WO2006134703A1

WO2006134703A1 - Method of data correction in liquid chromatography

Info

Publication number: WO2006134703A1
Application number: PCT/JP2006/306907
Authority: WO
Inventors: Masaya Ono; Tesshi Yamada; Setsuo Hirohashi
Original assignee: Japan Health Sciences Foundation
Priority date: 2005-06-17
Filing date: 2006-03-31
Publication date: 2006-12-21
Also published as: JP5119405B2; JP2012198245A; JPWO2006134703A1; JP5760253B2

Abstract

A method of data correction characterized by including the concentration gradient inducing step of applying a solution containing a test substance to liquid chromatography to thereby induce a concentration gradient of the solution over a given period of time; the measuring step of obtaining retention time data and mass/charge ratio data correlated to each other with respect to each of components of the test substance separated and dissolved out during the concentration gradient inducing step; and the correction step of carrying out correction by correlating two-dimensional data consisting of the retention time data and mass/charge ratio data obtained in the measuring step to a standard two-dimensional data determined in advance.

Description

Specification

Data correction method for liquid chromatography

Technical field

[0001] The present invention relates to a data processing method for liquid chromatography in proteome analysis, and more particularly to a data processing method for ultra-low flow liquid chromatography.

Background art

[0002] Liquid chromatography utilizes a specific affinity between a resin loaded in a column and a substance in the solution to create a stepwise concentration gradient of the solution, thereby concentrating a specific solution concentration. This is a method of releasing a substance from a resin at a degree.

[0003] The concentration gradient is created as expressed as a function of time (as it varies with time). Therefore, it is possible to identify (specify) a substance by grasping the retention time during which the substance is released. Therefore, the reproducibility of retention time is the most important for identifying substances in liquid chromatography. Here, when the flow rate of the solution to the column is high, relatively good reproducibility is obtained. The flow rate of the solution to the column is low! / In this case, the reproducibility is good! /

[0004] In addition, according to the proteomics technique using a mass spectrometer, the protein of an organism can be quantitatively identified. For this reason, the proteomics method has begun to be widely applied in the fields of medicine and biology. Among the proteomics methods, a nanoLC / MS system that combines ultra-low flow liquid chromatography and an accurate mass spectrometer has recently attracted attention. According to this apparatus, it is possible to identify a huge amount of protein from a small amount of sample. More specifically, a substance that has been subjected to intensive separation by ultra-low flow liquid chromatography is identified by accurately measuring the mass of each substance with an accurate mass spectrometer. It is.

[0005] However, a method for performing protein identification based on retention time data obtained from ultra-low flow liquid chromatography has not yet been put into practical use. For example, the retention time data and mass data obtained from the nanoLC / MS system force are expanded in two dimensions and the coordinates are used. The attempt to create a protein map has already been published in 2002 (Lipton MS, et al. Proceeding of the National Academy of Sciences. 99: 11049, 2002), but such an attempt has been developed to the practical stage. The published paper has not yet appeared.

[0006] In addition, Zhang H, et aU, “Molecular & Cellular Proteomics. 4: 144, 2005” describes that two-dimensionally expanded retention time data and mass data were “corrected”. There is no description as to whether the correction was made specifically.

[0007] And, regarding the point of how to compare the two-dimensionally developed retention time data and mass data between a plurality of sample solutions, any paper has been published. Absent.

Non-Patent Document 1: Proceeding of the National Academy of Sciences. 99: 11049, 2002 (Lip ton MS, et al)

Non-Patent Document 2: Molecular & Cellular Proteomics. 4: 144, 2005 (Lipton MS, et al) Invention Disclosure

Problems to be solved by the invention

[0008] The present invention has been made in consideration of the above points, and it is possible to confirm the substantially high reproducibility by correcting the data of the retention time that may differ for each inspection (each measurement). It is an object of the present invention to provide a liquid chromatography data correction method that can be performed, particularly, an ultra-low flow liquid chromatography correction method.

[0009] Further, the present invention provides a liquid chromatography data analysis method capable of realizing high-precision data comparison of a plurality of solutions to be tested, in particular, an ultra-low flow liquid chromatography analysis method. The purpose is to do.

Means for solving the problem

[0010] The present invention provides a concentration gradient generation step of flowing a solution containing a substance to be inspected through liquid chromatography to generate a concentration gradient of the solution over a predetermined time, and the separation and elution during the concentration gradient generation step Each component of the inspected substance is a two-dimensional measurement process in which retention time data and mass-to-charge ratio data are associated with each other, and the retention time data and mass-to-charge ratio data obtained in the measurement process. And a correction step of correcting the data by correlating with the standard two-dimensional data obtained in advance. It is.

[0011] According to the present inventors, in the liquid chromatography method, as will be described later, with respect to the retention time data to be measured, for example, a deviation of 79 seconds on average and 192 seconds at maximum can occur. However, according to the present invention, it is possible to confirm substantially high reproducibility by correlating two two-dimensional data. In other words, by correlating two two-dimensional data, it is possible to superimpose and evaluate the characteristics of the two. Therefore, regardless of the difference in the absolute value of the measurement data itself, High reproducibility can be confirmed. This makes it possible to identify even expression differences between different samples, for example. Therefore, the present invention is expected to be developed in the future, and will make a very important contribution to the field of proteomics.

[0012] The target for correlating the two-dimensional data obtained by the measurement is not limited to the standard two-dimensional data obtained in advance. For example, it is of course possible to correlate two two-dimensional data obtained by measurement.

[0013] That is, the present invention includes a first concentration gradient generation step of generating a concentration gradient of the first solution over a predetermined time by flowing a first solution containing the first analyte in liquid chromatography. A first measurement step for associating retention time data and mass-to-charge ratio data for each component of the first analyte to be separated and eluted during the first concentration gradient generation step; and the liquid chromatography. A second concentration gradient generating step for generating a concentration gradient of the second solution over a predetermined time by flowing a second solution containing the second substance to be inspected; and the second concentration gradient separated and eluted during the second concentration gradient generating step. (2) For each component of the substance to be inspected, the second measurement step obtained by associating the retention time data with the mass to charge ratio data, and the retention time data and the mass to charge ratio data obtained in the second measurement step. Two-dimensional data is converted into the first measurement step And a correction step of correcting by correlating with the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in (1).

[0014] Also in the present invention, it is possible to confirm substantially high reproducibility by correlating two two-dimensional data. In other words, by correlating two two-dimensional data, it is possible to superimpose and evaluate the characteristics of the two, so that the evaluation of the measurement data can be performed regardless of the difference in the absolute value of the measurement data itself. Confirmation of reproducibility can do.

[0015] Furthermore, the two correlated two-dimensional data may have been obtained using different liquid chromatography.

[0016] That is, the present invention provides a first concentration gradient generation step for generating a concentration gradient of the first solution over a predetermined time by flowing the first solution containing the first substance to be tested through the first liquid chromatography. A first measurement step for obtaining retention time data and mass-to-charge ratio data in association with each component of the first analyte to be separated and eluted during the first concentration gradient generation step, and a second liquid chromatography The second solution containing the second substance to be tested is flowed into the second concentration gradient generating step for generating a concentration gradient of the second solution over a predetermined time and separated and eluted during the second concentration gradient generating step. For each component of the second substance to be inspected, a second measurement step in which retention time data and mass-to-charge ratio data are associated with each other, and retention time data and mass-to-charge ratio obtained in the second measurement step 2D data with the first measurement A data correction method characterized by comprising a correction step of correcting by the this correlating a two-dimensional data of the obtained retention time data and mass-to-charge ratio data in extent.

[0017] Also in the present invention, it is possible to confirm substantially high reproducibility by correlating two two-dimensional data. In other words, by correlating two two-dimensional data, it is possible to superimpose and evaluate the characteristics of the two, so that the evaluation of the measurement data can be performed regardless of the difference in the absolute value of the measurement data itself. With high! Reproducibility can be confirmed.

[0018] According to the present invention, it is possible to correct the measured holding time deviation by the correction process, and therefore, the use of powerful ultra-low flow rate liquid chromatography data that has not been practically used in the past is used. Becomes realistic. Specifically, in the concentration gradient generating step, a solution containing a substance to be inspected in liquid chromatography can be flowed at a flow rate of 500 nlZmin or less, particularly preferably at a flow rate of about 200 nlZmin.

[0019] Preferably, in the correction step, a dynamic search for searching for an optimum corresponding position using a two-dimensional grid coordinate with each cycle being a cycle number of two two-dimensional data (ascending order number with respect to holding time). Algorithms are being used.

[0020] More specifically, the dynamic algorithm is, for example, n cycles of one two-dimensional data. The Pearson product-moment correlation coefficient between the mass-to-charge ratio (mass spectrum) A (n) in the second eye and the mass-to-charge ratio B (n) in the m-th cycle of the other two-dimensional data is R (A (n ), B (m)), the gap penalty is g, the total number of cycles of one 2D data is N, and the total number of cycles of the other 2D data is M, the two 2D data Two-dimensional lattice coordinates L (i, j) with the cycle number as each axis

L (i, j) = max (L (i- 1, j) + g,

L (i, G) + g ゝ

L (i-1, j-l) + R (A (n), B (m)))

(I = l, ···, NJ = 1, ···, M) and L = argmax (k, 1), ((k = N, l = l, ···, M) and (k = l, ···, N, 1 = M))

Starting from (k, 1) = V

o

A coordinate array represented by L = argmax (V), (V = V- (1, 1), V _- (0, 1), V_one (1, 0)) is determined.

[0021] More preferably, the dynamic algorithm is V = V out of L = argmax (V).

i i +1 i

+ After extracting only the coordinates that satisfy (1, 1), the cycle number is converted to the retention time, and the curve obtained by spline interpolation or polynomial regression is determined as the correction function.

[0022] Further, the present invention provides a data correction method having any one of the above-described characteristics for each of a plurality of solutions (or second solutions) each including a plurality of substances to be inspected (or second substances to be inspected). The two-dimensional data of the data correction process for carrying out the process and the retention time data corrected by the data correction method and the mass-to-charge ratio data are stored as the retention time data of each solution (or each second solution) for a certain mass charge. A data development process that develops two-dimensional image data arranged in parallel, and a same peak extraction process that extracts the same peak of retention time data based on the two-dimensional image data It is an analysis method.

[0023] According to the present invention, the same peak of retention time data is extracted from a plurality of solutions (or second solutions) each containing a plurality of substances to be inspected (or second substances to be inspected). The data characteristics of this solution (or second solution) can be analyzed effectively. This As a result, it can be expected to significantly promote the development of disease markers such as tumor markers.

[0024] For example, for a plurality of solutions (or second solutions) each containing a plurality of test substances (or second test substances) belonging to the first group, and a plurality of test substances belonging to the second group For the multiple solutions (or the second solution) each containing (or the second test substance), the data analysis method described above is performed (the data analysis step), and the first group of multiple solutions (or the second solution) Compare the same peak of retention time data obtained from the same peak of retention time data obtained from multiple solutions (or 2nd solution) of the second group, and there is a significant difference between them. If there is a significant difference, the difference can be used as a “marker”.

[0025] Here, usually, the same peak extraction step includes a peak detection step for detecting a peak for retention time data of each solution (or each second solution), and each solution detected in the peak detection step ( Or the same peak specifying step for specifying the correspondence between peaks of each second solution).

[0026] Preferably, the same peak specifying step includes a candidate peak extracting step of extracting a candidate peak included within a predetermined holding time width, and the candidate peak extracting step in a certain solution (or second solution). When there are one or more extracted candidate peaks, one candidate peak for the solution (or the second solution) is selected and extracted in the candidate peak extraction step in a certain solution (or the second solution). If there is no candidate peak, there is no candidate peak for the solution (or second solution), and the selected candidate peak score (total intensity) for each of the combinations of candidate peak selections. The combination of the score calculation step for calculating the selection and the selection of the candidate peak that provides the maximum score among the scores obtained in the score calculation step correspond to each other. And a peak identification step that identifies the peak as a peak.

[0027] In this case, more preferably, the same peak specifying step further includes a data dividing step of dividing retention time data into sections by the same peak specified in the peak specifying step after the peak specifying step. The candidate peak extraction step, the score calculation for the retention time data divided in the data division step The process, the peak identifying process, and the data dividing process are recursively repeated.

[0028] Further, for example, the candidate peak extracting step is performed with each peak as a reference and an allowable deviation range width from the peak as a predetermined holding time width. For example, the allowable deviation range is 0.7 min on the + side.

[0029] In this case, in order to reduce the number of calculations (computation burden), the ratio of the solution without the candidate peak (or the second solution) extracted in the candidate peak extraction step falls below a predetermined minimum detection rate. It is preferable that the execution of the same peak specifying step based on the peak is terminated. In this case, preferably, the minimum detection rate is set to 0.1 to 0.4 (if it exceeds 0.5, it becomes difficult to specify a significant difference between the two groups).

[0030] Preferably, in the data development step, the retention time data of each solution (or each second solution) is developed into two-dimensional image data arranged in parallel for each unit mass charge. The same peak extracting step extracts the same peak of the retention time data for each unit mass charge based on the two-dimensional image data.

[0031] Further, the present invention provides a concentration gradient generating step for generating a concentration gradient of the solution over a predetermined time by flowing a solution containing a substance to be inspected in liquid chromatography, and separation during the concentration gradient generating step. A measurement step for associating retention time data and mass-to-charge ratio data for each component of the eluted substance to be inspected, and a data correction apparatus for a liquid chromatography method comprising: The two-dimensional data cycle is corrected by correlating the two-dimensional data of the retention time data and mass-to-charge ratio data obtained by correlating with the standard two-dimensional data obtained in advance. Dynamic algorithms (software) that search for the best corresponding position using two-dimensional grid coordinates with numbers (ascending order with respect to holding time) as axes are now used! / A data correction apparatus according to claim Rukoto.

[0032] The correction device or each element means of the correction device can be realized by a computer system.

[0033] Further, a program for realizing them in a computer system and a computer-readable recording medium recording the program are also subject to protection in this case. Here, the recording medium includes a network that propagates various signals in addition to a medium that can be recognized as a single unit such as a flexible disk.

[0035] Further, the present invention provides a concentration gradient generation step in which a plurality of solutions each containing a plurality of substances to be tested are flowed through the same or different liquid chromatography and a concentration gradient of each solution is generated over a predetermined time. And a measurement step obtained by associating retention time data with mass-to-charge ratio data for each solution for each component of the substance to be inspected separated and eluted during the concentration gradient generation step, and obtained by the measurement step. Data development step of developing the two-dimensional data of the obtained retention time data and mass-to-charge ratio data into two-dimensional image data in which the retention time data of each solution is arranged in parallel for a certain mass charge, and the two-dimensional image It is a data analysis method characterized by comprising the same peak extraction step of extracting the same peak of retention time data based on the data.

[0036] At the time of the filing of the present application, if the data correction method proposed in "Japanese Patent Application 2005-1775 47", which is the basis for claiming priority, is not used, The same peak extraction process of extracting the same peak of retention time data based on this cannot be performed in practice (the correspondence between peaks cannot be specified as shown in FIG. 16). However, if the accuracy of the data measurement method is improved in the future, the data analysis method proposed in this application without using the data correction method proposed in “Japanese Patent Application 2005-177547” is independent. Can be used. That is, according to the present invention,

By extracting the same peak of retention time data from a plurality of solutions (or second solutions) each containing a plurality of substances to be inspected (or second substances to be inspected), the plurality of solutions (or second solutions) ) Data characteristics can be effectively analyzed, and the development of disease markers such as tumor markers can be significantly promoted.

[0037] Further, the present invention provides a concentration gradient generation step in which a plurality of solutions each containing a plurality of substances to be tested are flowed through the same or different liquid chromatography, and a concentration gradient of each solution is generated over a predetermined time. And a measuring step for associating retention time data and mass-to-charge ratio data for each solution for each component of the substance to be inspected separated and eluted during the concentration gradient generation step. A data analysis apparatus for the method, wherein the retention time data obtained in the measurement step and the mass-to-charge ratio data are 2 A data development device that develops two-dimensional image data in which the retention time data of each solution is arranged in parallel for a certain mass charge, and extracts the same peak of the retention time data based on the two-dimensional image data. A data analysis device comprising the same peak extraction device.

[0038] The data analysis device or each element means of the data analysis device can be realized by a computer system.

[0039] Further, a program for realizing them in a computer system and a computer-readable recording medium recording the program are also subject to protection in this case.

[0040] Here, the recording medium includes a network that propagates various signals in addition to a medium that can be recognized as a single unit such as a flexible disk.

Brief Description of Drawings

FIG. 1 is a flowchart showing an outline of one embodiment of the present invention.

FIG. 2 is a schematic diagram showing the concept of a dynamic algorithm according to an embodiment of the present invention.

FIG. 3 is a schematic diagram showing the concept of a dynamic algorithm according to an embodiment of the present invention.

FIG. 4 is a schematic diagram illustrating the operation of a dynamic algorithm according to an embodiment of the present invention.

FIG. 5 is a schematic diagram illustrating the operation of a dynamic algorithm according to an embodiment of the present invention.

FIG. 6 is a schematic diagram for explaining the operation of a dynamic algorithm according to an embodiment of the present invention.

FIG. 7 is a graph showing an example of correction of measurement data (an example of difference).

[Fig. 8] Graph showing the reproducibility of corrected data.

FIG. 9 is a flowchart showing an outline of the second embodiment of the present invention.

[FIG. 10] An example of 2D image data in which the retention time data of each sample is arranged in the vertical axis direction.

FIG. 11a is a graph showing the concept of a baseline correction process.

[Fig. 1 lb] A graph showing the concept of the smoothing process.

FIG. 11c is a graph showing the concept of the peak detection process.

[Fig. 12a] Example of 2D image data.

[FIG. 12b] Image data indicating peaks detected from the two-dimensional image data in FIG. 12a.

[Fig. 13] Schematic flowchart showing the same peak extraction process (same peak extraction algorithm) of the present embodiment _n [Fig. 14a] Example of two two-dimensional images including the same peak with a significant difference.

[Fig. 14b] ROC curve of Fig. 14a.

[FIG. 14c] Peak intensity distribution diagram of FIG. 14a.

[Figure 15a] Examples of two sets of two-dimensional images containing two identical peaks with significant differences.

FIG. 15b is a peak intensity distribution diagram of FIG. 15a.

[Fig.16] Example of 2D image without data correction.

BEST MODE FOR CARRYING OUT THE INVENTION

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

First, the substance to be inspected and the solution used in the present embodiment will be described.

[0044] As a test substance, a DLD1 human colon cancer cell line (Honda et al. Gastroenterology 2005; 128: 51-62) into which actinin 4 (ACTN 4) whose expression can be controlled with tetracycline was introduced was used. In normal culture, ACTN4 is expressed (DLDl Tet-off A CTN4), but 0.01 to 0.1 μg / ml of doxocycline (Dox) suppresses ACTN4 expression (DLDl Tet- on ACTN4). Each cell solution of DLDl Tet—off ACTN4 and DLDl Tet-on ACTN 4 was adjusted to a concentration of 3 mgZml.

[0045] Next, 100 1 of each cell solution of DLDl Tet-off ACTN4 and DLDl Tet-on ACTN4 was collected, and protein concentration was performed by acetone precipitation. Then, add 5 M UREA to 101, 1M NH HCO to 2.51, trypsin to 3.3 μg, and then add 50 μ1 with purified water.

4 3

did.

[0046] Then, after digestion reaction at 37 ° C for 20 hours, 50 μl of acetonitrile was added, centrifuged at 17400 G for 10 minutes, the supernatant was transferred to another tube, and dried using a speed bag. . Then, it was dissolved in 0.1% formic acid 50 1 to obtain a sample (solution) for measurement (Figure 1: STEP 1).

On the other hand, Splitless Nano HPLC System (KYA, Tokyo) was used as ultra-low flow liquid chromatography. Otadecyl group introduced into high-purity silica gel with a particle size of 3 m and a pore size of 120 A, and then the remaining silanol groups are end-capped to the limit. 0.5 mm and 1 mm long ones were used for the trap columns (HiQ sil, KYA, Tokyo). [0048] Then, 10 μl of the sample was taken, and at a very low flow rate of 200 nlZmin, a continuous concentration gradient was obtained over 60 minutes from 0.1% formic acid to 0.1% formic acid 80% acetonitrile. Generated (Figure 1: STEP2). In the meantime, each component was separated and eluted (Figure 1: STEP3).

[0049] QTOF Ultima (Waters, MA, USA) was used for mass analysis of each component, and measurement was performed for 60 minutes in the centroid format with a scan time of 1 second in the range from 250 to 16 OOMZZ. It was broken. For each of DLD1 Tet-off ACTN4 and DLD1 Tet-on ACTN4, data was collected in duplicate (data was collected twice) (Figure 1: STEP4).

[0050] In the two-dimensional display of the data, the mass-to-charge ratio was converted to the maximum value for each range of lmZz (mass to charge ratio) and output in wiff format. The analysis target range is limited to mass charge specific power 00 ~: L000mZz, retention time (RT) 1 ~ 1800sec, and intensity 200 is replaced with 1 ~ 255 gray scale. Displayed.

[0051] In the present embodiment, data correction was performed by correlating the data of two times (total of four times) each of which two sample forces were collected (Fig. 1: STEP5).

First, a method (algorithm) for obtaining a correction function used for data correction, which is adopted in the present embodiment, will be described.

[0053] In the present embodiment, the reference side (standard) two-dimensional data is A, the correction target two-dimensional data is B, and the sum of the mass spectrum correlation coefficients at each holding time is the maximum. A correction function is derived such that

[0054] First, in order to improve the execution speed and ensure redundancy for mass measurement errors, the ion intensity of the mass spectrum at each RT (retention time) is converted into a representative value for each lmZz interval.

[0055] Next, using the two-dimensional grid coordinates with the cycle numbers (ascending numbers for each RT) of the two two-dimensional data A and B as the respective axes, the dynamic algorithm as described below is used to optimize A route search that provides a corresponding position is performed.

[0056] The dynamic algorithm of the present embodiment uses the mass-to-charge ratio (mass spectrum) A (n) in the n-th cycle of one two-dimensional data and the mass-to-charge ratio in the m-th cycle of the other two-dimensional data. The Pearson product moment correlation coefficient between B (n) and R (A (n), B (m)), and the gap penalty as g (typically 0.5), Let N be the total number of cycles of the dimension data When the total number of cycles of the other two-dimensional data is M, the two-dimensional lattice coordinates L (i, j) with the cycle numbers of the two two-dimensional data as axes are

L (i, j) = max (L (i- 1, j) + g

L (i, G) + g

L (i-1, j-l) + R (A (n), B (m)))

(I = l, ···, N j = l, ···, M).

[0057] Then, L = argmax (k, 1), ((k = N 1 = 1, ···, M) and (k = l, ···, N 1 so as to correspond to the optimum corresponding position = M)) gives the coordinates (k, 1) = V as the starting point

0

[0058] Furthermore, the dynamic algorithm of the present embodiment is such that V = V among L = argmax (V).

i i +1 i

+ After extracting only the coordinates satisfying (1, 1), the coordinate system is RT-converted from the cycle number, and the curve obtained by spline interpolation or polynomial regression is used as the correction function.

The above algorithm will be described below with reference to the drawings.

As shown in FIG. 2, the target data “A” is a figure (data string) distorted in the Y-axis direction with respect to the reference data “A”. In this case, this algorithm corresponds to the algorithm for ^obtaining f ^_1 (y,).

[0061] FIG. 3 shows a plane formed with y and y 'as axes corresponding to FIG. In FIG. 3, the line force connecting the square marks is a line indicating the y-y 'corresponding position in this case. In this algorithm, y-y and the corresponding position (route) are obtained.

FIG. 4 corresponds to the example of FIGS. 2 and 3, and the Pearson product moment correlation coefficient R (

A (n), B (m)) is determined, and the gap penalty is set to -0.5, and L (i, j) is actually calculated using this algorithm.

[0063] Once L (i, j) is obtained for all (i, j), as shown in Fig. 5, L = argmax (k, 1)

, ((k = N l = l,..., M) and (k = l,..., N 1 = M)) gives the coordinates (k, 1) = V

o The coordinate array specified as the starting point and represented by L = argmax (V), (V = V — (1, 1), V-(0, 1), V one (1, 0)) is determined .

[0064] Finally, as shown in Fig. 6, among L = argmax (V), V = V + (1, 1) is satisfied.

i i + 1 i After extracting only the coordinates, the cycle number is converted into the retention time, and the curve is obtained by spline interpolation or polynomial regression. This curve is a correction function to be obtained.

[0065] With the algorithm as described above, three of the four two-dimensional data obtained with the two sample forces in the present embodiment were corrected.

[0066] For data correction, the knock ground was first calculated using the RollingBall Algorithm (from "NIH Image J") (Radius = 50). Then, the threshold at each point is set as (background value + data value) Z2, and adjacent spots with values above the threshold are integrated, peak intensity = sum of Intensity, peak mz = MZ centroid, peak rt = RT centroid As a result, a peak was detected. Then, the detected peak was fitted to the liquid chromatography corrected data (the retention time data after applying the correction function obtained by the above algorithm), and obtained (corrected). Two 2D data of the same sample are superimposed and “integrated” (averaged). In this “integrated” treatment, the allowable ranges of variation in mass-to-charge ratio and elution time (retention time) were set to ± 2, ± 20, respectively. In addition, when multiple peaks exist, the higher Intensity was adopted, and MZ and RT were averaged. The pair was detected from the data of each sample obtained in this way, and when the pair was detected, the ratio of the peak intensities was calculated (Fig. 1: STEP 6).

[0067] The results are summarized below.

[0068] As described above, data was collected in duplicate for each of DLD1 Tet-off ACTN4 and DLD1 Tet-on ACTN4 (hereinafter referred to as Offl, Off2, Onl, On2). Among these four data, one of DLD1 Tet-off ACTN4 data (Off 1) is the master data (standard data), and the other three data are corrected by the correction function obtained by the above algorithm. .

[0069] As shown in Figure 7, using Off 1 data as a reference, Off2 before correction has a difference (variation) of 190 sec at maximum and 71.9 sec on average, and Onl before correction has a maximum of 192 sec There was a difference (variation) of 130.2 seconds on average, and On2 before correction had a difference (variation) of 80 seconds at maximum and 36.3 seconds on average (in Fig. 7, the deviation in the X-axis direction was different) (Variation)). However, as a result of correction using the correction function obtained by the above algorithm, as shown in Fig. 8, a high correlation coefficient of 0.92 is obtained for DLD1 Tet-off ACTN4 (Offl, Off2) of the same sample. Similarly, a high correlation coefficient of 0.94 could be obtained even with DLDl Tet-on ACTN4 (Onl, On2) of the same sample.

[0070] The value obtained by averaging Off 1 and corrected Off 2 was used as the representative value of DLDl Tet-off ACTN4, and was obtained by averaging corrected Onl and corrected On2. As a representative value of DLDl Te t-on ACTN4, between those two groups, there is a strong expression of peak intensity force S 10 or more and no expression in one, or the ratio of peak intensity is 3 or more However, force S203peak was recognized (DLDl Tet-off ACTN4 predominance was 107peak, DLDl Tet-on A CTN4 predominance was 96peak). These peaks were found to vary depending on the expression of actinin 4 (ACTN4), that is, it was possible to identify actinin 4 (protein).

[0071] In this way, the measurement data itself is corrected by correlating multiple measurement data (Off 1 and Off 2 or Onl and On2) of the same type of sample for which it is difficult to find reproducibility. Therefore, it can be regarded as a data group of the same type having a certain degree of reproducibility, and by making the values obtained by averaging them as representative values, it becomes possible to perform identification or diagnosis with higher accuracy. . In other words, by correlating two or more two-dimensional data, it becomes possible to superimpose their characteristics for evaluation, and high reproducibility can be obtained when evaluating measurement data regardless of the difference in measurement data itself. This makes it possible to compare with many samples, for example, patient serum. This significantly increases the possibility of developing markers for different diseases.

[0072] It should be noted that the data correction processing described above can be normally performed by a data correction apparatus that can be configured by various computer systems. Here, a program for realizing the data correction apparatus on a computer system and a computer-readable recording medium recording the program are also subject to protection in this case.

[0073] Further, when the data correction apparatus is realized by a program such as an OS (second program) operating on the computer system, a program including various instructions for controlling the program such as the OS and the program are recorded. The recorded media are also subject to protection in this case.

Here, the recording medium includes a network that propagates various types of signals in addition to a medium that can be recognized as a single unit such as a flexible disk. [0075] Next, a second embodiment of the present invention will be described with reference to the drawings.

[0077] In the present embodiment, as test substances, plasma of 18 patients with spleen cancer (Group 1) and plasma of 19 patients with no cancer (Group 2) were used 10 1 respectively. Then, a glycoprotein fraction adsorbed to 100 1 of concanavalin A was extracted with these forces (this treatment is not an essential treatment, but is preferably performed in terms of sensitivity adjustment and point).

[0078] Then, 5M UREA was added to each of the glycoprotein fractions 10 1 and 1M NH HCO.

4 3 was adjusted to 2.51, and trypsin was adjusted to 3.3 / z g calorie.

[0079] Then, after digestion reaction at 37 ° C for 20 hours, 50 μl of acetonitrile was added, centrifuged at 17400 G for 10 minutes, the supernatant was transferred to another tube, and dried using a speed bag . Then, it was dissolved in 0.1% formic acid 50 1 to obtain a sample (solution) for measurement (Figure 9: STEP21

) o

On the other hand, Splitless Nano HPLC System (KYA, Tokyo) was used as ultra-low flow liquid chromatography. A high-purity silica gel with a particle size of 3 m and a pore size of 120 A, after introducing otadecyl groups, the remaining silanol groups were end-capped to the limit. 0.5 mm and 1 mm long ones were used for the trap columns (HiQ sil, KYA, Tokyo).

[0081] Then, 10 1 of the sample was taken, and a continuous concentration gradient was generated over 60 minutes from 0.1% formic acid power to 0.1% formic acid 80% acetonitrile with an ultra-low flow rate of 200 nlZmin (Fig. 9: STEP22). In the meantime, each component was separated and eluted (Figure 9: STEP23).

[0082] QTOF Ultima (Waters, MA, USA) was used for mass analysis of each component, and the measurement was performed for 60 minutes in the centroid format with a scan time of 1 second in the range from 250 to 16 OOMZZ. It was broken. For each sample (solution), data was collected by triplicate (data was collected three times) (Figure 9: STEP24).

[0083] In the two-dimensional display of the data, the mass-to-charge ratio was converted to the maximum value for each range of lmZz (mass to charge ratio) and output in wiff format. The analysis target range is limited to mass charge specific power 00 ~: L000mZz, retention time (RT) 1 ~ 1800sec, and intensity 200 is replaced with 1 ~ 255 gray scale. Displayed. [0084] In the present embodiment, the data correction was performed by correlating the data of 3 times (total of 111 times) each of which 37 sample forces were collected (FIG. 9: STEP25).

[0085] Also in the present embodiment, according to the algorithm described with respect to the above-described embodiment, the reference (standard) two-dimensional data is A, and the two-dimensional data to be corrected is B. A correction function was derived that maximized the sum of the mass spectral correlation coefficients at the retention time, and the resulting correction function was multiplied by the retention time data for each sample. Here, the data based on the plasma of a non-carcologist was used as the reference (standard) 2D data A.

[0086] With respect to the two-dimensional data after correction (retention time data-mass-to-charge ratio data) obtained in this way, in this embodiment, each lmZz is measured for each lmZz by a data analysis device configured by a computer system. Sample data retention time data was expanded into two-dimensional image data arranged in parallel (Figure 9: STEP26).

An example of 2D image data is shown in FIG. In Fig. 10, the horizontal axis is the holding time (RT:

Each sample data is arranged in the vertical axis direction. Figure 10 shows 2D image data for 86 3mZz!

[0088] Then, the same peak of retention time data in the two-dimensional image data was extracted by the data analysis apparatus according to the newly developed same peak extraction algorithm (FIG. 9: STEP27).

Here, the same peak extraction algorithm according to the present embodiment includes a baseline correction step (FIG. 11 (a)) and a smoothing step (FIG. 11 (b)) for the retention time data of each sample data. And a peak detection step (FIG. 11 (c)).

[0090] Baseline correction is a process for correcting the inclination and undulation of the baseline that occurs in the spectrum waveform due to the effect of light scattering on the sample. Smoothing is a process that removes noise by taking a weighted average using a Gaussian function (see Equation 1). These processes are often used as a processing method for data analysis.

[Number 1]

0 = area, a _x = center, a ₂ = width (std dev.)

[0091] Also, in the peak detection step of the present embodiment, the peak detection accuracy is improved by calculating the signal Z noise ratio for each data point.

An example of an image showing the detected peak is shown in FIG. 12 (b). Fig. 12 (a) is an example of 2D image data, and Fig. 12 (b) is image data showing a peak in which the data force of Fig. 12 (a) is also detected.

[0093] Then, the same peak extraction algorithm of the present embodiment includes the same peak specifying step of specifying the correspondence between peaks of each sample data detected in the peak detecting step.

As shown in FIG. 13, in this same peak specifying step, an allowable deviation range width from the peak is set as a predetermined holding time width on the basis of each peak, and it is included in the holding time width. It has a candidate peak extraction process to extract complementary peaks (Fig. 13: STEP31). The allowable deviation range width is, for example, 0.7 min on the + side.

Then, in the same peak specifying step, when there are one or more candidate peaks extracted within the holding time width in a certain sample data, one candidate peak for the sample data is selected and a certain sample is selected. If there is no candidate peak extracted within the retention time width in the data, the candidate peak is assumed to be none according to the sample data, and the selection of the candidate peak (over all sample data) is performed. Each of the all combinations has a score calculation step for calculating the score (total intensity) of the selected candidate peak (FIG. 13: STEP 32).

Here, if there is no candidate peak extracted within the holding time width and the ratio of sample data falls below a predetermined minimum detection rate, at that time, the same as described above for the holding time width. It is preferable that the peak identification process is completed! /. In such a case, the same peak should not be specified within the retention time width. The minimum detection rate is usually set between 0.1 and 0.4 (if it exceeds 0.5, the significance between the two groups It may be difficult to identify the difference).

[0097] Of the scores obtained in the score calculation process, the combinatorial power of selecting the candidate peak that provides the maximum score is identified as the same peak corresponding to each other (peak identification process) (Figure 13: STEP33) . Then, a process for complementing the peak is performed on the sample data for which the corresponding same peak is not recognized (with no extracted force) (FIG. 13: STEP 34).

[0098] After that, the retention time data is divided into sections by the same specified and supplemented peak (data dividing step) (FIG. 13: STEP35). The candidate peak extraction step, the score calculation step, the peak identification step, and the data division step are recursively repeated for the retention time data (both) divided in the data division step. (Figure 13: STEP36).

[0099] 105457 identical peaks were identified by the same peak extraction algorithm of the present embodiment as described above. For these same peaks, the same peak based on the plasma of the 18 patients with splenic cancer (Group 1) and the same peak based on the plasma of the 19 non-cancer-bearing patients (Group 2) were compared. If there are significant differences, those differences can be used as “markers” (Figure 9: STEP28).

[0100] Specifically, in the case of the present embodiment, the average peak intensity was 10 or more and a significant difference of 0.0001 or less was shown by the U test between the spleen cancer patient group and the non-carried group. There were 109 peaks (80% for the spleen cancer patient group, 29 peaks for the non-cancer-bearing group). In addition, 32 peaks having an area under the R OC curve of 0.9 or more were observed.

[0101] For one of these 32 peaks, Fig. 14 (a) to Fig. 14 (c) show a two-dimensional image (Fig. 14 (a)), ROC curve (Fig. 14 (b)), The peak intensity distribution diagram (Fig. 14 (c)) is shown.

[0102] The 32 peaks were analyzed with two factors using SVM. After cross validation, there are three combinations of peaks with a discrimination rate of 100% (sensitivity 100%, specificity 100%), and a discrimination rate of 97% (sensitivity 100%, specificity 95%, there is Or, there were 28 peak combinations with 94% sensitivity and 100% specificity. Fig. 15 (a) shows a peak image with a discrimination rate of 100% (sensitivity 100%, specificity 100%) by combination, and Fig. 15 (b) shows the spleen divided using the peak. Peak strength between cancer patients and non-cancer carriers Degree distribution.

[0103] Thus, according to the present embodiment, the same peak of retention time data is extracted from the solution containing the plasma of the spleen cancer patient, and the solution force retention time data including the plasma of the non-cancer carrier is extracted. By extracting the same peak and comparing the two, development of a spleen cancer marker can be promoted.

It should be noted that the data analysis process described above can be normally performed by a data analysis apparatus that can be configured by various computer systems. Here, a program for realizing the data analysis apparatus on a computer system and a computer-readable recording medium recording the program are also subject to protection in this case.

[0105] Furthermore, when the data analysis device is realized by a program such as an OS (second program) that runs on a computer system, a program including various instructions for controlling the program such as the OS and the program are recorded. The recorded media are also subject to protection in this case.

Here, the recording medium includes not only a flexible disk or the like that can be recognized as a single unit, but also a network that propagates various signals.

[0107] The above embodiment is intended to develop a spleen cancer marker, but the present invention is not limited to this. Extracting the same peak of multiple solution force retention time data, each containing multiple analytes, can effectively identify (analyze) the data characteristics of the multiple solutions. Can be expected to promote the development of

[0108] At the time of filing this application, if the data correction method proposed in "Japanese Patent Application 2005-177547" which is the basis of the priority claim in this application is not used, the retention time data of each sample data is arranged. Since the two-dimensional image data is in a state as shown in FIG. 16, if the same peak of the retention time data is extracted based on the two-dimensional image data! Virtually impossible. However, if the accuracy of the data measurement method is improved in the future, only the data analysis method may be used alone without using the data correction method.

Claims

The scope of the claims

[1] A concentration gradient generation step of flowing a solution containing a substance to be tested in liquid chromatography and generating a concentration gradient of the solution over a predetermined time;

A measurement step for obtaining retention time data and mass-to-charge ratio data in association with each component of the test substance separated and eluted during the concentration gradient generation step;

A correction step for correcting the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the measurement step by correlating with the standard two-dimensional data obtained in advance;

A data correction method comprising:

[2] A first concentration gradient generating step for generating a concentration gradient of the first solution over a predetermined time by flowing a first solution containing the first analyte in liquid chromatography;

A first measurement step of associating retention time data and mass-to-charge ratio data for each component of the first analyte to be separated and eluted during the first concentration gradient generation step; A second concentration gradient generating step of generating a concentration gradient of the second solution over a predetermined time by flowing a second solution containing the second test substance;

A second measurement step for obtaining retention time data and mass-to-charge ratio data in association with each component of the second analyte to be separated and eluted during the second concentration gradient generation step; Correlate the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the measurement process with the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the first measurement process. Correction process to correct by,

A data correction method comprising:

[3] a first concentration gradient generating step for generating a concentration gradient of the first solution over a predetermined time by flowing a first solution containing the first analyte in the first liquid chromatography;

A first measurement step for associating retention time data and mass-to-charge ratio data for each component of the first analyte to be separated and eluted during the first concentration gradient generation step; and a second liquid chromatography A second concentration gradient generating step of flowing a second solution containing the second substance to be inspected to generate a concentration gradient of the second solution over a predetermined time;

For each component of the second analyte to be separated and eluted during the second concentration gradient generation step, Then, the second measurement step obtained by associating the retention time data and the mass-to-charge ratio data, and the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the second measurement step, A correction step for correcting by correlating with the two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the first measurement step;

A data correction method comprising:

[4] In the concentration gradient generating step, the liquid force containing the test substance in the liquid chromatography

The flow rate is 500nlZmin or less!

The data correction method according to claim 1, wherein the data correction method is a data correction method.

[5] In the correction process, a dynamic algorithm is used to search for an optimal corresponding position using a two-dimensional grid coordinate with each cycle being the cycle number of two two-dimensional data (ascending order with respect to holding time). It looks like! /

[6] The dynamic algorithm is:

Pearson product phase relationship between the mass-to-charge ratio (mass spectrum) A (n) in the n-th cycle of one two-dimensional data and the mass-to-charge ratio B (n) in the m-th cycle of the other two-dimensional data Let R (A (n), B (m))

Let the gap pennality be g,

Let N be the total number of cycles for one 2D data,

When the total number of cycles of the other two-dimensional data is M,

Two-dimensional lattice coordinates L (i, j) with the cycle numbers of two two-dimensional data as axes, L (i, j) = max (L (i- 1, j) + g,

L (i, G) + g ゝ

L (i-1, j-l) + R (A (n), B (m)))

(I = l, ···, N, j = l, ···, M),

L = argmax (k, 1), ((k = N, l = l, ... ', Μ) and (k = l, ..., N, 1 = M )) Gives the coordinates (k, 1) = V

0

L = argmax (V), (V = V-(1, 1), V_-(0, 1), V_ one (1, 0)) The data correction method according to claim 5, wherein:

[7] The dynamic algorithm satisfies V = V + (1, 1) out of L = argmax (V).

i i + 1 i

7. The data correction according to claim 6, wherein after extracting only the coordinates, the cycle number is converted into a holding time, and a curve obtained by spline interpolation or polynomial regression is determined as a correction function. Method.

[8] A data correction step of performing the data correction method according to any one of claims 1 to 7 for each of a plurality of solutions or second solutions each containing a plurality of substances to be inspected or a second substance to be inspected. ,

The two-dimensional data of the retention time data and mass-to-charge ratio data corrected by the data correction method is converted into two-dimensional image data in which the retention time data of each solution or each second solution is arranged in parallel for a certain mass charge. Data development process to be deployed,

The same peak extraction step of extracting the same peak of the retention time data based on the two-dimensional image data;

A data analysis method characterized by comprising:

[9] The same peak extraction step includes:

A peak detection step for detecting a peak for the retention time data of each solution or each second solution;

The same peak identifying step for identifying the correspondence between the peaks of each solution or each second solution detected in the peak detecting step;

Contains

The data analysis method according to claim 8, wherein:

[10] The same peak identifying step includes:

If there are more than one candidate peak extracted in the candidate peak extraction step in a solution or second solution, and a candidate peak extraction step for extracting candidate peaks included within a predetermined holding time width, the solution Alternatively, one candidate peak for the second solution is selected, and if there is no candidate peak extracted in the candidate peak extraction step in a certain solution or the second solution, the candidate solution or the second solution Therefore, there is no candidate peak! For each of all combinations of candidate peak selections, A score calculation process for calculating the score (total strength) of the

Among the scores obtained in the score calculation step, a peak identification step for identifying the combination of candidate peak selections that provide the maximum score as the same corresponding peak,

have

The data analysis method according to claim 9, wherein:

[11] The same peak specifying step, after the peak specifying step,

A data dividing step for dividing the holding time data into segments by the same peak specified in the peak specifying step

Further including

For the retention time data divided in the data division step, the candidate peak extraction step, the score calculation step, the peak identification step, and the data division step are repeated recursively! ,

The data analysis method according to claim 10, wherein:

[12] The candidate peak extraction step is performed using each peak as a reference and an allowable deviation range width from the peak as a predetermined holding time width.

The data analysis method according to claim 10 or 11, wherein:

[13] The allowable deviation range width is 0.7 min on the + side.

The data analysis method according to claim 12, wherein:

[14] When the ratio of the solution without the candidate peak extracted in the candidate peak extraction step or the second solution falls below a predetermined minimum detection rate, the identification of the same peak based on the peak is performed. Implementation of the process has ended! /

14. The data analysis method according to claim 12 or 13, wherein:

[15] The minimum detection rate is set to 0.1 to 0.4

15. The data analysis method according to claim 14, wherein:

[16] In the data development step, for each unit mass charge, the retention time data of each solution or each second solution is developed into two-dimensional image data arranged in parallel.

The same peak extraction step is performed based on the two-dimensional image data. In addition, the same peak of retention time data is extracted

The data analysis method according to any one of claims 8 to 15, wherein

[17] For a plurality of solutions or second solutions each containing a plurality of test substances or a second test substance belonging to the first group, and a plurality of test substances or a second test substance belonging to the second group A data analysis step of performing the data analysis method according to any one of claims 8 to 16 for a plurality of solutions or second solutions each containing

Compare the same peak of retention time data obtained for multiple solutions or second solution force of the first group with the same peak of retention time data obtained for multiple solutions of the second group or second solution force. A verification process to verify whether there is a significant difference between the two,

A data comparison method characterized by comprising:

[18] A concentration gradient generating step for generating a concentration gradient of the solution over a predetermined time by flowing a solution containing the substance to be tested through liquid chromatography;

A data correction apparatus for a liquid chromatography method comprising:

The two-dimensional data of the retention time data and the mass-to-charge ratio data obtained in the measurement step is corrected by correlating with the standard two-dimensional data obtained in advance.

A dynamic algorithm that searches for the optimal corresponding position using a two-dimensional grid coordinate with each cycle as the cycle number of two two-dimensional data (ascending order number with respect to holding time) is used.

A data correction apparatus characterized by that.

[19] The dynamic algorithm is:

Pearson product phase relationship between the mass-to-charge ratio (mass spectrum) A (n) in the n-th cycle of one two-dimensional data and the mass-to-charge ratio B (n) in the m-th cycle of the other two-dimensional data Let the number be R (A (n), B (m))

Let the gap pennality be g,

Let N be the total number of cycles for one 2D data, When the total number of cycles of the other two-dimensional data is M,

L (i, G) + g ゝ

L (i-1, j-l) + R (A (n), B (m)))

(I = l, ···, N, j = l, ···, M),

0

L = argmax (V), (V = V-(1, 1), V_-(0, 1), V_ one (1, 0))

The data correction device according to claim 18, wherein

[20] The dynamic algorithm satisfies V = V + (1, 1) among L = argmax (V).

i i + 1 i

20. The data correction according to claim 19, wherein after extracting only the coordinates, the cycle number is converted into a holding time, and a curve obtained by spline interpolation or polynomial regression is determined as a correction function. apparatus.

21. A program that is executed by a computer system including at least one computer and causes the computer system to implement the data correction device according to any one of claims 18 to 20.

[22] instructions for controlling a second program running on a computer system including at least one computer are included;

21. A program that is executed by the computer system, controls the second program, and causes the computer system to implement the data correction apparatus according to claim 18.

[23] A concentration gradient generating step of generating a concentration gradient of each solution over a predetermined time by flowing a plurality of solutions each containing a plurality of test substances in the same or different liquid chromatography;

For each component of the inspected substance separated and eluted during the concentration gradient generation step, a measurement step for associating retention time data and mass-to-charge ratio data for each solution; and Data expansion that expands the two-dimensional data of the retention time data and mass-to-charge ratio data obtained in the measurement process into two-dimensional image data in which the retention time data of each solution is arranged in parallel for a certain mass charge Process,

A data analysis method characterized by comprising:

[24] A concentration gradient generating step of generating a concentration gradient of each solution over a predetermined time by flowing a plurality of solutions each containing a plurality of test substances in the same or different liquid chromatography,

A measurement step of associating retention time data and mass-to-charge ratio data for each solution for each component of the analyte to be separated and eluted during the concentration gradient generation step, and a liquid chromatography method comprising: A data analysis device for

Data expansion that expands the two-dimensional data of the retention time data and mass-to-charge ratio data obtained in the measurement process into two-dimensional image data in which the retention time data of each solution is arranged in parallel for a certain mass charge Equipment,

The same peak extraction device for extracting the same peak of the retention time data based on the two-dimensional image data;

A data analysis apparatus comprising:

[25] The same peak extraction device comprises:

Is supposed to run

25. The data analysis apparatus according to claim 24.

[26] The same peak specifying step,

A candidate peak extraction step for extracting candidate peaks included within a predetermined holding time width, and a candidate peak extracted in the candidate peak extraction step in a solution or a second solution. If there is more than one mark, select one candidate peak for the solution or the second solution, and there is no candidate peak extracted in the candidate peak extraction step over a certain solution or the second solution. In this case, there is no candidate peak in the solution or the second solution, and as a result, the score (total intensity) of the selected candidate peak for each of all combinations of the candidate peak selections. A score calculation step for calculating

Among the scores obtained in the score calculation step, a peak identification step for identifying combinations of candidate peaks that provide the maximum score as the same corresponding peaks;

have

26. The data analysis apparatus according to claim 25.

[27] The same peak identification step, after the peak identification step,

Further including

27. The data analysis apparatus according to claim 26.

[28] A program which is executed by a computer system including at least one computer and causes the computer system to implement the data analysis device according to any one of claims 24 to 27.

[29] includes instructions for controlling a second program running on a computer system including at least one computer;

A program that is executed by the computer system to control the second program to cause the computer system to realize the data analysis apparatus according to any one of claims 24 to 27.