KR20190005034A - Apparatus and method for quantification analysis of sample using liquid chromatograph-mass spectrometry - Google Patents

Abstract

A quantitative sample analysis apparatus includes: a spectrum recognition unit which receives the liquid chromatography-mass spectrometry spectrum of a sample including a first material attached with a first label material and a second material attached with a second label material having a different mass from the first label material; and a peptide spectrum match (PSM) analysis unit which extracts the peptide spectrum match spectrum corresponding to the first label material from the mass spectrometry spectrum received from the spectrum recognition unit, specifies the template of the first material by using the PSM of the first label material, calculates the predicted retention time (RT) shift of the first and second label materials, specifies the template of the second material by using the predicted RT shift, uses the template of the first and second materials to reconfigure the signals of the first and second materials for perform a quantitative peptide level analysis on the sample. The quantitative sample analysis apparatus can apply machine learning to accurately and efficiently predict the retention time shift caused by the substitution of the hydrogen in the label material with the deuterium (2H or D).

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for analyzing a sample using a liquid chromatograph mass spectrometry,

Embodiments relate to an apparatus and a method for analyzing a sample in a liquid chromatograph-mass spectrometry (LC-MS), and more particularly, to an apparatus and method for analyzing a sample by LC- The present invention relates to a technique for predicting a shift of retention time caused by the use of a substance and thereby improving the multiplexing of quantitative measurement by accurately separating signals of adjacent label substances.

Liquid Chromatograph-Mass Spectrometry (LC-MS) is a method in which a target material is separated into components by passing them through a column or the like in a liquid state, ionizes the respective components, and mass- As a technique for separating different substances, it can be used for identification of proteins and the like. When the ionization process for mass spectrometry in LC-MS is performed in a tandem manner, it is referred to as LC-MS / MS.

Among the LC-MS / MS techniques, quantification of the labeled target substance is performed by pre-binding a label substance known to the mass to charge ratio to the target substance and specifying the spectrum of the label substance in the obtained spectrum There is a way. In this case, depending on the degree of ionization of the substance used as a label, the quantitation technique may be performed by using a MS1-based quantitation technique based on a molecular-based label corresponding to a precursor, (MS2-based) quantification technique.

1 is a conceptual diagram for explaining an MS1-based quantitation technique using a parent molecule-based label.

Referring to FIG. 1, different types of label materials 110 and 120 are attached to different target materials 11 and 12 to be detected, respectively. Analysis of the sample containing the target materials 11 and 12 by the LC-MS / MS method revealed retention time due to LC, mass-to-charge ratio (m / z) The graphs 1100 and 1200 corresponding to the fragment ions are obtained in the three-dimensional space having the axis of the intensity as the axis. At this time, the target materials 11 and 12 corresponding to the respective graphs can be specified using the mass to charge ratio of the label materials 110 and 120, and thus the sample can be quantified.

However, since the label material may contain isotopes whose atomic amounts are slightly different, the actual mass spectrometry spectrum has a form as shown in Fig.

Referring to FIG. 2, the mass spectrometry spectrum of the label material 110 includes a plurality of graphs 1101-1103 with slightly different mass to charge ratios, and the mass spectrometry spectrum of the label material 120 is such that the mass- Lt; RTI ID = 0.0 > 1201-1203. ≪ / RTI > Because the mass spectrometric spectrum is distributed over a certain area of the mass to charge ratio, it is generally preferable to use a substance having a mass difference of at least a certain level between the label materials 110 and 120, for example, 6 Da (Da) to be.

For example, a paper titled "Stable-Isotope Dimethyl Labeling for Quantitative Proteomics" (Jue-Liang Hsu et al., Anal. Chem., 2003, 75 (24), pp 6843-6852) It is possible to substitute one or more hydrogen atoms with deuterium ( ² H, or D (Deuterium)) in a label material in the form of a peptide consisting of nitrogen (N) and hydrogen (H) Lt; / RTI > At this time, as shown in FIG. 2, a labeling width is defined by a difference in mass between the used label materials. The narrower the labeling width, the greater the multiplexing of the quantitative measurement.

However, if the mass difference between the label materials is to be reduced, the graphs in which the mass to charge ratios are adjacent to each other may overlap, which serves as a limitation on the multiplicity of the quantitative measurement. 3A is a graph showing a superposition of spectra generated when a label material having a small mass difference is used, wherein three kinds of label materials having a mass difference of 2Da are used, and a sample includes three kinds of isotopes having a mass difference of 1Da .

3A shows a mass spectrometry spectrum corresponding to a different kind of label material, and graphs 301-303 show mass spectrometry spectra of a label made of a basic peptide material, and graphs 311- 313) shows a mass spectrometry spectrum of a label substance in which two hydrogen atoms are replaced with deuterium in the basic label substance, and a graph (321-323) shows a mass spectrometry spectrum of a label substance in which four hydrogen atoms are replaced with deuterium in the basic label substance . As shown, it can be seen that the two graphs 303, 311 and 313, 312, which correspond to the label material of which the mass to charge ratio (m / z) of each spectrum are adjacent to each other, are overlapped with each other.

In order to solve the above problem, analysis apparatuses having a resolution several tens of times higher may be used, but this is not efficient due to excessive cost, and the intensity of the measurement signal is reduced even if the resolution is increased.

Further, there is a problem that a label material produced by replacing hydrogen with deuterium has a slight shift in the retention time in the LC as compared with the material before replacement. FIG. 3B is a graph showing a retention time shift occurring when a label material in which hydrogen is replaced with deuterium is used. As shown in FIG. 3B, as hydrogen is replaced with deuterium, each label material slightly shifts in retention time in the LC, thereby failing to detect the mass spectrometry spectrum corresponding to the label substance The intensity of the signal of the mass spectrometry spectrum can be detected incorrectly.

For example, if the label material 30 having a specific mass to charge ratio is specified on the graph 302 of the mass spectrometry spectrum, the experimenter knows in advance the mass difference between the used label materials, so that the specified label material 30 ) To find the spectrum of the other label material in the region of the graph where the mass-to-charge ratio has increased by the corresponding mass difference. However, due to the deviation of the retention time, the graphs 311 to 313 are shifted in one direction on the axis corresponding to the retention time, as compared with the predicted area, so that the signal strength of the graphs 311 to 313 is inaccurately measured Or the detection of the graphs 311-313 may not be performed.

FIGS. 3c and 3d are graphs showing that spectra corresponding to different label materials are overlapped with each other with a retention time shift. FIG. 3c shows the case where the mass of the label materials increases by 2Da and four isotopes are included.

As shown in FIG. 3C, the retention time shifts between the graphs 301-304, 311-314, and 321-324 corresponding to different label materials, respectively. The graphs 303 and 304 and the graphs 311 and 312 corresponding to different materials are overlapped with each other and the graphs 313 and 314 and the graphs 321 and 322 corresponding to different materials are overlapped with each other . Graphs superimposed on each other are detected in the form of graphs 331 and 332 and graphs 333 and 334 shown in FIG. 3D, and it is impossible to specify a signal corresponding to each material through such graphs.

FIG. 3E is a conceptual diagram illustrating a quantitative analysis process using a conventional protein quantitation tool having the above limitations.

Referring to FIG. 3E, quadrangles 341-347 shown in phantom represent pairs of mass spectra of samples labeled with a labeling material having a mass difference, and rectangles 351 and 352 shown in solid lines Indicates the actually detected pair of spectra. As shown, many spectra except the quadrangle 351 have not found the spectral pair itself due to the shift of the retention time and the overlap of the graph. On the other hand, in the case of the quadrangle 352, due to the deviation of the retention time, it was erroneously detected to be paired with the spectra of other materials than the original spectral pair.

&Quot; Stable-Isotope Dimethyl Labeling for Quantitative Proteomics ", Jue-Liang Hsu et al., Anal. Chem., 2003, 75 (24), pp 6843-6852

According to an aspect of the present invention, there is provided a method of separating deuterium ( ² ) of hydrogen (H) from a label material in a liquid chromatograph-mass spectrometry (LC-MS) H, or D (Deuterium)) substitution, which can accurately discriminate signals of adjacent substances with masses, and a method and apparatus for quantitatively analyzing a sample, which can predict the shift of retention time caused by substitution , And a computer program for executing the above method.

The apparatus for quantitatively analyzing a sample according to an embodiment includes a sample containing a first substance on which a first label substance is attached and a second substance on which a second label substance having a different mass from the first label substance is attached A spectral recognition unit configured to receive a Liquid Chromatograph-Mass Spectrometry (LC-MS) spectrum; Extracting a Peptide Spectrum Match (PSM) corresponding to the first label material from the mass spectrometry spectrum received in the spectrum recognition unit, and using the PSM of the first label material, template of the second label material and calculating a predicted retention time (RT) shift of the first label material and the second label material, and calculating a predicted retention time And a PSM analyzer configured to specify a template and to perform PSM level quantitative analysis of the sample by reconstructing signals of the first material and the second material using the template of the first material and the second material.

The apparatus for quantitatively analyzing a sample according to an embodiment further includes a protein analyzing unit configured to perform protein level quantitative analysis of the sample using the result of the quantitative analysis of the peptide matching spectrum for one or more substances.

In one embodiment, the PSM analyzing unit includes an RT deviation estimator for calculating the predicted RT deviation. Wherein the RT deviation predicting unit comprises: a feature extracting unit configured to extract one or more predetermined features from the mass spectrometry spectrum; And a machine learning unit configured to apply the one or more features to a rule obtained through machine learning using a training set of a mass spectrometry as an input value.

In one embodiment, the at least one characteristic is selected from the group consisting of: the number of deuterium substituted on the second label material; the normalized RT of the mass spectrometry spectrum; the peptide sequence length of the first label material; The ratio of the labeled location of the labeling substance, or the normalized peak width of the mass spectrometry spectrum.

In one embodiment, the PSM analyzer may determine a template of the first material using a normal distribution curve obtained from the PSM corresponding to the first label material and a previously known isotope distribution profile of the elements constituting the first material. And an isotope separation section configured to detect the presence of the isotope.

In one embodiment, the PSM analyzing unit calculates a relative mass of the first material and the second material using a template of the first material and a template of the second material, thereby obtaining a liquid chromatograph mass spectrometry spectrum And a reconstruction unit configured to reconstruct the signal of the first material and the signal of the second material.

The quantitative analysis method according to one embodiment is characterized in that the quantitative analysis apparatus includes a sample containing a first substance to which a first label substance is attached and a second substance to which a second label substance having a different mass from the first label substance is attached 0.0 > LC-MS < / RTI > The quantitative analysis apparatus comprising: calculating a predicted RT deviation of the first label material and the second label material from the mass spectrometry spectrum; Extracting a peptide matching spectrum corresponding to the first label substance from the mass spectrometry spectrum; Identifying the template of the first substance using the peptide matching spectrum of the extracted first label substance; The quantitative analysis apparatus further comprising: specifying a template of the second material by using the template of the first material and the predicted RT deviation; The quantitative analysis apparatus comprising: reconstructing a signal of the first material and the second material using a template of the first material and the second material; And the quantitative analysis apparatus performing PSM level quantitative analysis of the sample using the reconstructed signals of the first material and the second material.

The quantitative analysis method according to an embodiment may further include the step of analyzing the protein level of the sample using the quantitative analysis results of the PSM level for one or more substances.

In one embodiment, the first label material is a peptide comprising carbon, nitrogen and hydrogen, and the second label material is selected from the group consisting of deuterium ( ² H, or D (Deuterium )). ≪ / RTI >

In one embodiment, the second label material comprises five or more peptides that increase in mass sequentially from the mass of the first label material depending on the number of deuterium substituted.

In one embodiment, the step of calculating the predicted RT deviation further comprises: extracting one or more predetermined features from the mass spectrometry spectrum; And calculating the predicted RT deviation by applying the at least one feature to a rule obtained through machine learning using a training set of a mass spectrometry spectrum as an input value.

In one embodiment, the at least one characteristic is selected from the group consisting of: the number of deuterium substituted on the second label material; the normalized RT of the mass spectrometry spectrum; the peptide sequence length of the first label material; The ratio of the labeled location of the labeling substance, or the normalized peak width of the mass spectrometry spectrum.

In one embodiment, the step of specifying the template of the first material comprises using a normal distribution curve obtained from the peptide matching spectrum corresponding to the first label material and a previously known isotope distribution of the elements constituting the first material And specifying a template of the first material.

In one embodiment, the step of reconstructing the signals of the first material and the second material comprises the steps of: using a template of the first material and a template of the second material, .

The computer program according to one embodiment is for executing a method of quantitative analysis of a sample according to the embodiments described above in combination with hardware and can be stored in a computer-readable medium.

According to an apparatus and method for quantitatively analyzing a sample according to an aspect of the present invention, in a label substance, deuterium ( ² ) of hydrogen (H) in a liquid chromatograph-mass spectrometry (LC- H, or D (Deuterium)) substitution can be efficiently predicted by applying machine learning, and as a result, deuterium atoms can be generated from the label material So that the number of kinds of usable label materials can be increased.

In addition, according to an apparatus and method for quantitative analysis of a sample according to an aspect of the present invention, the signal of each substance is predicted from a template specified using a peptide matching spectrum (PSM) It is possible to accurately separate signals of each material even if the mass difference is about 4 daltons (Da) or even 2 daltons (Da), for example. Therefore, expensive equipments for increasing the resolution There is an advantage that the multiplexing of the quantitative measurement can be improved without using it.

FIG. 1 is a conceptual diagram for explaining an MS1-based quantitation technique using a label based on a parent molecule.
2 is a graph of a mass spectrometry spectrum obtained using a label material containing isotopes.
FIG. 3A is a graph showing superposition of spectra that occurs when using label materials with small mass difference. FIG.
3B is a graph showing a retention time shift occurring when a label material in which hydrogen (H) is replaced with deuterium ( ² H, or D (Deuterium)).
3C and 3D are graphs showing that spectra corresponding to different materials are overlapped with each other with a retention time shift.
3E is a conceptual diagram illustrating a quantitative analysis process using a conventional protein quantitation tool.
4 is a block diagram of a quantitative analysis apparatus for a sample according to an embodiment.
5A is a graph showing the log-normal fitting for the mass spectrometry spectrum.
5B is a graph illustrating a process of specifying a mass spectrometry spectrum of another label material using the predicted retention time deviation.
6A to 6C are graphs illustrating a process of separating mass spectrometry spectra according to one embodiment and performing quantitation.
7 is a graph showing a normalization process of the retention time.
8 is a graph showing the normalized retention time.
9 is a graph showing the tendency of retention time shift in repeated experiments on the same label material as the same peptide.
10 is a graph showing the retention time deviation with respect to the number of deuterium substituted on the label material.
11 is a graph showing the retention time deviation with respect to the normalized retention time.
12 is a graph showing the retention time shift with respect to the peptide sequence length.
13 is a graph showing the retention time shift with respect to the ratio of amino acids indicated by labels.
14 is a graph showing the retention time with respect to the peak width of the mass spectrometry spectrum.
15 is a graph showing errors according to various regression models.
16A and 16B are diagrams for explaining support vector regression (SVR) in the regression model.
17A and 17B are graphs showing the performance when SVR is used as a regression model.
Figure 18 shows exemplary peptides that can be utilized as labeling material when using the quantitative assay method of a sample according to one embodiment.
FIGS. 19A to 19H are graphs showing quantitation results according to one embodiment according to the ratio of labeled proteins to theoretical values. FIG.
FIGS. 20A and 20B are graphs showing the results of quantitation according to one embodiment according to the ratio of the labeled proteins to the quantification results according to the prior art.
21 is a flowchart of a quantitative analysis method of a sample according to an embodiment.

The following detailed description of the invention refers to the accompanying drawings, which illustrate, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with an embodiment.

It is also to be understood that the position or arrangement of the individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained. In the drawings, like reference numerals refer to the same or similar functions throughout the several views.

Some embodiments of the present invention will now be described in more detail with reference to the drawings.

4 is a block diagram of a quantitative analysis apparatus for a sample according to an embodiment.

Referring to FIG. 4, the apparatus for quantitatively analyzing a sample according to the present embodiment includes a spectrum recognition unit 41 and a peptide spectrum match (PSM) analysis unit 42. In one embodiment, the apparatus for quantitatively analyzing a sample further includes a protein analyzer 43. The apparatus for quantitative analysis of a sample according to an embodiment may be entirely hardware, or partly hardware, and partly software. That is, the sample quantitative analysis device and each unit included therein may be collectively referred to as a device for storing data of a specific format and contents, or a device for sending and receiving data by an electronic communication method, and related software. The hardware may be a data processing device comprising a CPU or other processor. Also, the software driven by the hardware may refer to a running process, an object, an executable, a thread of execution, a program, and the like.

In this specification, each of the parts 41-43 constituting the quantitative analysis apparatus of a sample is not necessarily intended to refer to a separate component that is physically separated. That is, in FIG. 1, each part of the quantitative analysis apparatus of a sample is shown as a separate block that is distinguished from each other, but it is a functional classification of the quantitative analysis apparatus of the sample by the operation performed thereby. Depending on the embodiment, some or all of the parts described above may be integrated into one and the same device, or one or more parts may be implemented as separate devices physically separated from one another. For example, each component may be components that are communicatively coupled to one another in a distributed computing environment.

The spectrum recognizer 41 is configured to receive a Liquid Chromatograph-Mass Spectrometry (LC-MS) spectrum of a sample for which a shift of a retention time (RT) is predicted . At this time, the samples are represented by two or more label materials having different masses from each other. For example, the sample may comprise a first material to which a first label material is attached and a second material to which a second label material with a different mass from the first label material is attached. In this case, the first and second substances may be proteins, and the first and second labeled substances may be peptides having a predetermined composition.

In one embodiment, the spectrum recognition unit 41 recognizes the PSM corresponding to the peptide from the mass spectrometry spectrum. Recognition can be, but is not limited to, the MS-GF + method, which compares the MS / MS spectrum to the peptides of the protein sequence database. The recognized PSM corresponds not only to the peptide sequence but also to the label substance labeled thereon.

The PSM analyzing unit 42 performs an analysis on the PSM recognized by the spectrum recognizing unit 41 to analyze a substance (for example, a first substance) to which a specific label substance (for example, a first label substance) A probability density function having a log-normal distribution is obtained through log-normal fitting on the signal of the material, and the template of the material is determined by multiplying the obtained probability density function by the isotope distribution profile of the substance , The PSM predicts the RT deviation occurring in the signal of the recognized label material and the other label material (e.g., the second label material) having a mass difference, and determines the RT deviation based on the template of the first material and the predicted RT deviation of the above- The template of the substance to which the label substance is attached (for example, the second substance) is specified, and a PSM level quantitative analysis is performed to reconstruct a signal of each substance based on the specified templates.

Therefore, in the present specification, the term "template" refers to a signal shape of a material obtained at a mass to charge ratio (m / z) corresponding to each material, which is predetermined in consideration of an isotope distribution profile of the material. The amount of each substance in the sample can be calculated by the ratio of the signal of the actually measured mass spectrometry to the template of each substance.

Specifically, the signal obtained by LC-MS / MS has an intensity with respect to the mass-to-charge ratio (m / z) and the RT value, which corresponds to the corresponding RT of the eluted ion having the corresponding m / Lt; / RTI > The signal obtained for the particular ion eluted forms a peak in the longitudinal form along the RT axis at a defined interval on the RT axis at a particular m / z, the area under this peak being proportional to the abundance of that ion. The process of reconstructing the signal of the label material according to the embodiments of the present invention will now be described.

The charge state of the ion is the same, and the signal corresponding to the first substance is referred to as P ₁ and the signal corresponding to the second substance is referred to as P ₂ . At this time, it is assumed that P ₁ has m / z of 1001, an RT range of 3 to 6, a P ₂ of m / z of 1005, and an RT range of 1 to 4. In this case, if the peptide has three isotopes and the ratio thereof is 1: r ₁ : r ₂ , the template matrix, which is a cluster signal of the isotopes of the first substance, can be expressed by the following equation (1) The template matrix, which is a cluster signal of isotopes, can be expressed by the following equation (2).

[Equation 1]

&Quot; (2) "

The i-th row in each of the matrices of Equations 1 and 2 corresponds to the m / z value (i + 1000), and the j-th row corresponds to the RT. Also, x _{t in} Equation (1) and y _t in Equation (2) represent monoisotopic signal intensity at the corresponding m / z and RT when the total signal intensity of the target signal is 1, respectively. At this time, the entire signal including the signals of the first and second materials may be defined as Equation (3), and q ₁ and q ₂ represent the total intensity of the signal having the corresponding m / z.

&Quot; (3) "

The PSM analyzing unit 42 reconstructs the signal of the substance to which the label material is attached through the log regular fitting using the PSM of the label material as a probability density function having a lognormal distribution and calculates the isotope distribution of the substance in the probability density function The templates T ₁ and T ₂ of the above equation (3) can be obtained. In this specification, a probability density function having a log normal distribution is also referred to simply as a normal distribution curve.

5A is a graph showing log-normal fittings for the mass spectrometry spectrum.

Referring to FIG. 5A, it is assumed that a PSM to be analyzed consists of one or more graphs 500-506. The PSM analyzer 42 detects the apex of the signal intensity from the graph 500 in which the PSM corresponding to the label material is detected for lognormal fitting. The vertex is determined as the inflection point of the first generated signal intensity as a result of searching for the signal corresponding to the label material from the specified RT in the direction of increasing the signal intensity. By performing the lognormal fitting using the vertex, The component x _t of T ₁ in equation (3) and its RT range can be determined.

Specifically, an input value corresponding to the graph 500 is a vector (z ₀ ; z ₁ ;..., Z _p ) having an ion signal intensity value at each position on the RT axis, If p is the RT value corresponding to the peak of the intensity, the probability density function at position x on the RT axis can be defined as lnN (x; μ; σ ² ). Where mu is an arbitrary positive number as a position parameter, and sigma is a scale parameter. In this case, the input values are extracted between e corresponding to the smallest RT with e> p and signal strength ze ≤ zp / 10, and s corresponding to the largest RT with s <p and signal strength z _s ≥ z _p / 100 , And a function can be defined with respect to the RT value t between s and e as shown in Equation (4). This is a probability density function with a lognormal distribution starting at s and irregardless of the μ value and having apexes at p, and the shape of the curve is determined by the μ value.

&Quot; (4) &quot;

In this case, PSM analysis section 42 is going to change the value vector μ _{(z s; z s + 1} ; ...; z e -1; z e) and the vector _(μ f (z _s); f _μ ( _{z s + 1); ...;} f μ (z e-1); f μ (z e)) ( determine the cosine (cosine) values between shown) to 510-513, and the cosine of the largest Determine the value of μ. A cosine value is substituted for the largest value vector μ _{(f μ (z s);} f μ (z s + 1); ...; f μ (z e-1); f μ (z e)) , the graph Is output as a normal distribution function corresponding to the normal distribution 500.

On the other hand, log normal fitting can not be performed for the graphs 502 and 504 in which the label material is not detected in the above manner. For this reason, the PSM analyzing unit 42 includes the RT deviation predicting unit 421 to calculate a predicted value of the RT deviation between the first label material and the second label material, and calculates a normal distribution curve fitted to the first label material 510) is shifted by the predicted value of the RT deviation to reconstruct the signal of the second label material as a normal distribution curve. By this process, the component y _t and its RT range of T ₂ in equations (1) and (2) can be determined.

5B is a graph illustrating a process of specifying a mass spectrometry spectrum of another label material using the predicted RT deviation.

Referring to FIG. 5B, it is assumed that a normal distribution function 600 corresponding to the graph 500 is obtained by the PSM analyzer 42, and a signal corresponding to the label material is obtained from the graph 500. At this time, since the PSM analyzer 42 can know the mass-to-charge ratio (m / z) at which the signal of the other label material is obtained by using the mass difference between the label materials, the graph 502 , 504). In addition, the PSM analyzer 42 calculates an RT deviation corresponding to each of the graphs 502 and 504 based on one or more characteristics extracted to predict the RT deviation. Normal distribution curves 620 and 640 corresponding to the respective graphs 502 and 504 can be obtained from the graphs 502 and 504 using the calculated RT deviation according to the same fitting method as the preprocessing. The machine learning process for the PSM analysis unit 42 to predict the RT deviation will be described later with reference to FIG. 5 to FIG.

 6A to 6C are graphs illustrating a process of separating mass spectrometry spectra according to one embodiment and performing quantitation.

Referring to FIG. 6A, spectra are obtained using different label materials 110, 120, and 130 as described above with reference to FIG. 3D. As a result, graphs 301 and 302 corresponding to different label materials and graphs 323, 324, and a graph in the form of graphs 331-334 in which the signals of different label materials are superimposed, and the signal corresponding to the label material from the graph 301 is specified. At this time, the PSM analyzer 42 can determine the domain 610 having the m / z range and the RT range of the corresponding substance through the log normal fitting process for the signal of the label material. The m / z range is known from the information on isotopes, and the RT range is obtained through logarithmic fitting.

Next, the PSM analyzing unit 42 uses the RT deviation estimated value calculated by the RT deviation predicting unit 421 to calculate the intensity of the label material in which the signals are detected and the domains 620, 630, and may detect a signal from the corresponding domain 620, 630. At this time, the signals detected in each of the domains 620 and 630 still have a form in which signals of different peptides are overlapped due to isotopes.

6B, in order to solve the above problem, in the PSM analyzing unit 42, signals superimposed through templates reflecting the isotope distribution are converted into signals 701-704, 801-804, 901 corresponding to different peptides -904). Specifically, the isotope separation section 422 of the PSM analysis section 42 calculates the isotope distribution profile through the ion composition of the label-labeled substance. For example, if the ionic composition of C Speaking ₁₀ O _30, isotopic distribution of the carbon (C), known is represented as a vector of n-th component represents the bunporyang the n-th isotopes be represented as (0.9893, 0.0107) and , And oxygen (O) can be expressed in the same way (0.99757, 0.00038, 0.00205).

At this time, the ion composition of C ₁₀ O ₃₀ to calculate the isotopic distributions of the peptides, isotope separation unit 422 is reported to be the same as that of 40 the peptides virtual element α, the isotope distribution of the α The profile can be determined as shown in Equation (5) by adding up isotope distribution of carbon (C) and isotope distribution of oxygen (O) in the peptide proportional to each element ratio.

&Quot; (5) &quot;

Next, the ratio of the signal intensity of the nth isotope of the peptide can be determined by a Directed Acyclic Graph (DAG) algorithm using the above isotope distribution profile. Specifically, if there are two types of isotopes, each isotope is represented by edges e ₁ and e ₂ in the DAG algorithm, where e ₁ is the n-1th node for any n, n is to connect the second node, e ₂ is to connect the n-2-th node to the n-th node. Assuming a path consisting of l ₁ e ₁ and l ₂ e ₂ , the path strength can be expressed by a probability mass function of Equation (6).

&Quot; (6) &quot;

In Equation (6), m is the number of trials, and p ₁ , p ₂ , and p ₃ are probability distributions of the respective isotopes. When the result of Equation (5) Is defined by the following equation (7).

&Quot; (7) &quot;

Based on Equation (7), it is possible to collect all paths from the 0th node to the nth node by the path search algorithm of the DAG, where the signal intensity of the nth isotope is the sum of the strengths of all the paths collected as above . This intensity is normalized so that the zeroth century is one.

6C, the reconstructing unit 423 of the PSM analyzing unit 42 determines whether or not the RT deviation estimated by the RT deviation predicting unit 421 and the predicted value calculated by the isotope separating unit 422 by the above- The signal of each material is reconstructed using a template corresponding to each specified material reflecting the intensity distribution of the element. Specifically, the reconstruction unit 423 calculates weights q ₁ and q ₂ for each template that minimizes the error between the signal intensity measured based on the above-described Equation (3) and the signal intensity calculated by the prediction.

이라 하고 예측값을 적용한 템플릿 신호 행렬을

및

라 할 경우, 재구성부(423)는 하기 수학식 8이 최소가 되는 q₁ 및 q₂의 값을 계산할 수 있다. q₁ 및 q₂을 얻기 위한 계산 과정은 통상의 기술자에게 용이하게 이해될 수 있으므로 자세한 설명을 생략한다. For example, if the measured signal strength is

And the template signal matrix to which the predicted value is applied

And

, The reconstructing unit 423 can calculate the values of q ₁ and q ₂ that satisfy Equation (8) below. The calculation process for obtaining q ₁ and q ₂ can be easily understood by a person skilled in the art, and therefore, a detailed description thereof will be omitted.

&Quot; (8) &quot;

In one embodiment, the reconstruction unit 423 may consider noise due to other ions co-eluted together in the process of reconstructing the signal. If the matrix representing the shape and position of noise is C _j and the amount of noise is w _j , the noise-considered signal can be expressed as Equation (9). In Equation (9), L represents the number of label substances.

&Quot; (9) &quot;

The noise component of Equation (9) can be obtained by log-normal fitting using the peak of signal intensity in other parts of the signal except for the signal of the labeled peptide, and is normalized so that the maximum size is 1.

On the other hand, when the reconstructing unit 423 considers noise, a difference between the measured signal and the reconstructed signal can be defined according to Equation (10).

&Quot; (10) &quot;

At this time, since the peak signal intensity excluding the noise can be defined as the following Equation (11), the Signal-to-Noise Ratio (SNR) is defined as Equation (12). The reconstructing unit 423 can exclude from the target that the SNR of the PSM to be analyzed is less than a predetermined threshold value for the sake of accuracy. If there are a plurality of PSMs having the same domain, the reconstructing unit 423 may target PSMs having the lowest PSM and / or the lowest q-value of the spectrum E-value.

&Quot; (11) &quot;

&Quot; (12) &quot;

The protein analyzing unit 43 performs analysis at the protein level using the relative quantities (e.g., q ₁ and q ₂ ) of each labeled peptide obtained by the PSM analyzing unit 42. One PSM may be uniquely matched to a particular protein, or it may be unclear as to which protein it is from as a sequence shared by several proteins. In order to solve this uncertainty, the protein analyzer 43 selects the PSM based on the assumption that the protein ratio of the sample and the PSM ratio matching therewith are constant.

Specifically, the PSM and the protein are represented by the node sets U and V of the bipartite graph, the edges between the nodes are denoted by E, and the graphs G (U, V, E) The edge (u, v) between the node v to which it belongs is assumed to be connected when the PSM and protein are matched. At this time, the signal strength of the node u is given by the quantitative analysis of the PSM analysis unit 42 as shown in the following equation (13).

&Quot; (13) &quot;

Let u (v) denote the set of PSM nodes connected to the protein node v and let Q ^v denote the sum of the vectors q ^u belonging to U (v). If the ratio of the proteins and the corresponding PSM ratios are constant, q ^u And Q ^v should have a similar direction. The protein analyzing unit 43 selects the edges so as to show the similarity of such directions. The protein analyzing unit 43 analyzes the vector q ^u and the vector Q ^v Calculate the cosine value between q and ^u , which corresponds to the degree of coincidence of q ^u and Q ^v .

The protein analyzing unit 43 updates U (v) by removing the edges whose cosine value calculated as above is less than a preset threshold value (for example, 0.8) for U (v) having a size of 1 or more, ^v . The above procedure is repeated for all protein nodes v, so that it is possible to exclude nodes corresponding to an inaccurate or singular value PSM. The above process can be termed per-protein-pruning.

Next, the protein analyzing unit 43 analyzes each PSM node u. When the node u is connected to a plurality of protein nodes v, the corresponding PSM is shared by various proteins. At this time, the protein analyzing unit 43 calculates the vector q ^u and the vector Q ^v Leaving only the edge at which the cosine value between them becomes maximum. The above procedure is performed for all nodes u. The above process can be termed per-peptide-pruning.

The above-described pruning process for each protein or each peptide can be repeatedly performed multiple times (for example, 10 times) alternately. After this process, a protein node with a number of matched PSMs less than a preset threshold (e.g., 2) may be excluded. The vector Q ^v containing the remaining protein node v is the result of quantitative analysis of the final determined protein level. Through the above process, it is possible to obtain information on the proportion of each component of the protein used as the sample.

Hereinafter, a process of determining at least one characteristic having a high correlation with the RT deviation through machine learning in order to predict the RT deviation among the concrete operations of the RT deviation prediction unit 421 of the PSM analysis unit 42 will be described .

7 is a graph showing the RT normalization process.

7, an LC-MS / MS spectrum of a target material is obtained by a plurality of measurements using a plurality of label materials having different masses and an absolute value of RT measured therefrom It can be expressed as the occurrence frequency for each time between 0 and 125 minutes. In this case, if the time interval between the RT ₂₀ corresponding to the 20 th RT and the RT ₈₀ corresponding to the 80 th RT from the smallest absolute value to the largest is defined as an RT span, the normalized RT value is expressed as [(RT- RT ₂₀ ) / RT span].

8 is a graph showing the normalized RT.

8, RT ₂₀ and RT span were 27 min and 62 min, respectively, and RT at the signal point of the label material shown in dot form was 94 min and the normalized RT value was 1.08. The peak width of the signal intensity graph for RT was 1 minute, and the normalized peak width was 0.01613 when normalized by RT span. The RT deviation between the label materials was 0.4 minutes, which was normalized by dividing by RT span and was 0.0065.

Figure 9 is a graph showing the tendency of RT deviation in repeated experiments on the same labeling substance as the same peptide.

9, the correlation coefficient of each test result is 0.803690, and the root-mean-square error (RMSE) is 0.001312, which is less than 1/10 of the peak width. This implies that the RT deviation has a certain tendency in spite of repeated execution, suggesting the possibility of predicting the RT bias of the other samples from the RT deviation for one sample.

The present inventors paid attention to the above points and conducted various experiments in order to extract features affecting the RT deviation based on the normalized RT. We tested various features with a training set of 22,000 PSMs (also referred to as mass spectrometry spectra) obtained by three LC-MS / MS runs and selected features with a linear correlation with the RT deviation Respectively. In one embodiment, the selected feature comprises the number of deuterium (D) substituted for the other labeling material in contrast to the peptide corresponding to the base labeling material, and the normalized RT of the ion signal corresponding to the target material.

In one embodiment, the selected feature further comprises a peptide sequence length of the target material, a ratio of the labeled location in the target material, and / or a normalized RT width of the ion signal corresponding to the target material. The ratio of the position indicated by the label refers to the value obtained by dividing the number of positions at which the labeling substance molecules are attached in the peptide divided by the length of the peptide. The molecule of the label substance may be an amino acid such as lysine, but is not limited thereto, and other amino acids such as arginine or tyrosine may be used.

10 is a graph showing the number of deuterium substituted for labeled substance as one of the features for machine learning of the RT deviation. As shown in FIG. 10, the normalized RT deviation value appeared to be related to the number of deuterium (D) substituted in the label material, and the distance correlation was the highest at 0.57. Correlation distances are used herein to refer to similarity across linear and / or non-linear associations between selected features and RT shifts, as described in Gabor J. Szekely et al., &Quot; Measuring and testing dependence by correlation of distances & Annals of Statistics, 35 (6): 2769-2794).

Figure 11 is a graph showing the RT deviation versus normalized RT as one of the other features for machine learning. Referring to FIG. 11, the RT deviation has a correlation with the normalized RT value, and the RT deviation tends to decrease at the beginning and the end of the liquid chromatography. To be visually conspicuous, the number of deuterium substituted in the label material was set to 20, which is also the same in Figs. 12 to 15 described later. The correlation distance between the normalized RT value and the RT deviation was 0.33.

Figure 12 is a graph showing the RT deviation versus peptide sequence length as one of the other features for machine learning. Referring to FIG. 12, the longer the length of the peptide corresponding to the target substance, the smaller the RT deviation decreases. It can be understood that the effect of deuterium (D) substitution is relatively small when the peptide is long, and the correlation distance between the peptide sequence length and the RT shift is 0.43.

Figure 13 is a graph showing the RT deviation versus amino acid (lysine) ratio labeled as one of the other features for machine learning. Referring to FIG. 13, the higher the ratio of lysine was, the more the RT deviation increased. The correlation distance between the lysine ratio and the RT deviation was 0.40.

14 is a graph showing the RT deviation as one of the features for machine learning with respect to the peak width of the mass spectrometry spectrum. In this specification, the peak width is defined as follows. In a normal distribution curve corresponding to each mass-to-charge ratio (m / z), in a specific RT time interval, the signal intensity is 1/100 or more of the signal intensity maximum value of the corresponding normal distribution curve, When the intensity is less than 1/100 of the maximum value described above, the specific RT time interval is defined as a peak width. The correlation distance between the peak width and the RT deviation was 0.25.

The RT deviation predicting section 421 calculates a predicted value of the RT deviation by performing a machine learning algorithm using each characteristic having a correlation with the RT deviation as an input value, as shown in FIGS. 10 to 14. FIG.

Specifically, from the training set of the mass spectrometry spectrum in which the protein corresponding to the target substance, the kind of the peptide used as the label, and the mass difference of the isotope of the target substance are known, 4211 extracts the respective feature values described above. Next, the machine learning unit 4212 of the RT deviation predicting unit 421 defines the relationship between each feature value and the RT deviation by a machine learning algorithm in which each feature value and the RT deviation measured from the training set are input values Generate a rule and apply these rules to the sample, not to the training set, to calculate the predicted value of the RT deviation occurring in the mass spectral spectrum of the sample.

In the embodiments, the types of rules that define the relationship between the feature values and the RT deviation may differ depending on the machine learning algorithm and can not be collectively defined. In embodiments, the machine learning unit 4212 may include a linear regression algorithm, a K Nearest Neighbor (KNN) algorithm, a Radius Neighbors algorithm, a Neural Network algorithm, A Decision Tree algorithm, a kernel ridge regression (KRR) algorithm, a support vector regression (SVR) algorithm, or other suitable machine learning algorithms.

In applying the algorithm, each feature value can be applied in a rescaled or standardized form. In the case of the rescale, each feature value is converted into a rescaled feature value x 'according to the following equation (14) so that the minimum value is -1 and the maximum value is 1 when the feature value is x.

&Quot; (14) &quot;

In the case of normalization, each characteristic value is converted into a standardized characteristic value x 'according to Equation (15) such that the average of the characteristic values becomes 0 and the standard deviation becomes 1. In Equation 15,? Denotes the standard deviation of the population before the standardization.

&Quot; (15) &quot;

15 is a graph showing the error according to various regression models. Referring to FIG. 15, machine learning is performed using all of the five features described above with reference to FIGS. 10 to 14, and in each of the regression models, a 10-fold cross validation The RMSE between the RT deviation value and the actually measured RT deviation value was calculated. The SVR algorithm with standardized feature values showed the best RMSE value of 0.0016.

In the case of the SVR algorithm, assuming that the data source value has the variables x and y, only the limitation that all the data should fall within the tolerance range (for example, denoted by?) Is ignored, The model is created in the form of y = wx + b. Then, by calculating the values of w and b that define y = wx + b such that all the data belongs to the error range and the width (m) between the data closest to the model value is the maximum, Linear combination form. For example, the model as shown in FIG. 16A is a good explanation of the relationship between the variables, whereas the model as shown in FIG. 16B determines that the relationship between the variables is not well explained. In the present embodiment, the relation between the feature values and the RT deviation can be defined and machine learning can be performed.

Meanwhile, in one embodiment, the machine learning may be performed through the SVR algorithm to which a kernel function is applied in order to linearly approximate the relationship between nonlinear variables. When a kernel function called Gaussian Radial Basis Function is applied to a data source and machine learning is performed by the SVR algorithm, the dimension of the data can be increased, and the complex relationship between the variables can be included in the model .

Since the specific operation of the SVR algorithm can be easily understood by those skilled in the art, detailed description thereof will be omitted in order to clarify the gist of the present invention.

17A and 17B are graphs showing the performance when SVR is used as a regression model. Referring to FIG. 17A, when the normalized peak width average is 0.016, the RMSE value of the estimated RT deviation is only 0.0016 corresponding to 1/10 of the estimated RT deviation, so that the accuracy of the prediction for the RT deviation can be confirmed. Also, referring to FIG. 17B, the correlation coefficient between the predicted RT deviation and the actually measured RT deviation was 0.76242, indicating a high correlation.

Further, the present inventors compared the results of RT deviation prediction based on the results of the machine learning in the case of using the five characteristics individually and the case of using all of the five characteristics described above with reference to Figs. 10 to 14, Table 1 shows. The deviations in the RMSE were relatively small when used in combination with the individual use of the five features as shown in Table 1. On the other hand, correlation coefficient showed a high correlation with the results of machine learning using all the features together.

Forecast type Features Used RMSE Correlation coefficient Single Feature Usage Number of deuterium (D) 2.2 x 10 ^-3 0.47 Normalized RT 2.4 × 10 ^-3 0.24 Peptide sequence length 2.4 × 10 ^-3 0.21 Percent of labeled amino acids 2.3 x 10 ^-3 0.33 Peak width 2.4 × 10 ^-3 0.32 Use all features All of the above features 1.6 x 10 ^-3 0.77

The results shown in Table 1 show that the value of the RT deviation in the mass spectrometry spectrum is relatively small in comparison with the total width of the spectrum and that in the mass spectrometry spectrum in which the RT deviation is largely shown, This means that it is possible to accurately predict the RT deviation through machine learning.

According to the quantitative analysis apparatus for samples according to the embodiments of the present invention described above, the RT deviation generated by the substitution of deuterium ( ² H, or D (Deuterium)) of hydrogen (H) So that it can be predicted efficiently. It is also possible to accurately separate the signals of each labeling material, even if the mass difference is, for example, about 4 daltons (Da) or even 2 daltons (Da) It is possible to improve the multiplexing of the quantitative measurement without using it.

Figure 18 shows exemplary peptides that can be utilized as labeling material when using the quantitative assay method of a sample according to one embodiment.

Referring to FIG. 18, peptides having the structure of H ₂ N-peptide-K-NH ₂ are reacted with sodium cyanoborohydride (NaBH ₃ CN) and aldehyde as shown at the top, (Da) from the basic label substance by replacing some hydrogen (H) atoms with deuterium (D) in the reactive substance as a basic label substance with a peptide structure having an ethyl group bonded thereto Five additional peptide structures for labeling substances can be obtained. Using the peptides shown in FIG. 18, six-plexing can be performed using six label materials at a time. However, by way of example, it is also possible to measure using 8 or 10 label substances by increasing the kinds of peptides by changing the reactants. The number of available peptides can be one or more arbitrary numbers, It is not limited.

FIGS. 19A to 19H are graphs showing quantitation results according to one embodiment according to the ratio of labeled proteins to theoretical values. FIG.

19A to 19H show quantitative analysis results of a sample using six kinds of peptide constructs as shown in FIG. 18 as a labeling substance. In FIG. 19A, the ratio of the protein to which each labeling substance was attached was 1: 1: 1: 19: 1: 1.2: 1: 1.2: 1: 1.2, and the ratio is 1: 4: 1: 4: 1: 4 in FIG. 19: 1: 8: 1: 8: 1: 8, and the ratio is 1: 10: 1: 10: 19H, the ratio is 1: 50: 1: 50: 1: 50.

Each graph shows the relative ratio of the other label material signals to the signal of label material No. 2 in terms of a log 2 fold change and the dashed line represents the theoretical signal position. As shown in the figure, the quantitative analysis results of the present example follow the theoretical values even when the ratios of the labeled proteins of the respective label materials vary widely.

20A and 20B are graphs showing the results of quantitation according to one embodiment according to the ratios of the labeled proteins to quantitative results according to the prior art.

20A and 20B, MaxQuant, which is an LC-MS / MS spectrum-based quantitation tool, is used. Simultaneously measuring up to three types of label materials is possible in Maxquant. Therefore, Three types of peptides were used as the labeling substance, one mass of which increased by 4 daltons (Da). No overlap of signals due to isotopes was made. 20A and 20B, each graph represents the relative ratio of the other label material signals to the signal of the label material No. 2 by the log 2 fold change, and the dotted line represents the theoretical signal position .

FIG. 20A is a result of quantitative analysis in the case where the ratio of labeled proteins of three kinds of label materials is 1: 1: 1, and FIG. 20B is a result of quantitative analysis in the case where the ratio is 1:10:20. As shown, in the case of the conventional technology, a relatively accurate result is obtained when the protein ratio is 1: 1: 1, but the error increases from the theoretical value while the ratio increases to 1:10:20, , It can be seen that the quantitative analysis follows the theoretical value even when the ratio increases.

21 is a flowchart of a quantitative analysis method of a sample according to an embodiment.

Referring to FIG. 21, an LC-MS / MS spectrum of a sample is received (S1), and an analysis target spectrum among the received spectra can be specified (S2). The process of specifying the spectrum to be analyzed includes a process of recognizing PSM from the spectrum, for example, but not limited to, the MS-GF + method in which the MS / MS spectrum is compared with the peptide in the protein sequence database.

Next, a signal corresponding to the first label substance is extracted from the recognized PSM (S3), and a template corresponding to the signal of the first substance labeled with the label substance is obtained through log normal fitting using the extracted signal (S4). At this time, the template may reflect the known isotope distribution profile of the first material. That is, the template of the corresponding substance can be obtained by multiplying the fitted normal distribution function for the recognized PSM by the isotope distribution profile of the substance.

Further, based on the known mass difference of the first label material and the second label material, a predicted value of the RT deviation between the first label material and the second label material is calculated. The step of calculating the predicted value includes extracting at least one feature for the machine learning from the mass spectral analysis (S5), and applying the extracted feature to the machine learning algorithm such as SVR to predict the RT deviation (S6) .

Next, using the predicted values of the reconstructed PSM and RT deviation of the first material, the template of the second substance labeled with the second label material can be specified (S7).

Next, the analysis of reconstructing the signals of each of the first and second substances using the template of the identified first substance and the template of the second substance can be performed to derive a PSM level quantitative analysis result on the sample (S8). The process may include calculating a relative amount of the first material and the second material by comparing the measured signal with each template. Further, analysis results for one or more PSMs matched to the same protein may be collected to perform protein level quantitative analysis of the sample (S9).

The operation of the quantitative analysis method of the sample according to the embodiments described above can be at least partially implemented in a computer program and recorded in a computer-readable recording medium. A program for implementing the operation of the quantitative analysis method of the sample according to the embodiments is recorded, and the computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like. The computer readable recording medium may also be distributed over a networked computer system so that computer readable code is stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present embodiment may be easily understood by those skilled in the art to which this embodiment belongs.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims. You will understand.

Claims (17)

A spectral recognition unit configured to receive a liquid chromatograph mass spectrometry spectrum of a sample including a first substance to which a first label substance is attached and a second substance to which a second label substance having a different mass from the first label substance is attached; And
Extracting a peptide matching spectrum corresponding to the first labeling substance from the mass spectrometry spectrum received in the spectrum recognizing unit, specifying a template of the first substance using the peptide matching spectrum of the first labeling substance, The method of claim 1, further comprising: calculating a predicted retention time deviation of the first label material and the second label material, specifying a template of the second material using the predicted retention time deviation, And a peptide matching spectrum analyzer configured to perform quantitative analysis of a peptide matching spectrum level of the sample by reconstructing a signal of the first substance and the second substance using a template.
The method according to claim 1,
And a protein analyzing unit configured to perform protein level quantitative analysis of the sample using the peptide matching spectral level quantitative analysis results for one or more substances.
The method according to claim 1,
Wherein the first label material is a peptide comprising carbon, nitrogen and hydrogen,
Wherein the second label material is a peptide in which one or more hydrogens of the first label material are replaced by deuterium.
The method of claim 3,
Wherein the second label material comprises five or more peptides that increase in mass sequentially from the mass of the first label material depending on the number of deuterium substituted.
The method according to claim 1,
Wherein the peptide matching spectrum analyzing unit includes a retention time deviation predicting unit for calculating the predicted retention time deviation,
The retention time shift predicting unit predicts,
A feature extraction unit configured to extract one or more features predetermined from the mass spectrometry spectrum; And
And a machine learning unit configured to apply the one or more features to a rule obtained through machine learning using a training set of a mass spectrometry spectrum as an input value.
6. The method of claim 5,
Wherein said at least one characteristic comprises at least one of the number of deuterium substituted on said second label material, the normalized retention time of said mass spectrometry spectrum, the peptide sequence length of said first label material, A ratio of labeled sites or a normalized peak width of said mass spectrometry spectrum.
6. The method of claim 5,
The peptide matching spectrum analyzing unit may determine a template of the first substance by using a normal distribution curve obtained from the peptide matching spectrum corresponding to the first label substance and a previously known isotope distribution profile of the elements constituting the first substance And a quantitative analysis device for the sample including the constituted isotope separation section.
6. The method of claim 5,
Wherein the peptide matching spectrum analyzing unit analyzes the liquid chromatography mass spectrometry spectrum of the sample by calculating the relative mass of the first substance and the second substance using the template of the first substance and the template of the second substance, And a reconstruction unit configured to reconstruct a signal of the first material and a signal of the second material.
A quantitative analysis apparatus is provided for receiving a liquid chromatograph mass spectrometry spectrum of a sample comprising a first substance to which a first label substance is attached and a second substance to which a second label substance having a different mass from the first label substance is attached step;
The quantitative analysis apparatus comprising: calculating a predicted retention time deviation of the first label material and the second label material from the mass spectrometry spectrum;
Extracting a peptide matching spectrum corresponding to the first label substance from the mass spectrometry spectrum;
Identifying the template of the first substance using the peptide matching spectrum of the extracted first label substance;
The quantitative analysis apparatus comprising: specifying a template of the second material by using the template of the first material and the predicted retention time deviation;
The quantitative analysis apparatus comprising: reconstructing a signal of the first material and the second material using a template of the first material and the second material; And
Wherein the quantitative analysis apparatus comprises performing quantitative analysis of the peptide matching spectrum level of the sample using the reconstructed signals of the first substance and the second substance.
10. The method of claim 9,
Wherein the quantitative analysis apparatus further comprises analyzing the protein level of the sample using the result of the quantitative analysis of the peptide matching spectrum for one or more substances.
10. The method of claim 9,
Wherein the first label material is a peptide comprising carbon, nitrogen and hydrogen,
Wherein the second label material is a peptide in which one or more hydrogens of the first label material are replaced by deuterium.
12. The method of claim 11,
Wherein the second label material comprises five or more peptides that increase in mass sequentially from the mass of the first label material depending on the number of deuterium substituted.
10. The method of claim 9,
Wherein the step of calculating the predicted retention time deviation comprises:
Extracting one or more predetermined features from the mass spectrometry spectrum; And
Calculating the predicted retention time deviation by applying the one or more features to a rule obtained through machine learning using a training set of a mass spectrometry spectrum as an input value.
14. The method of claim 13,
Wherein said at least one characteristic comprises at least one of the number of deuterium substituted on said second label material, the normalized retention time of said mass spectrometry spectrum, the peptide sequence length of said first label material, A ratio of labeled sites or a normalized peak width of said mass spectrometry spectrum.
10. The method of claim 9,
Wherein the step of specifying the template of the first material comprises the steps of: using a normal distribution curve obtained from the peptide matching spectrum corresponding to the first label material and a previously known isotope distribution profile of the elements constituting the first material, The method comprising the steps of:
10. The method of claim 9,
Wherein reconfiguring the signals of the first material and the second material comprises calculating a relative amount of the first material and the second material using a template of the first material and a template of the second material A method for quantitative analysis of a sample.
17. A computer program stored in a medium, coupled to hardware, for executing a method of quantitative analysis of a sample according to any one of claims 9 to 16.

Images