KR102265529B1 - Method for predicting disease risk based on analysis of complex genetic information - Google Patents

Abstract

The present invention relates to a disease risk diagnosis method based on a complex genetic information network analysis, and the disease risk diagnosis method based on the complex genetic information network analysis developed according to the present invention can be performed in a small number through optimization or introduction of a learning method. It is possible to derive a stable correlation with disease from the combination of genetic information of , and provides a correlation of genetic information based on a network model. By using the correlation between genetic information and disease derived from the present invention, it is expected to be able to secure a diagnostic technology that satisfies the level of accuracy and economic feasibility that can be used in the actual medical industry.
In addition, it is expected that the biomarker derived from the present invention can be efficiently utilized for disease prognosis determination, etc. through the manufacture of medical devices including diagnostic chips and terminals and commercialization of disease diagnosis services.

Description

Method for predicting disease risk based on analysis of complex genetic information

The present invention relates to a disease risk diagnosis method based on complex genetic information network analysis.

Looking at the technological trends for disease diagnosis so far, microarray (microarray) using human-shared polymorphisms (single nucleotide polymorphism, gene copy number mutation, base insertion/deletion, etc.) of a specific gene or using expression information for the entire gene group ) or by measuring the expression change of genes and proteins using a protein chip, etc., research is being conducted in the direction of searching for genes related to specific diseases and studying the functions of genes.

However, existing research is focused on examining the relationship between a single type of sample and a disease, so the understanding of the relationship and interrelationship between various genetic information and diseases is lacking. In addition, due to the lack of a technique for analyzing the relationship between complex genetic information and disease, it is difficult to find a mutation specific for a new disease that has not been previously identified, and the accuracy of the diagnosis technique is also significantly low.

A technology for extracting biomarkers from genetic information is a method of extracting markers by statistically analyzing genetic information related to a disease. However, the conventional biomarker extraction technology is performed only within the range of existing information obtained in a bottom-up manner, and remains at the level of extracting markers based on some genetic information including genes, and There is a limit to the one-to-one relationship between diseases.

In addition, the disease diagnosis service based on the biomarker uses a method of performing the diagnosis service by calculating the degree of contribution of specific genetic information to the disease or trait. However, the conventional diagnostic service technology relies on deriving a simple relationship between one disease and one type of genetic information, and there is a problem in that it is not possible to perform a complex analysis between the disease and the genetic information, and as an additional variable, the passage of time There is a limit in that it is not possible to reflect changes in characteristics according to high-dimensional variables such as , treatment, and recurrence. For this reason, the accuracy of diagnosis is low, and there is a limitation in deriving different results depending on the type of service platform.

International Patent Publication No. 2014-052909.

In order to solve the above problems, in the present invention, disease state-specific information is derived from the relationship between complex genetic information, and optimization based on network models and artificial intelligence-based machine learning techniques are utilized to provide high An attempt was made to develop a biomarker for disease diagnosis with accuracy and a method for predicting disease risk.

To achieve this goal, the accuracy of diagnosis is improved by analyzing the relationship between complex and diverse genetic information and genetic information to understand the relationship between diseases, extracting the optimal combination of genetic information, and introducing an analysis technique based on a network model. An attempt was made to derive a biomarker that is high and economical.

The present invention is

extracting complex genetic information from samples of diseased patients and normal persons;

constructing a complex genetic information library by comparing and analyzing information between the complex genetic information;

deriving a disease state-specific biomarker by applying an optimization method or a learning method to the complex genetic information library; and

constructing a network model for predicting disease risk from the disease state-specific biomarkers and predicting the risk;

It provides a method of predicting disease risk through complex genetic information relationship analysis, including.

In addition, the present invention provides a disease state-specific biomarker derived through the disease risk prediction method through the complex genetic information relationship analysis.

The disease risk diagnosis method based on the complex genetic information network analysis in the blood developed according to the present invention can derive a stable correlation with the disease from a small number of combinations of genetic information through the introduction of a learning method, and a network model provides a correlation of genetic information based on By using the correlation between genetic information and disease derived from the present invention, it is expected that a diagnostic technology that satisfies the level of accuracy and economic feasibility that can be used in the actual medical industry is expected to be secured.

In addition, it is expected that the biomarker derived from the present invention can be efficiently utilized for disease prognosis determination, etc. through the manufacture of medical devices including diagnostic chips and terminals and commercialization of disease diagnosis services.

1 illustrates an example of a concept of a method for diagnosing disease risk based on a complex genetic information network analysis of the present invention.
2 shows the concept of step-by-step genetic information based on the gene expression process.
3 shows an example of a concept related to a method for deriving and verifying a disease state-specific biomarker using a learning method.
4 shows an example of characteristic modeling of disease state-specific biomarkers.
5 shows an example of a method for CNN analysis of protein expression data.
6 shows an example of an algorithm for predicting the risk of digestive cancer from mi-RNA information.
7 is a verification result using only basic CNN analysis.
8 is a result regarding the change of the learning result by extracting important mi-RNA candidate combinations.
9 is a result regarding the possibility of simultaneous screening and precise diagnosis confirmed for proteins.

Hereinafter, the disease risk diagnosis method based on the complex genetic information network analysis of the present invention will be described in detail with reference to the accompanying tables or drawings.

When drawings are described, they are provided as examples so that the spirit of the present invention can be sufficiently conveyed to those skilled in the art. Therefore, the present invention is not limited to the drawings presented and may be embodied in other forms, and the drawings may be exaggerated to clarify the spirit of the present invention.

At this time, if there is no other definition in the technical terms and scientific terms used, it has the meaning commonly understood by those of ordinary skill in the art to which this invention belongs, and the summary of the present invention in the following description and accompanying drawings Descriptions of known functions and configurations that may be unnecessarily obscure will be omitted.

In the present invention, "specimen sample" or "sample" refers to genetic information obtained for analysis, and is used in the same sense throughout the specification.

The present invention relates to a method for diagnosing disease risk based on analysis of a complex genetic information network in blood.

The present invention compares and analyzes general life phenomena and disease-related information based on the extracted complex genetic information, helps to understand the function of genetic information, and furthermore, derives disease state-specific biomarkers with high accuracy and A disease risk prediction model can be built.

In the present invention, in order to construct a disease state-specific biomarker derivation and disease risk prediction model, big data processing technology, artificial intelligence-based deep learning method, for example, machine learning technology ( machine learning method) may be used in combination.

Hereinafter, a method for predicting disease risk through complex genetic information relationship analysis will be described in detail.

the present invention

extracting complex genetic information from samples of diseased patients and normal persons;

constructing a complex genetic information library by comparing and analyzing information between the complex genetic information;

deriving a disease state-specific biomarker by applying an optimization method or a learning method to the complex genetic information library; and

constructing a network model for predicting disease risk from the disease state-specific biomarkers and predicting the risk;

It provides a method of predicting disease risk through complex genetic information relationship analysis, including.

Each step will be described in detail below.

First, the step of extracting complex genetic information from samples of diseased patients and normal persons will be described in detail.

In the step of extracting complex genetic information from a sample of a diseased patient or a normal person, information related to DNA, RNA, protein, etc. for the entire genome of the sample may be secured. The method of obtaining the information is not limited as long as it does not impair the purpose of the present invention, but it can be obtained from a genetic information database, etc., and more specifically, provided by the National Institutes of Health (NIH). A database can be used, and as a more specific example, information related to cancer can be obtained through the entire genome information provided by TCGA (The Cancer Genome Atlas) for each type of disease. As another example, information may be obtained by requesting genome sequencing of a patient's sample sample taken directly from a hospital or patient. As another example, a whole exome sequence set that plays a direct role in protein synthesis in a gene may be secured and used, but is not limited thereto.

In the present invention, the genome sequence information of the sample may have some changes depending on the type of the genetic information database, a device used for sequencing, a sequencing method, and the like. In addition, the genome sequence information is not limited as long as it does not impair the purpose of the present invention, but may be based on information provided by the human genome map revealed from the human genome project as an example.

In the present invention, the entire genome sequence information of samples from diseased patients and normal people can serve as the basis for detecting biomarkers according to the present invention, and cf-DNA, ct-DNA, etc. that can be obtained from such genome sequence information Analysis is performed based on differences in the genome sequence information of the sample, including DNA information, RNA expression information such as mRNA and mi-RNA, protein synthesis information, and the like. Although not limited among the entire genome sequence information, chromosome information, information related to the position of the nucleotide sequence in the chromosome, mutation information of the nucleotide sequence related to the addition, deletion or substitution of the nucleotide sequence, RNA information, protein expression information, protein 3 Information including dimensional structure and reliability may be mainly used for detection of disease diagnostic biomarkers.

In the present invention, in the analysis of the information included in the genome sequence information, addition or subtraction of information may be performed according to the type, version, and environment of the program used.

Next, the step of constructing a complex genetic information library by comparing and analyzing information between the complex genetic information will be described in detail.

In the step of constructing a complex genetic information library by comparing and analyzing the information between the complex genetic information, the complex relationship existing between the genetic information obtained in the step of extracting the complex genetic information from the samples of the disease patient and the normal person is analyzed to be related to the disease. It is possible to extract important genetic information and form a library.

In the present invention, the genetic information is not limited as long as it does not impair the purpose of the present invention, but for example, DNA information such as cf-DNA and ct-DNA related to the gene expression process, RNA expression information such as mRNA, mi-RNA, protein synthesis information (FIG. 2).

The extraction of important genetic information factors for analysis is not limited as long as it does not impair the purpose of the present invention, but may include the following process.

First, classification accuracy for a case in which a normal group and a disease group can be distinguished using a single genetic information factor can be extracted. The single genetic information factor is not limited to the type and quantity when it is possible to distinguish a normal group from a disease group only with the information, but for example, a single nucleotide polymorphism including a nucleotide sequence mutation related to addition, deletion or substitution of a nucleotide sequence mutations, gene copy number mutations, protein amino acid sequence polymorphism mutations, and the like, but are not limited thereto. For example, if mutations in the nucleotide sequence appear in common in the sample samples from the disease group and the mutation does not appear in the same sample in the sample from the normal group, the corresponding genetic information is identified, and positional information and mutation information of the nucleotide sequence are extracted. It is better to save it.

Next, by measuring the difference between the actual expression level and the reference amount for each genetic information factor, an on/off tag capable of determining whether the genetic information factor affects the selection of a disease group may be set. As an example of the method for the above setting, the reference value of the expression level for each step related to the important genetic gene expression process is Th ₁ , Th ₂ , Th ₃ , respectively, and the genetic information expression level increases or decreases by disease. In each case, it can be defined and used as an increase reference value (Th ₁ ^up , Th ₂ ^up , Th ₃ ^up ) and a decrease reference value (Th ₁ ^down , Th ₂ ^down , Th ₃ ^{down ).} By using the variables defined as described above, it is possible to extract genetic information whose expression level is changed due to a disease while satisfying each expression level criterion for the obtained specimen sample. At this time, if necessary, the nucleotide sequence information of the corresponding genetic information can be obtained and used, and single nucleotide polymorphic mutations including nucleotide sequence mutations related to addition, deletion or substitution of the above-mentioned nucleotide sequences, gene copy number mutations, etc. information on mutations in the nucleotide sequence caused by disease may be extracted by extracting mutants of , but is not limited thereto.

In the step illustrated in FIG. 2 using the extracted genetic information factor, the correlation between the complex genetic information is identified by analyzing the expression level change between the genetic information corresponding to different stages, the nucleotide sequence variation, etc. to construct a library. and can be used for the derivation of biomarkers thereafter.

[Table 1] Example of Genetic Information Relationship Analysis Library Construction

As an example of library construction through genetic information relationship analysis, as shown in Table 1 above, it was observed that mi-RNA1 _{was expressed below the reference value Th 2} ^down in male patients with stage 1 gastric cancer and male patients with stage 2 liver cancer, and at the same time protein5 was reduced to the reference value. When expressed above _{Th 3} ^up; When mi-RNA5 SNP1 was found in stage 1 gastric cancer male and gastric cancer stage 1 female, but no relationship with other specific genetic information was found; It is possible to record the genetic information for the library, etc.

The information analyzed through the above method may be converted into a specific platform, that is, in the form of the same frame as shown in Table 1 above, for example, and stored or managed.

Next, a step of deriving a disease state-specific biomarker by applying an optimization method or a learning method to the complex genetic information library will be described in detail.

In the step of deriving a disease state-specific biomarker by applying an optimization method or a learning method to the complex genetic information library, the complex genetic information library constructed through the above method is used as an optimization method or It is possible to derive biomarkers specific for a disease state by analysis through a learning method.

The method of extracting a disease state-specific biomarker candidate is not limited as long as it does not impair the purpose of the present invention, but from the derived complex genetic information library, the relationship between the same genetic information in the sample sample and the library in the disease state to be confirmed It can be confirmed whether it is established and selected as a candidate group for deriving a disease state-specific biomarker by extracting the increase/decrease relationship of genetic information, mutation information of the nucleotide sequence, and the number of genetic information from the confirmed genetic information. In the selection of the candidate group, it is preferable that the disease state-specific biomarker maximizes the accuracy of indicating the disease state, while minimizing the number of genetic information to be considered. After defining in the form of multi-variable optimization of a function, a disease state-specific marker may be derived through application of a mathematical algorithm, but is not limited thereto.

A mathematical algorithm for optimizing the multivariate function can be introduced and used without limitation as long as it is a method capable of solving the problem of the multivariate function. For example, a simulated annealing technique, a genetic algorithm, a tap search technique, a simulation Tied evolution, a stochastic evolution technique, etc. may be mentioned, and a genetic algorithm may be preferably used. In the case of extracting the disease state-specific biomarker through the above method, it is not necessary to complete the entire process, and the best solution may be used by stopping while finding the optimal solution.

The genetic algorithm is based on the basic theory of biogenetics in the natural world, and is a parallel and global search algorithm that expresses possible solutions to a problem as a data structure in a fixed form, and then gradually transforms them to create better solutions. . Here, the data structure representing the solutions is a gene, and the process of creating better solutions by modifying them can be expressed as evolution. In other words, the genetic algorithm can be called a simulated evolution search algorithm to find a solution x that optimizes an unknown function Y = f(x). A genetic algorithm is closer to an approach to solving a problem rather than an algorithm to solve a specific problem, and can be applied to all problems that can be expressed in a form that can be used in a genetic algorithm. In general, when a problem is too complex to be calculated, it is preferable because it can be approached as a method to obtain an answer close to the optimal solution even if the actual optimal solution cannot be obtained through a genetic algorithm.

As an example of a method for deriving a disease state-specific biomarker, as shown in FIG. 3 , a learning sample to be analyzed and a verification sample for verifying the accuracy of the learning sample may be provided. The analysis may include only genetic information specific to the disease state, but is not limited thereto, and as performed in an embodiment of the present invention, the analysis target library is arbitrarily separated into a learning sample and a verification sample to perform learning and repeating the learning process several times to improve accuracy.

When the size of the library is large, it is preferable to perform a process to reduce the complexity because it is difficult to calculate the classification accuracy for all subsets and the complexity increases. If the size of the library is N, the number of all subsets is 2^N cases. Accordingly, as the size of the library increases, it is difficult to calculate classification accuracy for all subsets and the complexity increases. In order to solve this problem, it is necessary to reduce the complexity by using, for example, a heuristic algorithm. As an example, when the size of the subset is N, the possibility of the marker is checked and only the case with the highest probability is prioritized and the size of the set is gradually reduced, the overall case for the marker to be investigated is reduced to N(N+1)/2.

The selection of genetic information, which is a variable for optimizing the multivariate function, is not limited as long as it does not impair the purpose of the present invention, but for example, genetic information may be arbitrarily selected according to the heuristic algorithm, and preferably, genetic information having maximum accuracy You can choose any combination of information. For example, if there is a characteristic that the genetic information mi-RNA1 and ct-DNA5 increase at the same time, information related to the increase and decrease of each expression level of mi-RNA1 and ct-DNA5 is used for learning and converted into two features. It can be used, and whether the mi-RNA1 and ct-DNA5 simultaneously increase characteristics exist in the sample can be used as a feature in learning.

In the present invention, the artificial intelligence-based learning method used for the learning for biomarker derivation is not limited in its type as long as it does not impair the purpose of the present invention, but as an example, a neural network , deep learning, etc. may be used, and an example corresponding to the neural network network may include a convolutional neural network (CNN), a recurrent neural network (RNN), and the like, and as an example, in an embodiment of the present invention CNN can be used, but it is not limited thereto, and an appropriate learning technique can be selected and used according to the characteristics of the obtained data and biomarkers.

In the present invention, preferably, a process for verifying the performance of the disease state-specific biomarker derived through the above method may be further performed. For this, it is better to verify the accuracy of the derived biomarker by applying the derived disease state-specific biomarker to a sample not used for biomarker detection or a normal sample, and then calculating the classification accuracy.

Next, the step of constructing a network model for predicting disease risk from the disease state-specific biomarker and predicting the risk will be described in detail.

In the step of constructing a network model for predicting disease risk from disease state-specific biomarkers, a disease derived using a complex genetic information library obtained through analysis of the relationship between the complex genetic information and an optimization technique or a learning method State changes such as the invention, progression, and recurrence of a disease can be constructed in the form of a network from the state-specific biomarkers.

The method for constructing the network is not limited as long as it does not impair the purpose of the present invention, but using the genetic information library constructed through the method, changes in information in disease state-specific biomarkers derived according to changes in a specific disease state may include a method for analyzing As an example of the analysis, as shown in FIG. 4 , the discontinuous expression change of ct-DNA1 or mi-RNA5, which is genetic information, can be tracked and modeled in the form of a mathematical function. The form of the mathematical function is not particularly limited, but for example, it is preferable to select a regression function that can approximately satisfy the data of discontinuous expression change.

The regression analysis method used to construct a regression function is largely divided into simple regression analysis and multiple regression analysis. Simple regression analysis is used to analyze the relationship between one dependent variable and one independent variable, and multiple regression analysis is one dependent variable analysis. It can be used to elucidate the relationship between a variable and several independent variables. The expression change illustrated in FIG. 4 is composed of one dependent variable and one independent variable, and a regression function can be obtained through simple regression analysis. For example, the expression of ct-DNA1 in FIG. 4 is an exponential function (exponential function). ), the expression of mi-RNA5 can be modeled as a regression function in the form of a step function.

After mathematically modeling the characteristics of disease state-specific biomarkers through the above method, genetic information, which is a network model for disease risk prediction made of genetic information, to track the change process of genetic information according to changes in the major state of the disease A genetic-information relation network model can be established.

The form of the genetic information relationship network model is not limited as long as it does not impair the purpose of the present invention, but a static disease network to which only correlations between complex genetic information are applied or individual-specific genetics such as time course, habits, etc. It may be in the form of a dynamic disease network in which information is added as a variable, preferably in the form of a dynamic disease network. By using the above type of network model, it is preferable to track continuously changing genetic information characteristics and to diagnose and predict diseases.

In the present invention, the accuracy of the genetic information relation network model, which is a network model for predicting the biomarker and the disease risk, is not limited, but may be evaluated using the following indicators.

- Sensitivity: A measurement index that evaluates whether a patient with a disease is properly classified. It can be defined as TP/(TP+FN) to prevent misdiagnosis-based diagnosis failure, where TP is a disease-causing patient The number of cases classified as , FN is the number of cases in which a diseased patient is classified as normal. For a network model for predicting biomarkers and disease risk, preferably 95% or more, more preferably 99% or more, and most preferably 99.9% or more, can increase the test cost and commercialization possibility, and provide key genetic information related to many diseases. It is good to increase the number of cases that can be confirmed with one inspection by using

- Specificity: A measurement index that evaluates whether a normal person is properly classified. It is defined as TN/(TN+FP) to prevent unnecessary follow-up tests due to a false diagnosis of a disease, where TN is a measure of classifying normal people as normal. The number of cases, FP, is the number of cases in which a normal person is classified as a diseased patient. For network models for biomarker and disease risk prediction, preferably 90% or more, more preferably 95% or more, and most preferably 99% or more can increase the cost of testing and commercialization, and provide key genetic information related to many diseases. It is good to increase the number of cases that can be confirmed with one inspection by using

In predicting disease risk, sensitivity or specificity may be used, or a combination of the two may be used. Of these, sensitivity is more important than specificity in predicting disease risk, so it is better to use them together.

In the present invention, the disease may be any disease as long as it is possible to derive a biomarker, for example, may be a disease requiring rapid diagnosis, such as cancer, and more specifically, the cancer is bladder cancer (Bladder urothelial carcinoma), breast cancer (Breast invasive carcinoma), cervical cancer (Cervical and endocervical cancers), colorectal cancer, colon cancer (Colon adenocarcinoma), esophageal carcinoma (Esophageal carcinoma), Glioblastoma multiforme, Head and Neck squamous cell carcinoma, atrophic Kidney Chromophobe, Kidney renal clear cell carcinoma, Kidney renal papillary cell carcinoma, Acute Myeloid Leukemia, Benign brain tumor (Brain Lower Grade) Glioma, Liver hepatocellular carcinoma, Lung adenocarcinoma, Lung squamous cell carcinoma, Ovarian serous cystadenocarcinoma, Pancreatic adenocarcinoma, Pheochromocytoma and Paraganglioma), prostate cancer (Prostate adenocarcinoma), rectal cancer (Rectum adenocarcinoma), sarcoma (Sarcoma), malignant melanoma (Skin Cutaneous Melanoma), stomach cancer (Stomach adenocarcinoma), testicular cancer (Testicular Germ Cell Tumors), Thyroid carcinoma, Thymoma and Uterine Corpus Endomet rial Carcinoma), preferably bladder cancer, breast cancer, colorectal cancer, colon cancer, cervical cancer, liver cancer, lung adenocarcinoma, anoxic renal cell carcinoma, clear cell type renal cell carcinoma, papillary epithelial cancer It may be one or more selected from the group consisting of cell cancer, serous ovarian cancer, prostate cancer, lung squamous cell carcinoma and gastric cancer, and more preferably, may be one or more selected from the group consisting of breast cancer, colorectal cancer and gastric cancer. It is not limiting.

In addition, the present invention provides a disease state-specific biomarker derived through the disease risk prediction method through the complex genetic information relationship analysis.

It is expected that the biomarker derived from the present invention can be efficiently utilized for disease prognosis determination, etc. through the manufacture of medical devices including diagnostic chips and terminals and commercialization of disease diagnosis services.

Hereinafter, the content of the present invention will be described in more detail through examples. The examples are only for explaining the present invention in more detail, and the scope of the present invention is not limited thereto.

[Experimental material]

1. The following mi-RNA related data were obtained and used.

1) GSE54397

It was provided and used from the research database of Professor Na-Young Kim, Seoul National University.

In the database, 3523 types of mi-RNA data among the microarray data of normal tissue and cancer tissue samples for 16 gastric cancer patients were provided and used.

2) GSE61741

It was provided from the database of Saarland University and used.

In the database, among microarray data of a total of 1049 blood samples including cancer patients and normal persons, there are 13, 29, and 94 samples of gastric cancer patients, colorectal cancer patients and normal persons, respectively, for a total of 136 samples of 848 types of mi-RNAs. Data was provided and used.

3) TCGA NGS data

It was provided from the TCGA database and used.

From the database, 45 and 446 samples of 491 normal and cancerous tissues of gastric cancer patients were downloaded, respectively, and miRNA and NGS read information was obtained.

A total of 211 mi-RNA data were used.

2. Data other than the following mi-RNA was obtained and used.

1) TCGA protein expression array data

It is a database of protein expression levels for patients with breast cancer, thyroid cancer, liver cancer, kidney cancer, lung cancer, and normal patients.

Breast cancer disease, breast cancer normal, thyroid cancer, liver cancer, kidney cancer 1 (Kidney Renal Clear Cell Carcinoma), kidney cancer 2 (Kidney Renal Papillary Cell Carcinoma), kidney cancer 3 (Kidney Chromophobe), lung cancer 1 (Lung Adenocarcinoma) Lung Squamous Cell Carcinoma (Lung Squamous Cell Carcinoma) data were obtained,

The number of samples is 1078, 45, 426, 183, 478, 215, 63, 365, and 327, respectively.

There is expression data for about 200 proteins in each sample, and only 146 protein data commonly present in all samples were extracted and used.

[Table 2] Examples of mi-RNA and protein expression data

[Example 1] Application of learning techniques to protein data

For protein data, a relationship between input data was derived using the CNN method of FIG. 5, passed through a fully connected layer, and finally classified using softmax.

[Example 2] Accuracy prediction

Learning was performed by applying the same CNN network method as in Example 1.

1. GSE54397

Learning was performed with 22 samples out of 32 samples, and verification was performed with the remaining 10 samples.

2. GSE61741

Learning was performed with 106 samples out of 136 samples, and validation was performed with the remaining 30 samples.

3. TCGA NGS data

Learning was performed with 391 samples out of 491 samples, and validation was performed with the remaining 100 samples.

The verification results are shown in FIGS. 7 and 8 .

According to the results of FIG. 7, as the learning progresses, the GSE54397 model (tissue, microarray data) shows 100% classification accuracy, GSE61741 (blood, microarray data) is about 96.67%, and the TCGA NGS data is about 99%. It was confirmed that all of them showed very high accuracy of 95% or more by converging to the classification accuracy.

According to FIG. 8, it was confirmed that the sensitivity as the extraction process of important mi-RNAs converges to 1 for 848 mi-RNAs and 30 optimal mi-RNA extraction (BEST) cases as learning progresses. In the case of specificity, fluctuations progressed near 1 while learning was progressing for 848 mi-RNAs, and convergence to more than 0.95 for 30 optimal mi-RNA extraction (BEST) cases. .

[Example 3] Derivation of biomarkers using clinical data

The data for three diseases of breast cancer, stomach cancer, and colorectal cancer were all obtained from the database of the TCGA (The Cancer Genome Atlas) project, which has been in progress since 2006 at the NIH in the United States. The detailed database names used to secure each disease data are as follows.

Breast cancer: TCGA-BRCA, stomach cancer: TCGA-STAD, colorectal cancer: TCGA-COAD.

Among them, for mi-RNA genetic information data, 30 types of biomarkers were derived for each cancer type through learning through the CNN network method in the same manner as in the above example.

The results are shown in Table 3 below.

Optimal mi-RNA biomarker for each type of cancer (BEST) Cancer type mi-RNA biomarkers breast cancer 'hsa-mir-30d', 'hsa-mir-145', 'hsa-mir-425', 'hsa-mir-203a', 'hsa-mir-452', 'hsa-mir-378a', 'hsa' -mir-455', 'hsa-mir-100','hsa-mir-199b', 'hsa-mir-205', 'hsa-mir-542', 'hsa-mir-532','hsa-mir -625', 'hsa-mir-200c', 'hsa-mir-183', 'hsa-mir-22', 'hsa-mir-451a', 'hsa-mir-30a', 'hsa-mir-30e' ', 'hsa-mir-148a', 'hsa-mir-143', 'hsa-mir-375', 'hsa-mir-584', 'hsa-mir-379', 'hsa-mir-10a', 'hsa-mir-182', 'hsa-mir-21', 'hsa-mir-486-1', 'hsa-mir-486-2', 'hsa-mir-10b' colorectal cancer 'hsa-mir-6086', 'hsa-mir-3118-1', 'hsa-mir-1321', 'hsa-mir-548f-5', hsa-let-7c', 'hsa-mir-4752' , 'hsa-mir-183', 'hsa-mir-29a', 'hsa-mir-30e', 'hsa-mir-486-1', 'hsa-mir-194-1', 'hsa-mir- 194-2', 'hsa-mir-30a', 'hsa-mir-28', 'hsa-mir-25', 'hsa-mir-486-2', 'hsa-mir-182', 'hsa- mir-30d', 'hsa-mir-203a', 'hsa-mir-10b', 'hsa-mir-148a', 'hsa-mir-145', 'hsa-mir-378a', 'hsa-mir- 143', 'hsa-mir-22', 'hsa-mir-10a', 'hsa-mir-200c', 'hsa-mir-21', 'hsa-mir-192', 'hsa-mir-375' stomach cancer 'hsa-mir-500b', 'hsa-mir-496', 'hsa-mir-2392', 'hsa-mir-5739', 'hsa-mir-4540', 'hsa-mir-6749', 'hsa' -mir-1915', 'hsa-mir-202', 'hsa-mir-2467', 'hsa-mir-27b', 'hsa-mir-583', 'hsa-mir-374c', 'hsa-mir' -219b', 'hsa-mir-299', 'hsa-mir-142', 'hsa-mir-30d', 'hsa-mir-3074', 'hsa-mir-147b', 'hsa-mir-5009' ', 'hsa-mir-624', 'hsa-mir-181d', 'hsa-mir-489', 'hsa-mir-581', 'hsa-mir-29b-2', 'hsa-mir-541' ', 'hsa-mir-485', 'hsa-mir-4519', 'hsa-mir-20b', 'hsa-mir-486-1', 'hsa-mir-527'

From the above, there are 11 types of mi-RNA biomarkers common to breast cancer, colorectal cancer and gastric cancer, which can be interpreted as biomarkers having common characteristics in the three cancers.

Biomarkers common to three cancers Cancer type Common mi-RNA biomarkers Breast, colon and stomach cancer 'hsa-mir-143', 'hsa-mir-148a', 'hsa-mir-182', 'hsa-mir-203a', 'hsa-mir-21', 'hsa-mir-22', 'hsa' -mir-30a', 'hsa-mir-30e', 'hsa-mir-375', 'hsa-mir-486-1', 'hsa-mir-486-2'

Among the biomarkers derived from the data of Example 3 through the analysis method of the present invention, from the biomarkers common to three cancers, it is known that the hsa-mir-486 family has a relationship with other cancers, and hsa The -mir-375 family is known to be associated with circulating in the body, and the hsa-mir-30 family is known to be associated with the inhibition of cancer.

That is, from the method of the present invention, it was confirmed that the biomarker previously identified as a factor related to cancer was accurately extracted as a major factor, and it was also confirmed that the conclusion was correct with respect to the existing known results.

In addition, it can be seen that individual cancer-specific biomarkers other than the above-described biomarkers are novel biomarkers for individual cancer diagnosis.

[Example 4] Prediction of accuracy of biomarkers derived using clinical data

The results of the disease risk prediction calculation using the biomarkers derived in the same way as in Examples 1 and 2 are as follows.

The measurement results of each sensitivity and specificity were performed through 100-fold cross validation technique so that they become general results for the algorithm itself, not specific results for a specific training set.

The risk prediction algorithm consists of a convolutional neural network, and consists of 7 convolutional layers and 4 fully connected layers.

All convolutional layers are composed of 1-dimensional filters, the first layer is 20 by 1, the second layer is 10 by 1, and the third and subsequent layers use a filter of 3 by 1.

Padding used 'Valid' technique.

The fully connected layer consists of 1024, 512, 256, and 128 nodes, and the last is configured to classify the disease probability using the readout layer and softmax activation.

The results are shown in Table 5.

Disease risk prediction results by cancer type Cancer type sensitivity specificity breast cancer 98.0% 95.5% colorectal cancer 99.3% 96.0% stomach cancer 99.0% 96.2%

From the above results, the disease risk prediction method through the complex genetic information relationship analysis through the present invention not only provides a biomarker with high accuracy of 95% or more, but also exhibits sensitivity and specificity of 95% or more. From this, through the present invention, the test cost can be reduced and the commercialization possibility can be increased, and it is confirmed that the number of cases that can be confirmed with a single test by utilizing the main genetic information related to a number of diseases increases, thereby satisfying the commercially available level of accuracy and economic feasibility The present invention was completed by showing that it can be used as a diagnostic technology.

Claims (14)

extracting complex genetic information from samples of diseased patients and normal persons;
constructing a complex genetic information library by comparing and analyzing information between the complex genetic information;
deriving a disease state-specific biomarker by applying an optimization method or a learning method to the complex genetic information library; and
constructing a network model for predicting disease risk from the disease state-specific biomarkers and predicting the risk;
As a disease risk prediction method through complex genetic information relationship analysis,
The complex genetic information library is derived and constructed through statistical analysis or optimization method,
When constructing the complex genetic information library, by measuring the difference between the actual expression level and the reference amount for each genetic information factor, setting an on/off tag that can determine whether the genetic information factor affects the selection of disease groups A method of predicting disease risk by analyzing the relationship between phosphorus and complex genetic information. The method of claim 1,
The complex genetic information is any one or two or more expression or synthesis information selected from the group consisting of DNA, RNA and protein, disease risk prediction method through complex genetic information relationship analysis. delete delete The method of claim 1,
The on/off tag setting is
_{a) Th 1} , Th ₂ , and Th ₃ are the expression levels for each step related to important genetic gene expression processes, respectively, and when the expression level of genetic information increases or decreases due to disease, the reference value of increase (Th ₁ ^up , defining variables as Th ₂ ^up , Th ₃ ^up ) and reduction thresholds (Th ₁ ^down , Th ₂ ^down , Th ₃ ^{down );} and
b) extracting genetic information whose expression level is changed by disease while satisfying each expression level criterion with respect to the specimen sample using the variable;
A method for predicting disease risk through complex genetic information relationship analysis, characterized in that it comprises a.
6. The method of claim 5,
extracting the mutant containing single-nucleotide polymorphic mutation or gene copy number mutation including addition, deletion or substitution of nucleotide sequence by securing the nucleotide sequence information of the genetic information when extracting the genetic information;
Disease risk prediction method through complex genetic information relationship analysis, characterized in that it further comprises. The method of claim 1,
By analyzing the relationship between the complex genetic information and the disease existing in the complex genetic information library through an optimization technique or a learning technique, biomarkers that can be used for disease analysis are derived, through complex genetic information relationship analysis How to predict disease risk. The method of claim 1,
A method for predicting disease risk through complex genetic information relationship analysis, characterized in that a static disease network model is constructed based on the disease state-specific biomarker. The method of claim 1,
A method of predicting disease risk through complex genetic information relationship analysis, characterized in that a network model for disease risk prediction is built and a dynamic disease network model is constructed in the step of predicting the risk. 8. The method of claim 7,
The optimization technique is a method for predicting disease risk through complex genetic information relationship analysis, characterized in that selected from the group consisting of a simulated annealing technique, a genetic algorithm, a tap search technique, a simulated evolution, and a stochastic evolution technique. 8. The method of claim 7,
The learning technique is a method for predicting disease risk through complex genetic information relationship analysis, characterized in that selected from the group consisting of a neural network and deep learning. 12. The method of claim 11,
The neural network network is a method for predicting disease risk through complex genetic information relationship analysis, characterized in that selected from the group consisting of a convolutional neural network (CNN) and a recurrent neural network (RNN). The method of claim 1,
The method of predicting disease risk through complex genetic information relationship analysis, characterized in that the accuracy of the network model for predicting the disease risk is 95% or more in sensitivity and 90% or more in specificity. As a disease state-specific biomarker derived through the disease risk prediction method through the complex genetic information relationship analysis of any one of claims 1, 2 and 5 to 13,
The disease state is breast cancer, colorectal cancer or stomach cancer,
The breast cancer-specific biomarkers are hsa-mir-30d, hsa-mir-145', hsa-mir-425, hsa-mir-203a, hsa-mir-452, hsa-mir-378a, hsa-mir-455, hsa-mir-100, hsa-mir-199b, hsa-mir-205, hsa-mir-542, hsa-mir-532, hsa-mir-625, hsa-mir-200c, hsa-mir-183, hsa- mir-22,hsa-mir-451a, hsa-mir-30a, hsa-mir-30e, hsa-mir-148a,hsa-mir-143, hsa-mir-375, hsa-mir-584, hsa-mir- 379, hsa-mir-10a, hsa-mir-182, hsa-mir-21, hsa-mir-486-1, hsa-mir-486-2, and any one selected from the group consisting of hsa-mir-10b or two or more,
The colorectal cancer-specific biomarkers are hsa-mir-6086, hsa-mir-3118-1, hsa-mir-1321, hsa-mir-548f-5, hsa-let-7c, hsa-mir-4752, hsa- mir-183, hsa-mir-29a, hsa-mir-30e, hsa-mir-486-1, hsa-mir-194-1, hsa-mir-194-2, hsa-mir-30a, hsa-mir- 28, hsa-mir-25, hsa-mir-486-2, hsa-mir-182, hsa-mir-30d, hsa-mir-203a, hsa-mir-10b, hsa-mir-148a, hsa-mir- 145, hsa-mir-378a, hsa-mir-143, hsa-mir-22, hsa-mir-10a, hsa-mir-200c, hsa-mir-21, hsa-mir-192, and hsa-mir-375 Any one or two or more selected from the group consisting of
The gastric cancer-specific biomarkers are hsa-mir-500b, hsa-mir-496, hsa-mir-2392, hsa-mir-5739, hsa-mir-4540, hsa-mir-6749, hsa-mir-1915, hsa -mir-202, hsa-mir-2467, hsa-mir-27b, hsa-mir-583, hsa-mir-374c, hsa-mir-219b, hsa-mir-299, hsa-mir-142, hsa-mir -30d, hsa-mir-3074, hsa-mir-147b, hsa-mir-5009, hsa-mir-624, hsa-mir-181d, hsa-mir-489, hsa-mir-581, hsa-mir-29b -2, any one selected from the group consisting of hsa-mir-541, hsa-mir-485, hsa-mir-4519, hsa-mir-20b, hsa-mir-486-1, and hsa-mir-527; two or more disease state specific biomarkers.

Images