Data processing method and data processing system

Abstract

A data processing method and a data processing system for separating peaks of a plurality of components overlapping on a three-dimensional chromatogram including a chromatogram and a spectrum of a sample obtained by chromatographic analysis using actual data of the three-dimensional chromatogram, the data processing method including: an adjustment target peak obtaining step of obtaining a plurality of adjustment target peaks by applying a peak model function prepared in advance for approximating a waveform of the chromatogram to the chromatogram; and an adjustment target peak adjusting step of setting the plurality of adjustment target peaks obtained in the adjustment target peak obtaining step as initial values before adjustment, and repeatedly performing adjustment of the adjustment target peaks until pseudo data of the three-dimensional chromatogram of the sample obtained by synthesizing the adjustment target peaks after adjustment is similar to the actual data.

Description

Technical Field

[0001] This invention relates to a data processing method and system using three-dimensional chromatographic data. Background Technology

[0002] In liquid chromatographs (LC) using multi-channel detectors such as photodiode array (PDA) detectors, the absorbance spectra of samples dissolved from the analytical column are continuously acquired, thereby obtaining three-dimensional chromatographic data with three dimensions: time, wavelength, and signal intensity (absorbance).

[0003] When using liquid chromatography to quantify a target component in a sample, the wavelength with the highest absorbance of the target component is generally used to generate the chromatogram, and the peak area of ​​the target component is determined on the chromatogram for quantification. However, sometimes the sample contains impurities other than the target component, and the peaks of these impurities may overlap with the peak of the target component on the chromatogram. In such cases, with multiple peaks overlapping, it is impossible to determine the peak area of ​​either the target component or the impurities, thus making it impossible to obtain a quantitative result. Therefore, it is necessary to separate the multiple components with overlapping peaks on the chromatogram.

[0004] As an algorithm for separating multiple overlapping peaks (referred to as a peak separation algorithm), the following algorithm is known: by applying peak model functions such as the Exponential Modified Gaussian (EMG) function to the actual chromatographic waveform, the individual chromatograms and spectra of multiple components with overlapping peaks are inferred (see Patent Document 1).

[0005] [Existing Technical Documents]

[0006] [Patent Literature]

[0007] [Patent Document 1] International Publication No. 2016 / 035167 Summary of the Invention

[0008] [The problem the invention aims to solve]

[0009] In peak separation algorithms using peak model functions (hereinafter referred to as model-based algorithms), there exist algorithms that utilize modified EMG functions that can also represent the tailing or leading characteristics of actual peak waveforms. These modified EMG functions can accurately reproduce the actual chromatographic waveform shape, resulting in highly accurate quantification of each component. However, it is known that when multiple component peaks with a very large relative concentration ratio of 100:0.05 overlap on the chromatogram, applying a model-based algorithm to separate these peaks can lead to a situation where the peak area values ​​of low-concentration components differ significantly from the actual peak area values.

[0010] The present invention was made in view of the aforementioned problems, and its object is to enable high-precision quantification of components with relatively low concentrations even when the peaks of multiple components with extremely large relative concentration ratios in a sample overlap on the chromatogram.

[0011] [Technical means to solve the problem]

[0012] In model-based algorithms, a pre-prepared peak model function is applied to the actual chromatographic waveform while its parameters (e.g., height, broadening) are adjusted. This allows for the inference of the shape or size of multiple overlapping peaks on the chromatogram. Therefore, the inferred peak shapes of each component are constrained by the peak model function. Consequently, a small deviation occurs between the area values ​​of each peak obtained through the peak model function and the actual area values ​​of each peak. This is considered to be the reason for the decreased accuracy in quantifying relatively low-concentration components when separating and quantifying peaks of multiple components with extremely large relative concentration ratios, such as 100:0.05.

[0013] Furthermore, there are also peak separation algorithms that do not use peak model functions (hereinafter referred to as model-free algorithms). Representative model-free algorithms include those utilizing matrix factorization (NMF). A model-free algorithm using matrix factorization works as follows: it mathematically separates the original 3D chromatogram into a specified number of peaks and fine-tunes each peak to synthesize a pseudo-3D chromatogram that approximates the original 3D chromatogram. Because the peak shapes are not constrained by peak model functions, this model-free algorithm offers high freedom in peak separation, allowing the synthesized pseudo-3D chromatogram to closely approximate the original 3D chromatogram. However, while not constrained by peak model functions, it lacks information about the peak shapes, meaning the shapes of the separated peaks may differ significantly from the actual peak shapes. Therefore, the reproducibility of peak separation results based on model-based algorithms is often worse than that based on model-based algorithms.

[0014] The inventors of this application, focusing on the degree of freedom in peak separation using model-free algorithms, conceived of supplementing model-based algorithms with model-free algorithms. Furthermore, the inventors of this application have realized that when peak separation results obtained using model-based algorithms are used as initial data and fine-tuned based on model-free algorithms, good separation accuracy can be achieved even in cases of peak overlap among multiple components with extremely large relative concentration ratios, such as 100:0.05. This invention is based on this realization.

[0015] The data processing method of the present invention uses actual data of three-dimensional chromatography (including chromatogram and spectrum) obtained by chromatographic analysis of a sample to separate the peaks of multiple components overlapping on the chromatogram. The data processing method includes: a target peak acquisition step, in which a multiple target peaks are obtained by applying a pre-prepared peak model function to the chromatogram to approximate the waveform of the chromatogram; a speculative data generation step, in which the multiple target peaks obtained in the target peak acquisition step are synthesized to generate speculative chromatographic and spectral data respectively related to the multiple components; and a target peak adjustment step, in which the speculative data generated in the speculative data generation step is set as the initial value before adjustment, and the target peaks are repeatedly adjusted until the pseudo data of the three-dimensional chromatogram of the sample obtained by synthesizing the adjusted target peaks is similar to the actual data.

[0016] The data processing system of the present invention includes: a storage unit storing actual data of a three-dimensional chromatogram of a sample obtained by chromatographic analysis, comprising chromatogram and spectrum, and a plurality of pre-prepared peak model functions; and a data processing unit configured to process the actual data of the three-dimensional chromatogram using the plurality of peak model functions, thereby performing a process to separate the peaks of a plurality of components overlapping on the chromatogram, the data processing unit being configured to perform: an adjustment target peak acquisition step, in which a plurality of adjustment target peaks are obtained by applying the plurality of peak model functions stored in the storage unit to the chromatogram; and an adjustment target peak adjustment step, in which the adjustment target peaks obtained in the adjustment target peak acquisition step are set as initial values ​​before adjustment, and the adjustment target peaks are repeatedly adjusted until pseudo data of the three-dimensional chromatogram of the sample obtained by synthesizing the adjusted adjustment target peaks is similar to the actual data.

[0017] In the data processing method and system of this invention, firstly, a model-based algorithm is used to obtain the adjustment target peaks for each of the multiple components inferred to overlap on the chromatogram. Because these adjustment target peaks are constrained by the peak model function, although they are not completely identical to the actual peaks of each component, they can be said to be approximating the actual peaks to some extent. These adjustment target peaks with a certain degree of approximation are obtained and set as the initial data before adjustment. Then, an adjustment based on a model-free algorithm is performed, i.e., an adjustment that is free from the constraints of the peak model function. When attempting to separate the peaks of multiple components from the actual data of a three-dimensional chromatogram using a model-free algorithm initially, it is difficult to obtain accurate inference results because peak separation is performed without any information about the peak shapes of each component. On the other hand, if the adjustment target peaks with a certain degree of approximation are set as the initial data for implementing the model-free algorithm, the adjustment target peaks are fine-tuned to be closer to the actual data, thereby improving the approximation of the inferred data relative to the actual data.

[0018] [The effects of the invention]

[0019] According to the data processing method and system of the present invention, after obtaining the adjustment target peaks of multiple components that overlap on the chromatogram using a model-based algorithm, a model-free algorithm is used to adjust the adjustment target peaks obtained by the model-based algorithm. Therefore, the adjustment target peaks are fine-tuned to be closer to the actual data, thereby improving the approximation of the inferred data relative to the actual data. As a result, even when the peaks of multiple components with extremely large relative concentration ratios in the sample overlap on the chromatogram, the quantification of components with relatively low concentrations can be performed with high precision. Attached Figure Description

[0020] Figure 1This is a block diagram that schematically represents one embodiment of a data processing system.

[0021] Figure 2 This is a flowchart that schematically illustrates a data processing method implemented using the data processing system described in the embodiment.

[0022] Figure 3 This is a flowchart illustrating a specific example of the data processing method.

[0023] Figure 4 (A) and Figure 4 (B) is a graph representing an example of the application of the peak model function in a peak model-based algorithm, where (A) represents the chromatogram of actual data at a certain wavelength, and (B) represents the state of applying the peak model function to the chromatogram.

[0024] Figure 5 (A) Figure 5 (C) is a diagram representing the peak separation process in the data processing method, (A) represents the chromatogram of the actual data at a certain wavelength, (B) represents the inferred data of the chromatogram of each component separated by the peak model using a type algorithm, and (C) represents the inferred data of the chromatogram of each component adjusted by matrix decomposition. Detailed Implementation

[0025] Hereinafter, embodiments of the chromatographic data processing method and data processing system of the present invention will be described with reference to the accompanying drawings.

[0026] Figure 1 The image shows an embodiment of a data processing system.

[0027] The data processing system 1 includes an actual data storage unit 2, a peak model storage unit 4, and a data processing unit 6. Analytical data acquired using the analytical apparatus 100 is input into the data processing system 1. The analytical apparatus 100 is configured to perform liquid chromatography analysis on a sample and acquire absorbance spectra at regular intervals. That is, the analytical data input from the analytical apparatus 100 to the data processing system 1 is three-dimensional chromatographic data, including both chromatograms and spectra. Hereinafter, the three-dimensional chromatographic data input from the analytical apparatus 100 to the data processing system 1 will be referred to as "actual data".

[0028] The actual data storage unit 2 is a storage area for storing the actual data of the three-dimensional chromatography acquired by the self-analyzing device 100. The actual data storage unit 2 can be implemented using a non-volatile flash memory or a hard disk drive, etc.

[0029] Peak model storage unit 4 stores multiple pre-prepared peak model functions. Examples of peak model functions include models based on a modified EMG function that combines Gaussian and exponential functions, configured to reproduce peak waveforms with tailing or leading-out characteristics, similar to actual peak waveforms observed in chromatography. Like the actual data storage unit 2, peak model storage unit 4 can be implemented using a non-volatile flash memory, a hard disk drive, or a database located on a network.

[0030] The data processing unit 6 processes the actual data of the three-dimensional chromatography stored in the actual data storage unit 2. The processing of the actual data using the data processing unit 6 includes quantitative processing, which quantifies the concentration of components contained in the sample based on the peak area values ​​on the chromatogram of the actual data, and peak separation processing, which separates peaks of multiple components when they overlap on the chromatogram of the actual data. The data processing unit 6 is a functional unit implemented by executing programs in a computer circuit including a central processing unit (CPU).

[0031] like Figure 2 As shown, the peak separation process performed by the data processing unit 6 includes: a first step (step 101), which uses an algorithm employing a peak model function (model-using algorithm) to generate inferred chromatographic and spectral data for multiple components with overlapping peaks; and a second step (step 102), which uses an algorithm without employing a peak model function (model-free algorithm) to adjust the inferred data. The model-using algorithm used in the first step can be a known algorithm, such as the algorithm disclosed in Patent Document 1 (International Publication No. 2016 / 035167). The peak-model-free algorithm used in the second step can also be a known algorithm, such as an algorithm using matrix factorization such as NMF (non-negative matrix factorization).

[0032] The more specific peak separation process is shown below. Figure 3 .

[0033] When peak separation processing begins, the data processing unit 6 first adjusts the parameters (height, broadening, etc.) of the peak model function while applying the peak model function to the chromatogram of the actual data, and obtains the applied peak model function as the peak to be adjusted (step 201). For example, the waveform of the chromatogram at a certain wavelength in the actual data is as follows: Figure 4 In the case of the waveform shown in (A), as Figure 4 As shown in (B), the chromatographic waveform is approximated by three peak model functions with adjusted parameters. The peak model functions applied to approximate the chromatographic waveform are the peaks to be adjusted. Step 201 is the first step in using a model-based algorithm.

[0034] The data processing unit 6 synthesizes the adjustment target peaks obtained in step 201 to generate pseudo data for the three-dimensional chromatogram of the sample (step 202), and calculates the similarity between the pseudo data and the actual data (step 203). "Similarity" simply means how similar the pseudo data is to the actual data. Therefore, there is no particular limitation on the method of calculating similarity; for example, the sum of the squares of the differences between the pseudo data values ​​and the actual data values ​​at each point of the three-dimensional chromatogram can be used as the similarity.

[0035] The data processing unit 6 uses matrix factorization to adjust the parameters of the target peak to achieve a good similarity, meaning the pseudo-data is closer to the actual data (step 205). Then, the data processing unit 6 synthesizes the adjusted target peaks to generate pseudo-data for three-dimensional chromatography (step 202), and evaluates the similarity of the generated pseudo-data to the actual data (steps 203 and 204). Steps 202 to 205 are repeated until the similarity of the pseudo-data to the actual data meets a predetermined condition, at which point the peak separation process ends (step 206: Yes). Prescribed conditions may include: a similarity lower than (or higher than) a pre-set threshold, or the similarity of the pseudo-data obtained by synthesizing the adjusted target peaks to the actual data converging to a certain value.

[0036] Steps 202 to 205 constitute the second step using a model-free algorithm. In this second step, the peaks separated in the first step are adjusted without being constrained by the shape of the peak model function. As a result, adjustments are made to the parts in the first step that were not entirely approximated to the actual data due to the constraints of the peak model function, and the size and shape of the peaks of the separated components are closer to reality.

[0037] Figure 5 (A) Figure 5 (C) represents an example of the peak separation state in each step of the peak separation process.

[0038] Figure 5 (A) is a portion of the chromatographic waveform at a certain wavelength of the actual data before peak separation processing is performed. When the first step of the model-based algorithm is applied to the actual data having said chromatogram, as... Figure 5 As shown in (B), by applying two peak model functions, two adjustment target peaks, P4 and P5, can be obtained. Furthermore, when the obtained adjustment target peaks are set as the initial data before adjustment and the second step of the algorithm using the model without using a specific type is implemented, as shown in (B),... Figure 5 As shown in (C), the shape and size of the target peaks P4 and P5 are adjusted.

[0039] As previously mentioned, when using only the model-based algorithm, i.e., performing only the first step to separate the peaks of two components with a very large relative concentration ratio of 100:0.05, the peak area of ​​the component with the relatively lower concentration can sometimes reach more than twice the actual peak area. In contrast, when the inventors set the peak separation result obtained in the first step as the initial data before adjustment and performed the second step using the model-free algorithm, they adjusted the shape and size of the peaks of the two components respectively, and confirmed that the peak area of ​​the component with the relatively lower concentration was close to the actual peak area.

[0040] The embodiments described above are merely illustrative of implementation methods for the data processing method and data processing system of the present invention. Implementation methods for the data processing method and data processing system of the present invention are as follows.

[0041] In one embodiment of the data processing method of the present invention, actual data of a three-dimensional chromatogram containing chromatograms and spectra obtained by chromatographic analysis of a sample is used to separate the peaks of multiple components overlapping on the chromatogram. The data processing method includes: an adjustment target peak acquisition step, in which a multiple adjustment target peaks are obtained by applying a peak model function prepared in advance to approximate the waveform of the chromatogram to the chromatogram; a speculative data generation step, in which the multiple adjustment target peaks obtained in the adjustment target peak acquisition step are synthesized to generate speculative chromatographic and spectral data respectively related to the multiple components; and an adjustment target peak adjustment step, in which the speculative data generated in the speculative data generation step is set as the initial value before adjustment, and the adjustment target peaks are repeatedly adjusted until the pseudo data of the three-dimensional chromatogram of the sample obtained by synthesizing the adjusted adjustment target peaks is similar to the actual data.

[0042] In a first aspect of one embodiment of the data processing method, a function combining a Gaussian function and an exponential function is used as the peak model function. According to this approach, a peak model function that takes into account the tailing or leading-out of the actual peak can be used, making the inferred shape of the separated peaks approximate the actual shape.

[0043] In a second aspect of one embodiment of the data processing method, matrix decomposition is used to perform the adjustment in the peak adjustment step. This second aspect can be combined with the first aspect.

[0044] As the matrix decomposition, nonnegative matrix factorization can be used.

[0045] In a third aspect of an embodiment of the data processing method, in the adjustment target peak adjustment step, the similarity between the pseudo data obtained by synthesizing the adjusted adjustment target peak and the actual data is calculated. When the similarity meets a preset benchmark, or when the similarity converges to a certain value, it is determined that the pseudo data is similar to the actual data, and the adjustment ends. The third aspect may be combined with the first aspect and / or the second aspect.

[0046] In one embodiment of the data processing system of the present invention, it includes: a storage unit storing actual data of a three-dimensional chromatogram of a sample obtained by chromatographic analysis, comprising chromatogram and spectrum, and a plurality of pre-prepared peak model functions; and a data processing unit configured to process the actual data of the three-dimensional chromatogram using the plurality of peak model functions, thereby performing a process to separate the peaks of a plurality of components overlapping on the chromatogram, the data processing unit being configured to perform: an adjustment target peak acquisition step, obtaining a plurality of adjustment target peaks by applying the plurality of peak model functions stored in the storage unit to the chromatogram; and an adjustment target peak adjustment step, setting the adjustment target peaks obtained in the adjustment target peak acquisition step as initial values ​​before adjustment, and repeatedly adjusting the adjustment target peaks until pseudo data of the three-dimensional chromatogram of the sample obtained by synthesizing the adjusted adjustment target peaks is similar to the actual data.

[0047] In a first aspect of one embodiment of the data processing system, the data processing unit is configured to use a function combining a Gaussian function and an exponential function as the peak model function. According to this configuration, a peak model function that takes into account the tailing or leading-out of the actual peak can be used, making the inferred shape of the separated peaks approximate the actual shape.

[0048] In a second aspect of one embodiment of the data processing system, the data processing unit uses matrix decomposition to perform the adjustment in the adjustment target peak adjustment step. This second aspect can be combined with the first aspect.

[0049] As the matrix decomposition, nonnegative matrix factorization can be used.

[0050] In a third aspect of an embodiment of the data processing system, the data processing unit is configured to: in the adjustment target peak adjustment step, calculate the similarity between the pseudo data obtained by synthesizing the adjusted adjustment target peak and the actual data; and terminate the adjustment when the similarity meets a preset benchmark or when the similarity converges to a certain value. The third aspect may be combined with the first aspect and / or the second aspect.

[0051] [Explanation of Symbols]

[0052] 1: Data Processing System

[0053] 2: Actual Data Storage Department

[0054] 4: Peak Model Storage Section

[0055] 6: Data Processing Department

[0056] 100: Analytical apparatus.

Claims

1. A data processing method, using a chromatogram obtained by chromatographic analysis of a sample, to separate peaks of multiple components overlapping on the chromatogram, characterized in that, The data processing method includes: The step of preparing a peak model function to approximate the waveform of the chromatogram; The adjustment of the target peak acquisition step involves adjusting the parameters of the peak model function and applying them to the chromatography to obtain multiple target peaks; and The adjustment of the target peaks involves setting multiple target peaks as initial values, and then using matrix decomposition to adjust the matrix factor parameters of the shape and size of the multiple target peaks to obtain a pseudo-chromatogram. The adjustment of the matrix factor parameters of the target peaks is repeated until the similarity between the pseudo-chromatogram and the chromatogram is lower than a predetermined threshold or converges to a certain value.

2. The data processing method according to claim 1, wherein the peak model function is a function composed of a Gaussian function and an exponential function.

3. The data processing method according to claim 1, wherein the matrix decomposition is a non-negative matrix factorization.

4. A data processing system, characterized in that, include: The storage section stores chromatographic data and peak model functions obtained from chromatographic analysis of the sample. as well as The data processing unit is configured to process the chromatogram using the peak model function, thereby separating the peaks of multiple components that overlap on the chromatogram. The data processing unit is configured to perform: The adjustment target peak acquisition step involves obtaining multiple adjustment target peaks by adjusting the peak model function stored in the storage unit and applying it to the chromatography; and The adjustment of the target peaks involves setting multiple target peaks as initial values, and then using matrix decomposition to adjust the matrix factor parameters of the shape and size of the multiple target peaks to obtain a pseudo-chromatogram. The adjustment of the matrix factor parameters of the target peaks is repeated until the similarity between the pseudo-chromatogram and the chromatogram is lower than a predetermined threshold or converges to a certain value.

5. The data processing system according to claim 4, wherein the data processing unit is configured to use a function composed of a Gaussian function and an exponential function as the peak model function.

6. The data processing system according to claim 4, wherein the matrix decomposition is a non-negative matrix factorization.

Images