JP5020491B2 - NMR data processing apparatus and method - Google Patents

Description

  The present invention relates to an apparatus and method for processing NMR (nuclear magnetic resonance) data, and is particularly suitable for use in multivariate analysis of a large number of NMR spectra, such as metabolomic analysis (metabonomics or metabolomics). Is.

  In recent years, with the progress of genome and proteome science, metabonomics (also referred to as metabolomics) that comprehensively analyzes the entire metabolite (metabolome or metabonome) has attracted attention. In particular, analysis methods for obtaining knowledge about disease states and non-diseases, for example, by performing pattern classification between samples by multivariate analysis using proton spectra (NMR spectra) of a large number of biological samples measured by an NMR apparatus Are employed in metabonomics.

  That is, the proton spectrum of a biological sample contains a variety of low-molecular metabolites and biopolymer compounds, so that many peaks overlap and show a complicated pattern. Moreover, there are many weak peaks, and it is difficult to distinguish them from noise. Therefore, it is difficult to identify and analyze the individual peaks included in the proton spectrum. Therefore, in metabonomics, the proton spectra of a large number of biological samples are measured, and statistical analysis such as multivariate analysis is performed using a large number of measured proton spectrum data. An analysis method of extracting a difference in ratio as a pattern (feature amount) is adopted. This method makes it possible to detect qualitative differences in metabolites and to detect changes over time.

  Incidentally, Patent Document 1 discloses a method for drug discovery, treatment of disease and diagnosis using metabonomics. However, Patent Document 1 does not disclose an NMR spectrum. Patent Documents 2 and 3 disclose techniques relating to processing of NMR spectra for metabonomics. Non-patent literature 1 discloses various methods for multivariate analysis of chemical data.

Special table 2003-530130 gazette Special table 2004-526130 gazette JP-T-2004-538559 “Computer Chemistry Series 3 Chemometrics Chemical Pattern Recognition and Multivariate Analysis” by Miyashita Yoshi and Shinichi Sasaki, Kyoritsu Shuppan

  A typical metabonomic process is to obtain FID (Free Induction Decay) data for a large number of samples output from the NMR instrument, to convert the FID signals of a large number of samples to NMR spectra, respectively, and , Including performing statistical processing represented by multivariate analysis on the NMR spectra of a large number of samples. Processing of complex NMR spectral data included in the metabonomics process is usually performed by a computer, but in order to obtain a reliable processing result, the accuracy of the NMR spectrum obtained from the FID signal is improved, and There is a need for data processing techniques to enable a large number of NMR spectra to be analyzed efficiently with a statistical processing program.

  Conventional processing techniques for improving the accuracy of NMR spectra include phase correction and baseline correction. Both phase correction and baseline correction can be performed automatically by the computer, but eventually the correction parameters can be adjusted manually by a person looking at the graph of the NMR spectrum displayed on the computer display screen. This is done by adding an operation to the computer to adjust the value to the most suitable value. However, this conventional correction method requires a long working time and is inefficient, and the appropriate level of the correction result depends on the person, so that it is difficult to always obtain an optimum NMR spectrum.

  In addition, in order to make it possible to analyze NMR spectra efficiently with a statistical processing program, conventionally, before inputting an NMR spectrum into a statistical processing program, bucket integration is applied to the NMR spectrum, which is smaller in data volume. A process of reducing to a histogram is performed. In other words, the observation range along the chemical shift axis (frequency axis) of the NMR spectrum is divided into a large number of buckets (chemical shift subsections) in steps of a constant width (typically 0.04 ppm). Integration of the intensity of the NMR spectrum (bucket integration) is performed, and as a result, a histogram consisting of a set of integral values of multiple buckets is obtained as a reduction of the NMR spectrum. Since the NMR spectrum of each sample is converted into a histogram, a total number of integral values corresponding to the number of samples × the number of buckets is obtained, and these integral values are input to the statistical processing program.

  However, the histogram obtained by the conventional bucket integration processing method does not necessarily reflect the characteristics of the NMR spectrum sufficiently well. That is, some important intensity peaks on the NMR spectrum may be dwarfed or diluted on the histogram. This phenomenon occurs when a chemical shift region in which one intensity peak is distributed is divided into a plurality of buckets in the bucket integration process, and as a result, the intensity peak is dispersed into a plurality of small integrated values.

  None of Patent Documents 1-3 provide a particularly useful technique for the above-described problems related to phase correction and baseline correction, and problems related to bucket integration. These problems exist not only in NMR spectral processing for metabonomics, but also in NMR spectral processing for other applications.

  In addition, regarding the multivariate analysis processing performed by the statistical processing program, there is no particular proposal in the above-described patent document.

  Here, as disclosed in Non-Patent Document 1, there are many types of analysis methods called multivariate analysis. One typical analysis method is PCA (Principal component Analysis). In PCA, in a multidimensional space consisting of coordinate axes of a large number of variables that are components of a sample, a few (for example, about 3) new coordinate axes in which the difference (dispersion) between samples is most prominent are defined, and The coordinate value of each sample on the new coordinate axis is calculated. Here, the new coordinate axis is called “principal component axis”, the variable along each principal component axis is called “principal component”, and the coordinate value on each principal component axis of each sample is “score” Called. Each principal component is defined by a linear linear expression of a large number of variables, and the coefficient of each variable term in the linear linear expression is called “loading”. The loading of each variable for a certain principal component represents the degree of contribution, that is, the weight of each variable to that principal component.

  By plotting a large number of samples on the principal component space based on such PCA results, it becomes easy to visually grasp the characteristics of these samples. In addition, it is possible to classify a large number of samples by using the arrangement of samples in the principal component space and the distance between them.

  Also, SIMCA (Soft Independent Modeling of Class Analogy) using PCA is considered useful in metabonomics. SIMCA performs PCA on different classes of samples for each class, determines principal components for each class, and evaluates how effective a principal component is to classify those classes be able to.

  By skillfully using such PCA and SIMCA analysis results, it will be possible to more easily and accurately achieve, for example, marker variable detection and other purposes by effectively utilizing the knowledge obtained from the sample. Be expected.

  However, in order to meet the above expectations, in particular how PCA and SIMCA are specifically implemented, how they relate, and in order to be highly useful in applications such as metabonomics, and As for the utilization technique such as how to display the analysis result to the user, there has not been provided a satisfactory technique. Therefore, multivariate analysis is still a threshold for non-statistical mathematicians, and is expected to be used in application fields such as metabonomics, but is not fully utilized to demonstrate its true value. is the current situation.

  Therefore, an object of the present invention is to increase the accuracy of NMR spectrum analysis processing.

  Another object is to increase the efficiency of the NMR spectrum analysis process.

  Another object is to make the characteristics of NMR spectrum peaks well reflected in the histogram when the NMR spectrum is reduced to a histogram by bucket integration.

  Yet another object is to provide a processing technique for effectively utilizing multivariate analysis in the analysis of NMR spectra of a large number of samples.

  Yet another object is to facilitate the use of multivariate analysis processes to derive more accurate results or conclusions in application fields such as metabonomics.

According to one aspect of the present invention, an NMR data processing apparatus includes a spectrum generating unit that generates target spectrum data indicating NMR characteristics of a sample, and a plurality of buckets having an unequal width with respect to the target spectrum data. And a means for storing or outputting the histogram data. The data reduction means reduces the target spectrum data to histogram data by executing bucket integration using a bucketset consisting of:

  When performing bucket integration of the target spectrum data, by using a bucket set with an appropriately set unequal width, the target spectrum data can be stored while maintaining the important peak information of the target spectrum data well. It can be reduced to a histogram that is a set of bucket integral data. Therefore, the accuracy of spectrum analysis performed using the histogram is improved.

  In a preferred embodiment, a non-uniform width bucket set is used so that none of the specified one or more peak areas is divided into multiple buckets in order to maintain the important peak information of the target spectrum data well. Is set. Furthermore, a non-uniform width bucket set is set so that none of the one or more peak areas detected from the target spectrum data is divided into a plurality of buckets.

  In a preferred embodiment, the NMR data processing apparatus is provided with bucket setting means for setting a bucket set having the non-uniform width as described above. The bucket setting means inputs a plurality of target spectrum data, generates projection spectrum data obtained by projecting the plurality of target spectrum data, detects a peak region from the projection spectrum data, and detects one or more detected ones or more Each bucket is set so that none of the peak areas are divided into a plurality of buckets. Further, the bucket setting means specifies one or more peak areas, and sets each bucket so that none of the specified one or more peak areas is divided into a plurality of buckets. Further, the bucket setting means can modify each bucket in response to a request input from the operator.

  In a preferred embodiment, absolute value differential spectrum data, which is a result of performing absolute value differentiation on NMR spectrum data obtained from FID data of a sample, is used as the target spectrum. Here, the absolute value differentiation of NMR spectrum data is the square root of the sum of the square of the frequency (chemical shift) derivative of the real part of the NMR spectrum and the square of the frequency (chemical shift) derivative of the imaginary part. By performing spectrum analysis using such absolute value differential spectrum data, it is possible to efficiently obtain an accurate analysis result without performing troublesome phase correction and baseline correction.

  An apparatus for processing NMR data according to another aspect of the present invention comprises means for acquiring FID data of a sample, means for generating NMR spectrum data from the FID data, and performing absolute value differentiation on the NMR spectrum data to obtain an absolute value. Means for generating differential spectrum data, and means for storing or outputting the absolute value differential spectrum data.

  According to still another aspect of the present invention, there is provided a statistical processing apparatus for NMR spectra, a multivariate data matrix input means for inputting a multivariate data matrix composed of NMR histogram data of a plurality of samples, and a PCA for the input multivariate data matrix. PCA analysis means for performing analysis calculation, PCA analysis result display means for receiving a plurality of types of PCA analysis result charts on a user interface screen in response to the result of the PCA analysis calculation, and a user request using the user interface screen When the user request input means for inputting and the input user request is to select one or more samples, the NMR spectrum data of the selected sample is input, and the input NMR spectrum is chemically converted. An NMR spectrum sequence for displaying an NMR spectrum chart on the shift axis on the user interface screen. And carrying means.

  The PCA analysis result chart includes at least a score plot that represents the score of the sample with respect to a plurality of principal components, a loading plot that represents loading with respect to the plurality of principal components of a predetermined plurality of chemical shift values as variables, and A contribution ratio / loading chart representing the contribution ratio or loading of a plurality of predetermined chemical shift values on the chemical shift axis is included. The contribution ratio / loading chart and the NMR spectrum chart are arranged and displayed in a direction perpendicular to the chemical shift axis so that the scales of the chemical shift axes coincide with each other.

  In a preferred embodiment, when the input user request is to enlarge the contribution ratio / loading chart, the PCA analysis result display means and the NMR spectrum cooperation means include the contribution ratio / loading chart and the NMR spectrum. The contribution ratio / loading chart and the NMR spectrum chart are enlarged and displayed so that the coincidence of the scale of the chemical shift axis with the spectrum chart is maintained.

  Further, in a preferred embodiment, when the input user request is to delete one or more selected samples, the PCA analysis means comprises NMR histogram data of samples other than the deleted samples. The PCA analysis calculation is executed again for the multivariate data matrix, and the PCA analysis result display means receives the result of the PCA analysis calculation executed again and displays the PCA analysis result chart again.

  Further, in a preferred embodiment, when the input user request is to delete the selected chemical shift value as one or more variables, the PCA analysis means includes a chemical other than the deleted chemical shift value. The PCA analysis calculation is executed again using the shift value, and the PCA analysis result display means receives the result of the PCA analysis calculation executed again and displays the PCA analysis result chart again.

  An NMR spectrum statistical processing apparatus according to still another aspect of the present invention comprises a multivariate data matrix input means for inputting a multivariate data matrix of a plurality of classes composed of NMR histogram data of a plurality of samples assigned to a number of classes. SIMCA analysis means for performing SIMCA analysis calculation for the input multivariate data matrix of the plurality of classes, and SIMCA analysis results for receiving a result of the SIMCA analysis calculation and displaying a plurality of types of SIMCA analysis result charts on a user interface screen Display means, user request input means for inputting a user request using the user interface screen, and when the input user request is to select one or more samples, the NMR spectrum of the selected sample Enter the data, and the NMR spectrum showing the entered NMR spectrum on the chemical shift axis. And a NMR spectrum linkage means for displaying the chart on the user interface screen.

  The SIMCA analysis result chart includes at least a Cooman plot representing the distance of the sample with respect to the plurality of classes, and a modeling ability representing the modeling power or discriminating power of a predetermined plurality of chemical shift values as variables on the chemical shift axis. / Discrimination power chart is included. The modeling force / discriminating power chart and the NMR spectrum chart are arranged and displayed in a direction orthogonal to the chemical shift axis so that the scales of the chemical shift axes coincide with each other. .

  In a preferred embodiment, when the input user request is for enlarging the modeling power / discrimination power chart, the SIMCA analysis result display means and the NMR spectrum cooperation means include the modeling power / discrimination power chart and The modeling power / discriminating power chart and the NMR spectrum chart are enlarged and displayed so that the coincidence of the scale of the chemical shift axis with the NMR spectrum chart is maintained.

  Also, in a preferred embodiment, when the input user request is to delete one or more selected samples, the SIMCA analysis means comprises NMR histogram data of samples other than the deleted samples. The SIMCA analysis calculation is executed again for a plurality of classes of multivariate data matrices, and the SIMCA analysis result display means receives the result of the SIMCA analysis calculation executed again and displays the SIMCA analysis result chart again.

  In a preferred embodiment, the SIMCA analysis is performed when the input user request is to select a chemical shift value that is a variable having the discriminating power or modeling power higher than a predetermined threshold and to request recalculation. The means re-executes the SIMCA analysis calculation using only the selected chemical shift value, and the SIMCA analysis result display means receives the result of the re-executed SIMCA analysis calculation, and displays the SIMCA analysis result chart. Display again.

  According to still another aspect of the present invention, there is provided a statistical processing apparatus for NMR spectra, PCA analysis means for performing PCA analysis calculation on a multivariate data matrix composed of NMR histogram data of a plurality of samples, and receiving the result of the PCA analysis calculation. , PCA analysis result display means to display multiple types of PCA analysis result charts on the user interface screen, and multivariate to input multiclass multivariate data matrix consisting of NMR histogram data of multiple samples assigned to multiple classes Data matrix input means, SIMCA analysis means for performing a SIMCA analysis calculation on the input multivariate data matrix of the plurality of classes, and receiving a result of the SIMCA analysis calculation, a plurality of types of SIMCA analysis result charts on a user interface screen The user request is input using the SIMCA analysis result display means to display and the user interface screen. User request input means.

  The PCA analysis result chart includes at least a score plot that represents the score of the sample with respect to a plurality of principal components, a loading plot that represents loading with respect to the plurality of principal components of a predetermined plurality of chemical shift values as variables, and A contribution ratio / loading chart representing the contribution ratio or loading of a plurality of predetermined chemical shift values on the chemical shift axis is included. The SIMCA analysis result chart includes at least a Cooman plot representing the distance of the sample with respect to the plurality of classes, and a modeling ability representing the modeling power or discriminating power of a predetermined plurality of chemical shift values as variables on the chemical shift axis. / Discrimination power chart is included.

  The user request input by the user request input means selects a chemical shift value that is a variable in which the discriminating power or modeling power displayed in the modeling power / discriminating power chart is higher than a predetermined threshold and re-executes the PCA analysis calculation. In the case of requesting execution, the PCA analysis means executes the PCA analysis calculation again using only the selected chemical shift value, and the PCA analysis result display means executes the PCA analysis calculation executed again. Then, the PCA analysis result chart is displayed again.

  In a preferred embodiment, the user request input by the user request input means selects a chemical shift value as a variable in which the discriminating power or modeling power displayed in the modeling power / discriminating power chart is higher than a predetermined threshold. When the PCA analysis calculation requesting the re-execution of the PCA analysis calculation with the variable added thereto is used, the PCA analysis means uses the variable formed by adding the selected chemical shift value to the current chemical shift value. The analysis calculation is executed again, and the PCA analysis result display means receives the result of the PCA analysis calculation executed again and displays the PCA analysis result chart again.

  In a preferred embodiment, the user request input by the user request input means is a chemical shift value that is a variable whose discrimination power or modeling power displayed in the modeling power / discrimination power chart is higher than a predetermined threshold. When selecting and requesting re-execution of the SIMCA analysis calculation, the SIMCA analysis means executes the PCA analysis calculation again using only the selected chemical shift value, and the SIMCA analysis result display means, In response to the result of the SIMCA analysis calculation executed again, the SIMCA analysis result chart is displayed again.

  The present invention further provides a method for processing NMR data based on the above-described principle, a computer program for processing NMR data, and a computer program for statistical processing of NMR spectra.

  According to the NMR data processing apparatus and method of the present invention, multivariate analysis can be effectively utilized in the analysis of NMR spectra of a large number of samples. Therefore, in an application field such as metabonomics, more accurate results or It is easier to draw conclusions.

  According to the NMR data processing apparatus and method of the present invention, the efficiency of NMR spectrum analysis processing is improved by applying absolute value differentiation to an NMR spectrum.

  According to the statistical processing apparatus for NMR spectra of the present invention, multivariate analysis can be effectively utilized in the analysis of NMR spectra of a large number of samples.

  According to the statistical processing apparatus for NMR spectrum of the present invention, by using this in an application field such as metabonomics, it becomes easy to draw a more accurate result or conclusion by utilizing multivariate analysis processing. .

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 shows the overall configuration and functions of an embodiment of an NMR data processing apparatus according to the present invention.

  As shown in FIG. 1, a computer system 100 as an NMR data processing apparatus according to the present invention is used in association with an NMR apparatus 102. The computer system 100 may be a programmed single computer machine (for example, a general-purpose personal computer, a workstation, or a mainframe computer machine), or may be composed of a plurality of computer machines. Instead of a computer system, one or more dedicated hardware circuits, or a combination of a dedicated hardware circuit and a computer system may be used as the NMR data processing apparatus.

  The computer system 100 includes a processor 104 and a storage device 106 for storing a computer program for the processor 104 and data processed by the processor 104. The processor 104 processes NMR data by executing a predetermined computer program stored in the storage device 306. If the purpose of processing the NMR data is different, the processing method is naturally different, but here, as an example of the processing of the NMR data, the NMR spectra of a large number of samples are used for the purpose of determining the marker component by metabonomics. The case of collecting and statistically analyzing those spectra will be described.

  The processing of the NMR data performed by the processor 104 is roughly divided into an FID data input process 108, an NMR spectrum process 110, a data contraction process 112, an integration result data storage / output process 114, and a statistical process 116. In the FID data input process 108, FID (Free Induction Decay) data of each sample output from the NMR apparatus 102 is input, whereby FID data of a large number of samples is collected and stored in the storage device 106. Is done. In the NMR spectrum processing 110, the FID data of each sample is converted into NMR spectrum data by Fourier transform, and the NMR spectrum data is further converted into “absolute differential calculation method” according to the principle of the present invention by “absolute differential calculation method”. The data is converted into Absolute Differential NMR Spectrum data, whereby “absolute differential NMR spectrum” data of a large number of samples are stored in the storage device 106. Here, “absolute differential method” and “absolute differential NMR spectrum” and “ The term “absolute value differential NMR spectrum” is a special term used for convenience of explanation in this specification, and will be described in detail later. “Absolute value differential NMR spectrum” is hereinafter abbreviated as “AD spectrum”. .

  In the next data reduction process 112, an “optimized bucket set” is calculated in a manner according to the principles of the present invention, and the “optimized bucket set” is used to “optimize” the AD spectrum of a large number of samples. "Bucket integration" is performed, thereby generating a histogram (a set of bucket integration data) that is the reduced data of the AD spectrum of a large number of samples. Here, the terms “optimized bucketset” and “optimized bucket integral” are also special terms used for convenience of explanation in this specification, and details thereof will be described later. In the integration result data storage / output process 114, histogram data (a set of bucket integration data) of a large number of samples generated by the data reduction process 112 is stored in the storage device 106, and externally as necessary. Is output. In the statistical processing 116, statistical processing according to the processing purpose, for example, detection multivariate analysis for determining a marker component, is performed using histogram data of a large number of samples. Data representing the result of the statistical processing is stored in the storage device 106 and is output to the outside as necessary.

  Here, since various known processing methods can be employed for the FID data input processing 108, the integration result data storage / output processing 114, and the statistical processing 116, detailed description thereof will be omitted. Hereinafter, the NMR spectrum processing 110 using the absolute value differentiation method and the data contraction 114 using the optimized bucket integration will be described in detail and specifically.

  FIG. 2 shows the flow of the NMR spectrum processing 110.

  As shown in FIG. 2, in step 120, the FID data of each sample acquired from the NMR device 102 and stored in the storage device 106 is read from the storage device 106. At step 122, window processing is performed on the FID data to improve the signal-to-noise ratio of the FID data and to remove ripples. In step 124, Fourier transform processing is performed on the FID data that has passed through the window processing 122, and the FID data is converted into NMR spectrum data. In step 126, an absolute value differential operation is performed on the NMR spectrum data, and the NMR spectrum data is converted into AD spectrum data (absolute value differential NMR spectrum data). In step 128, the NMR spectrum data and AD spectrum data are stored in the storage device 106, and output to the outside as necessary. For each of a large number of samples, the processing of steps 120 to 128 described above is performed, whereby the NMR spectrum data and AD spectrum data of the large number of samples are accumulated in the storage device 106.

  The absolute value differentiation operation in step 126 will be specifically described below.

The NMR spectrum, which is the target of absolute value differentiation, has the following formula when δ is a chemical shift and j is an imaginary symbol:
NMR spectrum = R (δ) + j · I (δ) (1)
Can be expressed as Here, R (δ) is the real part of the NMR spectrum and I (δ) is the imaginary part. And the AD spectrum which is the result of absolute value differentiation of the NMR spectrum is

で表現することができ（又は、ケミカルシフトδに代えて周波数ωで微分してもよい）。すなわち、これは、NMRスペクトルの実部R(δ)と虚部I(δ)それぞれの微分値の２乗和の平方根である。

(Or may be differentiated by frequency ω instead of chemical shift δ). That is, this is the square root of the square sum of the differential values of the real part R (δ) and the imaginary part I (δ) of the NMR spectrum.

  FIG. 3A shows the waveform of the real part of an example of an NMR spectrum, and FIG. 3B shows the waveform of an AD spectrum that is an absolute value derivative of the NMR spectrum.

  As can be seen from the real part 140 of the NMR spectrum shown in FIG. 3A, this NMR spectrum contains a phase shift (for example, the phase shift is markedly shown in the portion 142 of the solvent signal), and its base. The line component is not flat but distorted. The NMR spectra obtained from FID data almost without exception contain phase shifts and distorted baseline components. According to the prior art, the NMR spectrum needs to be phase corrected and baseline corrected.

  Phase correction is an operation for removing a phase shift. In the NMR spectrum including the phase shift, the peak waveforms in the real part R (δ) and the imaginary part I (δ) are waveforms as illustrated in FIG. 4A. In the phase correction, the phase shift is reduced and shaped into a waveform as illustrated in FIG. 4B. Baseline correction is correction for subtracting the baseline component from the NMR spectrum. The distorted baseline BL as illustrated in FIG. 5B is corrected to a zero and constant baseline BL as illustrated in FIG. 5B. However, the phase correction and the baseline correction need to be manually adjusted by the operator, which is time consuming and introduces a bias in the subjectivity and skill of the operator. Furthermore, even if it is difficult to adjust, it is difficult to achieve a sufficiently satisfactory result.

  On the other hand, as illustrated in FIG. 3B, in the AD spectrum 144, which is the absolute value differential result of the NMR spectrum, the phase shift is eliminated, and the baseline is substantially constant in the vicinity of zero. This means that phase correction and baseline correction are substantially unnecessary if the AD spectrum 144 is used. The reason will be specifically described.

The real part R (δ) and imaginary part I (δ) of the NMR spectrum are
R (δ) = A (δ) · cos (θ (δ)) (3)
I (δ) ＝ A (δ) ・ sin (θ (δ))… (4)
Can be expressed as Here, A (δ) is the amplitude, and θ (δ) is the phase. The phase θ (δ) is roughly a linear function of chemical shift δ,
θ (δ) = K · δ + L (5)
Where K and L are constant coefficients. Using these equations to calculate the absolute derivative, the AD spectrum is

に変換される。

Is converted to

  As can be seen from the above equation (6), the AD spectrum does not contain a component of phase θ (δ). In addition, the dominant part in the equation (6) is the first term, where the amplitude A (δ) is differentiated by the chemical shift δ, and the baseline component (this is the example of FIG. 5A). As can be seen from the above, the chemical shift δ is distorted with a gentle inclination), which is substantially zero. Therefore, the baseline component is reduced considerably in the AD spectrum. As described above, a high-quality spectrum can be obtained by a simple calculation called absolute value differentiation without performing phase correction and baseline correction.

  In addition, the NMR spectrum has a huge peak of a solvent (for example, light water) signal as indicated by reference numeral 142 in FIG. 3A. The base of the huge solvent peak spreads widely, which is also one of the main causes of spectrum accuracy degradation. However, in the AD spectrum, the base part of the solvent peak is also reduced well for the above reason as in the case of the base line, so that the influence of the solvent peak can be made extremely small.

  One drawback of the AD spectrum is that the absolute quantification of the NMR spectrum is lost, and in particular, the signal intensity of a broad peak is attenuated by differentiation. However, the relative relationship between multiple NMR spectra (for example, the relative relationship in terms of which chemical shift δ appears at the peak) is maintained in the AD spectrum. For the purpose of analyzing a large number of NMR spectra, the AD spectrum is very useful information.

  Referring again to FIG. 2, after the above-described AD spectrum is generated in step 126, the AD spectrum is stored in the storage device 106 in step 128. The process shown in FIG. 2 is repeated for each of the NMR spectra of a large number of samples necessary for statistical processing, and as a result, the AD spectra of the large number of samples are accumulated in the storage device 106. Thereafter, as shown in FIG. 1, a data reduction process 112 using optimized bucket integration is executed.

  FIG. 6 shows the flow of the data reduction process 112 using the optimized bucket integration.

  As shown in FIG. 6, in step 170, the reference point (δ reference point) (origin) of the chemical shift (δ) axis scale is automatically set for each of a large number of AD spectra accumulated in the storage device 106. Is done. The δ reference point can be set by a known method, whereby the point on the chemical shift axis where the peak of the same specific substance appears for any AD spectrum is set as the reference point.

  Thereafter, in step 171, a process of projecting a large number of AD spectra onto a common spectrum is performed. The spectrum projection process 171 is a process of calculating the maximum value or sum of all AD spectra for each point of the chemical shift δ after matching the scales of the chemical shift axes of those AD spectra. As a result, one projection AD spectrum is generated as the maximum value or the sum of many AD spectra, and all peaks included in all AD spectra appear in the projection AD spectrum.

  7A and 7B are diagrams for explaining the principle of the spectrum projection processing 171, and FIG. 8 shows the flow of the spectrum projection processing 171.

  In the spectrum projection process 171, as shown in FIG. 8, in step 190, one AD spectrum Sa is selected from a large number of AD spectra. For example, the spectrum Sa as shown in FIG. 7A is selected. This AD spectrum Sa is initially set to the projected AD spectrum Sp. In step 192, another AD spectrum Sb is selected. For example, an AD spectrum Sb as shown in FIG. 7B is selected.

In step 194, the chemical shift axis scales of both spectra Sb and Sp are matched (for example, the δ scale (chemical shift axis scale) of the spectrum Sb is set by the difference (rb-ra) between the reference points rb and ra of both spectra. ) And shifting the reference points rb and ra), the maximum value or sum of the intensities of both spectra Sb and Sp is calculated for each δ point. In particular,
Maximum value: if (Sp> Sb (rb-ra)) Sp = Sa else Sp = Sb (rb-ra)
Sum: Sp = Sp + Sb (rb-ra)
Is calculated every δ points. Thereby, as shown in FIG. 7C, all peaks included in the processed spectra Sa and Sb are projected onto the projected AD spectrum Sp. Note that only one of the maximum value and the sum may be calculated, or both the maximum value and the sum may be calculated as separate projected AD spectra.

  Step 196 repeats steps 192 and 194 described above for all AD spectra to be processed. When the iterative process is completed for all AD spectra, a projected AD spectrum Sp in which all the peaks included in all the AD spectra are projected is completed. In step 198, the projection AD spectrum Sp is stored in the storage device 106, and is output to the outside as necessary.

  Refer to FIG. 6 again. When the above-described spectral projection processing 171 ends, in step 172, the bucket integral of the projected AD spectrum is calculated.

  FIG. 9 is a diagram for explaining the principle of the bucket integration processing 172 of the projected AD spectrum, and FIG. 10 shows the flow of the bucket integration processing 172.

  In FIG. 9, reference numeral 200 indicates an integration range of one unit in which bucket integration is performed, and is referred to as “bucket”. The bucket 200 has a specified constant width W (for example, 0.04 ppm) in the bucket integration processing 172, but the width W is corrected and optimized in the subsequent processing. Reference number 202 indicates the full range of chemical shift δ at which bucket integration is to be performed, for example in the range of 0 ppm to about 16 ppm. The entire range 202 is divided into about 200 buckets 200, for example. A set of a large number of buckets 200 set in the entire range 202 in this way is hereinafter referred to as “bucket set”. Reference numeral 204 indicates a range in which bucket integration is not performed, and is referred to as “dark region”.

  As shown in FIG. 10, in the bucket integration process 172, in step 210, the entire range 202 shown in FIG. 9 is equally divided into a large number of constant width W buckets 200, and the projected AD spectrum for each bucket 200. The integral value of Sp (that is, the area of the region sandwiched between the projected AD spectrum Sp and the chemical shift axis in each bucket) is calculated. Thereby, a set of bucket integral data corresponding to the number of buckets 200 is obtained. A set of such a large number of bucket integration data is hereinafter referred to as an “integration data set”. This integral data set is a kind of histogram indicating the peak area for each bucket of the projected AD spectrum Sp.

  Thereafter, in step 212, the bucket corresponding to the dark region 204 and the bucket integration data of the bucket set are deleted from the bucket set and integration data set obtained in step 210. In step 214, the integral data set is normalized so that the total sum of the integral data sets becomes a predetermined value. Then, the normalized integral data set (histogram data) and bucket set information representing the positions (start point and end point) of all buckets included in the bucket set are stored in the storage device 106, and if necessary, Is output externally.

  Refer to FIG. 6 again. When the above-described bucket integration process 172 is completed, in step 174, the bucket automatic correction process based on the automatic integration block setting is performed. Here, various peaks on the projected AD spectrum Sp are automatically detected by a known automatic integration block setting process, and the position and width W of the bucket are set so that each detected peak is not divided by a plurality of buckets. Is fixed. In the following, the term “peak” is used not only to indicate a single isolated peak but also to indicate a plurality of peaks in a lump that can be regarded as a single peak.

  FIG. 11 is a diagram for explaining the principle of the bucket automatic correction processing 174 based on the automatic integration block setting, and FIG. 12 shows the flow of the bucket automatic correction processing 174.

  In this bucket automatic correction process 174, as shown in FIG. 12, in step 230, an automatic integration block setting process, which is a known technique, is performed on the projection AD spectrum Sp. As a result, various peaks on the projected AD spectrum Sp are automatically detected and integrated in a one-to-one relationship with the chemical shift δ region (hereinafter referred to as “peak region”) where the detected peak exists. A block (integration interval) is automatically set. For example, as shown in FIG. 11A, if there are two peaks on the AD spectrum Sp (eg, the left one is a cluster of three peaks and the right one is a single peak), the two A peak is automatically detected, and as shown in FIG. 11B, one integration block 220x is set in the left peak region, and another integration block 220y is set in the right peak region. Is done. Each of the integration blocks 220x and 220y covers the corresponding peak region, and one peak region is not divided by a plurality of integration blocks.

  On the other hand, as illustrated in FIG. 11C, the current buckets 200 a, 200 b, 200 c,..., As described above with reference to FIG. ), The plurality of buckets may divide one peak area. In the example shown in FIGS. 11A and 11C, the left peak region is divided by two buckets 200b and 200c. Since the integral data set (histogram) currently stored in the storage device 106 is calculated using such a bucket set, the divided peak information is distributed and diluted to different integral data. It is trapped and the accuracy is not good. Therefore, in order to solve this problem, the following steps described below are performed to modify the bucket set so as not to divide the peak.

  That is, as shown in FIG. 12, at step 232, the end point of the bucket closest to the start point of each integration block is searched from the bucket set information in the storage device 106, and the end point of the found bucket is the corresponding integration block. It is corrected to the same value as the starting point. Further, in step 234, the starting point of the bucket closest to the end point of each integration block is searched from the bucket set information in the storage device 106, and the starting point of the found bucket is corrected to the same value as the end point of the corresponding integration block. Is done. For example, in the example shown in FIGS. 11B, 11C, and 11D, the end point e (a) of the bucket 200a closest to the start point s (x) of the left integration block 200x is corrected to the same value as the start point s (x). In addition, the end point e (d) of the bucket 200d closest to the start point s (y) of the right integration block 200y is corrected to the same value as the start point s (y). Also, the starting point s (d) of the bucket 200d closest to the end point e (x) of the left integration block 200x is corrected to the same value as the end point e (x), and the end point e (y) of the right integration block 200y is changed to the end point e (y). The start point s (f) of the nearest bucket 200f is corrected to the same value as the end point e (y).

  After the correction of the start point / end point, in step 236, a plurality of buckets dividing one integration block are searched, and the plurality of buckets are integrated into one bucket. For example, in the example shown in FIGS. 11B, 11C, and 11D, after the start point / end point correction, only the left integration block 200x is divided by the two buckets 200b and 200c. 200c is integrated into one bucket 200bc.

  In step 238, the bucket integral is recalculated for all buckets modified in steps 232 to 236 described above. Then, the integration data of the corresponding bucket in the integration data set currently stored in the storage device 106 is replaced with the recalculated integration data. For example, in the example shown in FIG. 11D, since all of the buckets 200a, 200bc, 200d, 200e, and 200f have been modified, the bucket integral is recalculated for each of the buckets 200a, 200bc, 200d, 200e, and 200f. Thus, the recalculated integration data is replaced with the previous integration data.

  In step 239, the integration data set modified as described above and the modified bucket set information are stored in the storage device 106, and output to the outside as necessary. According to the modified bucket set information, none of the peak areas automatically detected from the projected AD spectrum Sp by the automatic integration block setting process is divided into multiple buckets (that is, one peak area). A bucketset is defined so that it is always covered by one bucket. Since the modified integration data set is based on the modified bucket set, the information of all detected peaks is not diluted and is therefore more accurate than the previous integration data set.

  However, the chemical shift δ interval (integration interval) set based on the peak automatically detected by the known automatic integration block setting process is not necessarily appropriate. Therefore, processing for further optimizing the bucket data set and the component data set is subsequently performed. That is, as shown in FIG. 6, after the bucket automatic correction processing 174 based on the above-described automatic integration block setting, in order to further optimize the bucket data set and the component data set, the bucket automatic correction processing 176 based on the designated peak information. Then, the bucket manual correction processing 178 is executed.

  First, the bucket automatic correction process 176 based on the designated peak information will be described.

  In this process 176, a specific peak on an already identified NMR spectrum (not an AD spectrum) is specified, and an integration block is determined so as to cover the peak area of the specified peak. Based on this, the currently stored bucketset is modified so that the peak area of the designated peak is not divided by multiple buckets. The specified peak is typically a peak that can be a marker in this spectral analysis application. Since the markers are different for different uses of spectrum analysis, the designated peak differs depending on the use.

As source data for designating a peak, peak definition data in accordance with a general presentation format (Bull. Chem. Soc. Japan, etc.) of NMR data defined in academic journals can be used. An example of peak definition data in such a format is
1H-NMR (CDCl3) δ: 8.06 (2H, t, J = 8.1 Hz, CH2), 7.24 (1H, s, CH3),…
It is something like that. In this data example, the first “1H-NMR” indicates the name of the measurement nucleus, the next “(CDCl 3)” indicates the name of the solvent, and the next “δ: 8.06” indicates the chemical shift at the center of the first peak. The next “2H” indicates the proton number of the first peak, the next “t” indicates the splitting pattern of the first peak, and the next “J = 8.1 Hz” indicates the first peak. The spin coupling constant of the peak is shown, and the next “CH2” is a comment regarding the first peak. The definition of the first peak is followed by the definition of the second peak, the definition of the third peak,. Here, there are various peak splitting patterns, but typical splitting patterns include, for example, the following five types of patterns:
“S”: single
"D": doublet
“T”: triplet
“Q”: qualtet
“M”: multiplet
It becomes a combination.

  FIG. 13A shows an example of a three splitting peak, and FIG. 13B shows an example of a four splitting peak.

  As shown in FIG. 13A, in the case of a three split peak, the center chemical shift δ (unit: ppm) and one spin coupling constant J (unit: Hz) are defined by the peak definition data in the format as described above. The Although illustration is omitted, in the case of the bisecting peak, similarly, the central chemical shift δ and one spin coupling constant J are defined. In the case of a single peak, the central chemical shift δ is defined, but the spin coupling constant J is not naturally defined. Further, as illustrated in FIG. 13B, in the case of a quadrilateral peak, a central chemical shift δ and two broad and narrow spin coupling constants Ja and Jb (unit: Hz) are defined. Although illustration is omitted, in the case of a multi-split (split of 5 or more) peak, a central chemical shift δ and a number of different spin coupling constants depending on the number of columns are defined. Note that in the case of a multiple split peak where the central chemical shift δ is difficult to determine, the peak may be defined in the range of the chemical shift δ, such as “δa−δb”.

As illustrated in FIGS. 13A and 13B, the peak region of each peak (the chemical shift δ section where the peak is located) is a numerical value included in the above-described peak definition data, in particular, the chemical shift δ at the center, the spin coupling constant J, Can be determined based on For example, in the case of the triple split peak shown in FIG. 13A, the start point s (x) and end point e (x) of the peak region 240x are:
s (x) = δ− (J + PW / 2)
e (x) = δ + (J + PW / 2)
Further, in the case of the quadrilateral peak shown in FIG. 13B, the start point s (y) and the end point e (y) of the peak region 240y are
s (y) = δ− (Jb + PW / 2)
e (y) = δ + (Jb + PW / 2)
Can be calculated. Here, “PW” refers to the peak width (unit: Hz), which differs depending on several conditions such as the resolution of the NMR apparatus and the state of the sample. Just set it up. For splitting patterns other than those shown, if the peak is defined using the central chemical shift δ and the spin coupling constant J (where J = 0 for a single peak), then A peak area can be determined in a similar manner. Further, when a peak is defined in the range of chemical shift δ such as “δa−δb”, the start point s and end point e of the peak region are
s = δa−PW / 2
e = δb + PW / 2
Can be calculated.

  Now, the NMR data processing apparatus 100 according to the present embodiment inputs peak definition data in the format as described above for the designated peak, and is necessary for determining the peak area of the designated peak based on the peak definition data. Information (hereinafter referred to as “designated peak information”) is automatically generated. The generated designated peak information can be registered in the storage device 106 so that it can be used at any time later.

  14A and 14B show examples of the data structure of designated peak information registered in the storage device 106. FIG.

  As shown in FIG. 14A, different designated peak information sets 250A, 250B, and 250C respectively corresponding to different uses of NMR spectrum analysis are registered in the storage device 106. For example, the peak that can be used as a marker differs between an application that attempts to determine a marker common to the blood of a person suffering from a specific disease and an application that attempts to determine a marker that is common to a specific type of food. Different peaks will be assigned to. Therefore, different designated peak information sets 250A, 250B, 250C are registered for each application. In each of the designated peak information sets 250A, 250B, 250C, peak designation information of each of one or more designated peaks is recorded. The peak designation information for each peak includes a peak ID for identifying the peak, and numerical data such as a chemical shift, a splitting pattern, and a spin coupling constant for calculating the peak area of the peak. These numerical data are automatically extracted by the NMR data processing apparatus 100 from the peak definition data in the format as described above input to the NMR data processing apparatus 100.

  Further, as illustrated in FIG. 14B, a peak width set 252 indicating a preset peak width PW is also registered in the storage device 106. As described above, when the resolution and other conditions are different, the peak width PW is different. Therefore, different peak widths PW are set in the peak width set 252 corresponding to different condition sets.

  As a modification, the designated peak information sets 250A, 250B, 250C illustrated in FIG. 14A are based on numerical data such as chemical shift, splitting pattern, and spin coupling constant, or instead of the numerical data. Data indicating the peak area calculated by the above calculation method (for example, numerical data of the start point and the end point) may be registered.

  Now, using the designated peak information as described above, the bucket automatic correction processing based on the designated peak information in step 176 shown in FIG. 6 is performed in the following procedure.

  FIG. 15 shows the flow of this bucket automatic correction processing 176.

  As shown in FIG. 15, in step 260, peak definition information is input for each of one or more arbitrarily designated peaks, and peak definition information for each designated peak is generated based on the peak definition information. Alternatively, the designated peak information set corresponding to the current use is selected from the designated peak information sets 250A, 250B, and 250C illustrated in FIG. 14A registered in the storage device 106 in advance, and the selected peak information set is selected. The specified peak information for each peak is read from the specified peak information set. Further, the optimum peak width data for the current analysis is read from the preset peak width data for each condition as illustrated in FIG. 14B.

  In step 262, the peak area of each designated peak is determined by the calculation method as described above based on the designated peak information and peak width data of each designated peak, and the peak area is set as an integration block as it is. . For example, in the case of the two types of peaks illustrated in FIGS. 13A and 13B, the peak areas 240x and 240y of these peaks are set as integration blocks, respectively.

  Thereafter, in step 264, the bucket set that has been corrected once is stored in the storage device 106 in the same procedure as steps 232 to 238 of the bucket correction processing 174 based on the automatic integration block setting shown in FIG. Further correction is made according to the integration block corresponding to the peak. By the second bucket correction, the peak area of any designated peak is not divided by a plurality of buckets (that is, the peak area of any designated peak always falls within any one bucket). In addition, for the bucket modified here, the bucket integral of the projection AD spectrum Sp is recalculated, and the integral data of the corresponding bucket stored in the storage device 106 is replaced by the recalculated bucket integral data. It is done.

  In step 239, the bucket set and the integral data set corrected twice in this way are stored in the storage device 106, and are output to the outside as necessary.

  Next, the bucket manual correction process in step 178 shown in FIG. 6 will be described. This process 178 is for allowing the operator to further modify the bucket manually if the above-described two automatic bucket modifications are not sufficient.

  16A and 16B show the principle of the manual correction processing 178 of the bucket, and FIG. 17 shows the flow of the manual correction processing 178 of the bucket.

  In this process 178, the projected AD spectrum Sp stored in the storage device 106 and the bucket set that has been subjected to the above-described two automatic corrections are displayed on the display screen of the NMR spectrum processing apparatus 100. The operator inputs a correction request for the start point or end point of an arbitrary bucket into the NMR spectrum processing apparatus 100 while viewing the display. For example, as illustrated in FIG. 16A, when a certain bucket 200j is to be corrected, the operator designates the bucket 200j and uses the start point as a request for correcting the end point, and sets a new start point s (j) and a new start point s (j). Enter the end point e (j). Then, as shown in FIG. 17, in step 270, the start point and end point of the designated bucket 200j are corrected to the same value as the start point s (j) and end point e (j) input from the operator. In step 272, the end point e (i) of the adjacent bucket on the start point s (j) side of the designated bucket is corrected to the same value as the start point s (j) of the designated bucket. In step 274, the end point of the designated bucket The start point s (k) of the adjacent bucket on the e (j) side is corrected to the same value as the end point e (j) of the designated bucket. Thereafter, in step 276, the bucket integral is recalculated for each modified bucket, and the integral data of the corresponding bucket stored in the storage device 106 is replaced with the recalculated integral data. In step 278, the manually corrected bucket set and the integration data set corrected as described above based on the bucket set are stored in the storage device 106, and output to the outside as necessary. The

  As described above, in Step 172 of FIG. 6, a uniform-width bucket set in which all buckets have a constant width (for example, 0.04 ppm) is initially set, and thereafter, three-stage buckets of Steps 174, 176, and 178 are performed. Corrections are made to complete a non-uniform bucket set with different widths depending on the bucket. Note that not all of the three correction stages of steps 174, 176, and 178 have to be executed, and any of the correction stages (for example, manual correction) may be omitted. In any case, the finally completed non-uniform width bucket set is called an “optimized bucket set”.

  FIG. 18 shows an example data structure of an optimized bucketset stored in the storage device 106.

  As illustrated in FIG. 18, different optimization bucket sets 280A, 280B, and 280C corresponding to different uses of spectrum analysis are stored in the storage device 106. This is because, if the use is different, the designated peak is different as described above, and the NMR spectrum itself to be processed is also different, so that naturally the optimized bucket set is different. Each of the optimized bucket sets 280A, 280B, and 280C includes an application ID indicating an application, numerical data indicating a peak width PW, numerical data indicating a reference point (origin) of the chemical shift axis scale, and the bucket set. The bucket number, start point and end point data of a large number of buckets are registered. Furthermore, a peak ID indicating the corresponding designated peak is also registered in the bucket corresponding to any designated peak.

  As can be seen from the description of the three-stage bucket correction process described above, in the optimized bucketset, not all buckets have a uniform width (for example, 0.04 ppm), but have unique unequal widths. And for all peaks automatically detected from the projected AD spectrum Sp, in addition to all the peaks specified by the specified peak information, the peak area of each peak must be in any one bucket. And no peak area is divided by multiple buckets. Furthermore, almost all the peaks on the many AD spectra to be analyzed are projected on the projected AD spectrum Sp, which is the basis for creating this optimized bucket set, and the peak area of those peaks is optimized. The bucket set is reflected. Furthermore, the peak areas of known designated peaks that are considered important for analytical applications are also reflected in the optimized bucket set. Therefore, using this optimized bucket set, each bucket integration of a large number of AD spectra to be analyzed is performed, so that information on almost all the peaks contained in each AD spectrum is well maintained. In this state, it is converted into an integral data set (histogram).

  Refer to FIG. 6 again. When the optimized bucket set is completed, an optimized bucket integration process of multiple AD spectra in step 180 is performed.

  FIG. 19 shows a flow of optimization bucket integration processing 180 for multiple AD spectra.

  As shown in FIG. 19, at step 290, an optimized bucket set is read from the storage device. In step 292, each of a large number of AD spectra to be analyzed is read from the storage device 106. In step 292, the chemical shift axis scale (δ scale) of the read AD spectrum is adjusted to the chemical shift axis scale (δ scale) of the optimized bucket set (eg, the reference point of the AD spectrum is The chemical shift axis scale of the AD spectrum is shifted to match the reference point.) Then, in step 296, bucket integration of the AD spectrum is executed using the optimized bucket set, and in step 298, an integration data set (histogram data) obtained by the bucket integration is stored in the storage device 106 and necessary. Depending on the output.

  In step 299, the above-described processing in steps 292 to 298 is repeated for all AD spectra to be analyzed. As a result, all AD spectra to be analyzed are reduced to integral data sets (histogram data) and stored in the storage device 106.

  Refer to FIG. 6 again. After that, in step 182, the optimized bucket integral data set (histogram data) of all AD spectra to be analyzed is arranged in a data format suitable for passing to the next statistical processing and stored in the storage device 106. And output to the outside as necessary.

  6 is executed, the data reduction processing 112 using the optimized bucket integration shown in FIG. 1 and the integration result data storage / output processing are performed. 114 is complete. Thereafter, the statistical processing 116 shown in FIG. 1 is executed. In this statistical process 116, an integrated data set (histogram data) of all AD spectra to be analyzed is read from the storage device 106, and a predetermined statistical process (for example, multivariate) is used by using these integrated data sets (histogram data). A principal component analysis which is one of analysis methods is executed, and an analysis result such as determination of a peak serving as a marker is generated and output. As a specific method of statistical processing, various known methods can be used. For example, as will be described in detail later, in the statistical processing according to the principle of the present invention, principal component analysis (PCA) and subspace method (Soft Independent Modeling of Class Analogy), both of which are multivariate analysis methods. : SIMCA). As already explained, the integrated data set (histogram data) of the AD spectrum, which is the material for statistical processing, contains information on important peaks on the AD spectrum well. Is highly accurate.

  FIG. 20 shows the overall configuration and function of another embodiment of the NMR data processing apparatus according to the present invention.

  The NMR data processing apparatus 300 is used for a different application from the NMR data processing apparatus 100 described above. That is, the use of the NMR data processing apparatus 100 described above is to analyze the NMR spectra of a large number of samples and obtain statistical data, whereas the use of the NMR data processing apparatus 300 is basically one. Analyzing the NMR spectrum of a sample to examine the nature or composition of the sample (eg, analyzing the NMR spectrum of a person's blood to see if the person is afflicted with a particular disease Or analyzing the NMR spectrum of a food and examining what substances it contains). This NMR data processing apparatus 300 can also be realized by using a programmed computer system in the same manner as the NMR data processing apparatus 100 described above.

  As shown in FIG. 20, the processor 304 of the NMR data processing device 300 executes a predetermined computer program stored in the storage device 306, thereby performing an FID data input processing 308, an NMR spectrum processing 310, an optimized bucket integration. A data reduction process 312, a histogram analysis process 314, and an analysis result data storage / output process 316 are performed. In addition to the computer program, the storage device 306 stores in advance optimization bucket sets 280A, 280B, and 280C for each application as illustrated in FIG. Further, the storage device 306 stores reference model data corresponding to the inspection purpose. The reference model data corresponding to the inspection purpose is standard data to be compared with the NMR data to be inspected when the inspection is determined. For example, if the application is to test a person's blood to see if the person is afflicted with a particular disease, the NMR data of blood samples of a large number of people with the particular illness Marker information determined by analysis by the NMR data processing apparatus 200 can be adopted as one of the reference model data.

  In the FID data input process 312, FID data of one sample to be examined output from the NMR apparatus 102 is input and stored in the storage device 306. In the NMR spectrum processing 310, the input FID data is processed in the same flow as shown in FIG. 2 to generate an AD spectrum that is an absolute value derivative of the FID data, and is stored in the storage device 306. Remembered.

  In the data reduction process 312, an optimization bucket integration using an optimization bucket set corresponding to an application stored in advance in the storage device 306 is performed on the AD spectrum to be inspected. The AD spectrum is reduced to an optimized integral data set (histogram). In the histogram analysis process 314, the integral data set (histogram) to be inspected is compared with reference model data corresponding to the use stored in advance in the storage device 306, and a determination is made based on the comparison result to generate an inspection result. The In the analysis result storage / output process 316, the generated inspection result is stored in the storage device 306 and output to the outside.

  FIG. 21 shows the flow of the data reduction processing 312 using the optimized bucket integration shown in FIG.

  In step 320, the AD spectrum to be examined is read from the storage device 306, and in step 322, an optimized bucket set corresponding to the application is read from the storage device 306. At step 324, the chemical shift axis scale (δ scale) of the AD spectrum to be examined is matched to that of the optimized bucket set. Thereafter, in step 326, an optimized bucket integration using the optimized bucket set is performed on the AD spectrum to be inspected to obtain an integral data set (histogram) to be inspected. In step 328, the integral data set (histogram) to be inspected is stored in the storage device 306 and output to the outside as necessary.

  FIG. 22 shows the flow of the histogram analysis processing 314 shown in FIG.

  In step 330, the integral data set (histogram) to be examined is read from the storage device 306. In step 332, reference model data corresponding to the inspection purpose is read from the storage device 306. In step 334, the integral data set (histogram) to be inspected is compared with the reference model data corresponding to the inspection purpose, and the inspection result is determined. In step 336, the inspection result is stored in the storage device 306 and output to the outside.

  Even in the NMR data processing for the sample inspection as described above, the optimization bucket integration using the optimization bucket set prepared in advance according to the application is performed on the spectrum, so the peak information that the spectrum has Is maintained well, and a highly accurate inspection result can be obtained. In addition, since an AD spectrum, which is an absolute value derivative of the NMR spectrum, is used for the analysis, troublesome phase correction and baseline correction become unnecessary, and processing efficiency is improved.

  Next, an NMR spectrum statistical processing apparatus according to an embodiment of the present invention will be described.

  The statistical processing apparatus according to an embodiment of the present invention is incorporated as the statistical processing unit 116 in the NMR data processing apparatus 100 shown in FIG. 1, and typically, the preprocessing part (see FIG. 1 and the processing units 108, 110, 112, and 114) in FIG. However, this is merely an example for explanation, and the NMR spectrum statistical processing apparatus according to the present invention can be configured as a single apparatus separated from the pre-processed part already described.

  FIG. 22 shows a functional configuration of an NMR spectrum statistical processing apparatus according to an embodiment of the present invention, that is, the statistical processing unit 116 shown in FIG.

  As illustrated in FIG. 22, the statistical processing unit 116 includes a multivariate analysis unit 340, a user request input unit 342, and a display control unit 344.

  The multivariate analysis unit 340 includes a PCA analysis unit 346 and a SIMCA analysis unit 348. The PCA analysis unit 346 inputs an integral data set (hereinafter referred to as “NMR histogram”) of NMR spectra of a large number of samples stored in the storage device 106, and performs multi-component analysis (Principal component Analysis: PCA). Perform a random analysis. The SIMCA analysis unit 348 inputs NMR histograms of a large number of samples stored in the storage device 106, and performs multivariate analysis using a subspace method (Soft Independent Modeling of Class Analogy: SIMCA) (hereinafter referred to as SIMCA). To do. Histogram data of a large number of samples as analysis targets that are input to the PCA analysis unit 346 and the SIMCA analysis unit 348 are recorded in the PCA table 350 in the storage device 106. The configuration of the PCA table 350 will be described later.

  In the storage device 106, there is a large number of sample NMR histogram data 356 written by the integration result data storage / output process 114 shown in FIG. Each of the PCA analysis unit 346 and the SIMCA analysis unit 348 extracts NMR histogram data of an arbitrarily large number of samples from the large number of sample NMR histogram data 356 as an analysis target, and registers them in the PCA table 350. Can do. The multivariate analysis unit 340 also updates the PCA table 350 in response to a request from the user. Also, the PCA analysis unit 346 and the SIMCA analysis unit 348 respectively have a PCA analysis result (for example, sample score and variable loading) and a SIMCA analysis result (distance to each class of sample, variable modeling ability and discriminating ability). And the like can be stored in the storage device 106 as analysis result data 352.

  As will be described later in detail, the SIMCA analysis unit 348 and the PCA analysis unit 346 can operate in cooperation so that the analysis result of the former is fed back to the latter analysis. Accordingly, the purpose of performing such multivariate analysis, for example, the purpose of finding a biomarker in metabonomics can be achieved more easily.

  The display control unit 344 receives the PCA analysis result (score, loading, etc.) from the PCA analysis unit 346, and receives the SIMCA analysis result (distance to each class, modeling power, discriminating power, etc.) from the SIMCA analysis unit 348. Then, a graphical user interface (hereinafter referred to as “UI”) 346 displaying the PCA analysis result and the SIMCA analysis result is created, and the UI 346 is displayed on the display device (not shown) of the computer system 100 (FIG. 1). To do. The display control unit 344 reads the NMR spectrum data of one or more samples selected by the user from the NMR spectrum data 350 of a large number of samples stored in the storage device 106, and the PCA analysis result is displayed on the UI 346. Along with the SIMCA analysis results, the NMR spectrum of the sample selected by the user is displayed. In this embodiment, the displayed NMR spectrum data is not the original NMR spectrum data, but is AD spectrum data obtained by differentiating the absolute value thereof. Hereinafter, this is simply referred to as “NMR spectrum data”. Call it.

  The display control unit 344 controls the display on the UI 346 so that the display of the PCA analysis result, the SIMCA analysis result on the UI 346, and the display of the NMR spectrum cooperate. As a result, the operator can easily visually understand the relationship between the multivariate analysis results and the NMR spectrum, and the purpose of such multivariate analysis, for example, to find biomarkers in metabonomics, Will be easily achieved. Furthermore, on the UI 346, various input buttons for the user to input various requests are also displayed. Details of the UI 346 will be described later.

  The user request input unit 342 receives various requests input to the UI from a mouse or a keyboard by the user, and passes the input requests to the multivariate analysis unit 340 and the display control unit 344 described above. The multivariate analysis unit 340 and the display control unit 344 perform various operations as described below in response to a request input from the user.

  FIG. 24 shows a configuration example of the PCA table 350 shown in FIG.

  As described above, in the PCA table 350, NMR histogram data of a large number of samples to be analyzed are recorded. That is, as shown in FIG. 24, the PCA table 350 records a bucket integration condition 400 related to the bucket integration performed when creating the histogram, and this includes the chemical shift δ on which the bucket integration was performed. 9 (corresponding to the entire range 202 shown in FIG. 9), the basic width of the bucket (corresponding to the width W shown in FIG. 9), whether or not the bucket width is adjusted by the automatic integration block setting, The total integrated value of the range and dark region (corresponding to the dark region 204 shown in FIG. 9) (in the example shown, only one dark region of 5.3 ppm to 4.3 ppm is registered. Dark region can be registered).

  In the PCA table 350, the number of samples to be analyzed 402, the number of many variables (that is, the number of buckets) 404 constituting each sample, and the name of the folder storing the NMR spectrum data of each sample are stored. A set of 406 and file name 408 (that is, the path name of the NMR spectrum data file of each sample) is recorded. Here, in FIG. 23 described above, for the convenience of illustration, the NMR spectrum data of a large number of samples are collectively shown in one block 354, but this block 354 actually shows the NMR spectrum data for each sample. It is a collection of a large number of stored spectral data files, and the path names (folder name 406 and file name 408) of the large number of spectral data files are recorded in the PCA table 350. In FIG. 24, only the path names of the spectrum data files of three samples are shown for the convenience of illustration, but actually, the path names of all the spectrum data files of the number of samples shown in the sample number 402 are shown. It is recorded in the PCA table 350.

  Furthermore, the PCA table 350 defines a plot type 410 and a plot color 412 that specify the shape and color of the mark for each sample plotted on the UI 346 screen. The plot type 410 specifies, for example, a value “0” as an X mark, a value “1” as an unfilled circle mark, a value “−1” as a circle mark filled in black,. The plot color 412 is defined by a set of lightness (density) values of the three primary colors red (R), green (G), and blue (B). As will be described later, a score plot which is one of the PCA analysis results and a Kuman plot which is one of the SIMCA analysis results are displayed on the UI 346. On the score plot and the Kuman plot, each sample is displayed as follows. A mark having a shape and a color designated in the PCA table 350 is displayed.

  Further, a sample status 414 for each sample is registered in the PCA table 350. The sample status 414 of each sample indicates whether or not the corresponding sample is included in the analysis target. For example, the value “0” means that the corresponding sample is included in the analysis target, and the value “1”. "Means that the corresponding sample has been excluded from the analysis target at the request of the user. As will be described later, all of the samples registered in the PCA table 350 are initially included in the analysis target (that is, the sample status is “0”), but the user can make a score plot on the UI 346. By selecting an arbitrary sample on the Couman plot above, the selected sample can be excluded from the analysis target, and the sample status of the sample thus excluded is changed to “1”. Samples with a sample status of “1” cannot be entered in the PCA analysis or SIMCA analysis.

  Furthermore, a multivariate data matrix 418 that is a substantial analysis target is registered in the PCA table 350. This multivariate data matrix 418 includes chemical shift values (which correspond to variables or descriptors in multivariate analysis) 418 of each bucket included in the integration total range, and NMR for each sample. Histogram data (ie, a set of integration data obtained by optimization bucket integration for each sample and the net part of the multivariate data matrix to be analyzed) 420, the chemical shift value of the starting point of each bucket 422, the chemical shift value 424 of the end point, and the status 426 of each bucket are registered. The status 426 of each bucket indicates whether or not the bucket (that is, the variable) is used for analysis. For example, the value “0” means that the bucket is used for analysis, and the value “ “1” means not used for analysis. As will be described later, the user selects only a variable (bucket) having a discrimination power higher than an arbitrary level based on the discrimination power that is one of SIMCA analysis results displayed on the UI 346, and the selected bucket. It is possible to request that PCA analysis and SIMCA analysis be performed again using only All buckets initially have a status of "0" and are used for analysis, but only a specific bucket is selected for later analysis as described above The status of the unselected bucket is changed to “1”, and the bucket having the status “1” is not used in the later analysis.

  PCA analysis and SIMCA analysis are executed using the PCA table 350 having the above configuration.

  FIG. 25 shows the functions of the PCA analyzer 346, SIMCA analyzer 348, and display controller 344 of the multivariate analyzer 340 shown in FIG. 23 in more detail.

  As illustrated in FIG. 25, the PCA analysis unit 346 includes a PCA model reading / adding / saving unit 360 and a PCA calculation unit 362. The PCA model reading / adding / storing unit 360 selects NMR histogram data of a large number of samples to be analyzed from among the NMR histogram data 356 of a large number of samples on the storage device 106, and A function for newly creating information on the PCA table based on the NMR histogram data, or selecting NMR histogram data for additional samples from the NMR histogram data 356 for a large number of samples on the storage device 106, and storing them in the PCA table. Has the ability to add to a random data matrix. Further, the PCA model reading / adding / storing unit 360 has a function of storing information on the PCA table created, added, or changed as the PCA table 350 in the storage device 106. The PCA calculation unit 362 calculates a PCA analysis by using the information in the PCA table prepared by the PCA model reading / adding / saving unit 360 to create a PCA model, and the PCA analysis result based on the PCA model (for example, , Output the score of each sample, loading of each variable (each bucket), etc.). In addition, since the calculation method of PCA analysis itself is known (for example, refer nonpatent literature 1), the detailed description about this is abbreviate | omitted in this specification.

  The SIMCA analysis unit 370 includes a PCA model reading / adding / saving unit 370 and a SIMCA calculation unit 370. The PCA model reading / adding / saving unit 370 reads the NMR histogram data of the samples assigned by the user to the first and second classes from the NMR histogram data 356 of the large number of samples on the storage device 106, and the first And the function to prepare the PCA table information of the second two classes, and the NMR histogram data of the sample assigned by the user to the test class from the NMR histogram data 356 of a large amount of samples on the storage device 106, and the test It has a function to prepare PCA table information for classes. Further, the PCA model reading / adding / saving unit 370 has a function of saving the information of the PCA table of each class in the storage device 106 as the PCA table 350 of each class. The SIMCA calculation unit 372 creates a PCA model for each class by performing a calculation of SIMCA analysis using the information of the PCA table of a plurality of classes prepared by the PCA model reading / adding / storing unit 370. SIMCA analysis results (for example, distance from each class of each sample, modeling power and discriminating power of each variable (each bucket), etc.) are output. The SIMCA analysis calculation includes a PCA calculation for each class, and the SIMCA calculation unit 372 uses the PCA calculation unit 362 of the PCA analysis unit 346 in order to perform the PCA calculation for each class. In addition, since the calculation method of SIMCA analysis itself is well-known (for example, refer nonpatent literature 1), the detailed description about this is abbreviate | omitted in this specification.

  The PCA analysis unit 346 and the SIMCA analysis unit 348 include an analysis data storage unit 364 and an analysis data storage unit 366. The analysis data storage unit 364 stores the PCA analysis result output from the PCA analysis unit 346 and the SIMCA analysis result output from the SIMCA analysis unit 348. The analysis data storage unit 366 stores the PCA analysis result and the SIMCA analysis result stored in the analysis data storage unit 364 as analysis result data 352 on the storage device 106.

  As will be described in detail later, in the SIMCA analysis result stored in the analysis data storage unit 364, information for selecting a variable (bucket or a chemical shift value corresponding thereto) having a discrimination power higher than a certain level is included. include. This variable selection information can be fed back to the PCA calculation unit 362 and the SIMCA calculation unit 372 through the analysis data storage unit 364. Then, each of the PCA calculation unit 362 and the SIMCA calculation unit 372 only selects a selected variable (that is, a bucket having a discrimination power higher than a certain level or a chemical shift value corresponding thereto) according to the fed back variable selection information. The PCA analysis and SIMCA analysis are performed again, and the corrected PCA analysis result and SIMCA analysis result can be output again. As a result, more desirable PCA analysis results and SIMCA analysis results can be obtained.

  The display control unit 344 includes a PCA analysis screen control unit 374, a SIMCA analysis screen control unit 376, and an NMR spectrum cooperation unit 378. The PCA analysis screen control unit 374 receives the PCA analysis result output from the PCA calculation unit 362 to the analysis data storage unit 364 and, as will be described in detail later, a plurality of types of PCA analysis results that are displayed on the UI 346. A chart (for example, a score plot, a loading plot, a contribution rate chart, and a loading chart) (hereinafter collectively referred to as “PCA analysis result chart”) is displayed, and selection, enlargement, deletion, and deletion from the user input to the UI 346 In response to a request such as initialization, the charts on UI 346 are changed. The SIMCA analysis screen control unit 376 receives a SIMCA analysis result output from the SIMCA calculation unit 372 to the analysis data storage unit 364 and, as will be described in detail later, a plurality of types of charts (for example, the SIMCA analysis result) (Cuman plot, modeling power chart and discriminating power chart) (hereinafter collectively referred to as “SIMCA analysis result chart”) are displayed on the UI 346, and are selected, enlarged, deleted, and initialized from the user input to the UI 346. In response to the request, the charts on the UI 346 are changed.

  The NMR spectrum cooperation unit 378 reads the NMR spectrum data of one or more samples selected by the user on the UI 346 from the NMR spectrum data 354 of a large number of samples on the storage device 106, and the selected sample A chart showing the NMR spectrum is displayed on the UI 346 side by side with the PCA analysis result chart or SIMCA analysis result chart described above.

  In the following, the function of each unit illustrated in FIG. 25 will be described with reference to a screen example of the UI 346.

  FIG. 26 shows a screen example of the UI 346 when the PCA analysis result is displayed. FIG. 27 shows a screen example of the UI 346 when the SIMCA analysis result is displayed.

  As shown in FIGS. 26 and 27, the screen 430 of the UI 346 (hereinafter referred to as “UI screen”) in this embodiment is displayed on another screen in which the PCA analysis result and the SIMCA analysis result can be switched alternately. However, this is only an example of the display method. In order to use a sufficiently large display device, the PCA analysis result and the SIMCA analysis result may be displayed on the same screen.

  As shown in FIGS. 26 and 27, the UI screen 430 has a control panel 432 for displaying either the PCA analysis result or the SIMCA analysis result. 114, a preprocessing control panel 434 for a user to input various requests for controlling preprocessing of 114, a PCA control panel 436 for a user to input various requests for controlling PCA analysis, and SIMCA analysis. A SIMCA control panel 438 for the user to input various requests for control is arranged as shown.

  When the user selects (for example, clicks) the “PCA” selection radio button at the top of the PCA control panel 436 on the UI screen 430, the PCA analysis screen shown in FIG. The control unit 374 operates. The PCA analysis screen control unit 374 receives the PCA analysis result from the analysis data storage unit 364, and a plurality of PCA analysis result charts, that is, a score plot 440, a loading plot 442, as shown in FIG. A contribution plot 444 and a loading plot are displayed. On the other hand, when the “SIMCA” selection radio button at the top of the SIMCA control panel 438 is operated, the SIMCA analysis screen control unit 376 shown in FIG. 25 operates in response thereto. The SIMCA analysis screen control unit 376 receives the SIMCA analysis result from the analysis data storage unit 364 and, as shown in FIG. 27 on the UI screen 430, a plurality of SIMCA analysis charts, that is, the Kuman plot 462 and the first class. A modeling power chart 464, a modeling power chart 466 for the second class, and a discrimination power chart 468 are displayed.

  Further, as shown in FIGS. 26 and 27, when displaying either the PCA analysis result or the SIMCA analysis result, the NMR spectrum cooperation unit 378 shown in FIG. 25 operates, and the NMR screen is displayed on the UI screen 430. A spectrum chart 450 is displayed. The NMR spectrum chart 450 displays a graph of the NMR spectrum of one or more samples selected by the user. Here, it should be noted that, when the PCA analysis result is displayed, as shown in FIG. 26, the NMR spectrum chart 450, the contribution ratio chart 444, and the loading chart 446 are each of the horizontal axis of the chemical shift axis. They are arranged in the direction of the vertical axis perpendicular to the chemical shift axis so that the scales coincide with each other. Similarly, when the SIMCA analysis result is displayed, as shown in FIG. 27, the NMR spectrum chart 450 and the discriminating power chart 468 are set so that the scales of the chemical shift axes, which are the horizontal axes, coincide with each other. And arranged in the vertical axis direction orthogonal to the chemical shift axis. As a result, the user can easily visually grasp the correlation between the contribution rate, the variable having high loading or discriminating power (the bucket or the chemical shift value corresponding thereto) and the peak of the NMR spectrum, for example, biomarker in metabonomics. It will be easier to determine.

  In the following, the operation in each case will be described more specifically by dividing into the case of performing PCA analysis and the case of performing SMCA analysis. First, a case where PCA analysis is performed will be described.

  First, when the “PCA” selection radio button in the PCA control panel 436 on the UI screen 430 shown in FIG. 26 is operated and then the “Open_Model” button is operated, the PCA model reading / adding shown in FIG. 25 is performed. The storage unit 360 operates to select and read NMR histogram data of a plurality of samples designated by the user from among a large number of sample NMR histogram data 356 in the storage device 106, as shown in FIG. Create PCA table information consisting of multivariate data matrix and other information. If the number of samples specified by the user is N and the number of variables (buckets) constituting the NMR spectrum of each sample is M, the multivariate data matrix included in the PCA table information is N × M. It becomes a data matrix.

  The PCA model reading / adding / saving unit 360 delivers the information of the created PCA table to the PCA calculation unit 362. The PCA calculation unit 362 performs a PCA analysis calculation using the information in the PCA table, and calculates a PCA model for the multivariate data matrix included in the information in the PCA table. Here, the number of principal components calculated in the PCA analysis calculation is input by the user in the input box placed on the right side of the “PCA” selection radio button at the top of the PCA control panel 436 on the UI screen 430. (In the example shown, “6” is set as the number of principal components). As a PCA analysis result based on the calculated PCA model, the PCA calculation unit 362 obtains the score of each sample with respect to each principal component, the variance of the score, the loading of each variable with respect to each principal component, and the contribution rate of each variable (total principal components) The sum of the absolute values of loading with respect to) is calculated. The calculated PCA analysis result is stored in the analysis data storage unit 364.

  The PCA analysis screen control unit 374 receives the PCA analysis result from the analysis data storage unit 364. As shown in FIG. 26, the PCA analysis screen control unit 374 creates a score plot 440 based on the score of each sample for each principal component and displays it on the UI screen 430, and loads each variable for each principal component. Based on the above, a loading plot 442 and a loading chart 446 are created and displayed on the UI screen 430, and a contribution rate chart 444 is created and displayed on the UI screen 430 based on the contribution rate of each variable.

  As shown in FIG. 26, in the score plot 440, the scores for the two principal components selected by the user are taken on the X axis and the Y axis, respectively, and the scores for the two principal components of each sample are plotted. The principal components corresponding to the X axis and the Y axis are principal components of numbers selected by the user from the “X” axis pull-down menu and the “Y” axis pre-down menu in the PCA control panel 436 (in the illustrated example, The X-axis is the “first” principal component (PCA1), and the Y-axis is the “second” principal component (PCA2). The shape and color of the mark of the sample to be plotted follows the plot type and plot color (information corresponding to reference numbers 410 and 412 in FIG. 24) for each sample in the PCA table information. Furthermore, the corresponding principal component numbers and their variances are also displayed on the X-axis and Y-axis (example: PC1 (39.8%)). In addition, the number of samples used in the calculation is displayed.

  On the score plot 440, the closer the position between the plurality of sample marks, the more similar the samples are characteristically to each other, while the farther the position, the more different. Thus, if there are several sample mark chunks separated from each other on the score plot 440, for example, the sample to be analyzed is divided into several distinct classes corresponding to each chunk. Judgment can be made.

  Further, the variance of the score in each principal component means that the larger the value, the higher the degree that the principal component contributes to the classification of samples. For example, if the sum of the variances of the principal components of the X and Y axes is greater than a certain value (for example, about 60%), the sample can be classified fairly clearly in the PCA model using those principal components. Judgment can be made.

  When the user checks the “Name” check box in the PCA control panel 436, the PCA analysis screen control unit 374 displays on the score plot 440 as shown in FIG. 28 (enlarged view of a partial area in the score plot 440). The file names of the NMR spectra of the corresponding samples are displayed in the vicinity of all the sample marks. Even when the mouse pointer is placed on each sample, the corresponding file name is automatically displayed on the upper right of the score plot 440. Thereby, the user can easily grasp which sample each sample mark on the score plot 440 specifically corresponds to.

  As shown in FIG. 26, in the loading plot 442, the loadings of the two principal components selected by the “X” axis pull-down menu and the “Y” axis pre-down menu in the PCA control panel 436 are respectively taken as the X axis and the Y axis. The loading of each variable (bucket or corresponding chemical shift value) for the two principal components is plotted with variable marks having a predetermined shape and color. In addition, the corresponding principal component application numbers and their variances are displayed on the X and Y axes (example: PC1 (39.8%)). The number of variables (buckets) used for the calculation is also displayed.

  On the loading plot 442, variables that are farther from the origin of the X and Y axes have a higher contribution to the principal components corresponding to the X and Y axes, that is, a higher degree of contribution to the classification of samples. Means that.

  When the user checks the “Name” check box in the PCA control panel 436, the PCA analysis screen control unit 374 displays on the loading plot 442 as shown in FIG. 29 (an enlarged view of a partial area in the loading plot 442). The chemical shift value (ppm) of the corresponding variable (bucket) is displayed in the vicinity of all the variable marks. Also, when the mouse pointer is placed on each variable mark, the corresponding chemical shift value is automatically displayed on the upper right of the loading plot 442. Accordingly, the user can easily grasp which variable (chemical shift value) each variable mark on the loading plot 442 specifically corresponds to.

  As shown in FIG. 26, in the contribution rate chart 444, the horizontal axis represents the chemical shift, and the sum of the absolute value of the loading and the residual component for the first, second and third principal components of each variable (chemical shift value). Is displayed in the form of a bar graph. The variance of the scores of the first, second, and third principal components and the variance of the residual component are also displayed. In the contribution rate chart 444, it can be determined that the higher the contribution rate (the height of the bar graph), the higher the degree of contribution to the sample classification.

  As shown in FIG. 26, in the loading chart 446, the horizontal axis indicates the chemical shift, and the loading of each variable for one principal component selected from the “X” axis pull-down menu in the PCA control panel 436 is shown in the form of a bar graph. Is displayed. Also, the variance of the score of the selected principal component is displayed on the X axis. In the loading chart 446, it can be determined that the higher the absolute value of loading (the height of the bar graph), the higher the degree of contribution to the principal component.

  In the score plot 440, loading plot 442, contribution rate plot 444, and loading chart 446 shown in FIG. 26, various operations such as selection, enlargement / reduction, deletion, and initialization can be performed.

  First, various operations of the score plot 440 will be described.

  As shown in FIG. 30, when any one or more sample marks on the score plot 440 are selected by dragging the mouse, the PCA analysis screen control unit 374 (FIG. 25) displays the selection range 470 on the score plot 440. To do. In cooperation with this, the NMR spectrum cooperation unit 378 (FIG. 25) operates to store the NMR spectrum data of the sample corresponding to the selected sample mark, and the path name and file name of the sample registered in the PCM table. Are read from the NMR spectrum data 354 in the storage device 106, and the NMR spectrum of the selected sample is displayed on the NMR spectrum chart 450. The NMR spectrum chart 450 also displays the file name of the displayed NMR spectrum.

  On the NMR spectrum chart 450, a plurality of NMR spectra are displayed side by side in the vertical axis direction orthogonal to the chemical shift axis so that the chemical shift axis scales coincide with each other, and are distinguished and displayed in different colors. Thus, the mutual identification and correlation can be seen at a glance. In addition, as shown in FIG. 26, each NMR spectrum on the NMR spectrum chart 450, the contribution ratio chart 444, and the loading chart 446 are aligned in the vertical axis direction with the chemical shift axis scales matched with each other. So that you can see the correlation between them at a glance.

  When the user drags and selects an arbitrary sample mark on the score plot 440 with the mouse while pressing the “CTRL” key on the keyboard, the PCA analysis screen control unit 374 adds the sample mark to the selection range, or If the sample mark is already selected, the sample mark is removed from the selection range. In cooperation with this, the NMR spectrum linking unit 378 additionally displays the NMR spectrum of the sample corresponding to the sample mark added to the selected range on the NMR spectrum chart 450, or the sample spectrum removed from the selected range. The NMR spectrum is erased from the NMR spectrum chart 450.

  As shown in FIG. 31A, regarding the sample mark selected on the score plot 440, when the user makes a deletion request such as pressing the “Del” key, the PCA analysis screen control unit 374, as shown in FIG. 31B, The sample mark is deleted from the score plot 440, and in cooperation with this, the NMR spectrum cooperation unit 378 deletes the NMR spectrum of the selected sample from the NMR spectrum chart 450. At the same time, the PCA calculation unit 362 changes the sample status (corresponding to the reference number 414 in FIG. 24) of the sample requested to be deleted from “0” (included in the analysis target) to “1” (analysis). And the PCA analysis calculation is performed again using only the NMR histogram data of the sample whose sample status is “0” (included in the analysis target), and the result of the recalculation is analyzed. The data is output to the data storage unit 364. The PCA analysis screen control unit 374 receives the result of the recalculation from the analysis data storage unit 364, and on the UI screen 430, the score plot 440, the loading plot 442, the contribution rate chart 444, and the loading lot 446 based on the result of the recalculation. (FIG. 31B shows an example of the score plot 440 displayed after recalculation).

  For example, on the score plot 440, if a sample that is considered to be an obstacle or does not contribute to the analysis purpose is displayed, the sample is excluded from the analysis target as described above, and the PCA calculation is performed again. Thus, a more desirable analysis result can be obtained.

  When the user drags and selects an arbitrary partial area on the score plot 440 with the mouse while pressing the “Shift” key on the keyboard, the PCA analysis screen control unit 374 enlarges the selected partial area and displays the score plot. 440 is displayed. Thereafter, when the user selects the enlarged screen of the score plot 440 and presses the “Home” key on the keyboard, the PCA analysis screen control unit 374 initializes the enlargement ratio on the score plot 440 to “1”. , Display all score ranges.

  When the user operates the “Reset_score” button on the PCM control panel 436, the PCA calculation unit 362 sets the sample status that has been “1” by the sample deletion function in the PCA table to the initial value “0”. The PCA analysis calculation is performed again based on the contents of the PCA table initialized as described above, and the PCA analysis screen control unit 374 generates a score plot 440, a loading plot 442, a contribution rate based on the result of the PCA analysis calculation. The chart 444 and the loading lot 446 are displayed again.

  Next, various operations of the loading plot 442 will be described.

  As shown in FIG. 32, when the user selects one or more variable marks in the loading plot 442 by dragging with the mouse, the PCA analysis screen control unit 374 displays the selection range 472 on the loading plot 442. . In conjunction with this, the PCA analysis screen control unit 374 makes the bar graph of the variable (chemical shift value) corresponding to the selected variable mark in the contribution rate chart 444 and the loading chart 446 different from the bar graphs of other variables. It is displayed in a predetermined selection color (in the figure, the bar graph of the selected variable is displayed in black). Thereby, identification and correlation of each variable (chemical shift value) on the loading plot 442, the contribution rate chart 444, and the loading chart 446 can be known at a glance.

  When the user selects an arbitrary variable mark on the loading plot 442 by dragging with the mouse while pressing the “CTRL” key of the keyboard, the PCA analysis screen control unit 374 adds the variable mark to the selection range, or If the variable mark is already selected, the variable mark is removed from the selection range. In conjunction with this, the PCA analysis screen control unit 374 additionally displays a bar graph of variables corresponding to the variable marks added to the selection range in the selection color on the contribution rate chart 444 and the loading chart 446, or , Cancel the selected color display of the bar graph of the variable that has been removed from the selection range.

  Also, as shown in FIG. 33B, when the user makes a deletion request such as pressing the “Del” key for the variable mark selected in a certain selection range 476 as described above, the PCA analysis screen control unit 374 The variable mark is deleted from the loading plot 440. At the same time, the PCA calculation unit 362 sets the status (corresponding to the reference number 426 in FIG. 24) of the variable (bucket) corresponding to the variable mark requested to be deleted to “0” (used for analysis). To “1” (not used for analysis), and only the integral value corresponding to the variable (bucket) whose status is “0” (used for analysis) in the NMR histogram data of the sample to be analyzed The PCA analysis calculation is performed again, and the result of the recalculation is output to the analysis data storage unit 364. The PCA analysis screen control unit 374 receives the result of the recalculation from the analysis data storage unit 364, and on the UI screen 430, the score plot 440, the loading plot 442, the contribution rate chart 444, and the loading lot 446 based on the result of the recalculation. (FIG. 33B shows an example of a loading plot 442 displayed after recalculation.)

  For example, on the loading plot 440, if a variable that is considered to be an obstacle or not contribute to the analysis purpose is displayed, the variable is excluded from the variables used for the analysis as described above, and then the PCA calculation is performed again. By performing the above, a more desirable analysis result can be obtained.

  If the user drags and selects an arbitrary variable mark on the loading plot 442 with the mouse while pressing the “Shift” key, the PCA analysis screen control unit 374 corresponds to the selection range on the loading plot 442. Magnify and display. Thereafter, when the user selects the enlarged loading plot 442 and presses the “Home” key on the keyboard, the PCA analysis screen control unit 374 initializes the enlargement ratio of the loading plot 442 to “1”. Displays the chemical shift range.

  When the user operates the “Reset loading” button on the PCM control panel 436, the PCA calculation unit 362 sets the status of the bucket that has been “1” by the sample deletion function of the PCA table contents to the initial value “0”. , The PCA analysis calculation is performed again based on the contents of the PCA table initialized in this way, and the PCA analysis screen control unit 374 performs score plot 440 and loading plot 442 based on the result of the PCA analysis calculation. The contribution rate chart 444 and the loading lot 446 are displayed again.

  Next, various operations of the contribution rate chart 444 and the loading chart 446 will be described.

  When the user selects an arbitrary variable (chemical shift value) range on the chemical shift axis of the contribution rate chart 444 or the loading chart 446 by dragging with the mouse, the PCA analysis screen control unit 374 selects a variable belonging to the selected range. The variable mark on the loading chart 442 and the bar graph on the contribution rate chart 444 and the loading chart 446 are displayed in a predetermined selection color that is distinguished from that of the variable outside the selection range. Thereby, identification and correlation of each variable (chemical shift value) on the loading plot 442, the contribution rate chart 444, and the loading chart 446 can be known at a glance.

  When the user makes a deletion request such as pressing the “Del” key for the variable (chemical shift value) range selected on the contribution rate chart 444 or the loading chart 446 as described above, the PCA analysis screen control unit In step 374, the variable mark of the variable belonging to the selected range is deleted from the loading plot 440. At the same time, the PCA calculation unit 362 changes the status (corresponding to the reference number 426 in FIG. 24) of variables (buckets) belonging to the selected range from “0” (used for analysis) to “1” (analysis). PCA analysis using only integration values corresponding to variables (buckets) whose status is “0” (used for analysis) in the NMR histogram data of the sample to be analyzed The calculation is performed again, and the result of the recalculation is output to the analysis data storage unit 364. The PCA analysis screen control unit 374 receives the result of the recalculation from the analysis data storage unit 364, and on the UI screen 430, the score plot 440, the loading plot 442, the contribution rate chart 444, and the loading lot 446 based on the result of the recalculation. Is displayed.

  For example, when there is a variable (chemical shift value) range on the contribution rate chart 444 or loading chart 446 that is considered to be a problem or not contributing to the analysis purpose, the variable in the range is analyzed as described above. A more desirable analysis result can be obtained by performing the PCA calculation again after excluding the variables to be used.

  Further, as shown in FIG. 34A, the user selects an arbitrary variable (chemical shift value) range on the chemical shift axis of the loading chart 446 or the contribution ratio chart 444 by dragging with the mouse while pressing the “Shift” key. Then, the PCA analysis screen control unit 374 enlarges and displays a portion corresponding to the selected variable (chemical shift value) range in the loading chart 446 and the contribution ratio chart 444, as shown in FIG. 34B. In cooperation with this, the NMR spectrum cooperation unit 378 enlarges and displays a portion corresponding to the selected variable (chemical shift value) range in the NMR spectrum chart 450. Therefore, the loading chart 446, the contribution ratio chart 444, and the NMR spectrum chart 450 are always maintained in a state where the chemical shift axis scales coincide with each other even when displayed at any magnification, and the correlation between them can be seen at a glance. It is like that.

  After the loading chart 446, the contribution ratio chart 444 and the NMR spectrum chart 450 are enlarged and displayed in this way, the user selects the loading chart 446 or the contribution ratio chart 444 that has been enlarged, and presses the “Home” key on the keyboard. When pressed, the PCA analysis screen control unit 374 and the NMR spectrum cooperation unit 378 initialize the enlargement ratios of the charts 446, 444, and 450 to “1” and display all the chemical shift ranges.

  When performing PCA analysis, various operations as described above can be performed on the UI screen 430. When the user operates the “Add_Model” button in the PCA control panel 436 at any stage in the process of proceeding with the PCA analysis while sequentially performing such operations, the PCA model reading / adding / saving unit 360 stores Select and read the NMR histogram data of the additional sample specified by the user from the NMR histogram data 356 of the large number of samples in the apparatus 106, and add this to the multivariate data matrix in the PCA table currently being processed. Then, the information of the PCA table to which the multivariate data matrix is added is passed to the PCA calculation unit 362. The PCA calculation unit 362 re-executes the PCA analysis calculation using information of the PCA table to which the multivariate data matrix is added. The PCA analysis screen control unit 374 displays the result of the re-executed PCA analysis calculation on the UI screen 430 again.

  In addition, when the user operates the “Save_model” button in the PCA control panel 436 at any stage of the process of proceeding with the PCA analysis as described above, the PCA model reading / adding / saving unit 360 performs processing at that time. Information on the PCA table including the multivariate data matrix therein is stored as the PCA table 350 in the storage device 106.

  The above is the operation of each part when performing PCA analysis. As can be seen from the above description, when the PCA analysis is performed, all information of the score plot 440, the loading plot 442, the contribution rate chart 444, the loading chart 446, and the NMR spectrum chart 450 is organically combined on the UI screen 430. When any operation is performed, the operation result is reflected in real time on all the related charts. In particular, it is easy to quickly re-execute a PCA analysis after deleting a sample or a variable from an analysis target, and derive a more accurate analysis result. On the UI screen 430, the NMR spectrum is also displayed in cooperation with the PCA analysis result. Therefore, the PCA analysis result and the shape of the NMR spectrum can be evaluated together, and it becomes easier to consider the meaning of the analysis result. After performing the necessary operations in the PCA analysis process, it becomes easy to quickly execute PCA reanalysis and spectrum reprocessing from the evaluation results to derive more accurate analysis results.

  Next, the operation when SIMCA analysis is performed will be described.

  That is, first, when the “Model 1” button is operated after the “SIMCA” selection radio button in the SIMCA control panel 438 on the UI screen 430 shown in FIG. The PCA model reading / adding / storing unit 370 in the SIMCA analysis unit 348 operates, and the NMR histogram data of a plurality of samples specified by the user is selected from the NMR histogram data 356 of a large number of samples in the storage device 106. Select and read to create PCA table information consisting of a multivariate data matrix and other information as shown in FIG. Information on the created PCA table is handled as PCA table information of the first class. Similarly, when the “Model 2” button is operated, the PCA model reading / adding / saving unit 370 selects the NMR histogram data of a plurality of samples designated by the user from the NMR histogram data 356 in the storage device 106. Select and read to create second class PCA table information.

  When starting the SIMCA analysis, at least first and second class PCA tables need to be defined. Further, if desired, the user can operate the “Test_model” button when the “Model1” button in the SIMCA control panel 438 is operated. When the “Test_model” button is operated, the PCA model reading / adding / saving unit 370 selects NMR histogram data of a plurality of samples designated by the user from the NMR histogram data 356 in the storage device 106. Read and create PCA table information of test class.

  The PCA model reading / adding / saving unit 370 delivers the information of the created first class and second class PCA tables to the SIMCA calculation unit 372. When a test class PCA table is further created, information on the test class PCA table is also delivered to the SIMCA calculation unit 372. The SIMCA calculation unit 372 calls the PCA calculation unit 362 using the information of the PCA table of those classes, calculates the PCA model for each class, and performs the calculation of the SIMCA analysis using the PCA model of those classes. The number of principal components of PCA models of various classes follows the number input by the user in the input box on the right side of the “SIMCA” selection radio button in the SIMCA control panel 438. The SIMCA calculation unit 372 calculates the SIMCA analysis results as follows: distance of each sample to each of the first and second classes, residual standard deviation (RSD) of each of the first and second classes, and each of the first and second classes. The modeling power of each variable (chemical shift value) and the discriminating power of each variable (chemical shift value) are calculated. The calculated SIMCA analysis result is stored in the analysis data storage unit 364.

  The SIMCA analysis screen control unit 374 receives the SIMCA analysis result from the analysis data storage unit 364. As shown in FIG. 27, the SIMCA analysis screen control unit 376 creates a Couman plot 460 based on the distance of each sample with respect to each of the first and second classes and displays it on the UI screen 430. Based on the modeling ability of each variable (chemical shift value) for each class of 2, a first class modeling ability chart 462 and a second class modeling ability chart 464 are created and displayed on the UI screen 430. Based on the discriminating power of each variable (chemical shift value), a discriminating power chart 466 is created and displayed on the UI screen 430.

  As shown in FIG. 27, in the Kuman plot 460, the distances from the first class and the second class are taken on the X axis and the Y axis, respectively, and the distances to the two classes of each sample are plotted. The shape and color of the mark of the sample to be plotted follows the plot type and plot color (information corresponding to reference numbers 410 and 412 in FIG. 24) for each sample in the PCA table information. In addition, on the Couman plot 460, the names and values (eg, RSD (1) 0.5449) of the residual standard deviation (RSD) for each of the first and second classes are displayed near the X and Y axes. In addition, straight lines (hereinafter referred to as “RSD” lines) 480 and 482 indicating the residual standard deviation (RSD) of each class are displayed.

  It can be determined that the sample is more clearly classified into the first class and the second class as the arrangement state of the sample marks on the Kuman plot 460 is closer to the following clear classification state. That is, the clear classification state is that the sample mark is divided into two chunks, one chunk is in the area between the first class RSD line 480 and the Y axis, and the other chunk is the second class RSD. It is in a state where it exists in the region between the line 482 and the X axis.

  When the user checks the “Name” check box in the SIMCA control panel 438, the SIMCA analysis screen control unit 374, as in the case of the score plot 440 described with reference to FIG. The file name of the NMR spectrum of the corresponding sample is displayed near the sample mark. Even when the mouse pointer is placed on each sample, the corresponding file name is automatically displayed on the upper right of the Kuman plot 460. Thereby, the user can easily grasp which sample each sample mark of the Kuman plot 460 specifically corresponds to.

  As shown in FIG. 27, in the modeling charts 462 and 464 of the first class and the second class, the chemical shift is taken on the horizontal axis, and the modeling power (this) for the first class and the second class of each variable (chemical shift value). Is an index for evaluating the effectiveness of each variable for describing the data structure of each class), and is displayed in the form of a bar graph. In the modeling charts 462 and 464 of each class, it can be determined that the variable having a higher modeling ability (height of the bar graph) is more effective in describing the data structure of each class.

  As shown in FIG. 27, in the discriminating power chart 466, the horizontal axis indicates the chemical shift, and the discriminating power of each variable (chemical shift value) (this is the classification of each variable for classification into the first and second classes). Is an indicator for evaluating effectiveness) in the form of a bar graph. In the discriminating power chart 466, it can be determined that the higher the discriminating power (the height of the bar graph), the higher the effectiveness for classification into the first and second classes.

  In the Cooman plot 460, the modeling charts 462 and 464, and the discriminant power chart 466 shown in FIG. 27, various operations such as selection, enlargement / reduction, deletion, and initialization can be performed.

  First, various operations of the Kuman plot 460 will be described.

  As shown in FIG. 35, when any one or more sample marks on the Couman plot 460 are selected by dragging the mouse, the SIMCA analysis screen control unit 376 displays the selection range 484 on the Couman plot 460. In cooperation with this, the NMR spectrum cooperation unit 378 operates, and the NMR spectrum data of the sample corresponding to the selected sample mark is used by using the path name and file name of the sample registered in the PCM table. The NMR spectrum data 354 in the storage device 106 is read, and the NMR spectrum of the selected sample is displayed on the NMR spectrum chart 450. The NMR spectrum chart 450 also displays the file name of the displayed NMR spectrum.

  On the NMR spectrum chart 450, a plurality of NMR spectra are displayed side by side in the vertical axis direction orthogonal to the chemical shift axis so that the chemical shift axis scales coincide with each other, and are distinguished and displayed in different colors. Thus, the mutual identification and correlation can be seen at a glance. In addition, as shown in FIG. 27, each NMR spectrum on the NMR spectrum chart 450 and the discriminating power chart 466 are displayed side by side in the vertical axis direction with the chemical shift axis scales matched. The correlation between them can be seen at a glance.

  When the user selects an arbitrary sample mark on the Couman plot 460 while dragging with the mouse while pressing the “CTRL” key on the keyboard, the SIMCA analysis screen control unit 376 adds the sample mark to the selection range, or If the sample mark is already selected, the sample mark is removed from the selection range. In cooperation with this, the NMR spectrum linking unit 378 additionally displays the NMR spectrum of the sample corresponding to the sample mark added to the selected range on the NMR spectrum chart 450, or the sample spectrum removed from the selected range. The NMR spectrum is erased from the NMR spectrum chart 450.

  As shown in FIG. 36A, when the user makes a deletion request such as pressing the “Del” key for the sample mark selected on the Kuman plot 460, the SIMCA analysis screen control unit 376, as shown in FIG. 36B, The sample mark is erased from the Couman plot 460, and in cooperation therewith, the NMR spectrum cooperation unit 378 erases the NMR spectrum of the selected sample from the NMR spectrum chart 450. At the same time, the SIMCA calculation unit 372 changes the sample status (corresponding to the reference number 414 in FIG. 24) of the corresponding class of the sample requested to be deleted from “0” (included in the analysis target) to “1”. ”(Not included in the analysis target), and the SIMCA analysis calculation is performed again using only the NMR histogram data of the sample whose sample status is“ 0 ”(included in the analysis target). The result is output to the analysis data storage unit 364. The SIMCA analysis screen control unit 376 receives the recalculation result from the analysis data storage unit 364 and displays the Couman plot 460, the modeling plots 462 and 464, and the discriminating power chart 466 based on the recalculation result on the UI screen 430. (FIG. 36B shows an example of the Couman plot 460 displayed after recalculation).

  For example, on the Couman plot 460, if a sample that is considered to be a problem or does not contribute to the analysis purpose is displayed, the sample is excluded from the analysis target as described above, and then the SIMCA calculation is performed again. Thus, a more desirable analysis result can be obtained.

  If the user drags and selects an arbitrary partial area on the Couman plot 460 with the mouse while pressing the “Shift” key on the keyboard, the SIMCA analysis screen control unit 376 enlarges the selected partial area and sets the Couman plot. 460 is displayed on the screen. Thereafter, when the user selects the enlarged screen of the Kuman plot 460 and presses the “Home” key on the keyboard, the SIMCA analysis screen control unit 376 initializes the enlargement ratio on the Kuman plot 460 to “1”. , Display the entire original display range.

  When the user operates the “Reset_score” button on the SIMCA control panel 438, the SIMCA calculation unit 372 sets the sample status that has been set to “1” in the PCA table of each class to the initial value “0”. Then, the SIMCA analysis calculation is performed again based on the contents of the PCA table thus initialized, and the SIMCA analysis screen control unit 376 performs the Kuman plot 460 and the modeling plot 462 based on the result of the SIMCA analysis calculation. 464 and the discriminating power chart 466 are displayed again.

  Next, various operations of the modeling power charts 462 and 464 and the discrimination power chart 466 will be described.

  As shown in FIG. 37, when the user drags an arbitrary height position in the vertical axis direction (that is, an arbitrary discriminating power value) as a threshold by dragging with the mouse on the discriminating power chart 466, SMCA analysis screen control is performed. The unit 376 displays a threshold line 486 indicating the designated threshold on the discriminating power chart 466, selects only variables having discriminating power higher than the designated threshold, and displays a bar graph of the selected variables Are displayed in a predetermined selection color different from other variables in all of the modeling power charts 462 and 464 and the discrimination power chart 466. At the same time, the analysis result storage unit 364 stores information that identifies the selected variable (that is, a variable having a higher discriminating power than the specified threshold value). Similarly, when the user designates a certain modeling force value as a threshold value on the first class or second class modeling force chart 462 or 464, a variable having a modeling ability higher than the threshold value is different from other variables. Are displayed in different selected colors, and information specifying the selected variable is stored in the analysis result storage unit 364. Thereby, the identification and correlation of the selected variables in the modeling power charts 462 and 464, the discriminating power chart 466 and the NMR spectrum chart 450 can be understood at a glance.

  After the variable having high discriminating power or modeling power is selected as described above, when the user operates the “Set_D.P.” Button in the SIMCA control panel 438, the SIMCA calculating unit 372 reads from the analysis data storage unit 364. Receives specific information on selected variables, changes the status of unselected variables in the PCA table information for each class from “0” (used for analysis) to “1” (not used for analysis), and The SIMCA calculation is performed again using only the selected variable. Subsequently, as shown in FIG. 38, the SIMCA analysis screen control unit 376 receives the result of the SIMCA recalculation, and displays the Coomman plot 460, the modeling plots 462 and 464, and the discriminating power chart 466 again based on the result. In this way, it is possible to obtain a more desirable analysis result by performing the SIMCA analysis again using only a variable having a high discrimination power or modeling power.

  When a user operates the “Set_D.P.” Button in the PCA control panel 436 after a variable having high discriminating power or modeling power is selected as described above, the PCA calculation unit 362 causes the analysis data storage unit to PCA table information for PCA analysis with multivariate data matrix of samples of all classes is created by receiving the specific information of the variable selected from 364 and integrating PCA table information of all classes used in SIMCA analysis Then, the status of the selected variable in the PCA table information for PCA analysis is set to “0” (used for analysis), and the status of the unselected variable is set to “1” (not used for analysis). Then, using only the selected variable, the PCA calculation unit 362 performs the PCA calculation again on the multivariate data matrix of the samples of all classes in the PCA table information for PCA analysis. Subsequently, the PCA analysis screen control unit 374 switches the UI screen 430 from the one for SIMCA analysis as shown in FIG. 27 to the one for PCA analysis as shown in FIG. A score plot 440, a loading plot 442, a contribution rate chart 444 and a loading chart 446 are displayed again. In this way, it is possible to obtain a more desirable analysis result by performing the PCA analysis again using only the variables having high discrimination power or modeling power.

  In addition, in the UI screen 430 for SIMCA analysis as shown in FIG. 27, while the user presses the “Shift” key, an arbitrary variable (chemical variable) on the chemical shift axis in the modeling force charts 462 and 464 or the discriminating power chart 466 is displayed. When the shift value) range is selected by dragging with the mouse, the SIMCA analysis screen control unit 376 selects a portion corresponding to the selected variable (chemical shift value) range in the modeling power charts 462 and 464 and the discriminating power chart 466. Enlarge and display. In cooperation with this, the NMR spectrum cooperation unit 378 enlarges and displays a portion corresponding to the selected variable (chemical shift value) range in the NMR spectrum chart 450. Therefore, the discriminating power chart 466 and the NMR spectrum chart 450 are always maintained in a state where the chemical shift axis scales coincide with each other even when displayed at any magnification, and the correlation between them can be seen at a glance. Yes.

  After the modeling power charts 462 and 464, the discriminating power chart 466 and the NMR spectrum chart 450 are enlarged and displayed, the user selects the enlarged modeling power charts 462 and 464 or the discriminating power chart 466, and the keyboard is selected. When the “Home” key is pressed, the SIMCA analysis screen control unit 376 and the NMR spectrum cooperation unit 378 initialize the magnifications of the charts 462, 464, and 466 to “1” and display all the chemical shift ranges. .

  The above is the operation of each unit when performing SIMCA analysis. As can be understood from the above description, when the SIMCA analysis is performed, all information of the Couman plot 460, the modeling plots 462 and 464, the discriminating power chart 466, and the NMR spectrum chart 450 is organically combined on the UI screen 430. When any operation is performed, the operation result is reflected in real time on all the related charts. In particular, it is easy to quickly re-execute SIMCA analysis after deleting a sample from the analysis target or selecting a variable, and to derive a more accurate analysis result. On the UI screen 430, an NMR spectrum is also displayed in cooperation with the SIMCA analysis result. Therefore, the SIMCA analysis result and the shape of the NMR spectrum can be evaluated together, and it becomes easier to consider the meaning of the analysis result.

  In addition, in the SIMCA analysis, two measures are calculated for each variable: modeling ability and discriminating ability. The discriminating power represents which variable is effective for classification for the two classes. Based on the discriminating power, variables that contribute to classification can be extracted, and other variables can be considered as noise that does not contribute much to classification. In this embodiment, as described above, the PCA analysis or SIMCA analysis can be executed again using only the variables effective for classification. This makes it possible to classify the samples more clearly.

  FIG. 39 shows a processing procedure for re-executing SIMCA analysis calculation or PCA analysis calculation using a variable having high discriminating power selected in SIMCA analysis.

  As shown in FIG. 39, when a SIMCA analysis is performed in step 500 and a threshold value of discrimination power is specified on the discrimination power chart 466 in step 502, a variable having discrimination power equal to or higher than the threshold value is determined in step 504. Extracted. Thereafter, when the “Set_D.P.” Button on the SIMCA control panel 438 is operated in step 506, the SIMCA analysis calculation is executed again using only the extracted variable in step 508, and the SIMCA analysis result is displayed. Is done.

  In addition, after step 504, when the “Set_D.P.” Button on the PCA control panel 436 is operated in step 510, in step 512, only the extracted variables are used for SIMCA analysis. The PCA analysis calculation is executed again for the sample, and the PCA analysis result is displayed. In addition, after step 504, when the “Add_D.P.” Button on the PCA control panel 436 is operated in step 514, it is extracted in step 516 as a variable to be used for the current analysis. After adding these variables, PCA analysis calculation is performed again for the sample that was subjected to SIMCA analysis using those variables, and the PCA analysis results are displayed.

  By the way, the variable addition function using the “Add_D.P.” Button described above can be used for the following purposes, for example. That is, when there are three or more classes, SIMCA analysis is performed with each combination of the two classes as the first and second classes to obtain a variable effective for classification, and the effective variable is defined as “Add_D” described above. .P. "Button can be added sequentially. That is, first, a SIMCA analysis is performed with the combination of the two classes selected first, and a valid variable is selected. Then, by operating the “Set_D.P.” Button on the PCA control panel 436, the selected variable is selected. Set the target. Next, after performing SIMCA analysis with another two class combination and selecting a valid variable, the selected variable is first displayed by operating the “ADD_D.P.” Button on the PCA control panel 436. Append to selected variable. Thereafter, SIMCA analysis is performed for each of the other two combinations of classes to select valid variables, and then the above “ADD_D.P.” Button operation is repeated. As a result, a variable that is effective for each class combination is set as a target of use, and only a variable that is not valid for any combination of classes is excluded from the target of use.

  The configuration, function, and operation of the NMR spectrum statistical processing apparatus (the statistical processing unit 116 illustrated in FIG. 1) according to an embodiment of the present invention have been described above with reference to FIGS. 23 to 39. Below, the usage example in the case of using this statistical processing apparatus of a NMR spectrum for a specific use is introduced.

  FIG. 40 shows a procedure for using this statistical processing apparatus for the purpose of determining a biomarker in metabonomics.

  As shown in FIG. 40, analysis begins at step 520. In this analysis, for example, if you want to determine a biomarker to determine whether a person is suffering from a particular disease based on a sample from that person, a sample from a person already suffering from the disease Goods and samples from healthy people are selected as samples for analysis. Different colors are specified as plot colors for each class. First, in step 522, PCA analysis is performed on a number of samples to be analyzed. In step 524, the sum of the variances of the scores of the first principal component (PC1) and the second principal component (PC2) shown in the score plot 440 or the like exceeds a certain threshold determined by the user, for example, 60%. It is determined by the user whether or not it exists. If the sum of the variances of the scores exceeds the threshold value of 60%, in step 526, the user clearly classifies the sample in the PCA model obtained by the PCA analysis performed in step 522. Judge. In that case, in step 528, the user selects a chemical shift value having a high contribution rate or a corresponding NMR spectrum peak while referring to the loading plot 442, the contribution rate chart 444, the loading chart 446, the NMR spectrum chart 450, and the like. Then, the biomarker is determined based on that.

  On the other hand, in step 524, if the sum of the variances of the scores is 60% or less, the user does not clearly classify the sample in the PCA model obtained by the PCA analysis performed in step 522. Then, the SIMCA analysis in step 532 is executed. In the SIMCA analysis, in step 534, the sample to be inspected is divided into first and second classes. For example, samples taken from people suffering from the disease are assigned to a first class, and samples taken from healthy people are assigned to a second class. When the SIMCA analysis is executed for these two classes and the SIMCA analysis result is displayed, the user sets a threshold value of the discrimination power on the discrimination power chart 466 in step 536, and the discrimination power equal to or higher than the threshold value is set. Select the variable you have. Thereafter, the user operates the “Set_D.P.” Button on the PCA control panel 436. Thereby, in step 522, the PCA analysis calculation is performed again on the same sample using only the selected variable. As a result, a PCA model in which the sample is more clearly dispersed than the result of the previous PCA analysis calculation is obtained, and the determination of the biomarker becomes easier.

  FIG. 41 shows an example of a use procedure when the NMR spectrum statistical processing apparatus (the statistical processing unit 116 shown in FIG. 1) according to an embodiment of the present invention is used for diagnosing the propensity of a test sample. .

  As shown in FIG. 41, diagnosis is started at step 540. This diagnosis is, for example, diagnosing whether or not a certain person has a specific disease. In step 542, two basic classes are set, the first and second classes. For example, samples taken from people who already have a disease are assigned to a first class, and samples taken from healthy people are assigned to a second class. In step 544, a sample collected from people to be diagnosed is set as a test class. Different colors are specified as plot colors for each class.

  Thereafter, in step 546, a PCA analysis is performed on the multivariate data matrix integrating the three classes of samples. Also, in step 548, SIMCA analysis is performed on these three classes of samples. In the SIMCA analysis, the distance to each of the first class and the second class of the sample of the test class is obtained and expressed on the Couman plot 440. In both the score plot 440 and the Kuman plot, the three classes of samples are displayed in different plot colors by class, so that their correlation can be seen at a glance.

  In step 550, the user refers to the PCA analysis result and the SIMCA analysis result, so that the test target people, the first class of people suffering from the disease, and the second class of health. From this, the person who is diagnosed has the disease, is healthy, or is undecided.

  As mentioned above, although embodiment of this invention was described, this embodiment is only the illustration for description of this invention, and is not the meaning which limits the scope of the present invention only to this embodiment. The present invention can be implemented in various other modes without departing from the gist thereof. For example, in the above embodiment, the score plot, loading, and Cooman plot, which are the results of multivariate analysis, are displayed in the form of a two-dimensional figure consisting of two coordinate axes on the UI screen, but instead of or in combination with it. It is also possible to form a three-dimensional figure composed of three coordinate axes so that the three-dimensional figure can be displayed from an arbitrary line-of-sight direction. Further, the present invention can be applied not only to metabonomics but also to other uses.

The block diagram which shows the whole structure and function of one Embodiment of the NMR data processing apparatus according to this invention. The flowchart which shows the flow of the NMR spectrum process 110. The figure which shows an example of the AD spectrum which is a result of having given an NMR value (real part) and absolute value differentiation to this. The figure explaining phase correction. The figure explaining baseline correction. The flowchart which shows the flow of the data reduction process 112 using an optimization bucket integral. The figure explaining the principle of the spectrum projection process 172. FIG. The flowchart which shows the flow of the spectrum projection process 172. The figure explaining the principle of the bucket integration process 172 of a projection AD spectrum. The flowchart which shows the flow of the bucket integration process 172. The figure explaining the principle of the automatic correction process 174 of the bucket based on an automatic integration block setting. The flowchart which shows the flow of the automatic correction process 174 of the bucket based on an automatic integration block setting. The figure which shows the example of the peak of 3 split and 4 split on a NMR spectrum. The figure which shows the data structure example of designation | designated peak information. The flowchart which shows the flow of the bucket automatic correction process 176 based on designated peak information. The figure which shows the principle of the manual correction process 178 of a bucket. The flowchart which shows the flow of the manual correction process 178 of a bucket. The figure which shows the data structure example of an optimization bucketset. The flow of the optimization bucket integration process 180 of many AD spectrum is shown. The block diagram which shows the whole structure and function of another embodiment of the NMR data processing apparatus according to this invention. The flowchart which shows the flow of the data reduction process 312 using an optimization bucket integral. The flowchart which shows the flow of the analysis process 314 of a histogram. The block diagram which shows the functional structure of the statistical processing apparatus of the NMR spectrum concerning one Embodiment of this invention, ie, the statistical processing part 116 shown in FIG. The figure which shows the structural example of the PCA table 350. FIG. 24 is a block diagram showing in more detail the functions of the PCA analyzer 346, SIMCA analyzer 348, and display controller 344 of the multivariate analyzer 340 shown in FIG. The figure which shows the example of a screen of UI346 when a PCA analysis result is displayed. The figure which shows the example of a screen of UI346 when a SIMCA analysis result is displayed. The enlarged view of the partial area | region in the score plot 440 for demonstrating the display function of a file name. The enlarged view of the partial area | region in the loading plot 442 for demonstrating the display function of a chemical shift value. The figure which shows the example of the score plot 440 and NMR spectrum yart 450 for demonstrating the plot selection function and the NMR spectrum cooperation function on the score plot 440. FIG. The figure which shows the example of the score plot 440 for demonstrating the deletion function of the plot on the score plot 440. FIG. The figure which shows the example of the loading plot 442, the contribution rate chart 444, and the loading chart 446 for demonstrating the plot selection function on the loading plot 442, the contribution rate chart 444 linked to it, and the display change function of the loading chart 446. The figure which shows the example of the loading plot 442 for demonstrating the deletion function of the plot on the loading plot 442. FIG. The figure which shows the example of the chart 450 of the contribution ratio chart 444, the loading chart 446, and NMR spectrum for demonstrating the expansion function of the contribution ratio chart 444 or the loading chart 446, and an NMR spectrum cooperation function. The figure which shows the example of the Kuart plot 460 and NMR spectrum Yart 450 for demonstrating the plot selection function on the Couman plot 460, and an NMR spectrum cooperation function. The figure which shows the example of the Cooman plot 460 for demonstrating the deletion function of the plot on the Cooman plot 460. FIG. The figure which shows the example of the modeling charts 462 and 464 and the discrimination power chart 466 for demonstrating the variable selection function in the modeling charts 462 and 464 and the discrimination power chart 466. The figure which shows the example of the UI screen 430 for demonstrating re-execution of the SIMCA analysis calculation after variable selection. The flowchart which shows the procedure of re-execution of SIMCA analysis calculation after variable selection and PCA analysis calculation. The flowchart which shows an example of the use procedure in the case of using the statistical processing apparatus (statistical processing part 116 shown in FIG. 1) of the NMR spectrum concerning one Embodiment of this invention for the use which determines a biomarker in metabonomics. The flowchart which shows an example of the use procedure in the case of using the statistical processing apparatus (statistical processing part 116 shown in FIG. 1) of the NMR spectrum concerning one Embodiment of this invention for the use which diagnoses the tendency of a test sample.

Explanation of symbols

100, 300 Computer system (NMR data processor)
102 NMR device 104, 304 Processor 106, 306 Storage device 108, 308 FID data input processing 110, 310 NMR spectrum processing 112, 312 Data reduction processing 126 Absolute value differentiation processing (AD spectrum generation processing)
140 NMR spectrum (real part)
144, Sa, Sb AD spectrum 171 Projection processing of many AD spectra 172 Bucket integration processing of projected AD spectrum 174 Automatic correction processing of bucket based on automatic integration block setting 176 Automatic correction processing of bucket based on specified peak information 178 Manual operation of bucket Correction processing 180 Multiple AD spectrum optimization bucket integration processing
Sp projected AD spectrum 200a-200f, 200i-200k bucket 220x, 220y automatically set integration block 240x, 240y peak area based on specified peak information 250A-250C specified peak information set 280A-280C optimized bucket set 340 multivariate analysis unit 342 User request input unit 344 Display control unit 346 PCA analysis unit 348 SIMCA analysis unit 350 PCA table 354 NMR spectrum data 356 NMR histogram data 360 PCA model reading / adding / saving unit 362 PCA calculation unit 364 Analysis data storage unit 366 Analysis result storage Unit 370 PCA model reading / addition / saving unit 372 SIMCA calculation unit 374 PCA analysis screen control unit 376 SIMCA analysis screen control unit 378 NMR spectrum linkage unit 430 UI screen 440 score plot 442 loading plot 4 44 Contribution chart 446 Loading chart 450 NMR spectrum chart 460 Cooman plot 462, 464 Modeling power plot 466 Discrimination power plot

Claims (10)

NMR spectrum data defined by the mathematical formula (1) indicating the NMR characteristics of the sample, and means for performing absolute value differentiation on the NMR spectral data to obtain the absolute value differential spectral data defined by the mathematical formula (2);
Data reduction means for reducing the absolute value differential spectrum data to histogram data by executing bucket integration using a bucket set consisting of a large number of buckets having unequal widths on the absolute value differential spectrum data. When,
Means for storing or outputting the histogram data;
When δ is a chemical shift, j is an imaginary number symbol, R (δ) is a real part of the NMR spectrum, and I (δ) is an imaginary part of the NMR spectrum, Equations (1) and (2) are respectively
NMR spectrum = R (δ) + j · I (δ) (1)

Because
The width of the bucket is determined according to the detected peak width in advance so that individual peak areas included in the absolute value differential spectrum are not divided into a plurality of buckets. An apparatus for processing NMR data. The apparatus of claim 1.
An NMR data processing apparatus in which the bucket set having the uneven width is set so that none of the designated one or more peak areas is divided into a plurality of buckets. The apparatus of claim 1.
An apparatus for processing NMR data in which the bucket set having an unequal width is set so that none of one or more peak areas detected from the absolute value differential spectrum data is divided into a plurality of buckets. The apparatus of claim 1.
An apparatus for processing NMR data, further comprising bucket setting means for setting a bucket set having an uneven width. The apparatus of claim 4.
The bucket setting means
A peak designating means for designating one or more peak areas;
An NMR data processing apparatus comprising: means for setting each bucket so that none of the specified one or more peak areas is divided into a plurality of buckets. The apparatus of claim 4.
The bucket setting means
Detecting means for detecting a peak region from the absolute value differential spectrum data;
An NMR data processing apparatus comprising: means for setting each bucket so that none of the detected one or more peak regions is divided into a plurality of buckets. The apparatus of claim 6.
The detection means receives a plurality of absolute value differential spectrum data, generates projection spectrum data on which the plurality of absolute value differential spectrum data is projected, and detects a peak region from the projection spectrum data. NMR data processing equipment. The apparatus of claim 4.
An apparatus for processing NMR data, wherein the bucket setting means has means for correcting each bucket in response to a request input from an operator. NMR spectrum data defined by the mathematical expression (1) indicating the NMR characteristics of the sample, and obtaining absolute value differential spectral data defined by the mathematical expression (2) by subjecting the NMR spectrum data to absolute value differentiation;
Reducing the absolute value differential spectrum data into histogram data by performing bucket integration using a bucket set consisting of multiple buckets having unequal widths on the absolute value differential spectrum data;
Storing or outputting the histogram data,
When δ is a chemical shift, j is an imaginary number symbol, R (δ) is a real part of the NMR spectrum, and I (δ) is an imaginary part of the NMR spectrum, Equations (1) and (2) are respectively
NMR spectrum = R (δ) + j · I (δ) (1)

Because
The width of the bucket is determined according to the detected peak width in advance so that individual peak areas included in the absolute value differential spectrum are not divided into a plurality of buckets. A method for processing NMR data. An NMR spectrum data defined by the mathematical formula (1) indicating the NMR characteristics of the sample and an absolute value differential spectral data defined by the mathematical formula (2) by performing absolute value differentiation on the NMR spectral data are obtained in a computer. A program part to be executed, and
A step of reducing the absolute value differential spectrum data to histogram data by performing bucket integration using a bucket set consisting of a large number of buckets having unequal widths on the absolute value differential spectrum data. A program part to be executed by
A program part for causing a computer to execute the step of storing or outputting the histogram data;
With
When δ is a chemical shift, j is an imaginary number symbol, R (δ) is a real part of the NMR spectrum, and I (δ) is an imaginary part of the NMR spectrum, Equations (1) and (2) are respectively
NMR spectrum = R (δ) + j · I (δ) (1)

Because
The width of the bucket is determined according to the detected peak width in advance so that individual peak areas included in the absolute value differential spectrum are not divided into a plurality of buckets. A machine-readable computer program for the processing of NMR data.

Images