KR20160089981A - Device and Method for evaluating performace of cancer biomarker - Google Patents

Abstract

According to an embodiment of the present invention, a device for evaluating the performance of a cancer biomarker comprises: a preprocessing module collecting a cancer expression dataset from a public DB and preprocessing the same; a database module rearranging the preprocessed cancer expression dataset to construct a new cancer information DB; and a performance evaluation module evaluating the performance of a cancer biomarker for each cancer type by using the constructed new cancer information DB. Therefore, the device can perform evaluation using a stored prediction model with any selected testing dataset, and leave-one-out cross-validation (LOOCV) evaluation using the selected dataset can be performed. In addition, the device can perform evaluation using a user-provided training dataset and a selected testing dataset and evaluate multi-marker using the selected training and testing datasets.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cancer biomarker,

The present invention relates to an apparatus and a method for evaluating the performance of cancer biomarkers, and more particularly, to an apparatus and a method for evaluating the performance of cancer biomarkers that evaluate single-and / or multi-genes as candidates for biomarkers .

There are over 200 types of cancer in more than 60 different tissues in the human body. These types of cancers are defined by the tissue, the type of primary site cells, and the gene transcriptional modifications that drive the progression of the tumor and affect the therapeutic response. Some cancers of many tissues have many similarities, such as therapeutic response, while some of the subtypes of cancer from the same tissue are usually quite different. The characteristics of these cancers depend on the molecular pattern of the single or multiple genes in the cancer species.

By analyzing the transcripts of molecular patterns for all cancer types, one can obtain the pathogenesis and therapeutic knowledge of one kind of cancer that can be applied to another. For example, ERBB2-HER2 is an amplification of a subset of glioblastoma and gastric cancer, severe endometrial cancer, bladder cancer and lung cancer. In some cases, these results are sensitive to HER2-targeted therapies similar to those previously observed in HER2-amplified breast cancer.

Despite this single success story (ERRB2), less than 100 of the myriad papers on tumor biomarkers received approval. For the purpose of increasing the clinical application to solve the problem of clinical usability, the bench-to-bedside of the biomarker (ie the system that enables the laboratory results to be applied to patients as soon as possible: from laboratory to clinical) Various standards and guidelines have been presented to make the development process difficult.

Of the 84 biomarkers approved since 1994, only 53 were used in one trial, 30 used for more than one test, and the average approval period was 15 years. In addition, despite the widespread use of several prognostic biomarkers (e.g., CA125 for ovarian cancer and CA19-9 for pancreatic cancer), the exact role of these proteins in the progression of each of these tumors is well known little.

Screening for diagnostic biomarkers for most common cancers generally requires a high standard of 95% specificity and 95% sensitivity. In general, a single biomarker can not satisfy such a high standard value, but a combination of biomarkers using a plurality of biomarkers ("biomarker panel") can satisfy this criterion.

Future human gene expression data in the form of "Big Data" will be stored persistently in repositories such as GEO, TCGA, ICGC, ArrayExpress and Pan-cancer initiatives. Therefore, there is a need for an integrated analysis method capable of identifying candidate biomarkers and biomarker panels using the stored human gene expression data.

1. US Patent No. 2009-0269856 2. Korean Patent Publication No. 2010-0131435

1. Nam S, et al. PATHOME: an algorithm for finding correctly differentially expressed subpathways. Oncogene 33, 4941-4951 (2014).

It is an object of the present invention to provide an apparatus and a method for evaluating the performance of cancer biomarkers that evaluate single-and / or multi-genes as candidates for cancer biomarkers.

In order to solve the problems of the present invention as described above,

A preprocessing module for collecting and pre-processing a cancer expression date set from an open database; A database module for rearranging the pre-processed cancer expression dating sets to construct a new cancer information DB; And a performance evaluation module for evaluating the performance of the cancer biomarker according to the cancer type using the constructed new cancer information DB. The present invention also provides an apparatus for evaluating the performance of an cancer biomarker.

In one embodiment of the present invention, the apparatus may further include a web-interface module for displaying a result of performance evaluation of the cancer biomarker evaluated for each type of cancer by mapping, graph or table.

In one embodiment of the present invention, the preprocessing module analyzes the collected cancer expression data sets and uses quantile normalization and robust multiple-array (RMA) normalization It can be normalizing.

In one embodiment of the present invention, the pre-processing module may use a normalized coefficient as an expression value for RNA expression data by RNA sequencing in a TCGA open DB.

In one embodiment of the present invention, the preprocessing module may be to detect outliers using within-group correlation and between-group correlation .

In one embodiment of the present invention, the preprocessing module may exclude the sample having the detected abnormal value according to the external input of the user.

In one embodiment of the present invention, the preprocessing module is designed as a new data set through analysis of diagnosis, prognosis and drug response information when the collected cancer expression date set includes clinical information and sample annotations It may be that it defines the exact arm type or subtype.

In one embodiment of the present invention, the database module may include the preprocessed expression data set and corresponding annotation data.

In one embodiment of the present invention, the database module may store all the expression data in the form of a user-customized index binary file.

In one embodiment of the present invention, the performance evaluation module includes a user-selected multi-marker (AUC), a balance accuracy (BA), sensitivity, specificity, positive predictive value (PPV) ), False positive rate (FPR), false positive rate (FDR), and F1 score.

In one embodiment of the present invention, the performance evaluation module is configured to determine a balance accuracy (BA) for all markers and a balance accuracy for all markers other than a single marker, in order to measure the contribution of a single marker to the performance of the multi- And to provide a difference between accuracy (BA).

In one embodiment of the present invention, the web-interface module comprises: an input layout for transferring user-selected multiple markers and queried parameters to the performance evaluation module; And a result explorer that provides table and graph visualization of the performance evaluation results.

In one embodiment of the invention, the input layout may be to select a set of pre-processed public data or a user-provided individual data set as a training data set.

In addition, the present invention relates to a method for detecting cancer, comprising collecting and pretreating a cancer expression dating set from an open DB; Rearranging the pretreated cancer expression date set to construct a new cancer information DB; And evaluating the performance of the cancer biomarker according to the cancer type using the constructed new cancer information DB.

In one embodiment of the present invention, the method may further include a performance evaluation result display step of displaying a performance evaluation result of the cancer biomarker evaluated for each type of cancer by a mapping, a graph or a table.

In one embodiment of the present invention, in the pre-processing step, the collected cancer expression data sets are analyzed and quantized normalization and robust multiple-array (RMA) normalization are used To normalize it.

In one embodiment of the present invention, the pre-processing may be to detect an outlier using within-group correlation and between-group correlation. have.

In one embodiment of the present invention, in the preprocessing step, when the collected cancer expression date set includes clinical information and sample annotations, it is designed as a new data set through analysis of diagnosis, prognosis, and drug response information To define the exact arm type or subtype.

In one embodiment of the present invention, in the step of constructing the new cancer information DB, all the expression data may be stored in the form of a user-customized index binary file.

In one embodiment of the present invention, in the performance evaluation step, in the case of user-selected multiple markers, the area under the curve (AUC), the balance accuracy (BA), the sensitivity, the specificity, the positive predictive value (PPV) NPV), false positive rate (FPR), false positive rate (FDR), and F1 score.

In one embodiment of the present invention, in the performance evaluation step, in order to measure the contribution of a single marker to the performance of the multi-marker, a balance accuracy (BA) for all the markers and a balance And to provide a difference between accuracy (BA).

In one embodiment of the present invention, in the performance evaluation result display step, evaluation of multiple markers using the selected training and testing data sets, evaluation using a stored prediction model for an arbitrary selected test data set, Evaluation of the rib-one-out cross-validation (LOOCV) using the user-provided training data set and the evaluation using the user-provided training data set and the selected testing data set.

In one embodiment of the present invention, the performance evaluation result display step may be to select a pre-processed public data set or a user-provided individual data set as a training data set.

An apparatus and method for evaluating the performance of an arm biomarker in accordance with the present invention can evaluate multiple markers using selected training and testing data sets and can be evaluated using a stored prediction model in any selected testing data set have. In addition, an apparatus and method for evaluating the performance of the cancer biomarkers of the present invention can evaluate rib-one-out cross-validation (LOOCV) using a selected data set and provide user- It is possible to perform evaluation using the test data set.

1 is a functional block diagram of an apparatus for evaluating the performance of an arm biomarker according to an embodiment of the present invention,
2 is a flowchart of a method for evaluating the performance of an arm biomarker according to an embodiment of the present invention,
FIG. 3 is a block diagram of a process for evaluating the performance of the cancer biomarker according to the embodiment of the present invention,
Figure 4 illustrates the overall process of evaluating the performance of a single-cancer marker candidate and / or multiple-cancer marker candidate from 18 tumor type data sets in accordance with an embodiment of the present invention,
Figures 5A-5C show CANES evaluation reports for seven multi-breast cancer biomarkers sorted by support vector machines and rib-one-out cross validation using test data sets and lung cancer data sets.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The terms used in the singular may also include a plurality of concepts . In addition, the preferred embodiments of the present invention will be described below, but it is needless to say that the technical idea of the present invention is not limited thereto and can be variously modified by those skilled in the art.

Over the last decade, a repository of cancer-related gene expression data in the form of "big data" forms has been continuously developed. Despite the great potential of these datasets, the cancer-related sector still has few biomarkers approved due to the lack of a consistent method of clinically evaluating single / multiple biomarkers.

Here, the present inventors have developed a comprehensive evaluation system called CANES (CANES-specific multi-gene) system in the initial stage of evaluating single- and / or multi-genes as biomarker candidates using various classification methods. marker Evaluation System ".

We have found that intuitively understandable, clinically valuable, web-based cancer-screening assays for biomarker assessment using 94,147 samples (cell lines, normal and cancer tissues) and a classification approach covering the entire 2,134 transcript data set, We introduce the specific multi-marker evaluation system (CANES). CANES measures the diagnostic and prognostic value of 18 cancer types by means of a support vector machine, random forest, neural network and a rigorous evaluation through classification and regression trees on single and multiple cancer markers. For user-provided multiple markers, CANES provides simplified evaluation results and graphical visualization. Finally, we demonstrated the utility of CANES by performing two analyzes of previously unexplored biomarker sets.

Hereinafter, an apparatus and method for evaluating the performance of an arm biomarker according to an embodiment of the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a functional block diagram of an apparatus for evaluating the performance of an arm biomarker according to an embodiment of the present invention, and FIG. 2 is a flowchart of a method of evaluating the performance of an arm biomarker according to an embodiment of the present invention.

1 and 2, an apparatus 100 for evaluating the performance of an arm biomarker according to an embodiment of the present invention includes a preprocessing module 110 for collecting and pre-processing cancer expression dating sets from an open DB, A database module 120 for rearranging cancer expression data sets to construct a new cancer information DB 120, and a performance evaluation module 130 for evaluating the performance of the cancer biomarkers for each type of cancer using the new cancer information DB . Preferably, the cancer biomarker performance evaluation apparatus 100 may further include a web-interface module 140 for mapping, graphing, or tabulating the results of performance evaluation of the cancer biomarkers evaluated for each type of cancer.

The present inventors referred to the cancer biomarker performance evaluation apparatus 100 as CANES 100, and hereinafter, CANES 100 refers to the performance evaluation apparatus 100.

We describe the potential of seven known breast cancer markers for lung cancer statistics that illustrate the utility of CANES to deliver specific cancer-type biomarkers to biomarkers of another cancer type to illustrate the utility of CANES (100) Revaluation.

FIG. 3 is a functional block diagram of a process for evaluating the performance of an cancer biomarker according to an embodiment of the present invention, and FIG. 4 is a block diagram of a single-cancer marker candidate from 18 tumor type data sets according to an embodiment of the present invention. 5A-C illustrate the overall process of evaluating the performance of a multi-cancer marker candidate material and / or a multi-cancer marker candidate, and FIGS. 5A-C illustrate the CANES evaluation report for seven multiple breast cancer biomarkers in support of test data sets and lung cancer data sets Vector machine and rib-one-out cross validation.

Hereinafter, an apparatus 100 for evaluating the performance of an arm biomarker according to an embodiment of the present invention will be described in more detail with reference to FIGS. 1 to 5C.

CANES (100) collected RNA molecular profiles from public databases and reclassified them into tumor types according to the mapping process and quality control process. Using public expression data sets, CANES 100 provides evaluation results for user-specificized multiple markers in various cancer types or studies. The CANES 100 shown in FIG. 1 has a preprocessing module 110, a database module 120, a performance evaluation module 130, and a web-interface module 140, which are four modules.

Hereinafter, the preprocessing module 110, the database module 120, the performance evaluation module 130, and the web-interface module 140, which are four modules included in the CANES 100, will be described in further detail.

1. For the preprocessing module 110,

Currently developed CANES (110) uses microarray data from two public repositories and two cancer consortia: GEO, ArrayExpress, TCGA, and ICGC. All expression data sets can be collected using the R package. In CANES pre-processing module 110, the set of expression data obtained from the published repository is analyzed and normalized using quantile normalization and robust multiple-array (RMA) normalization. For RNA expression data by RNA sequencing in the TCGA repository, the present inventors used normalized coefficients as the expression value. All data sets with a missing rate of more than 5% were excluded, and the remaining data sets with missing rates were replaced by Bioconductor's "impute" package. In order to detect outliers caused by errors in equipment, contamination of samples, mislabeling and misprocessing, the present inventors have found that within-group correlation and group- We performed anomaly detection using the between-group correlation. Since all detected abnormal values are displayed on the sample, users can use specific options to exclude samples of these abnormal values from their analysis. Moreover, if clinical information and sample annotations are available, they are analyzed in CANES database. To define the precise cancer type or subtype, the inventors have identified or reshaped the diagnosis, prognosis, and drug response as a design of the data set. All processed expression data is converted to a customized indexed binary file for fast preprocessing. The preprocessing module was implemented using Python and R.

2. For the database module 120,

The database module 120 includes pre-processed expression data sets and corresponding annotation data. Referring to Table 1, the preprocessed data set consists of gene expression data and annotation data for 94,147 samples. Gene expression data are processed as described below, obtained from these samples along with a widely used gene expression microarray platform and an RNA sequencing platform. The database module 120 is implemented using MySQL and Python. All expression data is stored as a user-customized index binary file. Table 1 below shows the number of samples in CANES.

3. For performance evaluation module 130,

The performance evaluation module 130 of CANES is implemented using classification methods such as SVM, RF, NN, and CART. Referring to Table 2, in the case of user-selected multiple markers, the module measures the area under the curve (AUC), the balance accuracy (BA), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) ), FDR (FDR), and F1 score. To measure the contribution of a single marker to the performance of a multi-marker, CANES provides an influence measure that is the difference between BA for all markers and BA for all markers other than a single marker. All of these operations can be performed by the following four different evaluation schemes, as shown in FIG.

1) Selected training and Testing  Multiplexing with Data Sets Marker  evaluation

CANES can perform predictive analysis using specific cancer types or studies. The user can create and store a prediction model for his / her own multi-marker list using the selected data set and the classification model. Graphs and interactive result layouts can be provided and saved.

2) any selected Testing  In data sets, evaluation using stored prediction models

CANES stores evaluation results, which can be used for different sets of testing data. For example, a user may store a prediction model using a breast cancer marker and a breast cancer data set, and then evaluate the stored model in a liver cancer data set.

3) Using the selected data set live - One-out cross-validation ( LOOCV )

To prevent overfitting by any particular training data set, CANES can evaluate multiple markers using LOOCV. In this evaluation scheme, CANES can also support the evaluation of individual markers in a multi-marker set by measuring the performance contribution of the multi-markers.

4) User-provided training data sets and selected Testing  Evaluating with Data Sets

By using CANES, a prediction model generated by a user-provided data set can be evaluated. The user data sets are uploaded through the web-interface module, used as training data sets that are preprocessed and then normalized and then applied different classification methods. Predictive models trained using your own data sets are tested using a data set that is independent of the public repository.

4. For web-interface module 140,

The web-interface module 140 of CANES consists of an input layout and a result explorer. The input layout is an interface that passes the user-selected multiple markers and queried parameters to the evaluation module. In the input layout, a user can input a genetic symbol, miRNA or probe ID set, and select a preprocessed public data set or a user-uploaded individual data set as a training data set. The result explorer provides a table and graph visualization of the evaluation results. The CANES web-interface module 140 is implemented using PHP along with the JQuery and CodeIgniter frameworks.

The main features of the CANES 100 are described in detail below.

Figures 1, 3, and 4 illustrate the overall performance of CANES 100 for evaluating the performance of single-cancer marker candidates and / or multi-cancer marker candidates from 18 tumor type datasets based on four evaluation schemes. FIG. The unique features of CANES are summarized and compared with other biomarker-related databases (Oncomine, IPA-biomarker (http://www.biomarker.com/)) in terms of tool-functions essential for biomarker evaluation described in Table 1. qiagen.com/ingenuity), cBioPortal).

In addition, 1) CANES provides survival analysis using the Kaplan-Meier plot and log-rank test, and 2) provides genes for miRNA markers as well as diagnostic or prognostic genes Perform biomarker assessments, and 3) provide pan-cancer assessment results for each single marker (see Table 3).

Breast cancer for lung cancer Marker  multiple- Marker  evaluation

We evaluated the multi-marker prediction power of known breast cancer markers for another cancer type among panels of 18 tumor types. One of the key features of CANES is to evaluate multi-markers for multi-cancer types. In this study, they assess lung cancer predictability of known breast cancer markers BRCA1, BRCA2, BRIP1, CHEK2, PALB2, RB1, and TP53 in lung cancer. Figure 5 shows the CANES assessment report for seven multiple breast cancer biomarkers as support data from a support vector machine and rib-one-out with 46 lung cancer tissues and 45 normal tissues as a test data set, And are categorized by leave-one-out cross-validation. Figure 5 shows a representative of the CANES performance report in a test data set. Seven multi-markers (see FIG. 5A) were evaluated in a multi-cancer type (see FIG. 5B), which showed high AUC, BA, SN and SP in lung cancer (see FIG. 5C). They show that these seven biomarkers can potentially be applied to lung cancer through a performance evaluation. In a previous report, ERBB2-HER2 has been reported to be overexpressed in a subset of glioblastoma and gastric serous endometrial cancer, bladder cancer and lung cancer. Pathological knowledge or therapeutic methods can be applied commonly to cancer types having the same transcript expression pattern. Thus, CANES provides a powerful predictor of biomarker assessment for cancer types.

In conclusion, CANES (100) is a benchmark for the success of target treatment, as well as a clinician for a better patient class, as well as a single or multi-marker for diagnosis and prognosis that can be used by researchers. It is a powerful tool to provide evaluation results.

Despite more than 1000 references, the actual number of clinically approved biomarkers is less than 100. For screening purposes, the prostate-specific antigen (PSA) is the only approved serum biomarker, and despite its use, the guidelines are still controversial. Although total genome and transcript sequencing are considered "personalized medicine" for patients diagnosed with a particular cancer, the cost / benefit of this massive analysis is still controversial. Furthermore, it is also suspected that this profile is used to confirm the presence of a diagnostic gene expression " signature ". Even the well-known diagnostic biomarkers such as cancer embryonic antigen (CEA, colon cancer), CA19-9 (pancreatic cancer), and CA-125 (ovarian cancer) have little or no known precise role of these markers in disease progression.

To address this issue, the inventors have designed and developed CANES, a simple and multi-biomarker assessment tool for simple, user-friendly, web-based applications for evaluating multiple markers for widely published cancer data sets . Furthermore, CANES is able to assess the performance of multiple biomarkers for a number of parameters (diagnosis, therapeutic response, survival rate, etc.) in areas of little clinical success. All evaluation results are provided with table and graph visualizations, and can be downloaded as high-quality PDF images and CSV-based text-based spreadsheet files. CANES is a powerful tool for evaluating multiple candidate material markers in an independent set of data for diagnosis or prognosis. At present, CANES not only makes available all publicly available microarray datasets, but also recently has an RNA-seq dataset for specific cancers made using the next generation sequencing technology in the TCGA database.

The usefulness of CANES exists further in addition to the above-mentioned examples. For example, CANES can be used as an evaluation of a marker (eg,>) for a particular cancer type (at least, for individuals belonging to the high risk group for this type of cancer) 95% sensitivity and > 95% specificity). The inventors of the present invention have made it easier to develop biomarkers with improved accuracy through the use of CANES. Moreover, from a research point of view, confirming that there is a strong association with a particular gene (s) and a particular tumor may help to understand the mechanistic understanding of its progression (and its possible inhibition).

In summary, the inventors have developed a new publicly available tool for research / development of a single biomarker or a set of multiple biomarkers for a particular cancer type. The tool is used by clinical and biomedical research communities to characterize specific cancer types, identify cancer pathways, and greatly improve the potential clinical utility (eg, diagnosis, prognosis, survival rate, etc.) of individual biomarkers . ≪ / RTI >

The list of abbreviations used in the present invention

AUC, area under the curve; BA, balanced accuracy; CART, classification and regression tree; GEO, Gene Expression Omnibus; FPR, false positive rate; FDR, false discovery rate; GC, stomach cancer; ICGC, international cancer genome consortium; LOOCV, rib-one-out cross-validation; NN, neural network; NPV, negative predictive value; PPV, positive predictive value; RF, random forest; RMA, multi-array average; SVM, support vector machine; TCGA, cancer genome atlas;

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. It will be possible. Therefore, the embodiments disclosed in the present invention and the accompanying drawings are intended to illustrate and not to limit the technical spirit of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings . The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention. In addition, claims that do not have an explicit citation in the claims may be combined to form an embodiment or included in a new claim by amendment after the application.

110: preprocessing module
120: Database module
130: performance evaluation module
140: Web-Interface Module

Claims (23)

A preprocessing module for collecting and pre-processing a cancer expression date set from an open database;
A database module for rearranging the pre-processed cancer expression dating sets to construct a new cancer information DB; And
A performance evaluation module for evaluating the performance of the cancer biomarker for each type of cancer using the established new cancer information DB;
Wherein said cancer biomarker is a cancer biomarker. The method according to claim 1,
And a web-interface module for displaying the result of performance evaluation of the cancer biomarker evaluated for each cancer type in a mapping, a graph or a table. The method according to claim 1,
Wherein the pre-processing module analyzes the collected cancer expression data sets and normalizes them using quantile normalization and robust multiple-array (RMA) normalization. ≪ RTI ID = 0.0 > . The method of claim 3,
Wherein the pre-processing module uses a normalized coefficient as an expression value for RNA expression data by RNA sequencing in a TCGA public DB. The method according to claim 1,
Wherein the pre-processing module detects an outlier using within-group correlation and between-group correlation. ≪ RTI ID = 0.0 > 8. < / RTI > 6. The method of claim 5,
Wherein the preprocessing module excludes the sample having the detected abnormal value according to an external input of the user. The method according to claim 1,
The preprocessing module may be designed with a new data set through analysis of diagnosis, prognosis and drug response information to define the correct cancer type or subtype if the collected cancer expression date set includes clinical information and sample annotations Wherein said cancer biomarker is a cancer biomarker. The method according to claim 1,
Wherein the database module comprises the pre-processed expression data set and corresponding annotation data. 9. The method of claim 8,
Wherein the database module stores all the expression data in the form of a user-customized index binary file. The method according to claim 1,
The performance evaluation module is designed to evaluate the performance of the user-selected multiple markers under the curve AUC, balance accuracy BA, sensitivity, specificity, positive predictive value PPV, negative predictive value NPV, false positive rate (FPR) (FDR), and an F1 score. The apparatus for evaluating the performance of a cancer biomarker according to claim 1, The method according to claim 1,
The performance evaluation module may be configured to provide a difference between a balance accuracy (BA) for all markers and a balance accuracy (BA) for all markers other than a single marker to measure the contribution of a single marker to the performance of the multi- Wherein said cancer biomarker is a cancer biomarker. 3. The method of claim 2,
The web-interface module comprises:
An input layout for transferring a user-selected multiple marker and a queried parameter to the performance evaluation module; And
And a result explorer providing a table and graph visualization of the performance evaluation result. 13. The method of claim 12,
Wherein the input layout selects a public data set pre-processed as a training data set or a user-provided individual data set. Collecting and pre-processing the cancer expression dating set from the public DB;
Rearranging the pretreated cancer expression date set to construct a new cancer information DB; And
Evaluating the performance of the cancer biomarker according to the type of cancer using the established new cancer information DB;
/ RTI > of cancer biomarkers. 15. The method of claim 14,
And displaying a performance evaluation result indicating the performance evaluation result of the cancer biomarker evaluated for each cancer type in a mapping, a graph or a table. 15. The method of claim 14,
Wherein the pre-processing step comprises analyzing the collected cancer expression data sets and normalizing them using quantile normalization and robust multiple-array (RMA) normalization. A method for evaluating the performance of a marker. 15. The method of claim 14,
Wherein the pre-processing step detects an outlier using within-group correlation and between-group correlation. The method according to claim 1, . 15. The method of claim 14,
In the pre-processing step, when the collected cancer expression date set includes clinical information and sample annotations, a new data set is designed through analysis of diagnosis, prognosis, and drug response information to specify the precise cancer type or subtype Wherein the method comprises the steps of: 15. The method of claim 14,
Wherein the step of constructing the new cancer information DB stores all the expression data in the form of a user-customized index binary file. 15. The method of claim 14,
(AUC), a balance accuracy (BA), a sensitivity, a specificity, a positive predictive value (PPV), a negative predictive value (NPV), a false positive rate (FPR) A detection rate (FDR), and an F1 score. ≪ Desc / Clms Page number 20 > 15. The method of claim 14,
In the performance evaluation step, a difference between the balance accuracy (BA) for all the markers and the balance accuracy (BA) for all the markers other than the single marker is measured in order to measure the contribution of the single marker to the performance of the multi- Wherein said cancer biomarker is a cancer biomarker. 22. The method according to claim 20 or 21,
In the performance evaluation result display step,
Evaluation of multiple markers using selected training and testing data sets, evaluation using stored predictive models for arbitrary selected testing data sets, and evaluation of rib-one-out cross-validation (LOOCV) using selected data sets Evaluating at least one of evaluation and evaluation using a user-provided training data set and a selected testing data set. 16. The method of claim 15,
Wherein the performance evaluation result display step selects a public data set pre-processed as a training data set or a user-provided individual data set.

Images