WO2021230604A1 - Method for diagnosing ovarian cancer by using machine learning prediction model and system applying same - Google Patents

Abstract

The present invention relates to a method for providing information for diagnosing ovarian cancer by using a machine learning prediction model, and a system applying same. According to the method, before surgery, ovarian cancer can be more quickly and accurately diagnosed by using a small amount of serum, thereby diagnosing ovarian cancer at an early stage, avoiding unnecessary surgery, and determining an optimal treatment plan.

Description

Ovarian cancer diagnosis method using machine learning prediction model and system applying the same

This application claims priority to Republic of Korea Patent Application No. 10-2020-0056152 filed on May 11, 2020, and the entire specification is a reference to the present application.

It relates to a method for providing information for diagnosing ovarian cancer using a machine learning prediction model and a system for applying the same.

Ovarian cancer is the leading cause of gynecological cancer death and the 5-year survival rate is as low as 44%. The poor prognosis of ovarian cancer is because there is no effective screening test for early detection of ovarian cancer, and because it is located in an anatomically difficult area, most patients are diagnosed at advanced stage 3 or 4. In patients with ovarian tumors, prior to surgery, distinguishing between malignant and benign is very important in avoiding unnecessary surgery and in determining the optimal treatment plan. Serum cancer antigen 125 (cancer antigen 125, CA125) is a representative tumor marker for ovarian cancer and is most commonly used for differential diagnosis before surgery along with the risk of ovarian malignancy algorithm (ROMA). However, in the case of CA125, since only about 50% of patients with low-stage (stage I or II) ovarian cancer show an elevated level, there is no sufficient effect in terms of sensitivity and specificity for early diagnosis of ovarian cancer.

On the other hand, it is known that there is a significant difference in the expression pattern of micro RNA (miR) in various solid cancer tissues including ovarian cancer compared to normal tissues, and this dysregulation of miR is involved in various stages of the carcinogenesis process. On the other hand, since blood miR reflects the miR expression status of cancer tissues, various previous studies attempted to develop ovarian cancer diagnostic biomarkers by analyzing the expression pattern of circulating free-circulating miR (miR) in the blood of ovarian cancer patients. there was However, since circulating free miR is degraded by RNAse present in blood, there is a problem of low stability. On the other hand, in the case of exsomal miR, it is packed in exosomes, so it is safe to avoid degradation by RNase, and cancer cells secrete a greater amount of exosomes than normal cells. It has the advantage of being able to reduce perturbation due to cell-derived miRs. Therefore, it is clinically very important to develop a biomarker with improved sensitivity and specificity for ovarian cancer diagnosis by overcoming the problem of circulating free miR.

One aspect is to generate data by measuring the expression level of the miRNA of one or more biomarkers isolated from the subject's sample; calculating an abnormal value from the miRNA expression level data; and generating cancer diagnosis information based on the abnormal value result. To provide an information providing method for diagnosis of ovarian cancer using a machine learning prediction model.

One aspect is to generate data by measuring the expression level of the miRNA of one or more biomarkers isolated from the subject's sample; calculating an abnormal value from the miRNA expression level data; and generating cancer diagnosis information based on the abnormal value result.

Another aspect is a receiving unit for receiving data measuring the miRNA expression level of one or more biomarkers isolated from the subject's sample; a calculation unit for calculating an abnormal value by applying the data to the learning data; and a generator for generating cancer diagnosis information based on the abnormal value result.

One aspect is to generate data by measuring the expression level of the miRNA of one or more biomarkers isolated from the subject's sample; calculating an abnormal value from the miRNA expression level data; and generating cancer diagnosis information based on the abnormal value result. Another aspect is to generate data by measuring the expression level of the miRNA of one or more biomarkers isolated from the subject's sample; calculating an abnormal value from the miRNA expression level data; and generating cancer diagnosis information based on the abnormal value result. The machine learning prediction model may be a decision tree model.

As used herein, the term "decision tree model" refers to a combination of predictable rules by analyzing data and representing a pattern existing between them as a combination of predictable rules. analysis method to be performed. The learning process of the decision tree is divided into two processes: recursive partitioning, which divides input variable regions into two, and pruning, which integrates detailed regions.

In addition, as used herein, the term “outlier value” means a very small value or a very large value that is far out of the range of measured data. For example, the outlier value is a value far out of the range of miRNA expression level data of one or more biomarkers isolated from a sample of an individual, and means a value in which the miRNA expression level is ideally high or ideally small compared to a control. do.

The subject is a subject for diagnosing ovarian cancer, for example, a subject for predicting the likelihood of ovarian cancer, a subject for diagnosing the state of ovarian cancer, a subject for determining a prognosis, preventing or treating ovarian cancer It means a target for determining the dosage of a drug, a target for determining a treatment method according to the progression of ovarian cancer, a target for predicting the possibility of recurrence after ovarian cancer surgery and chemotherapy. The target for predicting the likelihood of ovarian cancer is, for example, a target for predicting the likelihood of ovarian cancer before undergoing surgery for ovarian cancer, or a target in which a tumor is found by imaging tests such as ultrasound, CT imaging, etc. The target may be a normal person who wants to diagnose ovarian cancer at an early stage through a health checkup. The subject may be a vertebrate, specifically mammals, amphibians, reptiles, birds, etc., and more specifically, may be a mammal, for example, a human ( Homo sapiens ). The sample may include a sample such as tissue, cells, whole blood, serum, plasma, saliva, sputum, cerebrospinal fluid or urine isolated from a subject.

In one embodiment, the miRNA may be one or more selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c. In addition, the miRNA may be isolated from an exosome. In another embodiment, the step of obtaining data by measuring the protein expression level of the tumor marker (Cancer Antigen, CA) 125 may be further included.

As used herein, the term "miRNA (microRNA)" is a non-coding RNA having a length of 21 to 25 nucleotides, and is a post-transcriptionally gene by binding to the 3'-untranslated region of the target mRNA. Regulating expression results in mRNA degradation and/or transcriptional repression.

As used herein, the term “exosome” refers to nanoparticles having a diameter of 50 to 100 nm that exist in blood and are secreted by cells to exchange information between cells, and are derived from microvesicular bodies in cells. Exosomes are mostly secreted by proliferating cells and are found in a variety of body fluids, including blood, urine, saliva, and ascites. In addition, malignant cells secrete a greater amount of exosomes than normal cells, and the exosomes extracted from the tumor contain various molecular components including proteins, DNA, and RNA that reflect the tumor cells secreting them. Therefore, exosomes are useful as reliable serum biomarkers for ovarian cancer because they are abundant in body fluids that can be obtained non-invasively, and reflect the properties of cells of origin.

In the case of ovarian cancer tissue or blood circulation free (free-circulating) miR-145, the expression is reduced compared to normal tissue or blood, but exosome miR-145 is characterized by increased expression in the blood of ovarian cancer patients. In addition, the miR-145 can be usefully used as a single marker in predicting ovarian cancer, as it exhibits a sensitivity of 91.7%. In addition, when miR-145, miR-200c and CA125 are combined, the sensitivity is 100%. As a result, exosome miR-145 and/or miR-200c is up-regulated in ovarian cancer patients, resulting in ovarian cancer and benign lesions. It can be a biomarker to distinguish between the two and to overcome the limitations of CA125.

Measuring the expression level of miRNA is a process of determining the presence and expression level of miRNA in a sample of an individual for diagnosing ovarian cancer, and measuring the amount of miRNA. Analytical methods for this include reverse transcription polymerase reaction (RT-PCR), competitive reverse transcription polymerase reaction (Competitive RT-PCR), real-time reverse transcription polymerase reaction (Real-time RT-PCR), RNase protection assay (RPA; RNase protection) assay), Northern blotting, reverse transcription polymerase reaction (RT-PCR) such as DNA chip, and competitive reverse transcription polymerase reaction (Competitive RT-PCR). In addition, the expression level is the expression level of one or more miRNAs selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c, or the expression level of CA125 may be measured simultaneously or sequentially.

In one embodiment, the step of calculating an abnormal value from the expression level data of the miRNA comprises miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c. When the expression level of one or more miRNAs selected from the group or the expression level of CA125 is increased compared to the control group, it is determined as an abnormal value and calculated. Also, in another embodiment, calculating an abnormal value from the miRNA expression level data includes the following steps. Specifically, when the expression value of the first miRNA is expressed as X ₁ (in this case, X ₁ is a specific constant capable of separating the first miRNA from the second to n-th miRNAs). separating from the to nth miRNA; When the expression value of the isolated second miRNA is expressed as X ₂ (in this case, X ₂ is a specific constant capable of separating the second miRNA from the third to nth miRNA). Isolating from n miRNA; When the expression value of the j and isolated n-th niRNA is expressed as X _n (in this case, X _n is a specific constant that can be separated from the n-th to n+m miRNA.) It may include the step of isolating the m miRNA. In this case, the abnormal value is an expression value at which the nth miRNA can be separated from the n+1th to n+m miRNAs, and may be expressed as a specific constant.

In one embodiment, the generating cancer diagnostic information based on the abnormal value result comprises miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c. When the expression level of one or more miRNAs selected from the group or the expression level of CA125 is increased compared to the control group, it may be a step of determining ovarian cancer. Further, in another embodiment, the step comprises: determining that the expression value of the first miRNA is X ₁ or more to be positive for ovarian cancer, and measuring the expression value of the second to n-th miRNAs when the expression value of the first miRNA is less than _{X 1;} determining that the expression value of the second miRNA is _{greater than or equal to X 2 as} positive for ovarian cancer, and measuring the expression value of the third to n-th miRNAs when the expression value of the second miRNA is less than _{X 2;} ... ; and determining that the expression value of the n-th miRNA is _{greater than or equal to X n as} positive for ovarian cancer, and determining that _{the expression value is less than X n as} negative for ovarian cancer.

Another aspect is a receiving unit for receiving data measuring the miRNA expression level of one or more biomarkers isolated from the subject's sample; a calculation unit for calculating an abnormal value by applying the data to the learning data; and a generator for generating cancer diagnosis information based on the abnormal value result. In one embodiment, the system may further include a monitoring unit for monitoring the expression pattern of the miRNA of the biomarker.

4 and 5 are structural diagrams of an ovarian cancer diagnosis system according to an embodiment.

As shown in FIG. 4 , the ovarian cancer diagnosis system 100 according to an embodiment may include a miRNA expression level measurement data receiver 110 , a calculator 120 , and a generator 130 according to each function. . In addition, as shown in FIG. 5 , the ovarian cancer system according to an embodiment may further include a monitoring unit 140 . First, the receiver 110 receives miRNA expression level measurement data of one or more biomarkers isolated from a sample of an individual. Specific details of the miRNA are the same as described above. The measurement of the miRNA expression level is a process of determining the presence and expression level of miRNA in a sample of an individual in order to diagnose ovarian cancer, and measuring the amount of miRNA. Specific details of the miRNA expression level measurement are as described above.

The calculator 120 calculates and determines an abnormal value by applying the data obtained by measuring the miRNA expression level to the learning data. In one embodiment, the expression level of one or more miRNAs selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c or CA125 expression level If it increases by more than the reference value compared to the control group, it is determined to be an abnormal value and calculated. In another embodiment, when the expression value of the first miRNA is expressed as X ₁ (in this case, X ₁ is a specific constant capable of separating the first miRNA from the second to n-th miRNAs). The first miRNA can be separated from the second to nth miRNAs. For example, calculating and determining the abnormal value may be performed when the expression value of the first miRNA is expressed as X ₁ (in this case, X ₁ is the first miRNA can be separated from the second to n-th miRNAs) It is a specific constant.) separating the first miRNA from the second to n-th miRNAs; When the expression value of the isolated second miRNA is expressed as X ₂ (in this case, x is a specific constant capable of separating the second miRNA from the third to n-th miRNAs). separating from miRNA; When the expression value of the j and isolated n-th niRNA is expressed as X _n (in this case, X _n is a specific constant that can be separated from the n-th to n+m miRNA.) It is possible to calculate and determine an abnormal value including the step of separating from m miRNA.

The generator 130 generates cancer diagnosis information based on the abnormal value result. Specifically, when the miRNA expression level is determined to be an abnormal value and calculated, ovarian cancer diagnosis information of an individual is generated from the abnormal value result. Specific details of generating the diagnostic information are the same as described above. The ovarian cancer diagnosis information may be, for example, whether to additionally follow-up whether the individual has progressed to ovarian cancer or to determine whether to proceed with surgery after being diagnosed as having ovarian cancer. In addition, when ovarian cancer is diagnosed, it may be to determine whether to proceed with chemotherapy.

The monitoring unit 140 monitors the expression pattern of the biomarker miRNA. Specifically, by tracking the miRNA expression pattern of the biomarker after surgery and/or chemotherapy of an individual diagnosed with ovarian cancer in the generating unit 130, it is possible to confirm early whether the individual has ovarian cancer recurrence. . For example, when the miRNA expression level of the biomarker in the subject is determined to be an abnormal value and calculated, it may be determined that the subject's ovarian cancer has recurred.

6 is a conceptual diagram of a method for early diagnosis of ovarian cancer using a machine learning prediction model according to an embodiment.

As shown in FIG. 6 , the method and system for diagnosing ovarian cancer according to an embodiment measure the expression level of miRNA and serum cancer antigen 125 (cancer antigen 125, CA125) present in the subject's serum exosomes, and then the expression value According to this, it can be judged whether or not ovarian cancer is positive. For example, when the expression level of the miRNA and/or CA125 increases compared to a normal control, or when the expression level of the miRNA and/or CA125 is calculated as an abnormal value, it can be determined that the ovarian cancer is positive. For an individual determined to be positive for ovarian cancer, a treatment method according to the progression of ovarian cancer may be determined by performing surgery for diagnosis and treatment or chemotherapy after surgery. The recurrence of ovarian cancer can be determined by periodically measuring the expression levels of miRNA and CA126 in the same way as above for an individual who has undergone surgery and treatment. Specifically, when the expression level of the miRNA and/or CA125 of the individual increases compared to a normal control, or when the expression level of the miRNA and/or CA125 is calculated as an abnormal value, it is determined that ovarian cancer has recurred can do. On the other hand, when the expression level of the miRNA and/or CA125 does not increase (eg, the same, similar, or decreases) compared to the normal control or the expression level of the miRNA and/or CA125 is not judged to be an abnormal value, the ovarian It can be judged to be cancer-negative. For an individual determined to be negative for ovarian cancer, the progress or recurrence of ovarian cancer can be continuously monitored through follow-up. In addition, the expression level of the miRNA and/or CA125 is not increased (eg, the same, similar, or decreased) compared to the normal control in the subject who has completed the surgery and treatment, or the expression level of the miRNA and/or CA125 is abnormal If it is not determined as a value, it may be determined that ovarian cancer has been cured.

7 and 8 are diagrams of a diagnosis method and a diagnosis system for ovarian cancer to which a decision tree model is applied according to an embodiment. For example, after sequentially or simultaneously measuring the expression levels of miR-200c, miR-145, and miR-93 present in the serum exosomes of an individual using RT-PCR, it is possible to determine whether or not ovarian cancer is positive according to the expression value. can Specifically, first, if the expression value (Ct) of miR-200c is 0.895 or higher, it is judged to be positive for ovarian cancer. If it is less than 0.895, the expression values of isolated miR-145 and miR-93 are measured. If it is above, it can be determined that the ovarian cancer is benign. When the expression value of miR-145 is less than 1.225, the expression value of miR-93 is measured. When the expression value of miR-93 is 1.67 or more, it is determined that the ovarian cancer is positive, and when the expression value of the miR-93 is less than 1.225, it can be determined that the ovarian cancer is negative.

Ovarian nodules or masses are one of the chief complaints of many patients in the gynecological field. However, due to the anatomical location of the ovary, biopsy is impossible and the sensitivity of existing ovarian cancer tumor markers is relatively low. Therefore, the method according to one aspect uses a small amount of serum from an ovarian cancer patient prior to surgery and uses the expression levels of CA125, exosome miR-145, miR-200c, miR-93, miR-21, and combinations thereof. can be predicted by a simpler method, and the sensitivity, specificity, and accuracy are very superior to those of existing methods, which can help in the establishment of a treatment plan for patients with high ovarian cancer risk. In addition, it has the advantage of enabling early diagnosis of ovarian cancer in normal people, predicting the progress of ovarian cancer patients before surgery, and tracking the possibility of recurrence after ovarian cancer surgery and chemotherapy.

Another aspect is an ovary comprising an agent for measuring the expression level of one or more miRNAs selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c A composition for diagnosing cancer is provided. The miRNA may be isolated from an exosome. In one embodiment, the composition may further comprise an agent for measuring the expression level of serum cancer antigen (CA). The cancer antigen may be CA125. CA125 is a high molecular weight glycoprotein, which is increased in gynecological cancers such as ovarian cancer and endometrial cancer, and can be usefully used for ovarian cancer size, staging, prognosis determination, recurrence detection, and treatment effect monitoring. In another embodiment, the agent comprises an agent for measuring the expression level of miRNA of miR-200c, miR-145 and miR-93, or the expression level of miRNA of miR-200c, and miR-21; And it may be an agent for measuring the protein expression level of CA125.

Another aspect provides a kit for diagnosing ovarian cancer comprising the composition. Specific details of the composition are the same as described above.

The kit may be a kit for diagnosing ovarian cancer. Kit is used in the sense known in the art. The kit may include, for example, an agent for measuring the expression level of miRNA as described above and items necessary for a specific use thereof. In addition, it may include a reagent necessary for its method of use together with the agent for measuring the expression level of miRNA as described above.

The kit comprises miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c for measuring the expression level of one or more miRNAs selected from the group consisting of miRNA analysis. It may include one or more other components or devices suitable for the method.

In another aspect, measuring the expression level of one or more selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c in a sample isolated from the subject ; and comparing the expression level of the miRNA with the expression level of the corresponding miRNA in a control sample. It provides a method of providing information for diagnosis of ovarian cancer.

The method according to one embodiment comprises one or more expression levels selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c in a sample isolated from an individual. Including the step of measuring.

The method according to one embodiment comprises comparing the expression level of the miRNA with the expression level of the corresponding miRNA in a control sample. For example, when the expression level of the miRNA is increased compared to the control, the step of determining that the miRNA is ovarian cancer may be further included. The expression level of the miRNA may be increased in blood isolated from an individual, and specifically, the expression level in serum may be increased.

According to the method according to one aspect, it is possible to diagnose ovarian cancer more quickly and accurately before surgery using a small amount of serum, so that ovarian cancer can be diagnosed at an early stage, unnecessary surgery can be avoided, and an optimal treatment plan can be determined. In addition, since continuous monitoring is possible after surgery and after chemotherapy is completed, recurrence of ovarian cancer can be diagnosed early. In addition, when two or more biomarkers are combined, the sensitivity of predicting ovarian cancer is excellent, which can help in surgery and treatment planning for patients at high risk of ovarian cancer.

1 shows benign ovarian cysts (Benign), borderline ovarian tumors (BOT), high-grade serous ovarian carcinoma (HGSOC) and non-high-grade serous ovarian cancer. Graph showing the expression levels of serum exosomes miR-145(A), miR-200c(B), miR-21(C) and miR-93(D) measured by the qRT-PCT method in (Non-HGSCO) patients am. Expression levels of miR-145 and miR-200c were significantly different between HGSOC and non-cancer (positive/BOT) groups. On the other hand, miR-21 and miR-93 showed significant differences between the non-HGSOC and non-cancer (positive/BOT) groups. The horizontal line represents the median.

Figure 2 is a graph showing the difference in the expression of serum exosome miRNA between non-cancer (positive / BOT) and cancer (HGSOC / non-HGSOC) groups. Expression levels of miR-145(A), miR-200(B) and miR-93(D) were significantly higher in the cancer group compared to the non-cancer group. The horizontal line represents the median.

3 shows the ROC curves that differentiate between preoperative malignant (HGSOC/non-HGSOC) and non-cancerous (benign/BOT) with CA125, serum exosome miR-145, miR-200c, miR-21 and miR-93. .

4 is a structural diagram of an ovarian cancer diagnosis system according to an embodiment.

5 is a structural diagram of an ovarian cancer diagnosis system according to an embodiment.

6 is a conceptual diagram of a method for early diagnosis of ovarian cancer using a machine learning prediction model according to an embodiment.

7 is a conceptual diagram illustrating a method for diagnosing ovarian cancer using miR-200c, miR-145 and miR-93 according to an embodiment.

8 is a conceptual diagram illustrating a method for diagnosing ovarian cancer using miR-200c, miR-21 and CA125 according to an embodiment.

Hereinafter, preferred examples are presented to help the understanding of the present invention. However, the following examples are only provided for easier understanding of the present invention, and the contents of the present invention are not limited by the following examples.

[Example]

Example 1. Preparation of patient samples and confirmation of clinical characteristics

From October 2016 to April 2018, preoperative peripheral blood was collected from 48 patients with ovarian tumors of uncertain type scheduled for surgery at Bundang CHA Hospital. After the pathological diagnosis of ovarian tumors, the 48 patients were treated with benign, borderline ovarian tumors (BOT), high-grade serous ovarian carcinomas (HGSOC) and non-high-grade ovarian tumors. It was classified into a group of non-high-grade serous ovarian carcinomas (non-HGSOC). In addition, 20 serum samples collected before surgery from HGSOC patients were provided from the Biobank of Keimyung University Dongsan Hospital and the Biobank of Inje University Paik Hospital. Clinical data such as the patient's age at initial diagnosis, FIGO stage, resection metastasis, distant metastasis, local recurrence, or systemic recurrence were obtained from electronic charts and pathology reports. This study was approved by the institutional review committee of CHA Bundang Hospital, and informed consent was obtained from all patients.

The pathological diagnoses of the 68 patients are shown in Table 1 below: benign ovarian cysts (n=10), BOT (n=10), HGSOC (n=39), and non-HGSOC (n=9). In addition, the histological types of non-HGSOC patients were clear cell carcinoma (n=3), mucinous carcinoma (n=4), endometrioid carcinoma (n=l), and low-grade serous carcinoma (n=l).

		benign ovarian cyst (n=10)	BOT (n=10)	HGSOC (n=39)	non-HGSOC (n=9)	P
age	average (range)	57.2 (35-70)	41.6 (22~69)	58.1 (35-83)	43.7 (24~65)	<0.001
Preoperative CA125 (U/mL)	Median (range)	9.1 (5.3~53.6)	17.8 (5.3~274.1)	539.5 (20.6~19805)	41.6 (10.8~465.6)	<0.001
<35 U/mL	N(%)	6 (85.7)	6 (60.0)	1(2.6)	4 (44.4)	<0.001
≥35 U/mL		1 (14.3)	4 (40.0)	38 (97.4)	5 (55.6)
FIGO Ordnance	I	NA	NA	6 (15.4%)	7 (77.8%)	0.004
	II	NA	NA	4 (10.3%)	0(0)
	Ⅲ	NA	NA	11 (28.2%)	1 (11.1%)
	IV	NA	NA	18 (46.2%)	1 (11.1%)

As shown in Table 1, HGSOC patients had the highest average age among the four groups (p=0.001). Preoperative serum CA125 levels were significantly higher in the HGSOC group than in the positive, BOT, and non-HGSOC groups (p<0.001). Based on the standard value of 35 U/mL, which is generally accepted as the upper limit of the normal range, the elevation of CA125 was 97.4% in the HGSOC group, which was more significant than in the positive (14.3%), BOT (40.0%), and non-HGSOC (55.6%) groups. was very high The FIGO stage at diagnosis was significantly different between the HGSCO group and the non-HGSOC group. Most of the non-HGSOC groups were stage I (77.8%), whereas most of the HGSOC groups were stage III or IV (74.4%; p=0.004).

Example 2. Isolation and identification of exosomes from serum

2-1. Isolation of serum exosomes

The blood samples of 68 patients prepared in Example 1 were centrifuged at 3,000 rpm for 15 minutes at room temperature. The resulting supernatant represents serum, and the serum was immediately divided into aliquots of 500 μl and stored at -80°C. After thawing the frozen serum samples, cell debris was removed by centrifugation at 2,000 x g for 30 min. Then, the supernatant containing cell-free serum was combined with a 0.2-fold volume of Total Exosome Isolation Reagent (Thermo Scientific, Worcester, MA). The serum samples were incubated at 4°C for 10 min and then centrifuged at 10,000 x g for 10 min. After removing the fresh juice, the exosome pellet was resuspended in PBS.

2-2. Identification of serum exosomes

Western blotting was performed on the expression of CD9 and CD63, which are exosome-rich cell surface glycoproteins, and it was confirmed by using exosome (precipitation protocol) whether the extracted pellet was an exosome. Pro-Prep, iNtRON Biotechnology, South Korea) was used to extract total protein, and 20 μg of protein was separated by 10% SDS-PAGE and transferred to a polyvinylidene fluoride membrane (PVDF) (GE Health care, Piscataway, NJ). After blocking with 5% steam milk at room temperature for 1 hour, the membrane was mixed with a primary antibody (anti-p-actin 1:1000 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-CD9 1:1000 (Cell) Signaling, Danvers, MA, USA) or anti-CD63 1:1000 (Abeam, Cambridge, UK)) overnight at 4° C. followed by horseradish peroxidase-conjugated anti-mouse 1 :1000 or anti-rabbit secondary antibody 1:1000 (Novus Biologicals, Littleton, CO, USA) and incubated for 1 hour at room temperature After incubation, the membrane was washed and Western Blotting Luminol reagent (Western Blotting Luminol) Reagent) (Bio-Rad, Hercules, CA, USA) was used to identify the protein.

As a result, CD9 and CD63 were easily detected in the pellets extracted from the serum of ovarian cancer patients (HGSOC and non-HGSOC). That is, it can be confirmed that the exosomes are properly purified.

Example 3. Confirmation of expression of miRNA in serum exosomes

3-1. RNA extraction from serum exosomes

Total RNA including miRNA was extracted from the exosome pellet using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Briefly, 1.0 mL of Trizol reagent and 200 mL of chloroform were added to the serum sample. After centrifugation at 4 °C, 1,200 x g for 20 min, the supernatant was incubated with 500 µL of isopropanol at -20 °C for 10 min. The mixture was centrifuged at 4 °C, 1,200 x g for 10 min to remove the supernatant, and the RNA pellet was washed with 75% ethanol and then dissolved in 12 μl of RNAse-free water. The purity of the isolated RNA was determined at OD 260/280 using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Worcester, Mass.).

3-2. Confirmation of miRNA expression in serum exosomes

Seven miR candidates (miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b) known to have increased or decreased expression in ovarian cancer tissues compared to normal tissues by searching several published papers and miR-200c) using quantitative real-time PCR to evaluate the expression level of candidate miRNAs in serum exosomal RNA. Total RNA (2 ng) was reverse transcribed using the TaqMan microRNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA, USA) and microRNA-specific stem-loop primers (part of the TaqMan microRNA Assay Kit; Applied Biosystems). The mixture was incubated at 16° C. for 30 minutes, 42° C. for 30 minutes and 85° C. for 5 minutes. qRT-PCR for mRNA was performed according to the TaqMan miR Assay protocol (Applied Biosystems Carlsbad, CA, USA) using a Bio-Rad CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA). All PCR reactions were performed in duplicate. Relative gene expression values for the target miRNA were normalized to RNU48 and calculated using the ^{2-ΔCT method.}

Among the 7 serum exosomal miRNAs measured by qRT-PCR, no amplification curves were detected for miR-141, miR-200a, and miR-200b even after 40 PCR cycles or more, regardless of patient group. That is, it can be seen that the miRNAs are expressed at a very low level in the serum exosomes.

1 shows the expression levels of miR-21, miR-93, miR-145 and miR-200c detected using qRT-PCR in serum exosomes of positive, BOT, HGSOC and non-HGSOC groups.

As a result, as shown in Figure 1A, the mean expression level of miR-145 was significantly increased in the HGSOC group compared to the positive and BOT groups (47.7-fold and 9.9-fold, p<0.001 and p=0.001, respectively), and positive There was a significant increase (29.6 fold) in the non-HGSOC group compared with the group (p=0.001). However, positive vs. BOT, BOT vs. non-HGSOC and HGSCO vs. No significant difference was found in miR-145 levels between non-HGSOCs. In addition, as shown in Figure 1B, the mean expression level of miR-200c was significantly increased in the HGSOC group compared to the positive, BOT and non-HGSOC groups (46.7-fold, 34.4-fold, 25.5-fold, respectively, p<0.001, p <0.001 and p=0.001). As shown in FIGS. 1C and 1D , the expression of miR-21 in the non-HGSOC group increased 26.4-fold and 17.4-fold, respectively, and the expression of miR-93 increased by 5.8-fold and 5.5-fold, respectively, compared to the positive and BOT groups. . On the other hand, when compared with the HGSOC group, it was confirmed that there was no significant difference.

Figure 2 is a graph showing the difference in the expression of serum exosome miRNA between non-cancer (positive / BOT) and cancer (HGSOC / non-HGSOC) groups.

As shown in Figure 2, when four groups were simply divided into non-cancer (positive and BOT, n=20) and cancer (HGSOC and non-HGSOC, n=48), miR-145, miR-200c and miR It was confirmed that the expression level of -93 was significantly higher in the cancer group (p<0.001, p<0.001, and p=0.001, respectively) ( FIGS. 2A, 2B and 2D ). On the other hand, the expression level of miR-21 in the cancer group was slightly higher than that in the non-cancer group, but it was not statistically significant (p=0.120) ( FIG. 2C ).

Example 4. Confirmation of the effect of serum exosome miRNA as a diagnostic marker compared to serum CA125

ROC curve analysis was performed to compare the effects of serum CA125 (>35 U/mL) and serum exosome miRNA for differential diagnosis of ovarian tumors (positive/BOT vs. HGSOC/non-HGSOC). Specifically, ANOVA and Bonferroni post hoc t-test were used to compare patient ages between 4 disease groups. The Kruskal-Wallis and Mann-Whitney U tests were used to compare CA125 levels and expression levels of candidate miRNAs between the four disease groups. Fisher's exact test was used to compare FIGO staging between HGSOC and non-HGSOC groups. A receiver operating characteristic (ROC) curve was generated and the area under the ROC curve (AUC) was calculated to evaluate the diagnostic performance of the marker. Sensitivity, specificity and accuracy of ovarian cancer prediction are as follows using true positive (TP), true negative (TN), false negative (FN) and false positive (FP). Calculated as:

Diagnostic Sensitivity = TP/(TP+FN)

Specificity = TN/(TN + FP)

Accuracy = (TN + TP)/(TN+TP+FN+FP).

Statistical analysis was performed using SPSS software version 21.0 (Chicago, IL). Statistical differences were considered significant when the p-value was less than 0.05.

The sensitivity, specificity and accuracy of each miRNA and CA125 >35 U/mL in ovarian cancer prediction are shown in Table 2 below.

diagnostic markers	responsiveness	specificity	accuracy
single marker
CA125+	89.5%	70.6%	84.6%
miR-145+	91.7%	75.0%	86.8%
miR-200c+	72.9%	90.0%	77.9%
combination of two markers
CA125+ and/or miR-145+	97.9%	60.0%	86.8%
CA125+ and/or miR-200c+	93.8%	70.0%	86.8%
miR-145+ and/or miR-200c+	93.8%%	65.0%	85.3%
combination of 3 markers
CA125+ and miR-145+ and/or miR-200c+	100.0%	55.0%	86.8%

As shown in Table 2, it can be confirmed that miR-145 is the best single marker in sensitivity (91.7%) and accuracy (86.8%). miR-200c showed the highest specificity (90.0%) but relatively low sensitivity (72.9%). The combination of CA125 and miR-145 gave a sensitivity of 97.9% and an accuracy of 86.8%. On the other hand, the combination of the three markers CA125, miR-145 and miR-200c showed low specificity (55.0%) but perfect sensitivity (100%).

3 shows ROC curves for predicting tumors with CA125, serum exosomes miR-145, miR-200c, miR-21 and miR-93. As shown in FIG. 3 , the AUC values (95% confidence interval; p value) of CA125, miR-145, miR-200c, miR-21 and miR-93 were 0.801 (0.662-0.940; p<0.001), 0.910, respectively. (0.840-0.980; p<0.001), 0.802 (0.698-0.906; p<0.001), 0.585 (0.444-0.725; p=0.303), and 0.755 (0.620-0.890; p=0.002). AUC values of miR-145 and miR-200c were higher than those of CA125.

That is, serum exosome miR-145 and/or miR-200c can accurately and rapidly diagnose ovarian cancer before surgery by discriminating benign lesions of ovarian cancer.

Example 5. Evaluation of sensitivity, specificity and accuracy through machine learning predictive model

5-1. Assays with miR-200c, miR-145 and miR-93

According to the result of Example 3, sensitivity, specificity, and sensitivity, specificity, and Accuracy was evaluated. First, we measured the expression levels of candidate miRNAs in serum exosomal RNA using quantitative real-time PCR for miR-200c miR-145 and miR-93. Then, using the R package tree, a decision tree model that can best distinguish positive and negative from these three candidate miRNAs was constructed. The decision tree model constructed is shown in FIG. 7 . In the early stage of building the decision tree model, it was a model in which one more step was added in FIG. 7 , but the branch was removed in the pruning step because the last step did not show much significance in cancer prediction.

As shown in Figure 7, if the expression value (Ct value) of miR-200c confirmed in quantitative real-time PCR measurement is 0.895 or higher, it is judged to be ovarian cancer positive, and if it is less than 0.895, the expression value of the two isolated candidate miRNAs is again measured. If the expression value of miR-145 was 1.225 or higher, it was judged to be ovarian cancer-positive, and if it was lower than the expression value of miR-145, the expression value of miR-93 was measured again. Thereafter, when the expression value of miR-93 was 1.67 or higher, it was judged to be positive for ovarian cancer, and when it was lower, it was judged to be negative.

Based on the above results, the predictive power of the decision tree model was evaluated for a total of 68 people including normal (3), benign (7), BOT (10), HGSOC (39), and other cancer (9) groups. was evaluated, and the results are shown in Table 3 below.

the model's predictions	the actual value of the data		Sum
	positivity	voice
predicted as positive	45	3	48
prediction by voice	3	17	20

Then, based on the values in Table 3, diagnostic sensitivity, specificity, and accuracy were calculated according to the formula of Example 4. As a result, sensitivity of 94%, specificity of 95%, and accuracy of 91% were shown. That is, the decision tree model according to an aspect has excellent specificity, specificity, and accuracy, and thus may be usefully used for predicting ovarian cancer.

5-2. Analysis with miR-200c, miR-21 and CA125

According to the results of Example 3, sensitivity, specificity and accuracy were evaluated through a machine learning prediction model in the same manner as in Example 5-1 for miR-200c, miR-21 and CA125, which exhibit high expression in the cancer group. . First, we measured the expression levels of candidate miRNAs in serum exosomal RNA using quantitative real-time PCR for miR-200c, miR-21 and CA125. Thereafter, decision tree models for miR-200c, miR-21 and CA125 were constructed in the same manner as in Example 5-1, and are shown in FIG. 8 . In the early stage of building the decision tree model, miR-145 was also included, but there was a problem of overfitting the training data in the machine learning model, so a pruning step was added. Finally, a model as shown in FIG. 8 was constructed.

As shown in FIG. 8 , when the expression value of miR-200c confirmed in quantitative real-time PCR measurement was 0.85 or more, it was determined that the expression value of ovarian cancer was positive, and when it was less than 0.85, the expression values of miR-21 and CA125 were measured again. If the CA125 expression value was 53.6 or higher, it was judged to be positive for ovarian cancer, and if it was lower, the expression value of miR-21 was measured again. Thereafter, when the expression value of miR-21 was 1.02 or higher, it was judged to be positive for ovarian cancer, and when it was less than, it was judged to be negative.

Based on the above results, the predictive power of the decision tree model was evaluated for the same 68 persons as in Example 5-1, and the results are shown in Table 4 below.

the model's predictions	the actual value of the data		Sum
	positivity	voice
predicted as positive	48	5	53
prediction by voice	0	15	15

Then, based on the values in Table 4, diagnostic sensitivity, specificity, and accuracy were calculated according to the formula of Example 4. As a result, sensitivity of 100%, specificity of 75%, and accuracy of 93% were shown. That is, the decision tree model according to an aspect has very good sensitivity and accuracy by indicating 100% sensitivity, and thus may be usefully used for predicting ovarian cancer.

Claims (11)

1. generating data by measuring the expression level of miRNA of one or more biomarkers isolated from the subject's sample;

   calculating an abnormal value from the miRNA expression level data; and

   An information providing method for diagnosing ovarian cancer using a machine learning prediction model, comprising generating cancer diagnosis information based on the abnormal value result.
2. The method according to claim 1, wherein the miRNA is at least one selected from the group consisting of miR-21, miR-93, miR-141, miR-145, miR-200a, miR-200b and miR-200c.
3. The method according to claim 1, wherein the miRNA is isolated from an exosome.
4. The method according to claim 1, further comprising the step of obtaining data by measuring the protein expression level of the tumor marker (Cancer Antigen, CA) 125.
5. The method according to claim 1, wherein the calculating of the abnormal value is to generate information on whether the expression level of the one or more biomarkers is increased compared to the control.
6. The method according to claim 1, wherein the abnormal value is when the expression value of the n-th miRNA is expressed as X _n (where X _n is a constant at which the n-th miRNA can be separated from the n+1 to n+m miRNA). The method wherein the nth miRNA is separated from the n+1th to n+m miRNAs.
7. The method according to claim 1, wherein, in the cancer diagnosis information, when the expression value of the n-th miRNA is X _n or more, it is determined that the ovarian cancer is positive.
8. The method according to claim 1, wherein the machine learning prediction model is a decision tree model.
9. generating data by measuring the expression level of miRNA of one or more biomarkers isolated from the subject's sample;

   calculating an abnormal value from the miRNA expression level data; and

   A method of diagnosing ovarian cancer using a machine learning prediction model, comprising generating cancer diagnosis information based on the abnormal value result.
10. a receiver for receiving data obtained by measuring the miRNA expression level of one or more biomarkers isolated from a sample of an individual;

    a calculation unit for calculating an abnormal value by applying the data to the learning data; and

    An ovarian cancer diagnosis system using a machine learning prediction model, comprising a generator generating cancer diagnosis information based on the abnormal value result.
11. The system according to claim 10, further comprising a monitoring unit for monitoring the expression pattern of the miRNA of the biomarker.

Images