JP2006317198A - Data processor for analyzer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To form a calibration curve of higher precision without relying on the experience or skill of an analyst. <P>SOLUTION: A standard sample for calibration containing a plurality of sample components known in molecular weight is subjected to GPC analysis to collect data and the peaks appearing in a formed chromatogram are detected. If the holding times of the peak tops of the respective peaks are decided, the relation between the molecular weights and the holding times is held to a calibration point table (S11-S15). (M) calibration points are selected from this table (S16) while the calibration points are adapted to a plurality of prepared approximation formulae to calculate a coefficient and the respective approximation formulae are determined to calculate approximation errors (S17 and S18). After the respective approximation formulae are determined by variously changing the selective combinations of the calibration points, a list wherein the approximation formulae are arranged in the smaller order of the approximation errors is displayed as a calibration curve candidate (S21 and S22). The analyst selects the calibration curve from the list. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a data processing apparatus that processes data obtained by an analyzer such as a gel permeation chromatograph, and more specifically, a data processing apparatus having a function of supporting the creation of a calibration curve used in the data processing. About.

  Gel Permeation Chromatography (GPC), also called Size Exclusion Chromatography, is a kind of liquid chromatograph, which has pores (pores) in the packing material (gel) that is a stationary phase. The sample component molecules dissolved in the mobile phase are separated based on the difference in molecular size, and the molecular weight, molecular weight distribution, and the like are measured (see, for example, Patent Document 1).

  In such GPC analysis, in order to calculate the molecular weight of an unknown sample to be analyzed, a calibration curve indicating the relationship between the time when a peak appears on the chromatogram (retention time or elution time) and the molecular weight is used. . A conventional general calibration curve creation procedure is as follows.

  First, a calibration standard sample including one or more sample components having a known molecular weight is analyzed to obtain a chromatogram. The time (retention time) at which the peak top of a known sample component appears on the chromatogram is read, and a table in which the molecular weight is associated with the retention time is created from the result. In general, several different molecular weights are used. Using the molecular weight-retention time listed in the above table as a calibration point, an approximate expression is obtained from several different calibration points to create a calibration curve.

  In doing this, the analyst calibrates by applying the relationship shown in the above table to a few approximate equations that are considered appropriate in light of his experience, and adopting the one with the best correlation (small approximation error). Find a curve. In some cases, the calibration points may not be appropriate (for example, noise may be mixed when obtaining a certain calibration point). Therefore, the calibration points are appropriately excluded without using all the calibration points obtained by actual measurement. For example, a calibration curve with a smaller approximation error may be obtained by adjusting the combination. In any case, such work is done manually by the analyst, and all judgments such as what approximation formula to use and what calibration points to exclude are all determined by the analyst's experience and It is left to intuition.

  In general, in a normal liquid chromatograph or gas chromatograph, the calibration curve (calibration curve) is a straight line or can be approximated by a broken line or a low-order function even if it is not a straight line. Therefore, the variation of the calibration curve for each analyst is not so large, and in practice, it often does not cause a problem. However, in the GPC analysis, the shape of the calibration curve tends to be complicated, and it is often necessary to express it by a combination of a higher-order polynomial or a plurality of functions. For this reason, when the calibration curve is created manually as described above, even if the same calibration point is used, the variation of the calibration curve by the analyst increases. Therefore, the molecular weight accuracy of the target sample may vary. Even if there is a more appropriate approximate expression or combination of calibration points than the calibration curve adopted, it may be overlooked and the optimal analysis may not be possible.

JP-A-8-271494

  The present invention has been made to solve such a problem, and the object of the present invention is to improve analysis accuracy by creating a highly accurate calibration curve without depending on the experience and skill of the analyst. Another object of the present invention is to provide a data processing apparatus for an analyzer that can achieve good reproducibility.

The present invention, which has been made to solve the above-mentioned problems, is an analysis for deriving a predetermined index value relating to the target sample from data obtained by analyzing the target sample with an analyzer using a calibration curve prepared in advance. In the data processing device for equipment,
a) calibration point holding means for obtaining and storing a plurality of calibration points as a basis for creating a calibration curve based on data obtained by analyzing a calibration standard sample by the analyzer;
b) calibration point selection means for sequentially selecting and outputting all or a part of the plurality of calibration points stored in the calibration point holding means while changing the combination thereof;
c) for each of a plurality of approximate expression models prepared in advance, approximate expression determination means for determining an approximate expression by calculating a coefficient included in the approximate expression model using the calibration points output by the calibration point selection means; ,
d) error information calculating means for calculating an approximate error or a value corresponding to each approximate expression determined by the approximate expression determining means;
e) Information presentation means for presenting the approximate expression determined by the approximate expression determination means and the approximate error calculated by the error information calculation means or a value corresponding thereto as the calibration curve candidates;
It is characterized by having.

  Here, “a plurality of approximate expression models prepared in advance” can be, for example, polynomials having different orders from a lower order expression to a higher order expression, or a combination of a plurality of different functions.

  Further, the “approximation error or a value corresponding to it” only needs to be a value representing a deviation between the approximate expression and the calibration point from which the determination is made, and may include a correlation coefficient in addition to the approximation error. .

  In the data processing apparatus for an analyzer according to the present invention, when a plurality of calibration points are given by the calibration point selecting means, the approximate expression determining means, for each of the various approximate expression models prepared in advance as described above, An approximate expression is determined by calculating a coefficient included in the approximate expression model using the calibration points. Therefore, for example, when five types of approximate equations models are prepared, five approximate equations are determined for a certain combination of calibration points. Further, since the calibration point selection means sequentially selects and outputs all or a part of the plurality of calibration points while changing the combination, for example, five approximate expressions are respectively provided for each of the calibration points having different combinations as described above. To decide.

  Further, the error information calculation means calculates, for example, an approximation error for each approximate expression determined as described above. The information presenting means presents the approximate expression determined as described above and the corresponding approximate error, for example, in the form of a list on the monitor screen. The analyst looks at the information presented in this way and selects an approximate expression that is considered optimal as a calibration curve. Usually, an approximation formula having the smallest approximation error (high correlation coefficient) is selected as the calibration curve. Therefore, in the data processing apparatus for an analyzer according to the present invention, preferably, the information presenting means is configured so that the approximation error is in descending order. It is preferable that the approximate expression is presented. This allows the analyst to select the calibration curve more easily.

  The present invention is particularly useful when the calibration curve is not as simple as, for example, a straight line but is a relatively high-order polynomial or a combination of functions. An example of such an analyzer is a gel permeation chromatograph, in which case the index value is the molecular weight and the calibration curve shows the relationship between retention time and molecular weight.

  As described above, according to the data processing apparatus for an analyzer according to the present invention, the calibration curve that involves the selection of the order of the approximate expression, the selection of the combination of the calibration points, etc., which has been dependent on the experience and judgment of the analyst so far. Is automatically and exhaustively performed, eliminating variations in calibration curves due to differences in analyst experience and skills. As a result, a highly accurate analysis is always possible. In addition, in the past, an optimum calibration curve could not always be created. However, according to the data processing apparatus for an analyzer according to the present invention, calibration having the highest approximation to the analysis result of the calibration standard sample is possible. Since a curve can be obtained, it is possible to perform highly accurate analysis at that point.

  As an embodiment of the present invention, an example in which the data processing apparatus according to the present invention is applied to a gel permeation chromatograph (GPC) analyzer will be described below.

  FIG. 1 is a schematic configuration diagram of a GPC analyzer according to this embodiment. The GPC analyzer includes a mobile phase container 1, a liquid feed pump 2, an injector 3, a column 4, a differential refractive index detector 5, a data processing unit 6, and the like. The column 4 is packed with a GPC filler according to the type of target sample (for example, protein, polymer, etc.). The entity of the data processing unit 6 corresponding to the data processing apparatus according to the present invention is a personal computer. Various control and GPC analysis necessary for GPC analysis can be performed by installing and executing dedicated control / processing software in the personal computer. Data processing operations are achieved.

  The procedure for analyzing a target sample containing a component whose molecular weight is unknown in this GPC analyzer will be described with reference to FIG. First, GPC analysis is performed on the target sample to collect data (step S1). That is, the liquid feed pump 2 sucks the mobile phase from the mobile phase container 1 and sends it to the column 4 through the injector 3 at a substantially constant flow rate. The target sample is injected into the mobile phase by the injector 3 at a predetermined timing, and measurement with the differential refractive index detector 5 is started. The injected sample rides on the mobile phase and is introduced into the column 4 and remains in the column 4 for a time corresponding to the molecular weight of each sample component due to the influence of the pores of the filler in the column 4. A sample component having a large molecular size (generally a large molecular weight) is sequentially eluted from the column 4 and reaches the differential refractive index detector 5 to be detected.

  After the measurement is started, the value detected by the differential refractive index detector 5 is sent to the data processing unit 6 every moment, and the data processing unit 6 creates a chromatogram with the passage of time (step S2). FIG. 4 is an example of a chromatogram. When a sample component is eluted, a peak corresponding to the sample component appears. If such a chromatogram is acquired, the data processor 6 detects a peak appearing on the chromatogram by performing waveform processing on the chromatogram (step S3). In the example of FIG. 4, three peaks A, B, and C are detected.

  Then, the elution time (retention time) of the peak top of each peak detected is obtained (step S4). In the example of FIG. 4, the holding times of ta, tb, and tc are obtained for the three peaks of A, B, and C. Thereafter, referring to a calibration curve prepared in advance, the molecular weight is calculated from each of the retention times (step S5). Thereby, the molecular weights of the three sample components contained in the target sample are obtained.

  As described above, the data processing unit 6 holds the calibration curve in advance in order to calculate the molecular weight from the peak retention time. However, in the GPC analyzer of this embodiment, in order to easily create a highly accurate calibration curve. The data processing unit 6 includes a calibration curve creation support processing unit 7 as one of the functions. That is, the calibration curve creation support processing unit 7 is also a function realized by using predetermined hardware and hardware resources such as a CPU, ROM, and RAM of a personal computer.

  A procedure for creating a calibration curve in this GPC analyzer will be described with reference to FIG. To create a calibration curve, a calibration standard sample containing sample components of known molecular weight is used. Usually, one standard sample includes a plurality of sample components having known molecular weights, but when one standard sample is insufficient, a plurality of types of standard samples are prepared. The molecular weight of the sample component may be appropriately selected according to the molecular weight range of the target sample to be analyzed. Further, the number of sample components having different molecular weights can be determined as appropriate, but if the amount is too large, the amount of processing becomes enormous and it takes too much time. If the amount is too small, the accuracy of the calibration curve may be lowered. Therefore, the range is usually from about several to about several. Here, it is assumed that the number of sample components is 10, that is, n = 10.

  First, GPC analysis is performed on the calibration standard sample to collect data (step S11). Then, the data processing unit 6 creates a chromatogram based on the collected data (step S12), and detects a peak appearing on the chromatogram by performing waveform processing on the chromatogram (step S13). Then, the retention time of the peak top of each detected peak is obtained (step S14). The process up to this point is the same as in the analysis of the target sample described above.

  Assuming that the molecular weight of each of the ten sample components is M1,..., M10 and the peak top retention time is t1,..., T10, a set of molecular weight and retention time (t1, M1),. , M10) are calibration points, and these are collected and stored as a calibration point table (step S15). Ten calibration points held in the calibration point table serve as original data when a calibration curve is created. As shown in FIG. 5, each calibration point represents one point on a graph in which the horizontal axis represents the retention time and the vertical axis represents the molecular weight (normal logarithmic scale).

  Next, M (1 ≦ M ≦ N) calibration points are selected from the N calibration points held in the calibration point table. First, m = n, that is, all n calibration points are selected. Select (step S16). The calibration curve creation support processing unit 7 holds a plurality of predetermined approximate models. Specifically, for example, it has an approximate expression model such as a straight line, a broken line, a cubic expression, a cubic + hyperbola, a fifth order expression, a fifth order + hyperbola, a seventh order expression, and a seventh order + hyperbola. Each approximate expression model includes an undetermined coefficient, and when this coefficient is determined, the approximate expression is determined. Therefore, the calibration curve creation support processing unit 7 selects the approximate expression model as described above, and calculates the coefficient of each approximate expression model based on the plurality of calibration points received from the calibration point table (step S17). For example, first, an approximate expression of a straight line is calculated using n calibration points. When the coefficients are obtained and the approximate expression is determined, for example, an approximate error is calculated as a value indicating the correlation or closeness between the approximate expression and the calibration point (step S18).

  Next, it is determined whether approximate equations have been obtained for all approximate equation models (step S19). If there is an uncertain approximate equation model, the process returns to step S17. Therefore, by repeating the processes of steps S17 to S19, all approximate equation models prepared in advance as described above (straight line, broken line, cubic, cubic + hyperbola, fifth degree, fifth degree + hyperbola, seventh degree) , 7th order + hyperbola, etc.), approximate expressions when n calibration points are used are respectively obtained. FIG. 5 shows an example of a state in which different approximate expressions L1, L2, and L3 are created for the determined calibration points.

  After the approximation formulas for all approximation models are determined using n calibration points, it is then determined whether or not all selected combinations have been executed for n calibration points (step S20). If there is an unexecuted item, the process returns to step S16. That is, when the process of setting m = n is completed as described above, then m = n−1, that is, nine calibration points in this example are selected. In other words, one calibration point is treated as invalid data among ten calibration points in total. However, since 10 selected combinations can be considered depending on which one of the 10 calibration points is invalidated, each of them is processed in turn.

  For example, the first calibration point (t1, M1) is invalidated, and the process of step S18 is executed using the remaining nine calibration points (t2, M2),..., (T10, M10). If the approximate equations are determined for all approximate models using the calibration points, the second calibration point (t2, M2) is invalidated, and the remaining nine calibration points (t1, M1), ( The process of step S18 is repeated using t3, M3,..., (t10, M10). Further, when the process for m = n−1 is completed, the same process is repeated with m = n−2.

  However, in order to calculate the coefficient of the approximate expression model, the minimum number of calibration points required for each approximate expression model is determined. For example, in the 7th order equation, there are 8 coefficients, and 9 calibration points are required to calculate these 8 coefficients. Therefore, when m reaches 8, a 7th order approximate model cannot be used. Since the number of calibration points required increases as the order increases, the number of approximate models that can be used decreases as m is decreased from n in order. If the approximate equation model is a straight line, it is possible to draw a straight line if there are two calibration points. However, in reality, it is not meaningful to obtain an approximate equation by invalidating a large number of calibration points. . Therefore, it is practically necessary to limit the number of invalid calibration points to 2 or 3 (the lower limit p of m is 7 or 8) in advance.

  Then, if approximate equations are determined for all selected combinations of calibration points within a predetermined range and the approximate error of each approximate equation is obtained, the multiple approximate equations are used as calibration curve candidates in ascending order of approximate error. The arranged list is created (step S21) and displayed on the monitor screen (step S22). FIG. 6 is a diagram illustrating an example of a list of calibration curve candidates. When the analyst selects and instructs an appropriate approximate expression with reference to the displayed list (step S23), the selected approximate expression is stored in the storage unit as a calibration curve, and the target sample as described above is stored. Used for analysis. Usually, the highest order in the list, that is, the approximation formula with the smallest approximation error may be selected as the calibration curve.

  As described above, the GPC analyzer according to the present embodiment uses the function of the calibration curve creation support processing unit 7 provided in the data processing unit 6 to perform accuracy without depending on the experience or judgment of the analyst. A high calibration curve can be created automatically.

  The above embodiment is an example in which the data processing apparatus according to the present invention is applied to a GPC analyzer. However, a calibration curve is prepared in advance by measuring a standard sample, and an unknown sample is obtained using this calibration curve. The present invention can be generally applied to an analyzer that derives a target value related to the unknown sample from the measurement result. However, when it is determined that the calibration curve is always approximated by a straight line or the like, the usefulness is low, and the more complex the calibration curve is, the higher the usefulness is.

  In the above embodiment, calibration curve candidates are presented to the analyst, and finally, an approximation formula that the analyst thinks is most appropriate is selected as the calibration curve. It is also possible to automatically select a calibration curve that minimizes and store it in the internal memory.

  The above-described embodiment is an example of the present invention, and modifications, corrections, additions, and the like are appropriately included in the present invention within the scope of the present invention.

1 is a schematic configuration diagram of a GPC analyzer according to an embodiment of the present invention. The process flowchart at the time of the target sample measurement in this GPC analyzer. The processing flowchart at the time of the calibration curve preparation in this GPC analyzer. The figure which shows an example of the chromatogram acquired in this GPC analyzer. Explanatory drawing of a calibration curve creation assistance processing operation. The figure which shows an example of the calibration curve candidate list displayed as a result of a calibration curve creation assistance processing operation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Mobile phase container 2 ... Liquid feed pump 3 ... Injector 4 ... Column 5 ... Differential refractive index detector 6 ... Data processing part 7 ... Calibration curve preparation assistance processing part

Claims (3)

In a data processing apparatus for an analyzer that derives a predetermined index value related to the target sample from data obtained by analyzing the target sample with an analyzer using a calibration curve prepared in advance,
a) calibration point holding means for obtaining and storing a plurality of calibration points as a basis for creating a calibration curve based on data obtained by analyzing a calibration standard sample by the analyzer;
b) calibration point selection means for sequentially selecting and outputting all or a part of the plurality of calibration points stored in the calibration point holding means while changing the combination thereof;
c) for each of a plurality of approximate expression models prepared in advance, an approximate expression determining means for determining an approximate expression by calculating a coefficient included in the approximate expression model using the calibration points output by the calibration point selecting means; ,
d) error information calculating means for calculating an approximate error or a value corresponding to each approximate expression determined by the approximate expression determining means;
e) Information presentation means for presenting the approximate expression determined by the approximate expression determination means and the approximate error calculated by the error information calculation means or a value corresponding thereto as the calibration curve candidates;
A data processing apparatus for an analyzer, comprising:   The data processing apparatus for an analyzer according to claim 1, wherein the information presentation unit presents the approximation formulas in ascending order of approximation error. The analysis according to claim 1 or 2, wherein the analyzer is a gel permeation chromatograph, the index value is a molecular weight, and the calibration curve shows a relationship between a retention time and a molecular weight. Data processing device for equipment.

Images