JP2009066088A - Multivariable analyzer and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To support a researcher to comprehensively or complementarily analyze different types of data. <P>SOLUTION: A multivariable analyzer displays multivariable analysis results of one or more types of samples (data) on the same GUI. When a user specifies a specific sample 250 of a specific type on the GUI, this multivariable analyzer extracts and highlightedly displays other types of samples 252 and 254 having commonality with the specified sample 250, and detects and highlightedly displays marker variables 256, 258 and 259 unique to the samples 250, 252 and 254. When the user specifies the specific variable 256 of the specific type on the GUI, the multivariable analyzer extracts and highlightedly displays the sample 250 unique to the specific variable 256, extracts and highlightedly displays other types of samples 252 and 254 having the commonality with the sample 250, and detects and highlightedly displays the marker variables 258 and 259 unique to the samples 252 and 254. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an apparatus and a computer program for performing various analyzes mainly including multivariate analysis of data. The present invention can be applied to a wide variety of analysis applications, and examples thereof include analysis of life data for medicine, drug discovery, food or life science.

  Taking research such as medicine, drug discovery, food or life science as an example, there are various research fields such as phenome, metabonome, proteome, transcriptome, and genome. Naturally, the type of sample to be studied and the type of life data acquired from the sample (for example, the event represented by the data or the structure of the data) differ depending on the research field. Various primary analyzers for acquiring the desired life data from the sample first, such as NMR (nuclear magnetic resonance analyzer), MS (mass spectrometer), DNA array, MRI (magnetic resonance imaging) is there.

  By analyzing with NMR equipment, MS equipment and DNA array respectively, spectral data on different events (nuclear magnetic resonance, mass-to-charge ratio, DNA, etc.) of the constituent substances contained in the sample such as NMR spectrum, MS spectrum, DNA spectrum etc. are obtained. It is done. In addition, when using an MRI system, MRI images and a number of CSI spectrum data (a kind of NMR spectrum data) corresponding to a number of sites (voxels) in the MR image obtained by CSI (Chemical Shift Imaging) respectively. The MR image and CSI spectrum data have different data structures (the latter can be the target of multivariate analysis, but the latter cannot be normal).

  As a device for secondary analysis of life data using, for example, a statistical mathematics method, a device that has been configured to be suitable for analyzing a specific type of life data for a specific research field has been known. Yes. For example, Patent Document 1 discloses a data processing apparatus suitable for multivariate analysis of NMR spectra for metabonomics.

  Conventionally, much of the research in each field described above has been performed independently of research in other fields, where a specific type of life data is selected from the perspective unique to that field, and that specific type of life data is selected. Then, analysis and judgment are performed by a method based on the unique viewpoint of the field.

JP 2006-337354 A

  If the research field, research purpose, or researcher is different, the type (event or data structure) of life data to be handled will be different, or the analysis method or judgment method will be different. There are also variations in judgment due to differences in researchers. In research in each field, it is not often done to actively incorporate and utilize research results in other research fields, analysis results of other types of life data, or analysis results by other analysis methods. .

  However, the different levels of life phenomena studied in different fields, such as phenome, metabonome, proteome, transcriptome, and genome, do not exist independently in vivo, but are complex (and many Are connected to each other by (unknown) causality. Also, for the same, similar or related samples, there is some (and many unknown) association between life data obtained from different primary analyzers such as NMR instruments, MS instruments, DNA arrays and MRI instruments. Will exist.

  Therefore, if a single researcher (or a group of researchers) was able to analyze life data from different fields, or different types of life data, comprehensively or complementarily. For example, new markers with higher accuracy can be discovered, and unknown correlations between life phenomena in different fields or between different types of life data can be elucidated. Is expensive. This is not possible with a conventional analysis apparatus because the number of types of data to be handled is large and the number of data is enormous.

  Similar problems may exist in the analysis of data for other application fields other than life data.

  Accordingly, an object of the present invention is to assist researchers in analyzing different types of data comprehensively or in a complementary manner.

  Another object of the present invention is that researchers have obtained all the information obtained (for example, analysis results of life data, additional data obtained separately from the life data to be analyzed, and various other related knowledge). Etc.) to visually grasp the characteristics of the behavior of interest and the correlation between different fields or different types of information, and to make a comprehensive judgment.

A multivariate analyzer according to one aspect of the present invention is:
Target data storage means for storing the target data group, the target data group of the sample group, each target data having a plurality of variable values;
A calculation / control unit having a multivariate analysis unit that performs multivariate analysis using the target data group; and a feature extraction unit that performs a feature extraction process using the target data group;
Display means for visually displaying the result of the multivariate analysis and the result of the feature extraction processing on a graphical user interface (hereinafter abbreviated as GUI);
The multivariate analysis means of the calculation / control means generates a plurality of multivariate analysis charts representing characteristics of the sample group and characteristics of the plurality of variables found by the multivariate analysis,
The display means displays the plurality of charts on the GUI,
The feature extraction unit of the calculation / control unit specifies an arbitrary sample sub group in the sample group or an arbitrary variable in the plurality of variables, and the specified sample sub group is based on the target data group. Variables characteristic of the group, sample subgroups characteristic of the designated variable, or other sample groups if the target data group of the other sample groups is also stored in the target data storage means Detecting a sample subgroup associated with said designated sample subgroup,
The display means includes the designated sample subgroup or the designated variable and the detected characteristic variable, or the detected characteristic sample subgroup, or the detected related sample sub. Groups are highlighted on the GUI.

In one embodiment of the multivariate analyzer of the present invention,
The multivariate analysis performed by the multivariate analysis means includes principal component analysis,
In the plurality of multivariate analysis charts displayed on the GUI,
A. A sample plot chart in which each sample of the sample group is plotted at coordinates according to a score for a predetermined principal component in a coordinate space consisting of a predetermined principal component axis,
B. Variable plot chart in which the plurality of variables are plotted in coordinates according to loading for a predetermined principal component in a coordinate space consisting of a predetermined principal component axis;
C. A contribution chart illustrating contribution ratios of the plurality of variables for a predetermined principal component, and
D. a loading chart illustrating the sum of loadings for a given principal component of the plurality of variables;
At least two of the
The display means highlights and displays the designated sample subgroup or the detected related sample subgroup on the sample plot chart, and displays the detected marker variable in the variable plot. • Highlight and display on at least one of the chart, the contribution ratio chart, and the loading chart.

In one embodiment of the multivariate analysis apparatus of the present invention, the feature extraction unit of the calculation / control unit includes:
Sample designating means for designating an arbitrary sample subgroup from the sample group;
Marker variable detection means for detecting one or more characteristic variables common to the specified sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated sample subgroup and the detected marker variable on the GUI.

  In the above-described embodiment, the marker variable detecting unit further obtains a distribution of values of each variable in the entire target data group of the sample group, and distributes the values of each variable in the obtained entire target data group. And one or more characteristic variables common to the designated sample subgroup are detected based on the values of the variables of the target data of the designated sample subgroup.

  In the above embodiment, the multivariate analysis performed by the multivariate analysis unit includes principal component analysis, and the marker variable detection unit performs the principal component analysis on the detected marker variable. Filtering based on the contribution ratio obtained by the analysis is performed to narrow down to marker variables with higher accuracy.

In one embodiment of the multivariate analyzer of the present invention,
The feature extraction means of the calculation / control means is:
Marker variable designating means for designating an arbitrary variable among the plurality of variables as a marker variable;
Characteristic sample detection means for detecting one or more characteristic sample subgroups common to the designated marker variable based on the value of each variable of the target data group of the sample group;
The display means highlights and displays the designated marker variable and the detected characteristic sample subgroup on the GUI.

  In the above-described embodiment, the feature sample detection means further obtains a distribution of values of each of the specified marker variables in the entire target data group, and determines the specified in the entire target data group thus obtained. The characteristic sample sub-group is detected based on the distribution of the values of the marker variables and the values of the designated marker variables included in the target data of the samples.

  In the above embodiment, the multivariate analysis performed by the multivariate analysis means includes principal component analysis, and the feature sample detection means performs the principal component analysis on the designated marker variable. Filtering based on the contribution ratio obtained by the analysis is performed to narrow down to marker variables with higher accuracy.

In one embodiment of the multivariate analyzer of the present invention,
Further comprising additional data storage means for storing one or more additional data groups of the sample group;
The feature extraction means of the calculation / control means is:
Additional data designating means for designating an arbitrary additional data group from the one or more additional data groups;
Characteristic sample detection means for detecting one or more characteristic sample subgroups common to the designated type of additional data group from the sample group based on the value of the designated type of additional data group; ,
Marker variable detection that detects one or more characteristic variables common to the detected characteristic sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group Means,
Have
The display means highlights and displays the detected characteristic sample sub-group and the detected marker variable on the GUI.

In the above embodiment, further,
The feature sample detection means obtains a distribution of the value of the additional data in the entire additional data group of the specified type, obtains a distribution of the value of the additional data in the entire additional data group thus obtained, Detecting the characteristic sample subgroup based on the value of the additional data;
The marker variable detection means obtains a distribution of values of each variable in the entire target data group, and obtains a distribution of values of each variable in the obtained entire target data group and the detected characteristic sample sub One or more characteristic variables common to the designated sample subgroup are detected based on the value of each variable included in the target data of the group.

In one embodiment of the multivariate analyzer of the present invention,
Further comprising additional data storage means for storing one or more additional data groups of the sample group;
The multivariate analysis means of the calculation / control means performs at least a principal component analysis as the multivariate analysis, and based on a result of the principal component analysis, in a coordinate space composed of one or more principal component axes, A sample plot chart in which each sample of the sample group is plotted at coordinates according to a score for a predetermined main component is created as one of the multivariate analysis charts,
The display means displays the sample plot chart on the GUI;
The feature extraction means of the calculation / control means is:
Additional data designating means for designating an arbitrary principal component axis from the predetermined one or more principal component axes and designating an optional additional data group from the one or more additional data groups;
Based on the result of the principal component analysis and the value of the specified type of additional data group, the specified principal component axis in the sample plot chart is the value of the specified type of additional data group. And a principal component axis replacement means for creating a replacement sample / plot / chart that is replaced by a coordinate axis corresponding to
The display means displays the replacement sample / plot chart on the GUI.

In one embodiment of the multivariate analyzer of the present invention,
The target data storage means is
A first target data group of a first sample group, wherein each target data has a plurality of variable values;
A second target data group of the second sample group, each of the target data having a plurality of variable values, and storing the second target data group; and the first and second sample groups Between the samples that are common to each other,
The multivariate analysis means of the calculation / control means performs a first multivariate analysis using the first target data group and a second multivariate analysis using the second target data group, Creating one or more first charts showing the results of the first multivariate analysis and one or more second charts showing the results of the second multivariate analysis;
The display means displays the first chart and the second chart on the GUI,
The feature extraction means of the calculation / control means is:
Sample designating means for designating an arbitrary first sample sub-group from the first sample group;
An association for searching the second sample group for a second sample subgroup that is common to the designated first sample subgroup based on the association between the first and second sample groups Sample search means;
Have
The display means highlights and displays the designated first sample subgroup and the detected second sample subgroup on the GUI.

In the above embodiment, further,
The feature extraction means of the calculation / control means is:
Based on the value of each variable of the first target data group, the plurality of the first target data group have one or more characteristic variables common to the designated first sample subgroup. First marker variable detecting means for detecting as a first marker variable from among the variables of
Based on the value of each variable of the second target data group, the plurality of the second target data group has one or more characteristic variables common to the searched second sample subgroup. A second marker variable detecting means for detecting as a second marker variable from the variables of
The display means highlights and displays the detected first and second marker variables on the GUI.

In one embodiment of the multivariate analyzer of the present invention,
The target data storage means is
A first target data group of a first sample group, wherein each target data has a plurality of variable values;
A second target data group of the second sample group, each of the target data having a plurality of variable values, and storing the second target data group; and the first and second sample groups Between the samples that are common to each other,
The multivariate analysis means of the calculation / control means performs a first multivariate analysis using the first target data group and a second multivariate analysis using the second target data group, Creating one or more first charts showing the results of the first multivariate analysis and one or more second charts showing the results of the second multivariate analysis;
The display means displays the first chart and the second chart on the GUI,
The feature extraction means of the calculation / control means is:
Marker variable designating means for designating an arbitrary variable among the plurality of variables of the first target data group as a marker variable;
Based on the value of each variable of the first target data group, one or more characteristic first sample subgroups common to the designated marker variable are detected from the first sample group. A feature sample detection means;
An association for searching the second sample group for a second sample subgroup that is common to the detected first sample subgroup based on the association between the first and second sample groups Sample search means;
Have
The display means highlights and displays the designated marker variable, the detected first characteristic sample sub-group, and the searched second characteristic sample sub-group on the GUI. To do.

In the above embodiment, further,
The feature extraction means of the calculation / control means is:
Based on the value of each variable of the second target data group, the plurality of the second target data group has one or more characteristic variables common to the searched second sample subgroup. Second marker variable detecting means for detecting as a related marker variable from among the variables of
The display means includes not only the designated marker variable, the detected first characteristic sample subgroup and the searched second characteristic sample subgroup, but also the detected related marker variable. Is also highlighted on the GUI.

In one embodiment of the multivariate analysis apparatus of the present invention, the multivariate analysis apparatus further comprises another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Data portion specifying means for specifying an arbitrary data portion from among the plurality of data portions of the separate structure data;
An associated sample detection means for detecting a sample subgroup associated with the designated data portion from the sample group;
Marker variable detection means for detecting one or more characteristic variables common to the detected sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated data portion, the detected sample subgroup, and the searched marker variable on the GUI.

In one embodiment of the multivariate analyzer of the present invention,
Further comprising another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Sample designating means for designating an arbitrary sample subgroup from the sample group;
Related data portion detection means for detecting a data portion associated with the designated sample subgroup from the plurality of data portions of the separate structure data;
Marker variable detection means for detecting one or more characteristic variables common to the designated sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated sample subgroup, the detected data portion, and the searched marker variable on the GUI.

In one embodiment of the multivariate analyzer of the present invention,
Further comprising another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Marker variable designation means for designating an arbitrary variable as a marker variable from among the plurality of variables;
Feature sample detection means for detecting one or more characteristic sample subgroups common to the designated marker variable from the sample group based on the values of the variables of the target data group;
Related data portion detection means for detecting a data portion associated with the detected characteristic sample subgroup from the plurality of data portions of the different structure data;
Have
The display means highlights and displays the designated marker variable, the detected characteristic sample subgroup, and the detected data portion on the GUI.

  The present invention also provides a computer program for causing a computer to function as the above-described multivariate analyzer.

  According to the present invention, it becomes easy for researchers to analyze data of different kinds of events or dimensions comprehensively or complementarily.

  Hereinafter, a multivariate analyzer according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 shows the overall processing flow of the multivariate analyzer according to this embodiment.

In the multivariate analysis apparatus according to the present embodiment, at least a part of its functions is typically realized by a computer in which a computer program for causing the computer to perform the processing shown in FIG. 1 is installed. However, as a modification, at least a part of the functions of the multivariate analysis apparatus of the present invention may be realized by a dedicated word logic circuit.
As shown in FIG. 1, the multivariate analysis apparatus according to this embodiment can perform an original data input process 100, an original data process 102, a multivariate analysis process 104, and a feature extraction process 106.

  In the original data input processing 100, one or more types of external sensing devices (for example, in the case of life data, an NMR analyzer, an MS device, a DNA array, an MRI device, an X-ray CT imaging device, or various types The original data for a large number of samples obtained by using a biochemical examination apparatus or the like is input to the multivariate analysis apparatus.

  In the original data processing 102, based on the input original data of each sample, data for visually confirming the actual state represented by the original data (hereinafter referred to as “actual data”), multivariate analysis Direct target data (hereinafter referred to as “target data”) is generated. The actual data is, for example, NMR spectrum data, MS spectrum data, DNA spectrum data, MRI image data, X-ray CT image data, etc. On the other hand, the target data is, for example, by reducing the above-mentioned various spectrum data. This is histogram data composed of a predetermined number of variable values obtained. The actual data and target data of each sample are stored in this multivariate analyzer.

  As a specific process of the original data process 102, a known process can be used. For example, when the original data is FID (free induction decay) data obtained from an NMR apparatus, for example, processing for converting FID data into NMR spectrum data as actual data in a predetermined format by Fourier transform, and bucket integration To reduce the NMR spectrum data to NMR histogram data as target data in a predetermined format. In addition, for example, when the original data is spectrum data obtained from an MS device or a DNA array device, a process of converting the spectrum data into spectrum data as actual data in a predetermined format, and the spectrum data as target data in a predetermined format There is a process of reducing the data to histogram data. For example, when the original data is a set of an MRI image obtained from an MRI apparatus and a large number of CSI spectrum data (NMR spectrum data) corresponding to a large number of voxels in the MRI image, the MRI image is converted into a predetermined format. There is a process of converting into an MRI image as actual data and a process of converting CSI spectrum data for each voxel into CSI histogram data as target data in a predetermined format by bucket integration.

  In the multivariate analysis processing 104, multivariate analysis is performed using the target data group (histogram data group) of a large number of samples (sample groups) obtained in the data processing 102, and the analysis result is displayed on this display screen. The As a specific method of multivariate analysis, known methods such as PCA (principal component) analysis and SIMCA analysis can be used. For example, as will be described in detail later, a principal component analysis is performed, and based on the analysis results, a sample plot chart, a variable plot chart, a contribution ratio chart, a loading chart, a spectrum chart, etc. are created. Then, an operation of displaying them on the display screen of the multivariate analyzer can be performed.

  In the feature extraction processing 106, feature extraction analysis is performed based on the analysis result obtained in the multivariate analysis processing 104. This feature extraction processing 106 will be described in detail later.

  In the present embodiment, a single computer program performs all of the processes 100 to 106 described above. In particular, the original data processing 102 and the multivariate analysis processing 104 are performed by a single computer program, so that the multivariate analysis result by the multivariate analysis processing 104 and the actual data stored by the original data processing 102 (for example, as described above) Link to various spectral data, MRI image data, etc.), for example, when a PCA analysis is performed, a sample on the displayed sample plot chart When the user designates, the actual data of the sample (for example, NMR spectrum data) is read, and the image of the actual data (for example, the spectrum chart) is displayed together with the sample / plot chart, etc. , The user can simultaneously display the multivariate analysis results on the same display screen. Actual data of the desired samples associated with analysis results (i.e., real contents of the original data of the sample) can be confirmed by visually.

  Therefore, the multivariate analysis apparatus according to the present embodiment is suitable for the overall analysis of “mixture-based” samples. For example, various fields such as metabolite diagnosis, toxicity prediction, marker search drug discovery, combichem, food, polymer, etc. -Available in the market. Furthermore, the multivariate analyzer according to the present embodiment can be applied to other analysis purposes.

  In the following, the configuration and function of the multivariate analyzer according to the present embodiment will be described in more detail by taking as an example a case where it is used for analysis of life data in research fields such as medicine and life science.

The basic functions of this multivariate analyzer are as follows. In other words, this multivariate analysis apparatus can input any kind of life data, and with respect to the input life data, a multivariate analysis is used as a main tool, and a plurality of types of feature extraction analysis (feature extraction of FIG. 1) is performed. Processing 106) can be performed. For the multiple types of feature analysis extraction,
(1) Detection of feature variables and correlations based on analysis results of one event data,
(2) Detection of feature variables and correlations based on analysis results of multiple event data, and
(3) There is detection of feature variables and correlations based on the analysis results of data with different structures.

  The user's operation and display to the user for all the processes of reference numbers 100 to 106 shown in FIG. 1 are realized by using one user interface displayed on one display screen.

  FIG. 2 shows the basic configuration of this multivariate analyzer.

  As shown in FIG. 2, the multivariate analyzer 110 includes a calculation / control module 112 and a display module 114. The calculation / control module 112 performs processing as shown in FIG. The display module 114 creates a graphical user interface and displays it on the display screen, and displays the processing result of the calculation / control module 112 on the user interface. The calculation / control module 112 performs the requested processing in response to a request 115 input by a user operation with a mouse or a keyboard on the displayed graphical user interface.

  When the original data input process 100 and the original data process 102 shown in FIG. 1 are performed, the calculation / control module 112 performs various actual data (for example, NMR spectrum data, MS spectrum data, (DNA spectrum data, CSI spectrum data, MRI image data, etc.) 116 and 118 are stored in a storage device in the analysis device 110 in a multivariate manner. FIG. 2 shows two types of actual data 116 and 118. One type of actual data 116 has a predetermined standard data structure (for example, a structure called spectral data) (for example, a structure of spectral data). , NMR spectral data, MS spectral data, DNA spectral data, CSI spectral data, etc., and the other type of actual data 118 has other data structures (for example, MRI image data, X-ray CT image data) Etc.). In this embodiment, as one example, actual data of one event of one sample is managed as one file (in FIG. 2, one block given the reference number 116 or 118). For example, MNR spectrum data obtained by examining one sample with an NMR apparatus is managed as one file.

  Further, the calculation / control module 112 creates one or more analysis tables 120 when the original data processing 102 shown in FIG. 1 is performed, and stores them in the storage device in the multivariate analysis apparatus 110. In this embodiment, as an example, one analysis table 120 is created for one event. For example, one analysis table 120 is created for NMR and another analysis table 120 is created for MS. In each analysis table 120, a target data group related to one corresponding event is registered. The calculation / control module 112 performs a basic analysis operation using the analysis table 120.

  As indicated by dotted arrows in FIG. 2, each target data registered in each analysis table 120 and each corresponding actual data 116 or 118 are linked information ( (For example, directory names such as folder names and file names, or keywords such as identification codes). When accessing the certain analysis table 120, the calculation / control module 112 uses the link information of the individual target data registered in the analysis table 120 to use the individual actual data 116 corresponding to the individual target data. Or 118 can be accessed.

  In addition, between the plurality of analysis tables 120, the target data of the common sample (for example, the same sample or different samples obtained from the same sample source (for example, blood sample and urine sample obtained from the same person)) Are also associated with each other by assigning the same keyword to each. When accessing the plurality of analysis tables 120, the calculation / control module 112 searches the target data with the same keyword from the respective analysis tables 120 to find sample data that are common to each other. be able to.

  In response to the request 115 from the user interface, the calculation / control module 112 uses the target data group described in the analysis table 120 of the requested event, and features the multivariate analysis processing 104 shown in FIG. An extraction process 106 can be performed. The display module 114 is a result of the multivariate analysis process 104 by the calculation / control module 112 (for example, a sample plot chart, a variable plot chart, a spectrum chart, a contribution chart, a loading chart, etc.) and a feature extraction process 106. (Eg, feature variables and correlations) are visually displayed on the display screen.

  FIG. 3 shows the format of the analysis table 120.

  In the analysis table 120 shown in FIG. 3, the index / comment column 122 describes model names and arbitrary comments about the data group registered in the analysis table 120.

  In the data processing condition column 124, analysis conditions such as a range of variables to be subjected to multivariate analysis and a range not to be analyzed (for example, a chemical shift range in the case of an NMR histogram) are described.

  In the data analysis condition column 126, the “PCA” field has the model order in the PCA analysis, the “SIMCA” field has the model order in the SIMCA analysis, the “X” field has the order of the PC (principal component) axis displayed on the X axis, “Y "Field describes the degree of the PC axis displayed on the Y axis," Mean_centering "field indicates whether or not mean centering processing is performed, and" Auto_scaling "field indicates whether or not auto scaling processing is performed.

  In the data size column 128, the total number of samples (target data) registered in the analysis table 120 is described in the “Sample” field, and the total number of variables of the target data group is described in the “Variable” column.

  In the data scale column 130, the scale of each variable (the bucket range of bucket integration) included in the target data group registered in the analysis table 120 is described.

  In the actual data / file column 132, the actual data directory (save folder name and file name) corresponding to each target data registered in the analysis table 120 is described.

  The plot type field 134 describes the shape of the mark (dot) when displaying the mark (dot) of each sample (target data) on the display screen (for example, the sample plot chart of the PCA analysis result). In the color column 136, the display color of the mark is described.

  The keyword column 138 describes a keyword for identifying each sample (target data) registered in the analysis table 120.

  In the additional data column 140, additional data relating to each sample (target data) registered in the analysis table 120 (for example, various biochemical data (weight, blood pressure, etc.) about the sample obtained separately) is described. Is done.

  The plot status column 142 describes the state of each sample (target data) registered in the analysis table 120 (for example, “0” normally means “1” means deletion).

  In the target data column 144, target data (histogram data which is a column of variable values) registered in the analysis table 120 is described. In FIG. 3, one row corresponds to one sample of target data.

  Next, the configuration of a graphical user interface (GUI) displayed on the display screen by the multivariate analyzer 110 will be described.

As described above, the multiple types of feature analysis extraction performed by the multivariate analysis apparatus 110 include:
(1) Detection of feature variables and correlations based on analysis results of one event data,
(2) Detection of feature variables and correlations based on analysis results of multiple event data, and
(3) There is detection of feature variables and correlations based on the analysis results of data with different structures.

  FIG. 4 shows an example of a GUI in the case of performing “detection of feature variable and correlation based on analysis result of one event data”, and FIG. FIG. 6 shows an example of a GUI in the case of performing “detection of feature variables and correlations based on the analysis result of data having different structures”. In either case, PCA analysis is performed as a variable analysis.

  As shown in FIGS. 4 to 6, in any case, the GUI has a plurality of windows. A common window displayed in any case is a control panel 150 for a user to input a request to the multivariate analyzer 110. In the other windows, as will be described in detail below, the characteristics of the sample group determined by multivariate analysis (for example, if PCA analysis is performed, the score for each PC (principal component) of each sample) And charts showing the characteristics of the variables (for example, when PCA analysis is performed, the charts illustrating the loading and contribution ratio of each variable to each PC (principal component)) and the charts illustrating the spectra of specific samples Is done. All of these charts are created by the calculation / control module 112 shown in FIG. 2 and displayed in a plurality of windows on the GUI by the display module 114. The user can select which of these charts are displayed and which are not displayed.

  As shown in FIG. 4, when “detection of feature variables and correlations based on the analysis result of one event data” is performed using PCA analysis, a window included in the GUI includes, for example, a sample plot window 152 Variable plot window 154, contribution chart window 156, loading chart window 158, spectrum chart window 160, and view navigator window 162.

  The sample plot window 152 displays a sample plot chart created as a result of the PCA analysis. A sample plot chart is a coordinate space composed of coordinate axes (hereinafter referred to as PC axes) respectively corresponding to one or more and three or less PCs (principal components) specified by the user (in the example shown, 3 In the three-dimensional space of a PC axis), each sample is a chart in which coordinates are plotted at coordinates according to the scores for the above-described one or more and three or less PCs. When a three-dimensional sample / plot chart as illustrated in FIG. 4 is displayed, the multivariate analyzer 110 rotates the three-dimensional sample / plot chart around a plurality of PC axes. , Allowing the user to view the 3D spatial sample plot chart from any line of sight. Thereby, the user can grasp | ascertain exactly the mode of distribution of a sample that the whole sample group is divided into some subgroups, for example.

  The variable plot window 154 displays a variable plot chart created as a result of the PCA analysis. A variable plot chart is a coordinate space composed of coordinate axes (hereinafter referred to as PC axes) corresponding to one or more and three or less PCs designated by the user (the example shown is composed of two PC axes). In the (two-dimensional space), each variable is a chart in which coordinates are plotted at coordinates corresponding to the loading of the above one or more and three or less PCs.

  The contribution rate chart window 156 displays a contribution rate chart created as a result of the PCA analysis. The contribution rate chart is a graph (for example, a bar graph) in which the horizontal axis corresponds to the variable and the vertical axis corresponds to the contribution rate for each of the variables of 1 to 3 PCs specified by the user. It is a chart expressed in shape.

  The loading chart window 158 displays a loading chart created as a result of PCA analysis. The loading chart is the sum of the absolute value and the residual component of the loading for each variable of 1 to 3 PCs specified by the user, with the horizontal axis corresponding to the variable and the vertical axis representing loading. It is a chart represented in the form of a corresponding graph (for example, a bar graph).

  In the spectrum chart window 160, a spectrum chart is displayed. A spectrum chart is a chart representing actual data (spectrum) of one or more samples in the form of a graph (for example, a line graph) in which the horizontal axis corresponds to a variable and the vertical axis corresponds to a variable value.

  The contribution rate chart window 156, the loading chart window 158, and the spectrum chart window 160 are arranged so that the scales on the horizontal axis coincide with each other.

  In the view navigator window 162, to which part of the entire range the horizontal axis range (variable range) currently displayed in the contribution rate chart 156, loading chart 158, and spectrum chart 160 corresponds. Is displayed in the entire view of the spectrum displayed on the spectrum chart 160.

  As shown in FIG. 5, when “detection of feature variables and correlations based on the analysis result of data of multiple events” is performed using PCA analysis, each event has the same type as the window shown in FIG. 4. All or part of the window is displayed.

  For example, the GUI example shown in FIG. 5 corresponds to the case where the data of three types of events (A, B, C) of the same sample group or different sample groups is analyzed. Sample plot window 152 (152A, B, C), contribution chart window 156 (156A, B, C), spectrum chart window 160 (160A, B, C), and view navigator window 162 ( 162A, B, C) are displayed. The user can select what type of window set to display.

  As can be seen from the example of FIG. 5, the same type of windows are displayed for all the events, and they are arranged in the same positional relationship. When different event data is analyzed for the same sample group, the spectrum chart window 160 (160A, B, C) displays actual data of different events for one or more of the same samples selected by the user. Can do. For users, it is possible to compare the analysis results and actual data of multiple events, thereby finding the correlation between multiple events, or judging comprehensively or complementarily from the viewpoint of multiple events It becomes easy to find a marker with high accuracy.

  As shown in FIG. 6, when “detection of feature variables and correlations based on the analysis result of data of different structure” is performed, data of a standard structure (for example, spectral data) subjected to multivariate analysis is obtained. A window displaying the results of multivariate analysis (all or part of the same type of window as shown in FIG. 4) and the contents of actual data (for example, image data) of another structure that was not subject to multivariate analysis Is displayed.

  For example, the GUI example shown in FIG. 6 corresponds to the case where MRI image data and CSI spectrum data of a large number of voxels (each voxel is handled as a sample) in the MRI image are analyzed. Shows a separate structure data window 164 displaying MRI image data, a sample plot window 152 as a result of multivariate analysis (for example, PCA analysis) of CSI spectrum data of many voxels, a variable plot window 154, contribution A rate chart window 156, a loading chart window 158, a spectrum chart window 160, and a view navigator window 162 are displayed.

  When the user designates an arbitrary location in the MRI image data displayed in the separate structure data window 164 with the pointing device, the CSI spectrum data of the voxel (sample) corresponding to the designated location is displayed in the spectrum chart window. 160.

  It becomes easy for the user to find a correlation between data having different structures, or to find a marker with high accuracy by comprehensively or complementarily judging from the viewpoint of data having different structures.

Next, the three types of feature extraction functions described above, that is,
(1) “Detection of feature variables and correlations based on the analysis result of one event data” function,
(2) “Detection of feature variables and correlations based on analysis results of multiple event data” function, and
(3) Each of the “detection of feature variables and correlations based on the analysis result of data of different structure” functions will be described more specifically.

1 “Detection of Feature Variables and Correlation Based on Analysis Result of 1 Event Data” Function This function includes a plurality of sub-functions, and each sub-function depends on how the user operates the control panel 150. Selected. These sub-functions include “detection of marker variable by extracting feature variable of specific sample”, “detection of feature sample by specified marker variable”, “complementary analysis # 1 of analysis data and additional data”, “analysis Complementary analysis # 2 of data and additional data ”. Hereinafter, each sub-function will be described.

1.1 “Detection of Marker Variables by Extracting Feature Variables of Specific Samples” Sub-Function It is possible to see the trend of the entire analyzed sample group by statistical methods such as multivariate analysis. If this sample group seems to be divided into multiple sub-groups, one of the next challenges of the analysis process considered by the user is “what characteristics for each divided sample sub-group? Is there, in particular, which variables contribute to characterize the subgroups? ” To solve this problem, this sub-function separates the specified sample sub-group from other samples by comparing the data of the sample sub-group specified by the user with the data of other samples. Some feature variables that may contribute to the extraction are extracted, noise information is removed using the contribution rate, and feature variables with higher accuracy are extracted. This subfunction derives specific feature variables for the specified sample subgroup.

  FIG. 7 shows the flow of processing of this sub-function, and FIG. 8 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 7 will be described with reference to FIG. 8 together.

1.1.1 Step 170: Designation of Sample Subgroup The multivariate analysis apparatus 110 reads a target data group of a sample group of a certain event from the corresponding analysis table 120, and performs multivariate analysis such as PCA analysis. , And display the analysis result on one GUI. From the window displaying the sample group in the GUI, for example, the sample plot window 152 representing the sample plot chart obtained by the PCA analysis as shown in FIG. The device is used to specify any sample of interest (eg, a sub-group of samples that are separated from other samples) 178. Then, the multivariate analyzer 110 highlights the designated sample subgroup 178 on the sample plot window 152. Further, the multivariate analysis apparatus 110 refers to the analysis table 120 of the sample group, reads the actual data (spectrum data) of the specified sample using the link described in the analysis table 120, and As shown in FIG. 8, the spectrum of the sample designated based on the actual data is displayed in the spectrum window 160.

1.1.2 Step 172: Extracting Characteristic Variables Common to the Specified Sample Subgroup Several characteristic variables common to the samples specified in Step 170 are calculated by the multivariate analyzer 110. Here, a “characteristic variable (characteristic variable)” of a certain sample is determined that the value of the variable that the sample has is clearly larger or smaller than the value of the variable that the other sample has. Variable. As a method for discriminating such variables, various statistical methods can be employed. In this embodiment, a method using an average value as described below is employed.

  A group of n samples each having target data of d dimensions (that is, having d variables) can be represented as an n × d multivariate data matrix X as follows. Each row in the matrix X corresponds to the target data of each sample.

この行列Xにおけるｋ列目（k番目の変数）のサンプル群全体での平均値x_{k_avg}は次式で示される。

An average value x _{k_avg} in the entire sample group of the k-th column (k-th variable) in the matrix X is _expressed by the following equation.

ここで、ｋ列目（k番目の変数）のサンプル群全体内での最小値をx_{k_min}、最大値をx_{k_max}とする。また、特徴的か否かを判別するために設定される閾値レベルを、それぞれx_{k_min_th}、x_{k_max_th}とする。このサンプル群におけるｋ列目（k番目の変数）の値の分布は、例えば図９Aまたは図９Bに示されるようになる。

Here, the minimum value in the entire sample group of the k-th column (k-th variable) is x _{k_min} , and the maximum value is x _{k_max} . In addition, threshold levels set to determine whether or not they are characteristic are x _{k_min_th} and x _{k_max_th} , respectively. The distribution of the values in the k-th column (k-th variable) in this sample group is as shown in FIG. 9A or FIG. 9B, for example.

Variable to be extracted as the feature variable are those having an area value falling (called "feature space" following these regions) between the region or x _{K_max_th} and x _{k_max} between x _{K_min} and _x k_min_th. Depending on the variable, the distribution of the values may show a shape close to a normal distribution as shown in FIG. 9A, or may show a considerably biased shape as shown in FIG. 9B. Extracting feature variables means that a variable having a distribution clearly divided into groups concentrated in the feature region as shown in FIG. 9B is derived.

  As shown in FIG. 9B, when a specific variable belongs to any feature region in common to all of the samples (designated sample subgroups) 178 designated from the sample group, the variable is designated. Since it can be said that the feature variable is common to the sample subgroups, such a feature variable is extracted.

  Therefore, the process of checking whether the variable falls into any feature region in common with all the specified subgroups is executed for all rows (all variables) of the matrix X, and the check result is yes. The characteristic variables common to the designated sample subgroup are extracted by extracting the actual variables.

1.1.3 Step 174: Filtering by Contribution In this step 174, for each of the feature variables obtained in step 172, the variable differentiates the specified sample subgroup from the other samples. Evaluate whether it is a really meaningful variable using another parameter such as contribution rate. In particular, a random and accidentally selected variable such as noise is detected using a prior parameter such as a contribution rate obtained in advance, and is removed. Specifically, if the contribution rate of a certain variable is below a certain level, that variable is removed. As a result, only feature variables that are really meaningful (more accurate) in terms of the above are narrowed down. Such characteristic variables are hereinafter referred to as “marker variables”.

1.1.4 Step 176: Display of Feature Variable In this step 176, the marker variable extracted in step 174 is highlighted and displayed so as to be visually identifiable on the GUI. For example, as illustrated in FIG. 8, a band 180 of a predetermined color is attached only to the extracted marker variable through the spectrum chart window 160, the contribution chart window 156, and the loading chart window 158. Also, in the variable plot window 154, the marker variables are displayed as dots 182 of a different color from other variables. Thereby, the user can easily identify those marker variables visually on the GUI. Preferably, the marker variable is close to the maximum value shown in FIG. 9A (that is, the value of that variable that the specified sample subgroup has is clearly greater than that of the other samples) or close to the minimum value ( In other words, whether the value of the variable of the specified sample subgroup is clearly smaller than that of the other samples is displayed using different colors (for example, the color of the band 180 attached to the former is orange) The color of the band 180 attached to the latter is blue). As a result, the user can easily visually grasp the quantitative characteristics of the designated sample subgroup.

1.2 “Detection of Feature Samples Using Specified Marker Variables” Sub-Function The above-mentioned “Detection of Marker Variables by Extracting Feature Variables of Specific Samples” sub-function is a method for searching for marker variables common to specified sample sub-groups. . On the other hand, the “Sample feature distribution detection by significant difference test of specified marker” sub-function searches for which sample sub-group the specified marker variable has a specific influence by the opposite approach to the previous sub-function. It is a technique to do. In this sub-function, the designated marker variable is filtered by the contribution rate, and then the sample subgroup in which the designated marker variable is characteristically extracted is extracted.

  FIG. 10 shows the flow of processing of this sub-function, and FIG. 11 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 10 will be described with reference to FIG. 11 together.

1.2.1 Step 190: Designation of Marker Variable The multivariate analysis apparatus 110 reads a target data group of a certain sample group of a certain event from the corresponding analysis table 120, and performs multivariate analysis such as PCA analysis. And display the analysis result on one GUI. In a window in which variables are displayed on the GUI, such as the variable plot window 154, the user designates any one or more variables 198 of interest as marker variables (contribution plot window 156 or Marker variables may be specified in the loading chart window 158.) In the example shown in FIG. 11, the user is interested in one or more variables 198 that are isolated and isolated from other variables on the variable plot window 154, and the variable 198 is used as a marker variable. It is specified.

1.2.2 Step 192: Filtering by Contribution Rate For the variable specified in Step 190 above, whether or not the variable is meaningful is evaluated using another parameter such as the contribution rate. In particular, a variable that is randomly and accidentally specified, such as noise, is detected based on an evaluation result based on a contribution rate obtained in advance, and is excluded from the marker variable. Specifically, by excluding variables whose parameter values such as contribution rate are below a certain level, only variables that are truly meaningful are narrowed down as marker variables.

  Filtered marker variables are highlighted and displayed for visual identification on the GUI. For example, as illustrated in FIG. 11, a predetermined color band 200 is attached only to the marker variables through the spectrum chart window 160, the contribution chart window 156, and the loading chart window 158, and In the variable plot window 154, the marker variables are displayed with different colored dots from the other variables.

1.2.3 Step 194: Extraction of Feature Sample Subgroup Common to Designated Marker Variable One or more characteristic samples (sample subgroup) common to the marker variable determined in step 192 are detected. Here, a “characteristic sample” for a marker variable is determined to be that the value of the marker variable of that sample is clearly larger or smaller than the value of the marker variable of another sample. This is a sample. Although various statistical methods can be adopted as the determination method, this embodiment employs a method using an average value as described below.

That is, the distribution of the value of each marker variable in the entire sample group is calculated based on the entire target data group of the sample group. For a certain marker variable, for example, a value distribution as shown in FIG. 12 is obtained. Based on the distribution of the value of each marker variable, a feature region (between x _{k_min} and x _{k_min_th )} on the side of the minimum value x _{k_min} for that marker variable in the same manner as described with reference to FIGS. 9A and 9B. Region) and a feature region on the side of the maximum value x _{k_max} (a region between x _{k_max_th} and x _{k_max} ) are determined. Then, based on the value of the marker variable of the target data of each sample, the value of the marker variable enters one of the feature areas (that is, the value of the marker variable of the sample Samples 202, 204 are extracted (obviously large or small). This process is performed for all the marker variables, and by narrowing down only the samples in which all the marker variable values fall in any feature region, a feature sample subgroup common to those marker variables is extracted.

1.2.4 Step 196: Display of Feature Sample Subgroup Feature samples 202, 204 finally extracted in step 194 above are high on the GUI sample plot window 152 so that they can be visually identified. Written and displayed. Further, the actual data of the feature samples are read out, and their spectra are displayed on the spectrum chart window 160. Preferably, the former feature sample 202 and the latter feature sample can be visually discriminated whether the value of the designated marker variable of each feature sample 202, 204 is close to the maximum value or the minimum value. 204 is displayed in a different color. For example, on the sample plot window 152, the color of the table of the former feature sample 202 is orange, and the display color of the latter feature sample 204 is blue. Further, for example, on the sample plot window 152, the value of the marker variable of each of the feature samples 202 and 204 is also displayed. Thereby, the quantitative feature of the designated marker variable can be easily grasped visually.

1.3 “Complementary analysis # 1 of analysis data and additional data” sub-function Most conventional analysis methods are based on a single point of view (eg, using one primary analyzer for one event). Based on the data obtained from the analysis, a unique conclusion is being sought. However, more accurate conclusions can be drawn only by analyzing from many different viewpoints, collecting various related information and data, and interpreting the information (knowledge) comprehensively or complementarily. Moreover, new knowledge will be acquired by seeing the correlation between the information (knowledge) obtained from each viewpoint. This sub-function organically combines the results obtained from analysis data from a certain point of view with information (additional data) from one or more other points related to it, and their correlation is combined into a single GUI. It is a function that is presented so that it can be visually understood. This sub-function is a function that automatically extracts samples characteristic of specified additional data and automatically finds marker variables specific to those samples. Thereby, the user can easily grasp the correlation between the additional data and the variable visually. The additional data is registered in advance in the additional data column 140 in the analysis table 120 shown in FIG.

  FIG. 13 shows the flow of processing of this sub-function, and FIG. 14 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 13 will be described with reference to FIG. 14 together. In this description, a case where a certain kind of biochemical data is used as additional data is taken as an example.

1.3.1 Step 210: Specifying Additional Data (Biochemical Data) The multivariate analysis apparatus 110 reads a target data group of a certain sample group of a certain event from the corresponding analysis table 120, and performs multivariate analysis, For example, PCA analysis is performed, and the analysis result is displayed on one GUI. Using the additional data selection menu 220 on the control panel 150 of the GUI, the user selects any kind of additional data of interest (for example, weight difference data which is a kind of biochemical data).

1.3.2 Step 212: Extraction of Specific Sample Subgroup From the analysis table 120, the value of designated biochemical data of the entire sample group is read. A sample group for each of the designated biochemical data in the same manner as that performed for the designated marker variable in step 194 described in section 1.2.3 above (the method described with reference to FIG. 12). The overall distribution of values is calculated, and characteristic samples (sample subgroups) 220, 222 common to the specified biochemical data are extracted based on the distribution. The feature sample 220 whose designated biochemical data value is close to the maximum value and the feature sample 222 close to the minimum value are displayed on the sample plot window 152 in different colors so that they can be distinguished. Actual data of the feature samples 220 and 222 are read out, and their spectra are displayed on the spectrum chart window 160.

1.3.3 Step 214: Extraction of Feature Variables in Feature Sample Subgroup Characteristic common to the feature sample subgroup extracted in step 212 in the same manner as step 172 described in section 1.1.2. Variable is extracted.

1.3.4 Step 216: Filtering by Contribution Ratio Filtering by the same method as step 174 described in section 1.1.3 is applied to the feature variable extracted in step 214, and the marker variable is It is determined. The marker variable is highlighted and displayed by a band 224 of a predetermined color, for example, through the spectrum chart window 160, the contribution chart window 156, and the loading chart window 158 so as to be visually distinguishable.

1.3.5 Displaying marker variables In the same manner as in step 176 described in section 1.1.4, the marker variable is displayed in a different color from the other variables so that it can be visually identified. It is displayed on the window 154. These marker variables can be estimated to have a higher correlation with designated additional data (biochemical data) than other variables.

1.4 “Complementary analysis # 2 of analysis data and additional data” sub-function This sub-function is similar to the sub-function described in section 1.3, and results obtained from analysis data from a certain point of view. This is a function that organically combines information (additional data) from one or more other viewpoints related thereto and presents such correlation so that it can be visually understood by one GUI. This sub-function is a function that plots sample groups in the coordinate space with the PC (principal component) obtained by PC (principal component) analysis and the specified additional data as coordinate axes and displays them on the GUI. is there.

  FIG. 15 shows the flow of processing of this sub-function, and FIG. 16 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 15 will be described with reference to FIG. 16 together. In this description, a case where a certain kind of biochemical data is used as additional data is taken as an example.

1.4.1 Step 230: Designation of Additional Data (Biochemical Data) The multivariate analysis apparatus 110 reads a target data group of a certain sample group of a certain event from the corresponding analysis table 120, multivariate analysis, For example, PCA analysis is performed, and the analysis result is displayed on one GUI. The user uses the three coordinate (x, y, z) selection menu 234 on the GUI control panel 150 to select any type of additional data of interest (eg, biochemistry) on at least one arbitrary coordinate axis. Specify weight difference data, which is a kind of data), and specify any PC obtained by PC (principal component) analysis for the remaining coordinate axes. In the example shown in FIG. 15, a weight difference, which is one of biochemical data, is designated on the x axis, and the same PC is designated on the y axis and the z axis (different between the y axis and the z axis). Of course, you can specify a PC.)

1.4.2 Step 232: Replacement of PC Axis The multivariate analyzer 110 reads the data of the designated additional data of the sample group from the analysis table 120 of the sample group to be analyzed, and the sample plot window 152 Of the plurality of PC axes constituting the displayed coordinate space, the PC axis specified by the user is replaced with a coordinate axis (hereinafter referred to as an additional data axis) corresponding to the value of the additional data specified by the user. Then, a sample plot chart is created by re-plotting the sample group in the coordinate space composed of the additional data axis and the remaining PC axis, and the sample plot chart is converted into the sample plot. -Display in window 152. In the example illustrated in FIG. 15, the weight difference is designated on the x-axis, and the same PC is designated on the y-axis and the z-axis (that is, the coordinate axes are substantially two). The plot window 152 displays a two-dimensional sample plot chart (vertical axis is weight difference, horizontal axis is specified PC), but different components are specified for each of the three axes x, y, and z. Then, a three-dimensional sample / plot chart is displayed.

  By viewing this chart, the user can easily grasp the correlation between any additional data and any PC (or variable) visually. For example, in the sample plot window 152 shown in FIG. 15, there is an isolated sample subgroup 36 separated from other sample groups, and this sample subgroup 236 has a weight difference (vertical axis). It is gathered in the vicinity of the specific value and the specific value of the score of the specified PC (horizontal axis). This leads to the speculation that there may be a meaningful association between the weight difference and some variable characteristic of this sample subgroup 236. Therefore, the user may detect a marker variable common to the sample sub-group 236 using the sub-function of 1.1.

2 “Detection of feature variables and correlations based on analysis results of multiple event data” function This function includes “Detection of correlations between analysis data of different events and marker variables” sub-function and “Detection of different events in specified marker variables” "Extract Behavior" sub-function is included. Hereinafter, these sub-functions will be described.

2.1 “Correlation between analysis data of different events and marker variable detection” sub-function This sub-function uses multiple available analysis / evaluation methods to understand a phenomenon and Data is collected and analyzed, making it easy to explore the whole picture of the phenomenon. In general, there is no analysis method / evaluation method that is all-purpose, so researchers understand the pros and cons of each method and use the data (information) obtained by each method. It is processed and analyzed, and finally one conclusion (inference) is drawn in the researcher's head by combining information obtained by different methods comprehensively and complementarily. This sub-function is a function for displaying different event data (information) obtained from various methods so that the mutual correlation can be easily grasped on one GUI. In particular, this sub-function picks up samples (sample data) that are common between different events and detects and displays marker variables specific to samples that are common between these different events. As a result, the user can easily and comprehensively combine information obtained by different methods.

  FIG. 17 shows the flow of processing of this sub-function, and FIG. 18 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 17 will be described with reference to FIG. 18 together.

2.1.1 Step 240: Designation of Sample Subgroup The multivariate analysis apparatus 110 reads target data groups of a plurality of different events from the analysis table 120 of each event, and performs multivariate analysis, for example, PCA analysis for each event. , And the analysis results of these multiple events are displayed on one GUI. For example, on the GUI shown in FIG. 17, sample plot windows 152A to 152C and spectrum chart window 160A are obtained as PCA analysis results for three types of events AC from the viewpoint of metabolome, proteome, and genome. ~ 160C and contribution chart windows 156A to 156C etc. are displayed in the same arrangement throughout the three events.

  The user designates any sample (sample subgroup) 250 of interest on a window in which the sample group for any one event, for example, event A, is displayed, for example, the sample plot window 152A. Then, the designated sample subgroup 250 is highlighted on the sample plot window 152A of the event A in the same manner as in the step 170 described in 1.1.1 above, and the designated sample subgroup 250 is designated. The spectrum of the actual data of the sample subgroup 250 is displayed on the spectrum chart window 162A of the event A.

2.1.2 Step 242: Linking Different Event Data The multivariate analysis apparatus 110 uses the keywords assigned to each sample registered in the analysis table 120 of events A, B, and C (in short, different events A and B). , C, and other events B, C that are common to the data of the sample subgroup 250 specified for the event A in step 240 above. The data of the sample subgroups 252 and 254 are searched. Here, sample data “common” between different events is, for example, data obtained by analyzing different types of samples (eg, blood, cells, chromosomes) obtained from the same sample source (eg, the same person). Alternatively, it means data obtained by analyzing the same sample (for example, a mixture) with different primary analyzers (for example, NMR apparatus, MS apparatus, DNA array, etc.).

  Then, the multivariate analysis apparatus 110 highlights the data of the searched common sample subgroups 252 and 254 on the sample plot windows 152B and 152C of the respective events B and C, and The spectrum obtained from the actual data of the common sample subgroups 252 and 254 is displayed on the spectrum chart windows 160B and 160C of the respective events B and C. As a result, common sample sub-group data regarding different events A, B, and C are correlated with each other.

  In order to perform processing for associating sample data common to different events A, B, and C in this way, as shown in FIG. 19, analysis tables 120A, 120B, 120C for those events A, B, and C are used. The same character string or number is assigned as a keyword to common sample data. For example, in the example of FIG. 19, the sample A of the event A, the sample F of the event B, and the sample J of the event C are common samples.

  As illustrated by an arrow in FIG. 19, the sample name specified by the user is searched in the analysis table 120A of one event A, the keyword assigned to the searched sample name is read, and another event is detected. By examining the sample names to which the keywords are assigned from the analysis tables 120B and 120C for B and C, it is possible to identify samples that are common between different events A, B, and C.

2.1.3 Steps 244A, B, C to 248A, B, C: Detection and display of marker variables for each event In the same manner as described in the above section 1.1.2 to 1.1.4, For each of events A, B, and C, extraction of feature variables common to the specified sample subgroup (event A) and its associated sample subgroup (events B and C), filtering by contribution rate, and display of marker variables Is performed. As a result, as illustrated in FIG. 18, marker variables 256, 258, and 259 detected in those events A, B, and C are highlighted.

  Users can easily visually identify marker variables specific to samples that are common between different events A, B, and C, and have some meaningful association between the marker variables of events A, B, and C (eg causal It is easy to know the possibility that a relationship) exists. In this way, for example, the correlation between different levels of research fields such as metabolome, proteome and genome, or between data obtained by different types of primary analyzers such as NMR apparatus, MS apparatus, DNA array, etc. Research to obtain new knowledge can be performed more easily.

2.2 “Extracting behavior of different events in specified marker variables” sub-function This sub-function, like the sub-function in 2.1 above, is also available with multiple analysis / evaluation methods that can be used to understand a phenomenon. Makes it easy to collect and analyze data from the perspective of different events and explore the whole picture of the phenomenon. This sub-function, especially when a particular marker variable in a particular event is specified, picks up a sub-group of samples characterized by that marker variable from that particular event and other event sample groups, and Detect and display marker variables specific to foreign event samples. As a result, the user can easily and comprehensively combine information obtained by different methods.

  FIG. 20 shows the flow of processing of this sub-function, and FIG. 21 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 20 will be described with reference to FIG. 21 together.

2.2.1 Step 260: Designation of Marker Variable The multivariate analysis apparatus 110 reads the target data groups of a plurality of different events from the analysis table 120 of each event, as in the case of the subfunction of 2.1 above. Multivariate analysis is performed for each event, for example, PCA analysis, and the analysis results of these multiple events are displayed on one GUI. For example, on the GUI shown in FIG. 21, sample plot windows 152A to 152C, spectrum chart windows 160A to 160C, and contribution rate chart window 156A are obtained as PCA analysis results for three types of events A to C. ˜156C etc. are displayed in the same arrangement throughout the three events.

  The user designates any variable 280 of interest as a marker variable on a window in which variables are displayed for any one event, such as event A, for example, the contribution chart window 156A (variable plot window 154 or (The loading chart window 158 may be displayed, and the marker variable may be specified in any window.)

2.2.2 Steps 262 to 266: Extraction of Feature Sample Subgroups For the one event A, the designated marker variable is extracted in the same manner as described in the above sections 1.2.2 to 1.2.4. Filtering based on the contribution ratio of, extraction of feature sample subgroups common to marker variables narrowed down by filtering, and display of the extracted feature sample subgroups are performed. As illustrated in FIG. 21, the marker variable 280 narrowed down by the filtering by the contribution rate is highlighted so that it can be easily identified visually. Further, as illustrated in FIG. 21, a feature sample sub-group 282 having the marker variable 280 in the feature region close to the maximum value of the distribution of the marker variable 280, and the marker variable in the feature region close to the minimum value of the distribution. Feature sample sub-group 284 having 280 is extracted, and the two feature sample sub-groups 282, 284 are highlighted in different colors on the sample plot window 152A so that they can be visually distinguished. The Further, the spectrum obtained from the actual data of the feature sample subgroups 282 and 284 is displayed in the spectrum chart window 160A.

2.2.3 Step 268: Concatenation of Different Event Data The feature sample subgroups 282 and 284 extracted for event A in steps 262 to 264 above in the same manner as described in section 2.1.2 above. Data for the sample subgroups 286, 282; 290, 292 of other events B, C that are in common with the data are retrieved. The data of the retrieved common sample subgroups 286, 282; 290, 292 are similarly highlighted on the respective event B, C sample plot windows 152B, 152C. Further, spectra obtained from actual data of the sample subgroups 286, 282; 290, 292 having the commonality are displayed on the spectrum chart windows 160B, 160C of the respective events B, C. This correlates data for the same sample subgroup for different events A, B, C.

2.3.4 Steps 270B, C to 274B, C: Detection and Display of Marker Variables of Other Events In the same manner as described in Sections 1.1.2 to 1.1.4 above, in Step 268 above For each of the retrieved sub-groups 286 and 282 of other events B and C; 290 and 292, extraction of feature variables common to the sample sub-groups, filtering by contribution ratio, and display of marker variables are performed. . As a result, as illustrated in FIG. 21, marker variables 294 and 296 detected in other events B and C are highlighted.

  The user can easily know the possibility that some meaningful relationship (for example, a causal relationship) exists between the marker variables of the highlighted events A, B, and C.

3 “Detection of feature variables and correlations based on analysis results of data with different structure” This function is a further development of the above-mentioned concept of analysis of data on different events, and is handled as described above. Data having a structure suitable for multivariate analysis (or a phenomenon grasped from data having such a structure) and data having a different structure (for example, image data or data having such a different structure) By examining the correlation with the phenomenon) and extracting features, it becomes easy to obtain completely new information that could not be obtained so far. Below, as a specific example, CSI spectrum data (structure data suitable for multivariate analysis) obtained by the CSI (Chemical Shift Imaging) function of the MRI apparatus and MRI image data (data of another structure) obtained from the MRI apparatus ) Will be described. Here, the CSI spectrum data is roughly chemical spectrum data (a kind of NMR spectrum data) of each of a large number of sites (voxels) in the MRI image.

  In this example, the multivariate analysis apparatus 110 includes MRI image data (corresponding to the actual data 118 of another structure shown in FIG. 2) and spectrum data of multiple parts in the MRI image obtained by CSI (FIG. 2). Corresponding to the target data group in the analysis table 120 shown in FIG. 5), multivariate analysis, for example, PCA analysis, is applied to the input spectrum data of a large number of parts, and the multivariate analysis result and The correlation with the position information in the MRI image is derived.

  Conventionally, focusing on a specific substance / component extracted as a characteristic of a specific spectrum, evaluation is performed based on the number or increase / decrease. On the other hand, according to this function, not only the substance / component of each spectrum but also the pattern of the whole spectrum (for example, MRI image) represented by the separate structure data is recognized, and the feature is extracted from the whole pattern. This makes it possible to obtain a completely different interpretation. In particular, there is a possibility of identifying an abnormal site and investigating and diagnosing the cause of the abnormality, which have been difficult until now.

  This function includes a “correlation detection between different structure data # 1” sub-function, a “correlation detection between different structure data # 2” sub-function, and a “correlation detection between different structure data # 3” sub-function. There is. Hereinafter, these sub-functions will be described.

3.1 “Detection of Correlation between Different Structure Data # 1” Subfunction FIG. 22 shows the flow of processing of this subfunction, and FIG. 23 explains the processing of this subfunction using the GUI. Hereinafter, each step shown in FIG. 22 will be described with reference to FIG. 23 together.

3.1.1 Step 300: Designation of Specific Sample The multivariate analyzer 110 reads MRI image data (corresponding to the actual data 118 of another structure shown in FIG. 2) and creates an MRI image based on the MRI image data. And display this on the GUI. Further, the multivariate analysis apparatus 110 reads CSI histogram data groups (target data groups shown in FIG. 2) of a large number of regions, that is, voxels (hereinafter referred to as samples) in the MRI image displayed on the GUI, and performs CSI. Multivariate analysis of histogram data group, for example, PCA analysis is performed, and the analysis result is displayed on the same GUI. For example, on the GUI illustrated in FIG. 23, an MRI image is displayed in a separate structure data window 164, and a sample plot window 152, a variable plot window 154, a contribution chart window 156, and a loading chart are displayed. In the window 158, PCA analysis results of spectrum data of a large number of samples (voxels) in the MRI image are displayed.

  The user designates an arbitrary sample (voxel) 312 of interest on the MRI image displayed in the separate structure data window 164 of the GUI.

3.1.2 Step 302: Linking Different Structure Data The multivariate analysis apparatus 110 uses the position information (identification number) on the MRI image of the sample (voxel) specified in step 300 and the keyword of the analysis table 120. And the position of the designated sample (voxel) on the MRI image and the multivariate analysis result of the target data of the sample (voxel) are associated with each other and connected.

  That is, as illustrated in FIG. 24, in the CSI spectrum analysis table 120, position information (identification number) of each sample on the MRI image is registered in advance as a keyword of each sample (voxel). The multivariate analysis apparatus 110 calculates the position information (identification number) on the MRI image of the specified sample (voxel) from the position on the MRI image specified by the user, and assigns the position information as indicated by the arrows in FIG. The target data of the sample (voxel) having the keyword corresponding to the issued position information (identification number) is searched from the analysis table 120, and the position of the specified sample (voxel) on the MRI image is searched. The target data of the sample (voxel) is visually associated on the GUI.

  The above association is performed as follows, for example. That is, as illustrated in FIG. 23, the designated sample (voxel) 312 is highlighted and displayed on the MRI image on the separate structure data window 164, and on the sample plot window 152, The sample (target data) 314 searched as described above is highlighted and displayed. Further, CSI spectrum data (corresponding to the actual data 116 of the standard structure shown in FIG. 2) of the retrieved sample (voxel) 314 is read out, and based on this, the spectrum of the retrieved sample (voxel) 314 is obtained. Are generated and their spectral images are displayed in the spectrum chart window 160.

3.1.3 Steps 304-308: Detection and Display of Marker Variables Samples (voxels) retrieved in the same manner as described in the above-mentioned sections 1.1.2 to 1.1.4 (ie, Feature variables common to the designated sample sub-group) 314 are extracted, and filtering is performed on the contribution variables and the like, so that marker variables specific to the finally designated sample (voxel) are detected, The detected marker variable is highlighted and displayed as illustrated in FIG.

3.2 “Detection of Correlation Between Different Structure Data # 2” Sub-Function FIG. 25 shows the flow of processing of this sub-function, and FIG. 26 explains the processing of this sub-function using the GUI. The steps shown in FIG. 25 will be described below with reference to FIG.

3.2.1 Step 320: Designation of Specific Samples As in the case of “Detect Correlation between Different Structure Data # 1” sub-function in section 3.1 described above, MRI images and a number of samples in MRI images ( In the state where the multivariate analysis result of the target data group (CSI histogram data group) of the voxel is displayed on the GUI, the user can select any sample ( Voxel) 330 is designated.

3.2.2 Step 322: Linking Different Structure Data The multivariate analysis apparatus 110 uses the keyword on the analysis table 120 of the sample (voxel) 330 specified in step 320 and the position information (identification) on the MRI image. And the position of the sample (voxel) on the MRI image are linked to each other and linked to each other.

  That is, as illustrated in FIG. 27, a keyword of a designated sample (voxel) is read from the analysis table 120, and a sample (voxel) on the MRI image having position information (identification number) corresponding to the keyword is determined. Then, the target data of the designated sample (voxel) and the sample (voxel) on the determined MRI image are visually associated on the GUI.

  The above association is performed as follows, for example. That is, as illustrated in FIG. 26, the specified sample (target data) 330 is highlighted on the sample plot window 152 and displayed as described above on the separate structure data window 164. The sample (voxel) 332 on the determined MRI image is highlighted and displayed. Further, CSI spectrum data (corresponding to the actual data 116 of the standard structure shown in FIG. 2) corresponding to the designated sample (voxel) 330 is read out, and based on this, the spectrum of those samples (voxel) is read. Images are generated and their spectral images are displayed in the spectral chart window 160.

3.2.3 Steps 324 to 328: Detection and Display of Marker Variable Specified sample (target data) (specified) in the same manner as described in the above section 1.1.2 to 1.1.4 Feature variables common to the sample subgroups 330) are extracted, and filtering is performed on the contribution variables and the like, so that marker variables specific to the finally designated sample (target data) are detected, The detected marker variable is highlighted and displayed as illustrated in FIG.

3.3 “Detection of Correlation between Different Structure Data # 3” Sub-Function FIG. 28 shows the flow of processing of this sub-function, and FIG. 29 explains the processing of this sub-function using the GUI. Hereinafter, each step shown in FIG. 28 will be described with reference to FIG. 29 together.

3.3.1 Step 340: Specify Marker Variable Similar to the above-mentioned “Detect Correlation between Different Structure Data # 1” sub-function in section 3.1, MRI image and a number of samples in MRI image ( In the state where the multivariate analysis result of the target data group (CSI histogram data group) of the voxel is displayed on the GUI, the user can select any arbitrary one of interest on the variable plot window 154 of the GUI. The above variable 360 is designated as a marker variable (the marker variable may be designated in the contribution rate plot window 156 or the loading chart window 158).

3.3.2 Steps 342-346: Extraction and Display of Feature Sample Subgroups The specified marker variable 360 is processed in the same manner as described in Sections 1.2.2 to 1.2.3 above. Are filtered by the contribution rate to the feature sample sub-groups 364 and 366 (feature sample sub-groups 364 whose marker variable values are close to the minimum value and feature samples close to the maximum value). Subgroup 366) is extracted, and the extracted feature sample subgroups 364, 366 are highlighted and displayed on the sample plot window 152. At this time, the feature sample subgroup 364 whose marker variable value is close to the minimum value and the feature sample subgroup 366 close to the maximum value are displayed in different colors, for example, so that they can be distinguished at a glance. Further, the spectrum images of the extracted feature sample subgroups 364 and 366 are displayed on the spectrum chart window 160.

2.3.3 Step 348: Linking Different Structure Data MRI corresponding to the feature sample subgroups 364 and 366 extracted in step 346 in the same manner as described in the above section 3.2.2. Corresponding feature sample subgroups (voxel subgroups) 368, 370 on the image are determined, and the target data of the extracted feature sample subgroups 364, 366 and the feature samples on the determined MRI image Subgroups (subgroups of voxels) 368 and 370 are visually associated and linked on the GUI.

2.3.4 Step 350: Display of Related Sample Subgroups on Separate Structure Data As a result of the association in step 348 above, feature samples (voxel) subgroups (voxel subgroups) 368, 370 on the MRI image are , Highlighted on the separate structure data window 164 and displayed. At this time, the feature sample subgroup 368 whose marker variable value is close to the minimum value and the feature sample subgroup 370 close to the maximum value are displayed in different colors, for example, so that they can be distinguished at a glance.

  By utilizing the sub-functions of “detection of correlation between different structure data # 1” to “detection of correlation between different structure data # 3” as described above, the user can change multivariate data like various spectrum data. It is possible to visually recognize the correlation between structural data that can be analyzed and structural data that is not subject to multivariate analysis, such as image data. It is easier to discover and discover new knowledge through complementary and integrated observation and analysis.

  The preferred embodiment of the present invention has been described above, but this is an example for explaining the present invention, and the scope of the present invention is not limited to the embodiment. The present invention can be implemented in various modes different from the above-described embodiments without departing from the gist thereof.

  For example, the present invention can be applied to analysis of various data other than life data as exemplified in the embodiment.

The flowchart which shows the flow of the whole process of the multivariate analyzer which concerns on one Embodiment of this invention. The block diagram which shows the basic composition of this multivariate analyzer. The figure which shows the format of the analysis table 120. FIG. The figure which shows the example of GUI in the case of performing "detection of the characteristic variable and correlation based on the analysis result of 1 event data." The figure which shows the example of GUI in the case of performing "detection of the characteristic variable and correlation based on the analysis result of the data of multiple events". An example of a GUI in the case of performing “detection of feature variables and correlation based on analysis result of data of different structure” is shown. The flowchart which does not show the flow of a process of the "detection of a marker variable by extraction of the characteristic variable of a specific sample" subfunction. The figure explaining the process of the same subfunction using GUI. The figure which shows distribution of the variable value for demonstrating the process which extracts a characteristic variable. The flowchart which does not show the flow of the "detection of the feature sample by the designated marker variable" sub function processing. The figure explaining the process of the same subfunction using GUI. The figure which shows distribution of the variable value for demonstrating the process which extracts a feature sample. The flowchart which shows the flow of a process of the "complementary analysis # 1 of analysis data and additional data" subfunction. The figure explaining the process of the same subfunction using GUI. The flowchart which shows the flow of a process of a "complementary analysis # 2 of analysis data and additional data" subfunction. The figure explaining the process of the same subfunction using GUI. The flowchart which shows the flow of a process of the "detection of the correlation between the analysis data of different events, and a marker variable" subfunction. The figure explaining the process of the same subfunction using GUI. The figure explaining the process which searches the data of the sample which is common between different events. The flowchart which shows the flow of a process of the "extraction of the behavior by the different event in a designated marker variable" subfunction. The figure explaining the process of the same subfunction using GUI. The flowchart which shows the flow of a process of the "detection of correlation between different structure data # 1" subfunction. The figure explaining the process of the same subfunction using GUI. The figure explaining the process which associates different structure data. The flowchart which shows the flow of a process of the "detection of correlation between different structure data # 2" subfunction. The figure explaining the process of the same subfunction using GUI. The figure explaining the process which associates different structure data. The flowchart which shows the flow of a process of the "detection of correlation between different structure data # 3" subfunction. The figure explaining the process of the same subfunction using GUI.

Explanation of symbols

110 Multivariate Analyzer 112 Calculation / Control Module 114 Display Module 115 UI Request 116 Actual Data of Standard Structure 118 Actual Data of Different Structure 120 Analysis Table 150 Control Panel 152 Sample Plot Window 154 Variable Plot Window 156 Contribution Rate chart window 158 Loading chart window 160 Spectrum chart window 162 View navigator window 164 Separately structured data window

Claims (19)

Target data storage means for storing the target data group, the target data group of the sample group, each target data having a plurality of variable values;
A calculation / control unit having a multivariate analysis unit that performs multivariate analysis using the target data group; and a feature extraction unit that performs a feature extraction process using the target data group;
Display means for visually displaying the result of the multivariate analysis and the result of the feature extraction processing on a graphical user interface (hereinafter abbreviated as GUI);
The multivariate analysis means of the calculation / control means generates a plurality of multivariate analysis charts representing characteristics of the sample group and characteristics of the plurality of variables found by the multivariate analysis,
The display means displays the plurality of charts on the GUI,
The feature extraction unit of the calculation / control unit specifies an arbitrary sample sub group in the sample group or an arbitrary variable in the plurality of variables, and the specified sample sub group is based on the target data group. Variables characteristic of the group, sample subgroups characteristic of the designated variable, or other sample groups if the target data group of the other sample groups is also stored in the target data storage means Detecting a sample subgroup associated with said designated sample subgroup,
The display means includes the designated sample subgroup or the designated variable and the detected characteristic variable, or the detected characteristic sample subgroup, or the detected related sample sub. Groups are highlighted and displayed on the GUI,
Multivariate analyzer. The multivariate analyzer according to claim 1,
The multivariate analysis performed by the multivariate analysis means includes principal component analysis,
In the plurality of multivariate analysis charts displayed on the GUI,
A. A sample plot chart in which each sample of the sample group is plotted at coordinates according to a score for a predetermined principal component in a coordinate space consisting of a predetermined principal component axis,
B. Variable plot chart in which the plurality of variables are plotted in coordinates according to loading for a predetermined principal component in a coordinate space consisting of a predetermined principal component axis;
C. A contribution chart illustrating contribution ratios of the plurality of variables for a predetermined principal component, and
D. a loading chart illustrating the sum of loadings for a given principal component of the plurality of variables;
At least two of the
The display means highlights and displays the designated sample subgroup or the detected related sample subgroup on the sample plot chart, and displays the detected marker variable in the variable plot. Highlighting and displaying on at least one of the chart, the contribution chart, and the loading chart;
Multivariate analyzer. In the multivariate analyzer according to claim 1 or 2,
The feature extraction means of the calculation / control means is:
Sample designating means for designating an arbitrary sample subgroup from the sample group;
Marker variable detection means for detecting one or more characteristic variables common to the specified sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated sample subgroup and the detected marker variable on the GUI.
Multivariate analyzer. In the multivariate analyzer according to claim 3,
The marker variable detection means obtains the distribution of the values of each variable in the entire target data group of the sample group, the distribution of the values of each variable in the entire target data group thus obtained, and the designated sample sub Detecting one or more characteristic variables common to the designated sample subgroup based on the value of each variable of the target data of the group;
Multivariate analyzer. In the multivariate analyzer according to claim 3 or 4,
The multivariate analysis performed by the multivariate analysis means includes principal component analysis,
The marker variable detection means performs filtering based on the contribution ratio obtained by the principal component analysis on the detected marker variable, and narrows down to a marker variable with higher accuracy.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 5,
The feature extraction means of the calculation / control means is:
Marker variable designating means for designating an arbitrary variable among the plurality of variables as a marker variable;
Characteristic sample detection means for detecting one or more characteristic sample subgroups common to the designated marker variable based on the value of each variable of the target data group of the sample group;
The display means highlights and displays the designated marker variable and the detected characteristic sample subgroup on the GUI.
Multivariate analyzer. The multivariate analyzer according to claim 6,
The feature sample detection means obtains a distribution of each value of the designated marker variable in the entire target data group, and obtains a distribution of each value of the designated marker variable in the obtained whole target data group. And detecting the characteristic sample sub-group based on each of the designated marker variables of the target data of each sample.
Multivariate analyzer. In the multivariate analyzer according to claim 6 or 7,
The multivariate analysis performed by the multivariate analysis means includes principal component analysis,
The feature sample detection means performs filtering based on the contribution ratio obtained by the principal component analysis for the designated marker variable, and narrows down to a marker variable with higher accuracy.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 8,
Further comprising additional data storage means for storing one or more additional data groups of the sample group;
The feature extraction means of the calculation / control means is:
Additional data designating means for designating an arbitrary additional data group from the one or more additional data groups;
Characteristic sample detection means for detecting one or more characteristic sample subgroups common to the designated type of additional data group from the sample group based on the value of the designated type of additional data group; ,
Marker variable detection that detects one or more characteristic variables common to the detected characteristic sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group Means,
Have
The display means highlights and displays the detected characteristic sample sub-group and the detected marker variable on the GUI.
Multivariate analyzer. The multivariate analyzer according to claim 9, wherein
The feature sample detection means obtains a distribution of the value of the additional data in the entire additional data group of the specified type, obtains a distribution of the value of the additional data in the entire additional data group thus obtained, Detecting the characteristic sample subgroup based on the value of the additional data;
The marker variable detection means obtains a distribution of values of each variable in the entire target data group, and obtains a distribution of values of each variable in the obtained entire target data group and the detected characteristic sample sub Detecting one or more characteristic variables common to the designated sample subgroup based on the value of each variable of the target data of the group;
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 10,
Further comprising additional data storage means for storing one or more additional data groups of the sample group;
The multivariate analysis means of the calculation / control means performs at least a principal component analysis as the multivariate analysis, and based on a result of the principal component analysis, in a coordinate space composed of one or more principal component axes, A sample plot chart in which each sample of the sample group is plotted at coordinates according to a score for a predetermined main component is created as one of the multivariate analysis charts,
The display means displays the sample plot chart on the GUI;
The feature extraction means of the calculation / control means is:
Additional data designating means for designating an arbitrary principal component axis from the predetermined one or more principal component axes and designating an optional additional data group from the one or more additional data groups;
Based on the result of the principal component analysis and the value of the specified type of additional data group, the specified principal component axis in the sample plot chart is the value of the specified type of additional data group. And a principal component axis replacement means for creating a replacement sample / plot / chart that is replaced by a coordinate axis corresponding to
The display means displays the replacement sample / plot chart on the GUI.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 11,
The target data storage means is
A first target data group of a first sample group, wherein each target data has a plurality of variable values;
A second target data group of the second sample group, each of the target data having a plurality of variable values, and storing the second target data group; and the first and second sample groups Between the samples that are common to each other,
The multivariate analysis means of the calculation / control means performs a first multivariate analysis using the first target data group and a second multivariate analysis using the second target data group, Creating one or more first charts showing the results of the first multivariate analysis and one or more second charts showing the results of the second multivariate analysis;
The display means displays the first chart and the second chart on the GUI,
The feature extraction means of the calculation / control means is:
Sample designating means for designating an arbitrary first sample sub-group from the first sample group;
An association for searching the second sample group for a second sample subgroup that is common to the designated first sample subgroup based on the association between the first and second sample groups Sample search means;
Have
The display means highlights and displays the designated first sample subgroup and the detected second sample subgroup on the GUI.
Multivariate analyzer. The multivariate analyzer according to claim 12,
The feature extraction means of the calculation / control means is:
Based on the value of each variable of the first target data group, the plurality of the first target data group have one or more characteristic variables common to the designated first sample subgroup. First marker variable detecting means for detecting as a first marker variable from among the variables of
Based on the value of each variable of the second target data group, the plurality of the second target data group has one or more characteristic variables common to the searched second sample subgroup. A second marker variable detecting means for detecting as a second marker variable from the variables of
The display means highlights and displays the detected first and second marker variables on the GUI.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 13,
The target data storage means is
A first target data group of a first sample group, wherein each target data has a plurality of variable values;
A second target data group of the second sample group, each of the target data having a plurality of variable values, and storing the second target data group; and the first and second sample groups Between the samples that are common to each other,
The multivariate analysis means of the calculation / control means performs a first multivariate analysis using the first target data group and a second multivariate analysis using the second target data group, Creating one or more first charts showing the results of the first multivariate analysis and one or more second charts showing the results of the second multivariate analysis;
The display means displays the first chart and the second chart on the GUI,
The feature extraction means of the calculation / control means is:
Marker variable designating means for designating an arbitrary variable among the plurality of variables of the first target data group as a marker variable;
Based on the value of each variable of the first target data group, one or more characteristic first sample subgroups common to the designated marker variable are detected from the first sample group. A feature sample detection means;
An association for searching the second sample group for a second sample subgroup that is common to the detected first sample subgroup based on the association between the first and second sample groups Sample search means;
Have
The display means highlights and displays the designated marker variable, the detected first characteristic sample sub-group, and the searched second characteristic sample sub-group on the GUI. To
Multivariate analyzer. The multivariate analyzer according to claim 14,
The feature extraction means of the calculation / control means is:
Based on the value of each variable of the second target data group, the plurality of the second target data group has one or more characteristic variables common to the searched second sample subgroup. Second marker variable detecting means for detecting as a related marker variable from among the variables of
The display means includes not only the designated marker variable, the detected first characteristic sample subgroup and the searched second characteristic sample subgroup, but also the detected related marker variable. Is also highlighted and displayed on the GUI,
Multivariate analyzer. The multivariate analyzer according to claim 1,
Further comprising another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Data portion specifying means for specifying an arbitrary data portion from among the plurality of data portions of the separate structure data;
An associated sample detection means for detecting a sample subgroup associated with the designated data portion from the sample group;
Marker variable detection means for detecting one or more characteristic variables common to the detected sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated data portion, the detected sample subgroup, and the searched marker variable on the GUI.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 16,
Further comprising another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Sample designating means for designating an arbitrary sample subgroup from the sample group;
Related data portion detection means for detecting a data portion associated with the designated sample subgroup from the plurality of data portions of the separate structure data;
Marker variable detection means for detecting one or more characteristic variables common to the designated sample subgroup as marker variables from the plurality of variables based on the values of the variables of the target data group;
Have
The display means highlights and displays the designated sample subgroup, the detected data portion, and the searched marker variable on the GUI.
Multivariate analyzer. In the multivariate analyzer according to any one of claims 1 to 17,
Further comprising another structure data storage means for storing another structure data having a plurality of data portions associated with each sample of the sample group,
The calculation / control means further includes means for creating a separate structure data chart illustrating the separate structure data,
The display means displays not only the plurality of multivariate analysis charts but also the separate structure data chart on the GUI,
The feature extraction means of the calculation / control means is:
Marker variable designation means for designating an arbitrary variable as a marker variable from among the plurality of variables;
Feature sample detection means for detecting one or more characteristic sample subgroups common to the designated marker variable from the sample group based on the values of the variables of the target data group;
Related data portion detection means for detecting a data portion associated with the detected characteristic sample subgroup from the plurality of data portions of the different structure data;
Have
The display means highlights and displays the designated marker variable, the detected characteristic sample subgroup, and the detected data portion on the GUI.
Multivariate analyzer. Target data storage means for storing the target data group, the target data group of the sample group, each target data having a plurality of variable values;
A calculation / control unit having a multivariate analysis unit that performs multivariate analysis using the target data group; and a feature extraction unit that performs a feature extraction process using the target data group;
Display means for visually displaying the result of the multivariate analysis and the result of the feature extraction processing on a graphical user interface (hereinafter abbreviated as GUI);
The multivariate analysis means of the calculation / control means generates a plurality of multivariate analysis charts representing characteristics of the sample group and characteristics of the plurality of variables found by the multivariate analysis,
The display means displays the plurality of charts on the GUI,
The feature extraction unit of the calculation / control unit specifies an arbitrary sample sub group in the sample group or an arbitrary variable in the plurality of variables, and the specified sample sub group is based on the target data group. Variables characteristic of the group, sample subgroups characteristic of the designated variable, or other sample groups if the target data group of the other sample groups is also stored in the target data storage means Detecting a sample subgroup associated with said designated sample subgroup,
The display means includes the designated sample subgroup or the designated variable and the detected characteristic variable, or the detected characteristic sample subgroup, or the detected related sample sub. Groups are highlighted and displayed on the GUI,
A computer program that causes a computer to function as a multivariate analyzer.

Images