KR20210044400A - Method and apparatus for discovering biomarker for predicting cancer prognosis using heterogeneous platform of DNA methylation data - Google Patents

Abstract

The present invention relates to a biomarker discovery method for predicting the prognosis of cancer using DNA methylation data of a heterogeneous platform and a device thereof. According to an aspect, the method and the device increase the accuracy of a cancer prognosis prediction model using DNA methylation data of a heterogeneous platform. The method and the device can derive methylation markers related to the prognosis of cancer and develop a cancer prognosis prediction kit with higher accuracy.

Description

[Method and apparatus for discovering biomarker for predicting cancer prognosis using heterogeneous platform of DNA methylation data}

The present invention relates to a method and apparatus for discovering biomarkers for predicting cancer prognosis using DNA methylation data of heterogeneous platforms.

Epigenetics is a field that studies the regulation of gene expression in a state where the nucleotide sequence of DNA is unchanged. Epigenetics studies the regulation of gene expression through epigenetic mutations such as DNA methylation, miRNA or histone acetylation, methylation, phosphorylation and ubiquitination. Of these, DNA methylation is the most studied epigenetic mutation. Epigenetic mutations can lead to mutations in gene function and changes to tumor cells. Therefore, DNA methylation is associated with the expression or suppression of disease-regulating genes in cells, and recently, a number of studies on a method of diagnosing the risk or prognosis of cancer through DNA methylation measurement have been reported (Korean Patent Laid-Open No. 10-2014- 0022293), researches using machine learning are being conducted.

On the other hand, when creating a model using methylation data, if the number of methylation sites is overwhelmingly larger than the number of data, the training data is overtrained, resulting in excessively high accuracy for the training data that trains the model, and low accuracy for the test data. The problem of overfitting occurs. For this reason, only a few well-known or experimentally verified gene methylation are used as markers. In order to solve the overfitting problem, a larger number of data must be secured, but a large amount of data is incurred to secure a large number of data from the platform. Therefore, there is a need for a method of improving the accuracy of the model by solving the overfitting problem without increasing the cost.

It is to provide a method and apparatus for discovering methylation markers for predicting cancer prognosis using DNA methylation data from heterogeneous platforms. The technical problem to be achieved by the present embodiment is not limited to the technical problems as described above, and other technical problems may be inferred from the following embodiments.

One aspect is a method of discovering a methylation marker for predicting cancer prognosis by using machine learning by a device operated by at least one processor, wherein the at least one processor is important for predicting cancer prognosis among the first methylation data. A first step of extracting a CpG site having a significant correlation with a variable; A second step of selecting, by the at least one processor, data corresponding to the extracted CpG site from among the second methylation data as final input data for learning; And a third step of generating, by the at least one processor, a cancer prognosis prediction model using the final training input data.

In another aspect, prior to the first step, the at least one processor derives a variable by reducing the dimension of the first methylation data; Generating, by the at least one processor, a test model for predicting a prognosis of cancer through machine learning using the derived variable as input data for learning and clinical data of a cancer patient as output data for learning; And determining, by the at least one processor, the performance of the test model to determine one or more variables important for predicting the prognosis of cancer.

In another aspect, the first methylation data or the second methylation data may be array-based methylation data or bisulfite sequencing data.

In another aspect, the step of deriving the variable by reducing the dimension includes principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization NMF. ) May be performed using any one method selected from the group consisting of.

In another aspect, the step of generating the cancer prognosis prediction test model may be generated using a bootstrap algorithm.

In another aspect, in the step of deriving one or more variables important for predicting the prognosis of cancer, one or more variables important for predicting the prognosis of cancer may be derived from a model showing maximum performance by performing cross-validation.

Another aspect provides a computer-readable recording medium storing a program for executing the method on a computer.

Another aspect is a CpG site extraction unit for extracting a CpG site having a significant correlation with a variable important for predicting the prognosis of cancer from the first methylation data; A final training input data selection unit for selecting data corresponding to the extracted CpG site from among the second methylation data as final training input data; And a model generator for generating a cancer prognosis prediction model using the final training input data.

In yet another aspect, a variable derivation unit for deriving a variable by reducing the dimension of the first methylation data; A test model generator configured to generate a cancer prognosis prediction test model through machine learning by using the derived variables as input data for learning and clinical data of a cancer patient as output data for learning; And It may further include an important variable determining unit that evaluates the performance of the test model and determines one or more variables that are important for predicting the prognosis of cancer.

In another aspect, the first methylation data or the second methylation data may be array-based methylation data or bisulfite sequencing data.

In another aspect, the dimensional reduction is any one selected from the group consisting of principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization NMF. It can be done using one method.

In another aspect, the test model generation may be generated using a bootstrap algorithm.

In another aspect, the determination of one or more variables important for predicting the prognosis of cancer may perform cross-validation to determine one or more variables important for predicting the prognosis of cancer from a model showing maximum performance.

According to a method and apparatus according to an aspect, by improving the accuracy of a cancer prognosis prediction model using DNA methylation data from heterogeneous platforms, it is possible to derive methylation markers related to cancer prognosis prediction and cancer prognosis with higher accuracy. Predictive kits can be developed.

1 is a block diagram illustrating a hardware configuration of a computing device for discovering a methylation marker for predicting a prognosis of cancer using machine learning according to an exemplary embodiment.
2 is a block diagram showing a detailed hardware configuration of the processor of FIG. 1.
3 is a flowchart of a method of discovering methylation markers for predicting cancer prognosis using machine learning according to an exemplary embodiment.
4 is a graph illustrating a process of determining one or more variables important for predicting a prognosis of cancer through evaluation of a test model performance according to an exemplary embodiment.
FIG. 5 shows (A) a cancer prognosis prediction model generated using Bisulfite sequencing data as input data for learning, and (B) bisulfite sequencing for a CpG site extracted by processing array-based methylation data according to an embodiment. This is a graph comparing and evaluating the average prediction accuracy for each cancer stage of a cancer prognosis prediction model generated using data as input data for learning.

The terms used in the embodiments were selected from general terms that are currently widely used as possible while considering the functions in the embodiments, but this varies depending on the intention or precedent of a technician engaged in the relevant technical field, the emergence of new technologies, etc. I can. In addition, in a specific case, there are terms that are arbitrarily selected, and in this case, the meaning will be described in detail in the description of the corresponding embodiment. Accordingly, the terms used in the present embodiments should be defined based on the meaning of the term and the contents of the present embodiments, not a simple name of the term.

In the descriptions of the embodiments, when a part is said to be connected to another part, this includes not only the case that it is directly connected, but also the case that it is electrically connected with another component interposed therebetween. . In addition, when a part includes a certain component, it means that other components may be further included, rather than excluding other components unless specifically stated to the contrary. In addition, the terms "... unit" and "... module" described in the embodiments mean a unit that processes at least one function or operation, which is implemented in hardware or software, or by a combination of hardware and software. Can be implemented.

Terms such as "consisting" or "included" used in the present embodiments should not be construed as necessarily including all of the various constituent elements or various steps described in the specification, and some constituent elements or It should be construed that some steps may not be included, or may further include additional elements or steps.

The description of the following embodiments should not be construed as limiting the scope of the rights, and what those skilled in the art can easily infer should be construed as belonging to the scope of the embodiments. Hereinafter, embodiments for illustration only will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating a hardware configuration of a computing device for discovering a methylation marker for predicting a prognosis of cancer using machine learning according to an exemplary embodiment.

Referring to FIG. 1, the computing device 50 pre-processes/processes/analyzes the first methylation data 40, the second methylation data 41, and the clinical information data 42 of a cancer patient. A device for discovering a marker for predicting the prognosis of, may include a data interface 80, a memory 90, and a processor 100. Meanwhile, in the computing device 50 illustrated in FIG. 1, only components related to the present exemplary embodiment are shown in order to prevent the features of the present exemplary embodiment from being blurred. In addition to the elements, other general-purpose components may be further included.

The terms “cancer”, “tumor” or “malignant” generally refer to or describe the physiological condition of a mammal characterized by unregulated cell growth. Examples of cancer may include, but are not limited to, carcinoma, lymphoma, leukemia, blastoma and sarcoma.

The term "prognosis" refers to predicting the course and outcome of a disease, and may include predicting the recurrence progression of cancer.

The term "methylation" refers to a change in gene expression pattern by attaching a methyl group to a base constituting DNA, and may mean that it occurs in the cytosine of a specific CpG site of a specific gene.

The first methylation data 40 or the second methylation data 41 may be obtained experimentally. The methylation data can be obtained experimentally at any time point by methods or apparatus that are apparent to those of ordinary skill in the art. For example, PCR, methylation specific PCR, real time methylation specific PCR, MethyLight PCR, MehtyLight digital PCR, EpiTYPER, PCR using methylated DNA specific binding protein, quantitative PCR, DNA It may be performed by any one method selected from the group consisting of chip, pyro-sequencing, bisulfite sequencing, and microarray, but is not limited thereto.

Furthermore, the first methylation data or the second methylation data may be obtained from a public database (DB). For example, it may be obtained from a database (DB) already known in the art, such as National Center for Biotechnology Information (NCBI), Gene Expression Omnibus (GEO), European Bioinformatics Institute databases, or European Nucleotide Archive. In an embodiment, the first methylation data or the second methylation data may be obtained from published array-based methylation data or bisulfite sequencing data. However, since new DNA methylation data are continuously discovered and updated due to the development of gene analysis technology, the DNA methylation data to be described in this example is not limited to those that can be obtained from the public database (DB).

Array-based methylation data is one of Illumina's methods for evaluating the methylation level, and can be obtained by measuring the intensity of the methylated and non-methylated beads of the Illumina Infinium HumanMethylation 450 BeadChip.

Bisulfite sequencing data is converted to uracil by treating bisulfite prior to sequencing to confirm the methylation pattern.At this time, the methylated cytosine can be treated with bisulfite. Since it still remains cytosine, these differences can be used to assess the level of methylation.

The data interface 80 acquires first methylation data 40, second methylation data 41, and clinical information data 42 of cancer patients, which are experimentally measured from a biological sample or stored in a database (DB). do. That is, the data interface 110 may be implemented as hardware of a wired/wireless network interface for the computing device 50 to communicate with other external devices.

The memory 90 is hardware for storing data to be processed and results of processing in the computing device 50, memory chips such as random access memory (RAM), read only memory (ROM), or hard disk (HDD). drive), solid state drive (SSD), and the like. That is, the memory 90 may store the first methylation data 40, the second methylation data 41, and the clinical information data 42 of a cancer patient acquired by the data interface 80, and the processor ( 100) can also be stored for the methylation marker discovered.

The processor 100 generates a model for predicting cancer prognosis using the first methylation data 40, the second methylation data 41, and the cancer patient's clinical information data 42, and derives a methylation marker therefrom. It corresponds to the hardware for discovering methylation markers for. The processor 100 is a module implemented with one or more processing units, and may be implemented as a combination of a microprocessor having an array of a plurality of logic gates and a memory module in which a program executable in the microprocessor is stored. The processor 100 may be implemented in the form of a module of an application program.

The methylation marker discovered by the processor 100 is transmitted to an external device, for example, a display device, another computing device, or the like through the data interface 80, or an external network, for example, the Internet, a public database ( DB) can be transmitted.

FIG. 2 is a block diagram showing a detailed hardware configuration of the processor of FIG. 1.

2, the processor 100 includes a variable derivation unit 101, a test model generation unit 102, an important variable determination unit 103, a CpG site extraction unit 104, and an input data selection unit for final learning. (105), and a model generation unit (106) may be included. Meanwhile, in the processor 100 shown in FIG. 1, only the components related to the present embodiment are shown in order to prevent the features of the present embodiment from being blurred. Therefore, the processor 100 includes the components shown in FIG. In addition, other general-purpose components may be further included. The variable derivation unit 101, the test model generation unit 102, the important variable determination unit 103, the CpG site extraction unit 104, the final training input data selection unit 105, and the model generation unit 106 are respectively It is only divided into separate independent names according to the functions of the variable derivation unit 101, the test model generation unit 102, the important variable determination unit 103, the CpG site extraction unit 104, for final learning. The input data selection unit 105 and the model generation unit 106 may be implemented with one processor 100. Alternatively, a variable derivation unit 101, a test model generation unit 102, an important variable determination unit 103, a CpG site extraction unit 104, an input data selection unit for final learning 105, and a model generation unit 106 Each may correspond to one or more processing modules within the processor 100. Alternatively, a variable derivation unit 101, a test model generation unit 102, an important variable determination unit 103, a CpG site extraction unit 104, an input data selection unit for final learning 105, and a model generation unit 106 May correspond to a separate software algorithm unit classified according to each function. That is, in the processor 100, a variable derivation unit 101, a test model generation unit 102, an important variable determination unit 103, a CpG site extraction unit 104, a final learning input data selection unit 105, and The implementation form of the model generation unit 106 is not limited by any one.

The variable derivation unit 101 receives the first methylation data and derives a variable by reducing the dimension of the received first methylation data. The term “dimensional reduction” may refer to a preprocessing process of data that increases the density of data by compressing the data. When machine learning is performed using high-dimensional data, the learning speed may be slowed and the performance of the model may be degraded.The device according to an embodiment includes the variable derivation unit 101 to improve the learning speed and the performance of the model. It may be to let you do. In one embodiment, the dimension reduction method is any selected from the group consisting of principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization NMF. It can be done using one method. The dimensionality reduction method according to an exemplary embodiment may perform principal component analysis (PCA) to compress array-based methylation data to find an axis that maximizes the variance of the data, and finds the principal components; PCs) can be used to transform samples of high-dimensional space into low-dimensional space. In the present specification, the term “feature” may be used as a meaning including the term “Principle components (PCs)”.

The test model generation unit 102 receives the variable derived through the dimension reduction from the variable derivation unit 101 and uses it as input data for learning, receives clinical data of cancer patients, and uses it as output data for learning through machine learning. Generate a test model for predicting the prognosis of cancer. The test model generation may be generated using a bootstrap algorithm, but is not limited thereto.

The important variable determination unit 103 evaluates the performance of the test model generated by the test model generation unit 102 and determines one or more variables important for predicting the prognosis of cancer therefrom. Determination of one or more variables that are important for predicting the prognosis of cancer can be cross-validated to determine one or more variables that are important for predicting the prognosis of cancer from a model showing maximum performance by performing cross-validation.

In the creation of a test model and determination of important variables according to an embodiment, principal components derived by performing principal component analysis (PCA) on array-based methylation data are used as input data for learning, and clinical data of cancer patients are used. An important variable can be determined by generating a cancer prognosis prediction model using the training output data and evaluating the model's performance. When a prognosis prediction model is trained, if the number of features is larger than the number of samples, such as methylation data, the model may overfit, so the test model according to an embodiment is a constraint on the coefficient of the variable. It can be created using Elastic net, a feature regularization method that prevents the model from overfitting by adding. In the determination of important variables according to an embodiment, the coefficient of the model is 0 for variables that are not important for predicting prognosis by the Elastic net, so variables that appear more than 100 times as non-zero coefficients out of 200 cross-validations are important for predicting prognosis. It can be judged and selected by variables. 4 is a graph illustrating a process of determining one or more variables important for predicting a prognosis of cancer through evaluation of a test model performance according to an exemplary embodiment. As shown in FIG. 4, as a result of evaluating the average C-index obtained by performing 200 cross-validations on the test model generated by varying the number of principal components, the model using PC1 to PC50 as predictor variables is the most suitable model. As a result of 200 tests, the main components selected as the most frequently important variables were selected as important variables in predicting prognosis.

The CpG site extraction unit 104 receives the first methylation data, receives one or more variables determined as important variables for predicting the prognosis of cancer from the important variable determination unit 103, and receives the cancer prognosis among the first methylation data. CpG sites that are highly correlated with variables important for prediction are extracted. When performing the principal component analysis, each principal component is composed of the weighted sum of the CpG site, and the CpG site with a higher weight, that is, the CpG site with a higher correlation, means that it has a greater influence on the formation of the principal component, which is important for predicting prognosis. CpG sites that have a lot of influence on the active ingredient can be selected. In the extraction of CpG sites according to an embodiment, CpG sites having a correlation of 0.6 or more with each principal component important for predicting the prognosis of cancer determined through test model generation and evaluation among array-based methylation data may be selected.

The final training input data selection unit 105 receives the second methylation data, receives the CpG site extracted from the CpG site extraction unit 104, and receives data corresponding to the CpG site extracted from the second methylation data. It is selected as the final learning input data.

The model generation unit 106 receives methylation data from the final training input data selection unit 105 and uses it as training input data to generate a cancer prognosis prediction model.

In the selection of final training input data according to an embodiment, data corresponding to CpG sites extracted from array-based methylation data among bisulfite sequencing data are selected as final training input data to generate a prognosis prediction model, and 200 By measuring the area under the curve (AUC) of the model by performing cross validation twice, it is possible to evaluate whether prognosis can be predicted with only bisulfite sequencing data for the extracted CpG site. The prognostic prediction model can be generated by dividing the training data and the test data using the bootstrap technique, and applying the Elastic net.

FIG. 5 shows the average prediction accuracy for each stage of cancer by performing AUC analysis in order to confirm whether the model generated using methylation data of heterogeneous platforms has improved prediction accuracy compared to the model using methylation data of a single platform. It is a graph that was compared and evaluated. 5A is a case where methylation data of a single platform (bisulfite sequencing data) is used, and FIG. 5B is a bisulfite sequencing data for a CpG site extracted by processing methylation data (array-based methylation data) of a heterogeneous platform. ) Is the result of evaluating the average prediction accuracy of the cancer prognosis prediction model. After principal component analysis was performed on methylation data of each single platform or heterogeneous platform in FIGS. 5A and 5B, a model was generated using principal components having a large variance as predictor variables, and accuracy was measured. At this time, modeling was performed by increasing the number of principal components used by 10 units, and among them, the model with the largest AUC was determined as the optimal model for a specific stage of cancer. As a result, as shown in FIG. 5, when the methylation data of heterogeneous platforms is used as the final learning input data, the AUC value is approximately 0.16 in stage 1, 0.13 in stage 2, and stage compared to the case of using methylation data of a single platform. It was confirmed that the accuracy of the prognosis prediction model was improved by using the DNA methylation data of the heterogeneous platform as it increased by 0.06 in 3 and 0.19 in stage 4. Therefore, this means that the problem of overfitting in the prognosis prediction model is solved and methylation markers related to cancer prognosis can be discovered.

3 is a flowchart of a method of discovering methylation markers for predicting cancer prognosis using machine learning according to an exemplary embodiment. Referring to FIG. 3, a method of discovering a methylation marker for predicting a prognosis of cancer using machine learning includes steps processed in a time series by the computing device 50 described in the previous drawings. Accordingly, even if omitted below, the contents described in the previous drawings may be applied to the method of discovering methylation markers for predicting the prognosis of cancer using machine learning of FIG. 3.

In the first step, at least one processor may derive a variable by reducing the dimension of the first methylation data (S10). For example, the first methylation data may be array-based methylation data. In one embodiment, the method of deriving the variable by reducing the dimension is principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization (NMF). It may be performed using any one method selected from the group consisting of.

In the second step, at least one processor may generate a cancer prognosis prediction test model through machine learning by using the derived variable as training input data and the cancer patient's clinical data as training output data (S20). In one embodiment, the test model may be generated using a bootstrap algorithm.

In the third step, at least one processor may determine one or more variables important for predicting cancer prognosis by evaluating the performance of the test model (S30). In an embodiment, one or more variables important for predicting cancer prognosis may be derived from a model showing maximum performance by performing cross-validation.

In the fourth step, at least one processor may extract a CpG site having a significant correlation with a variable important for predicting the prognosis of the cancer from among the first methylation data (S40).

In the fifth step, at least one processor may select data corresponding to the extracted CpG site from among the second methylation data as final input data for learning (S50). For example, the second methylation data may be bisulfite sequencing data.

In the sixth step, at least one processor may generate a cancer prognosis prediction model using the final training input data (S60).

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed on a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium. In addition, the structure of the data used in the above-described embodiment of the present invention can be recorded on a computer-readable recording medium through various means. The computer-readable recording medium includes a storage medium such as a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.) and an optical reading medium (for example, a CD-ROM, a DVD, etc.).

40: first methylation data
41: second methylation data
42: Clinical information data of cancer patients
50: computing device
80: data interface
90: memory
100: processor
101: variable derivation unit
102: test model generation unit
103: important variable decision section
104: CpG site extraction unit
105: input data selection unit for final learning
106: model generation unit

Claims (13)

A method for discovering a methylation marker for predicting cancer prognosis by an apparatus operated by at least one processor using machine learning,
A first step of extracting, by the at least one processor, a CpG site having a significant correlation with a variable important for predicting cancer prognosis from the first methylation data;
A second step of selecting, by the at least one processor, data corresponding to the extracted CpG site from among the second methylation data as final input data for learning; And
A method for discovering methylation markers for predicting cancer prognosis using machine learning comprising a third step of generating, by the at least one processor, a cancer prognosis prediction model using the final training input data. The method according to claim 1, before the first step,
Deriving a variable by reducing, by the at least one processor, a dimension of the first methylation data;
Generating, by the at least one processor, a test model for predicting a prognosis of cancer through machine learning by using the derived variable as input data for learning and clinical data of a cancer patient as output data for learning; And
And determining, by the at least one processor, the performance of the test model to determine one or more variables important for predicting a prognosis of cancer. The method of claim 1, wherein the first methylation data or the second methylation data is array-based methylation data or bisulfite sequencing data. The method of claim 2, wherein the step of deriving the variable by reducing the dimension comprises principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization (NMF). A method that is performed using any one method selected from the group consisting of. The method of claim 2, wherein the generating of the test model for predicting prognosis of cancer is generated using a bootstrap algorithm. The method of claim 2, wherein the step of deriving one or more variables important for predicting the prognosis of cancer is to derive one or more variables important for predicting the prognosis of cancer from a model showing maximum performance by performing cross-validation. A computer-readable recording medium storing a program for executing the method of any one of claims 1 to 6 on a computer. A CpG site extraction unit for extracting a CpG site having a significant correlation with a variable important for predicting cancer prognosis from among the first methylation data;
A final training input data selection unit for selecting data corresponding to the extracted CpG site from among the second methylation data as final training input data; And
A model generation unit for generating a cancer prognosis prediction model using the final training input data; and a methylation marker discovery device for predicting cancer prognosis using machine learning. The method of claim 8,
A variable derivation unit for deriving a variable by reducing the dimension of the first methylation data;
A test model generator configured to generate a cancer prognosis prediction test model through machine learning by using the derived variables as input data for learning and clinical data of a cancer patient as output data for learning; And
The apparatus further comprises a critical variable determining unit for determining one or more variables important for predicting cancer prognosis by evaluating the performance of the test model. The apparatus of claim 8, wherein the first methylation data or the second methylation data is array-based methylation data or bisulfite sequencing data. The method of claim 8, wherein the dimensional reduction is any one selected from the group consisting of principal component analysis (PCA), singular value decomposition (SVD), and non-negative matrix factorization NMF. The device that is carried out using the method of. The apparatus of claim 8, wherein the test model generation is generated using a bootstrap algorithm. The apparatus of claim 8, wherein the determination of one or more variables important for predicting the prognosis of cancer is to determine one or more variables important for predicting the prognosis of cancer from a model showing maximum performance by performing cross-validation.

Images