JP6857597B2 - Biomarker search method for urinary biotransformers - Google Patents

Description

The present invention relates to methods, systems and programs for searching for urinary biotransformer markers. Specifically, the present invention relates to methods, systems and programs for searching for urinary biotransformation markers associated with a particular disease or condition, particularly cancer.

As a conventional cancer testing method, for example, there is a report of a cancer testing method using patient-derived biotransforms (Patent Document 1). The test is quick and easy. Specific steps are (1) acquisition of mass spectrum data of patient-derived metabolites by mass spectrometry (MS) analysis, (2) patient-derived metabolites, and patient-derived metabolites after surgical resection. It consists of a step of detecting the fluctuation of the mass spectrum by comparing the mass spectrum data of the above, and (3) a step of determining whether the sample derived from the patient is a sample derived from a cancer patient by detecting the fluctuation.

Japanese Unexamined Patent Publication No. 2010-266386

The feature of the conventional method is that the cancer testing method is not based on one biomarker but mainly on multi-marker testing, and is trying to reflect the pathological condition of the patient in more detail. In the mass spectrum data analysis method, cancer is determined by analyzing the disappearance, appearance time, retention time, and increase or decrease of peak intensity of a specific mass spectrum in the obtained mass spectrum data. The analysis model in this case is determined by multivariate analysis such as OPLS-DA.

In general, in the workflow of urinary metabolite analysis by Liquid Chromatograph / Mass Spectrometer (LC / MS), first, comprehensive analysis of urinary metabolites by LC / MS using multiple analysis modes is performed. , And extract metabolites that are significantly different between the groups (for example, healthy subjects and cancer patients). The plurality of analysis modes are a combination of a reverse phase mode and a hydrophilic interaction mode in a liquid chromatograph, and a positive / negative ionization mode in a mass spectrometer.

The data obtained by LC / MS are three-dimensional data of the elution time of each metabolite in the liquid chromatograph and the ionic mass and ionic strength of each metabolite in the mass spectrometer (see the upper part of FIG. 2). After preprocessing these data, multivariate analysis (principal component analysis (PCA)) while confirming metabolites that have a significant difference between the two groups (for example, healthy subjects and cancer patients) by a test. ), Partial Least Squares Discriminant Analysis (PLS-DA), OPLS-DA (Orthogonal Partial Least Squares Discriminant Analysis), Kernel Discriminant Analysis, etc.) , Extract biomarker candidates that are effective in distinguishing the two groups, and build a (cancer) test model. By such a method, biomarkers useful for analysis of sweetness components and functional components of agricultural products and foods, colorectal cancer tests using blood, and the like have been found so far.

However, for example, in the case of a cancer test using urinary metabolites, even if OPLS discriminant analysis method is used, it may not be effective for extraction of biomarkers. In general, the S plot is used for the extraction of biomarkers by the OPLS discriminant analysis method. The S plot shows the absolute value of the peak intensity difference between groups on the horizontal axis for the metabolites detected by comprehensive analysis by LC / MS (the larger the value, the larger the difference between groups, and it becomes a biomarker candidate). The figure plotted on the vertical axis as reliability (repetition accuracy, correlation coefficient between group differences and intragroup error, the closer it is to 1, there is no individual difference and there is a difference between groups), and it is used for biomarker candidate search. Often used. In the case of cancer analysis, in urinary metabolites, the difference in intensity between groups may not be so large, and the error within the group may be large. In addition, there are thousands of biomarker candidate metabolites and the structure is diverse. It is very difficult to narrow down the biomarker candidates with the S plot.

Therefore, it is an object of the present invention to provide an improved method for searching for biomarkers of urinary biotransformers.

Based on the above issues, we examined changes in the workflow for building a cancer checking model using urine samples. That is, instead of directly extracting important biomarker candidates by the S plot of the OPLS discriminant analysis method, the analysis flow is as follows:
(a) Comprehensive analysis results by LC / MS may be narrowed down by p-value for urinary metabolites that have a significant difference between healthy subjects and cancer patients by Wilcoxon's rank sum test. .. Normally, the significance level at this time is 5%.
(b) Although it is possible to narrow down the biomarker candidates that are important biomarkers in the above test, it is difficult to quantitatively evaluate them. Therefore, for the narrowed down biomarkers, random forest (RF: The importance of biotransforms is quantitatively evaluated using the Random Forest) method. The higher the value of importance, the more important the metabolite and the higher the ranking.
(c) Using the extracted biomarker group, calculate a discriminant formula for discriminant analysis (for example, OPLS / kernel discriminant analysis) to distinguish between healthy subjects and cancer patients, and construct a cancer test model.
(d) In some cases, prepare multiple cancer checking models with different numbers of biomarkers. For example, build a cancer checking model with 2, 10, and 30 biomarkers in descending order of importance.

In one aspect, the invention
(A) A step of subjecting a urine sample to a liquid chromatograph mass spectrometer (LC / MS) and analyzing urine biotransforms in the urine sample.
(B) A step of quantitatively evaluating the importance of the urinary metabolite by a random forest method based on the analysis data of the urinary metabolite and selecting a urinary metabolite having a high importance.
(C) A step of performing a discriminant analysis method using the analysis data of the selected urinary metabolites.
(D) Provided is a method for searching for a urinary biotransformer marker, which comprises a step of determining a urinary biotransformer associated with a specific disease or condition as a marker candidate based on the result of the discriminant analysis.

In one aspect, the invention
(A) A step of subjecting a urine sample to a liquid chromatograph mass spectrometer (LC / MS) and analyzing urinary metabolites in the urine sample.
(B) Based on the analysis data of the urinary biotransformers, the urinary biotransformers that were significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects were narrowed down by Wilcoxon's rank sum test. Above, the step of quantitatively evaluating the importance of the urinary metabolite by the random forest method and selecting the urinary metabolite having a high importance.
(C) A step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker.
(D) Provided, optionally, a method for searching for urinary biotransformer markers, which comprises the step of constructing a plurality of test models having different numbers of biomarkers based on the discriminant analysis result.

In another aspect, the invention
A counting unit for inputting analysis data of urinary biotransformers by a liquid chromatograph mass spectrometer (LC / MS),
A calculation unit that performs numerical analysis using the analysis data of urinary metabolites,
A determination unit that selects urinary metabolites based on the results of the analysis and / or determines urinary metabolites as marker candidates.
It is provided with an output processing unit that processes the determination result by the determination unit for output.
The calculation unit quantitatively evaluates the importance of the urinary metabolite by a random forest method using the analysis data of the urinary metabolite, and a discriminant analysis method using the analysis data of the urinary metabolite. And
The determination unit selects urinary metabolites having high importance based on the result of the evaluation by the random forest method, and determines the urinary metabolite as a marker candidate based on the result of the discriminant analysis method. It provides a system for searching for biotransformer markers.

In another aspect, the invention
A counting unit for inputting analysis data of urinary biotransformers by a liquid chromatograph mass spectrometer (LC / MS),
A calculation unit that performs numerical analysis using the analysis data of urinary metabolites,
A determination unit that selects urinary metabolites based on the results of the analysis and / or determines urinary metabolites as marker candidates.
It is provided with an output processing unit that processes the determination result by the determination unit for output.
The calculation unit uses the analysis data of the urinary metabolites, and in some cases narrows down the biomarker candidates by the Wilcoxon rank sum test, and then quantitatively determines the importance of the urinary metabolites by the random forest method. Evaluate and perform a discriminative analysis method using the urinary metabolite as a biomarker.
The determination unit selects a urinary metabolite having a high importance based on the result of the evaluation by the random forest method, determines the urinary metabolite as a marker candidate based on the result of the discriminant analysis method, and may determine it as a marker candidate. Provided is a system for searching for urinary biotransformer markers, which constructs a test model by discriminant analysis.

In yet another aspect, the present invention
On the computer
Based on the analysis data of urinary metabolites by liquid chromatograph mass spectrometer (LC / MS), the importance of the urinary metabolites is quantitatively evaluated by the random forest method, and the urinary metabolites of high importance are evaluated. Steps to select and
A step of performing a discriminant analysis method using the analysis data of the selected urinary biotransforms, and
Provided is a program for searching for a urinary biotransformer marker, which executes a step of determining a urinary biotransformer associated with a specific disease or condition as a marker candidate based on the result of the discriminant analysis.

In another aspect, the invention
On the computer
Based on the analysis data of urinary biotransformers by liquid chromatograph mass spectrometer (LC / MS), in some cases, urinary biotransformers significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects. After narrowing down by Wilcoxon's rank sum test, the importance of the urinary metabolite is quantitatively evaluated by the random forest method, and the step of selecting the urinary metabolite with high importance, and
The step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker, and
Optionally, a program for searching for urinary biotransformer markers is provided, which executes a step of constructing a plurality of test models having different numbers of biomarkers based on the discriminant analysis result.

Further features relating to the present invention will become apparent from the description and accompanying drawings herein.

According to the present invention, in the search for urinary biotransformer biomarkers, biomarker candidates can be effectively narrowed down, and urinary metabolites that are significantly related to a specific disease or condition can be quickly and reliably selected. Can be decided. Therefore, the present invention is useful for constructing a disease checking model, particularly a cancer checking model. The present invention is also useful in fields such as early detection of diseases or conditions and development of therapeutic agents.

An example of the analysis flow of the present invention is shown. A display example of the analysis flow shown in FIG. 1 and the obtained results is shown. An example of the analysis system of the present invention is shown. The analysis flow of the present invention is shown in detail. It is shown together with the analysis process for structurally unknown metabolites. An example of the database structure in the analysis system of the present invention is shown. An example of the total metabolite database structure in the analysis system of the present invention is shown. An example of Wilcoxon rank sum test table is shown. The metabolites are listed in order of p-value. It is a graph which shows an example of the importance evaluation by the random forest method. An example of important metabolites by the random forest method is shown. The metabolites are arranged in descending order of importance. An example of displaying the total number of metabolites detected by LC / MS is shown. An example of displaying the total number of metabolites narrowed down by the Wilcoxon rank sum test is shown. An example of OPLS discriminant analysis results when one biomarker is used is shown. An example of the discrimination result when one biomarker is used is shown. An example of OPLS discriminant analysis results when 6 biomarkers are used is shown. An example of the discrimination result when 6 biomarkers are used is shown.

The present invention provides methods, systems and programs for searching for biomarkers of urinary biotransformers. An example of the analysis flow of the present invention is shown in FIG. A display example of the result obtained by the analysis flow of FIG. 1 is shown in FIG. In addition, FIG. 4 shows a detailed analysis flow (upper side) for the cancer checking model together with an analysis process (lower side) for structurally unknown metabolites. In the analysis flow, after comprehensive analysis of urinary metabolites by liquid chromatograph mass spectrometer (LC / MS), data pretreatment is performed, and in some cases, biomarker candidates are narrowed down by Wilcoxon's rank sum test. , The importance is quantitatively evaluated by the random forest method, and specific by the discriminant analysis method (for example, OPLS discriminant analysis method or kernel discriminant analysis method) which is one of the multivariate analysis using the biomarker located at the top. Calculate predictive formulas to identify diseases or conditions and build test models. Urinary metabolites associated with a particular disease or condition are then determined as biomarker candidates. At this time, it is also preferable to construct a plurality of test models having different numbers of biomarkers so that which test model to use can be selected from the viewpoint of test accuracy and price.

Therefore, the method of searching for urinary biotransformer markers of the present disclosure is:
(A) A step of subjecting a urine sample to a liquid chromatograph mass spectrometer (LC / MS) and analyzing urinary metabolites in the urine sample.
(B) A step of quantitatively evaluating the importance of the urinary metabolite by a random forest (RF) method based on the analysis data of the urinary metabolite and selecting a urinary metabolite having a high importance.
(C) A step of performing a discriminant analysis method using the analysis data of the selected urinary metabolites.
(D) Includes the step of determining urinary metabolites associated with a particular disease or condition as marker candidates based on the results of the discriminant analysis.

In a more specific embodiment, the method of searching for urinary biotransformer markers of the present disclosure is:
(A) A step of subjecting a urine sample to a liquid chromatograph mass spectrometer (LC / MS) and analyzing urinary metabolites in the urine sample.
(B) Based on the analysis data of the urinary biotransformers, the urinary biotransformers that were significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects were narrowed down by Wilcoxon's rank sum test. Above, the step of quantitatively evaluating the importance of the urinary metabolite by the random forest (RF) method and selecting the urinary metabolite of high importance,
(C) A step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker.
(D) In some cases, a step of constructing a plurality of test models having different numbers of biomarkers based on the discriminant analysis result is included.

First, as shown in FIGS. 1 and 4, a urine sample is subjected to a liquid chromatograph mass spectrometer for comprehensive analysis. The urine sample means urine collected from a subject and a sample obtained by treating the urine (for example, urine deproteinized with methanol or urine to which a preservative such as toluene, xylene or hydrochloric acid is added). Subjects from which urine specimens are derived are those with a specific disease or condition, healthy subjects, subjects who had a specific disease or condition in the past but are now healthy, and are at high risk of developing a specific disease or condition. Includes subjects (with family history). The target species are humans and other mammals such as primates (monkeys, chimpanzees, etc.), domestic animals (cattle, horses, pigs, sheep, etc.), pet animals (dogs, cats, etc.), experimental animals. (Mice, rats, rabbits, etc.), and humans are preferred.

Liquid chromatograph mass spectrometer (LC / MS) analysis can be performed using a conventional liquid chromatograph mass spectrometer. The analysis mode is not particularly limited, and any of a reverse phase mode or a hydrophilic interaction mode in a liquid chromatograph and a positive / negative ionization mode in a mass spectrometer may be used. By performing a comprehensive analysis in a plurality of analysis modes, urinary metabolites can be comprehensively detected.

From this analysis, analysis data of urinary biotransformers is obtained. The analysis data includes, for example, mass spectra (retention time, mass, ionic strength, peak area value, etc. in a liquid chromatograph) measured in each urinary metabolite, and medical information of the target from which each urine sample is derived. included. Further, information on urinary metabolites (structure unknown and structure known metabolites), information on measurement conditions by LC / MS (analysis mode), and the like may be included. This analysis data may be displayed in a format such as a graph (for example, the upper part of FIG. 2) or a table (for example, FIG. 6).

The resulting analysis data is preferably preprocessed for subsequent steps. For example, when considering the ionic strength measured by LC / MS, first substitute the missing value with the minimum value in each substance data, and if there are multiple measurement dates, the measurement day is based on the median value of each biotransform. Correct the difference. After that, the osmotic pressure value by the freezing point depression method is used to standardize the difference between samples due to urine concentration, and the median value is set to 1.0 to standardize the data distribution difference between substances. In addition, it is expected that the concentration of the urine sample will differ depending on the physical condition of the subject. In order to correct this, a correction method based on the quantification of urinary creatinine or a correction method based on the concentration of urinary solute measured by freezing point depression, for example, a correction method using osmorality measurement can be applied. By the above correction, the normalized analysis data can be acquired.

Subsequently, before performing the discriminant analysis method for the urinary metabolites, the urinary metamorphosis is narrowed down. Although it is possible to subject all biotransforms to the random forest method and quantitatively evaluate their importance, Wilcoxon rank sum test is performed to narrow down the biotransforms in order to reduce the computational load. That is valid. This test method tests whether there is a significant difference in the data obtained between the groups, and the significance level is often set to 5% (p <0.05). The test result may be displayed in the form of a table (for example, FIG. 7) showing urinary metabolites and their p-values. This allows urinary biotransformers that change significantly in relation to a particular disease or condition to be selected to narrow down the urinary metabolites before proceeding to the next step, the random forest method.

The Random Forest (RF) method is a type of machine learning method, and its major feature is that the importance of features can be calculated. The evaluation data may be displayed in a format such as a graph (for example, FIG. 8) or a table (for example, FIG. 9) shown in order of importance. Metabolites with a high degree of importance, that is, metabolites at the top of the analysis results, are likely to be biomarkers.

The discriminant formula is calculated by the discriminant analysis method using the selected biomarker. As the discriminant analysis method, for example, OPLS discriminant analysis method, kernel discriminant analysis method and the like can be used. The OPLS discriminant analysis method is effective for discriminating differences between groups according to a specific disease or condition, and in a score plot in which each sample is plotted as one point, if it is separated on the horizontal axis, it is between groups. It indicates that there is a difference (the vertical axis represents the difference within the group) and greatly aids visual understanding (eg, the graph shown at the bottom of FIG. 2).

In this specification, the OPLS discriminant analysis method will be mainly described, but this is a method for performing linear data analysis. However, when the number of clinical cases increases and the data becomes complicated, it may be effective to perform linear analysis after moving to a higher-dimensional feature space, such as kernel discriminant analysis, and OPLS discriminant analysis method. It is not limited to.

When calculating a discriminant using urinary metabolites associated with a specific disease or condition as biomarkers, the discriminant differs depending on the number of biomarkers selected. One biomarker may be selected, or a plurality of biomarkers may be selected in combination. The combination of the plurality enhances the discriminating power of a specific disease or condition, and as a result, enhances the accuracy of the test. In this way, a checking model for a particular disease or condition is constructed. The analysis data, analysis results, inspection model, etc. up to this point may be collectively displayed in a table format (for example, FIG. 5).

Here, the specific disease or condition is not particularly limited as long as it is a disease or condition for which a biomarker is to be searched, that is, a disease or condition for which diagnosis, determination, risk determination, or the like is desired. Preferably, the disease or condition is cancer. Cancer refers to a disease, also called a malignant tumor or malignant neoplasm, which is characterized by autonomous growth, infiltration and metastasis, and cachexia. The cancer is not particularly limited, but is solid cancer (breast cancer, colon cancer, lung cancer, prostate cancer, stomach cancer, colorectal cancer, pancreatic cancer, kidney cancer, ovarian cancer, esophagus). Cancer, liver cancer, biliary tract cancer, bladder cancer, childhood cancer, etc.), sarcoma (osteosarcoma, chondrosarcoma, etc.), blood cancer (leukemia, malignant lymphoma, multiple myeloma, etc.) .. There are primary, metastatic, and recurrent cancers, and the malignancy varies depending on the type of cancer, and it is classified into stages according to the degree of progression and spread. The required treatment (treatment method) differs depending on the difference in primary, metastatic or recurrent disease, the difference in malignancy / stage, and the type of cancer. Therefore, if a urinary biotransformer marker related to a specific cancer type and / or condition can be searched for, it becomes possible to determine necessary treatments or make an early diagnosis by using the urinary biotransformer marker. .. Specific diseases or conditions include the presence or absence of cancer, the type of cancer, the degree of cancer progression, the malignancy of cancer, the stage of cancer, the prognosis of cancer, the therapeutic effect on cancer, and anticancer. Sensitivity to agents, but is not limited to this.

In addition, the method for searching for urinary biotransformation markers as described above can be easily and efficiently carried out by using an appropriate system or program. The system for searching for urinary biotransformer markers of the present disclosure is:
A counting unit for inputting analysis data of urinary biotransformers by a liquid chromatograph mass spectrometer (LC / MS),
A calculation unit that performs numerical analysis using the analysis data of urinary metabolites,
A determination unit that selects urinary metabolites based on the results of the analysis and / or determines urinary metabolites as marker candidates.
It is provided with an output processing unit that processes the determination result by the determination unit for output.
The calculation unit quantitatively evaluated the importance of the urinary metabolite by the random forest (RF) method using the analysis data of the urinary metabolite, and used the analysis data of the urinary metabolite. Perform a discriminant analysis method
The determination unit selects a urinary metabolite of high importance based on the result of the evaluation by the random forest (RF) method, and determines the urinary metabolite as a marker candidate based on the result of the discriminant analysis method. It is a thing.

In a more specific embodiment, the system for searching for urinary biotransformer markers of the present disclosure is:
A counting unit for inputting analysis data of urinary biotransformers by a liquid chromatograph mass spectrometer (LC / MS),
A calculation unit that performs numerical analysis using the analysis data of urinary metabolites,
A determination unit that selects urinary metabolites based on the results of the analysis and / or determines urinary metabolites as marker candidates.
It is provided with an output processing unit that processes the determination result by the determination unit for output.
The calculation unit uses the analysis data of the urinary metabolites, and in some cases narrows down the biomarker candidates by the Wilcoxon rank sum test, and then determines the importance of the urinary metabolites by the random forest (RF) method. Quantitatively evaluate and discriminate analysis using the urinary metatransformer as a biomarker.
The determination unit selects urinary metabolites having high importance based on the results of evaluation by the random forest (RF) method, and determines urinary metabolites as marker candidates based on the results of the discriminant analysis method. In some cases, an inspection model is constructed by a discriminant analysis method.

In the system of the present disclosure
The calculation unit performs a Wilcoxon rank sum test using the analysis data of the urinary metabolites.
The determination unit may narrow down urinary metabolites that are significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects based on the result of the Wilcoxon rank sum test.

Further, in the system of the present disclosure,
The calculation unit performs a discriminant analysis method using the selected urinary metabolite as a biomarker.
The determination unit may build an inspection model by the discriminant analysis method.

In the system of the present disclosure, preferably, the counting unit, the calculating unit, the determining unit, and the output processing unit are connected to each other so as to be able to operate in the memory so that the method as described above can be carried out. It is a communication-connected system. The system of the present disclosure may be connected or communicatively connected to a communication device, an input device, an output device, and / or a storage device. A specific configuration example of the analysis system is shown in FIG.

Analysis data of urinary metabolites by a liquid chromatograph mass spectrometer (LC / MS) is input to the counting unit. The analysis data may be input via a connected input device or may be input via a communication device. Alternatively, it may be connected to a liquid chromatograph mass spectrometer from which analysis data may be input. The input analysis data is the mass spectrum of urinary metabolites measured by a liquid chromatograph mass spectrometer (LC / MS) (retention time, mass, ion intensity, peak area value, etc. in the liquid chromatograph). Includes medical information of the subject from which the urine sample is derived. The analysis data may further include information on urinary metabolites (structure unknown and structure known metabolites), LC / MS measurement conditions (analysis mode), and the like. This analysis data may be displayed in a format such as a graph (for example, the upper part of FIG. 2) or a table (for example, FIG. 6) via an output device. The analysis data may also be stored in a storage device as a total metabolite table (FIG. 3).

The calculation unit processes the analysis data input to the counting unit or read from the storage device. Specifically, using analysis data of urinary metabolites, in some cases, biomarker candidates are narrowed down by Wilcoxon's rank sum test, and then the importance is quantitatively evaluated by the random forest (RF) method and / Alternatively, a discriminant analysis method is performed. In addition, in order to quantitatively evaluate the importance by the random forest method and identify a specific disease or condition by the extracted biomarkers, a plurality of discriminants having different numbers of biomarkers are calculated by the discriminant analysis method. These results may be displayed in the form of graphs, tables, etc. via the output device. Further, these results may be stored in the storage device as a random forest (RF) table, a discriminant analysis result (table), and a test table, respectively (FIG. 3).

The determination unit selects urinary metabolites and / or determines urinary metabolites as marker candidates based on the results of the analysis performed by the calculation unit. When the Wilcoxon rank sum test is performed by the calculation unit, the determination unit may narrow down the urinary metabolites based on the result of the Wilcoxon rank sum test. Further, the determination unit determines whether the predicted value calculated by the discriminant is higher or lower than the set threshold value for the unknown sample for a specific disease or condition (for example, for the presence or absence of a specific disease or condition). It is possible to judge.

The output processing unit processes the determination result by the determination unit for output. For example, the selected urinary metabolites and / or the urinary metabolites determined as marker candidates and / or the constructed test model are processed for visual display on an output device such as a display. Specifically, it is processed in the form of a table or graph (for example, FIGS. 8, 9, 12 and 14). Alternatively, it may be processed to output audio to an output device such as a speaker.

The storage device includes a total biotransformation table (list of biotransformers detected in each analysis), a test table (test results by Wilcoxon's rank sum test, ranked by p-value), and an RF table (importance). (Biotransformation table arranged in descending order of), discriminant analysis result (score plot), and test model (discriminant formula obtained by discriminant analysis method) are stored. Which test model is used depends on what kind of test is performed, that is, which test is performed to determine which disease or condition is to be performed. The score plot shows that the two groups are separated to the maximum on the horizontal axis, and if they are separated on the horizontal axis, there is a difference between the two groups, while the vertical axis is the difference within the group. (For example, the graph shown at the bottom of FIG. 2).

In addition, the programs for searching for urinary biotransformer markers disclosed in the present disclosure are:
On the computer
Based on the analysis data of urine biotransforms by liquid chromatograph mass spectrometer (LC / MS), the importance of the urine biotransforms is quantitatively evaluated by the random forest (RF) method, and the importance of urine is high. Steps to select medium biotransformers and
A step of performing a discriminant analysis method using the analysis data of the selected urinary biotransforms, and
Based on the results of the discriminant analysis, a step of determining urinary metabolites associated with a specific disease or condition as marker candidates is performed.

In a specific embodiment, the program for searching for urinary biotransformer markers of the present disclosure is:
On the computer
Based on the analysis data of urinary biotransformers by liquid chromatograph mass spectrometer (LC / MS), in some cases, urinary biotransformers significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects. After narrowing down by Wilcoxon's rank sum test, the importance of the urinary metabolites is quantitatively evaluated by the random forest (RF) method, and the step of selecting the urinary metabolites with high importance, and
The step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker, and
In some cases, based on the discriminant analysis result, a step of constructing a plurality of test models having different numbers of biomarkers is performed.

In the program of the present disclosure, the steps of selecting urinary biotransformers of high importance are significantly increased or decreased in patients with a specific disease or condition relative to healthy individuals using the analysis data of the urinary biotransformers. After narrowing down the urinary biotransformers that are present by Wilcoxon's rank sum test, the importance of the urinary biotransformers is quantitatively evaluated by the random forest (RF) method, and the urinary biotransformers with high importance are selected. It may be something to do.

The program of the present disclosure may further cause a computer to perform a step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker.

The program of the present disclosure, when read and executed on a computer, can carry out a method of searching for urinary biotransformer markers and can operate a system for searching for urinary metabolite markers. it can.

The program can be stored on a computer-readable medium such as a magnetic recording medium (hard disk drive), CD-ROM, CD-R, RAM and supplied to the computer, or supplied to the computer via wired or wireless communication. be able to.

Hereinafter, the present invention will be specifically described by exemplifying examples, but the present examples are provided merely for the purpose of explaining the present invention, and do not limit or limit the scope of the present invention. Absent.

[Example 1]
Specific calculation examples to which the method for searching for urinary metabolites of the present invention is applied are shown (FIGS. 6 to 15).
This is the result of using urine samples derived from healthy Caucasian subjects (15 cases) and breast cancer patients (15 cases). First, a comprehensive analysis of urinary biotransforms is performed by LC / MS. For example, analysis of urinary biotransforms by LC / MS gives the results shown in FIG. That is, 1325 kinds of biotransformers including fatty acids, phospholipids, sugars, amino acids, peptides, amines, organic acids, nucleic acids, etc., whose structures are known and unknown are detected. Here, the structure-unknown metabolite is a metabolite that was not hit by a mass spectrum search using a public database. The analysis results can be displayed as shown in the total metabolite database structure example (FIG. 6).

Although it is possible to subject all these biotransforms to the random forest method and quantitatively evaluate their importance, Wilcoxon rank sum test was performed to reduce the computational load in the random forest method. , It is effective to narrow down the biotransforms. This test method tests whether there is a significant difference in the data obtained between the groups (in this analysis, it corresponds to healthy subjects and breast cancer patients), but when the significance level is 5%, 374 It can be narrowed down to species (Fig. 11). Usually, the test results are presented in a table of urinary biotransformers and p-values as shown in FIG.

However, since the number is still large and the importance of individual metabolites cannot be quantitatively evaluated by the test alone, the importance is evaluated by the random forest method. The random forest method is a type of machine learning method and is characterized in that the importance of features can be calculated. The analysis result by the random forest method is shown in FIG. The horizontal axis of FIG. 8 is the importance, and the vertical axis corresponds to each metabolite, and the correspondence table is shown in FIG. In FIG. 9, X indicates a structure-unknown metabolite. Although the metabolites shown in FIG. 9 are biomarker candidates, it is desirable to exclude metabolites of exogenous substances (for example, drugs) from the biomarker candidates. For example, judging from the mass spectrum, it can be seen that the top three species are conjugates called glucuronides, so the most promising biomarker is XC.

Next, the discriminant is calculated by the OPLS discriminant analysis method using the biomarker. The OPLS discriminant analysis method is an effective method for discriminating differences between groups according to a specific disease or condition. Based on the above consideration, a one-dimensional score plot by the OPLS discriminant analysis method using XC as a biomarker is shown in FIG. The 15 cases on the left of the graph are healthy subjects, and the 15 cases on the right are breast cancer patients. FIG. 13 shows the discrimination result when the threshold value for distinguishing between a healthy subject and a breast cancer patient in the score is set to 0.65. In 14 cases, 15 healthy subjects can be estimated as healthy subjects, and the error rate at this time is 6.7%, whereas in 15 breast cancer patients, 15 cases can be estimated as breast cancer, and the error rate is 0%. .. This indicates that a promising biomarker can be extracted by the analysis flow shown in FIG.

In addition, for example, similar analysis was performed using the top 4, 5, 8, 9 and 11 metabolites (XC, XD, XH, XE, XI, XG) in the results of the random forest method. The result is as shown in FIG. Assuming that the threshold for distinguishing between healthy subjects and breast cancer patients in the score is 0.5, 15 healthy subjects can be estimated as healthy subjects and 15 breast cancer patients can be estimated as breast cancer, both of which are incorrect. The rate is 0% and the analysis accuracy is improved. Cancer testing using urinary metabolites is a major feature and merit that it can be finally used as a multibiomarker. This multibiomarker can also be easily extracted by the analysis flow shown in FIG.

Ultimately, whether to use a single biomarker or a multi-biomarker will be selected by the test side by presenting data on both analysis cost and analysis accuracy.

Using this cancer test model, a comprehensive analysis was performed on an unknown urine sample, and the discriminant obtained from the learning data (the sum of the values obtained by multiplying the intensity of the ion corresponding to the biomarker by a coefficient (including a constant)). )) By determining the numerical value, it is possible to determine whether the unknown urine sample is derived from a healthy person or a cancer patient. Alternatively, using urinary metabolites that are significantly different between cancer patients and healthy subjects as biomarkers, it is determined from LC / MS analysis data of urinary metabolites whether the urine sample is derived from healthy subjects or cancer patients. It is also possible.

Claims (15)

(A) A step of subjecting a urine sample to a liquid chromatograph mass spectrometer (LC / MS) and analyzing urinary metabolites in the urine sample.
(B) A step of quantitatively evaluating the importance of the urinary metabolite by a random forest method based on the analysis data of the urinary metabolite and selecting a urinary metabolite having a high importance.
(C) A step of performing a discriminant analysis method using the analysis data of the selected urinary metabolites.
(D) A method for searching for a urinary metametase marker, which comprises a step of determining a urinary metabolite associated with a specific disease or condition as a marker candidate based on the result of the discriminant analysis. In step (b), using the analysis data of the urinary metabolites, the urinary metabolites that are significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects are narrowed down by the Wilcoxon rank sum test. The method according to claim 1, wherein the importance of the urinary metabolite is quantitatively evaluated by a random forest method, and a urinary metabolite having a high importance is selected. The method according to claim 1, wherein the discriminant analysis method is an OPLS discriminant analysis method or a kernel discriminant analysis method. The method according to claim 1, wherein in step (c), a test model by a discriminant analysis method is constructed using the selected urinary metabolite as a biomarker. The method according to claim 4, further comprising a step of constructing a plurality of test models having different numbers of biomarkers based on the result of the discriminant analysis. Specific diseases or conditions include the presence or absence of cancer, the type of cancer, the degree of cancer progression, the malignancy of cancer, the stage of cancer, the prognosis of cancer, the therapeutic effect on cancer, and anticancer agents. The method of claim 1, wherein the method is selected from the group consisting of susceptibility to. A counting unit for inputting analysis data of urinary biotransformers by a liquid chromatograph mass spectrometer (LC / MS),
A calculation unit that performs numerical analysis using the analysis data of urinary metabolites,
A determination unit that selects urinary metabolites based on the results of the analysis and / or determines urinary metabolites as marker candidates.
It is provided with an output processing unit that processes the determination result by the determination unit for output.
The calculation unit quantitatively evaluates the importance of the urinary metabolite by a random forest method using the analysis data of the urinary metabolite, and a discriminant analysis method using the analysis data of the urinary metabolite. And
The determination unit selects urinary metabolites having high importance based on the result of the evaluation by the random forest method, and determines the urinary metabolite as a marker candidate based on the result of the discriminant analysis method. A system for searching for biotransformer markers. The calculation unit performs Wilcoxon rank sum test using the analysis data of the urinary metamorphosis.
7. The determination unit narrows down urinary metabolites that are significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects based on the result of the Wilcoxon rank sum test. The system described in. The system according to claim 7, wherein the discriminant analysis method is an OPLS discriminant analysis method or a kernel discriminant analysis method. The calculation unit performs a discriminant analysis method using the selected urinary metabolite as a biomarker.
The system according to claim 7, wherein the determination unit builds an inspection model by the discriminant analysis method. Urinary metabolite markers are the presence or absence of cancer, the type of cancer, the degree of cancer progression, the malignancy of cancer, the stage of cancer, the prognosis of cancer, the therapeutic effect on cancer, and anticancer agents. The system of claim 7, wherein the system is associated with at least one particular disease or condition selected from the group consisting of susceptibility to. On the computer
Based on the analysis data of urinary metabolites by liquid chromatograph mass spectrometer (LC / MS), the importance of the urinary metabolites is quantitatively evaluated by the random forest method, and the urinary metabolites of high importance are evaluated. Steps to select and
A step of performing a discriminant analysis method using the analysis data of the selected urinary biotransforms, and
A program for searching for a urinary biotransformer marker that executes a step of determining a urinary biotransformer associated with a specific disease or condition as a marker candidate based on the result of the discriminant analysis. The step of selecting urinary biotransformers of high importance is significantly increased or decreased in patients with a specific disease or condition with respect to healthy subjects using the analysis data of the urinary biotransformers. Is narrowed down by Wilcoxon's rank sum test, the importance of the urinary metabolite is quantitatively evaluated by the random forest method, and the urinary metabolite having high importance is selected. Described program. The program according to claim 12, wherein the discriminant analysis method is an OPLS discriminant analysis method or a kernel discriminant analysis method. The program according to claim 12, wherein the computer is further subjected to a step of constructing a test model by a discriminant analysis method using the selected urinary metabolite as a biomarker.

Images