distinguish the fold change in the probability (risk) of the unknown subjects having gastric cancer when compared to the gastric cancer incidence rate of the high-risk population (Figure 15D). For example, an unknown high-risk subject with a cancer risk score of 70 has a 14.6% probability of having gastric cancer, which is about 22 times higher than the average risk of the high-risk population.

[0199] By further applying two thresholds, 40 and 70, subjects can be defined into three categories based on the cancer risk score (FIG. 15C, D). Subjects with a cancer risk score in the range of 0-40 belong to the low cancer risk group, with a probability of 0.052% for the group to have gastric cancer, which is 13 times lower than the average cancer incidence rate of the high-risk population. Subjects with a cancer risk score in the range of 40-70 belong to the medium cancer risk group, with a probability of 1.4% for the group to have gastric cancer, which is two times higher than the average cancer incidence rate of the high-risk population. Subjects with a cancer risk score in the range of 70-100 belong to the high cancer risk group, with a probability of 12.2% for the group to have gastric cancer, which is 18 times higher than the average cancer incidence rate of the high-risk population.

Claims

[Claim 1] 1. A method for detecting gastric cancer in a subject, comprising: determining expression levels of at least four microRNAs (miRNAs) in a serum or plasma sample obtained from the subject, the at least four miRNAs being hsa-miR-21-5p, hsa-miR-142-5p, hsa-miR-181a-5p and hsa-miR-103a-3p, wherein downregulation of hsa-miR-181a-5p and hsa-miR-103a-3p and upregulation of hsa-miR-21-5p and hsa-miR-142-5p compared to a control is indicative of the subject having gastric cancer, the control being a subject without gastric cancer.

Images