Sequence analyzing method, sequence analyzing device, polymerization condition proposing device, and automatic synthesizing device

Abstract

Provided is a simple sequence analyzing method for macromolecules, for estimating the content, in a polymer, of polyads configured from an arrangement of a plurality of units, the method including: determining a number K of polyad variants; successively ionizing gas components generated by heating each of a reference sample and an estimation target sample, to obtain a data matrix including a two-dimensional mass spectrum; performing nonnegative matrix factorization of the data matrix to decompose the data matrix into a product of a basis spectra matrix and an intensity distribution matrix; performing nonnegative matrix factorization of the intensity distribution matrix to decompose the intensity distribution matrix into a product of a matrix representing mass fractions of model polymers consisting only of polyads and a matrix representing feature vectors; setting a K-1 dimensional simple substance encompassing all the feature vectors of the sample, with the feature vectors of the model polymers as end members; and estimating a content ratio of each polyad from a ratio of distances between the end members and the feature vectors of the estimation target sample.

Description

Sequence analysis method, sequence analysis device, polymerization condition proposing device, and automatic synthesis device

[0001] The present invention relates to a sequence analysis method, a sequence analysis device, a polymerization condition proposing device, and an automated synthesizer.

[0002] As is widely known in proteins, nucleic acid molecules, and the like, the physical properties of polymers are significantly affected by the sequence of units derived from monomers. Numerous sequence analysis methods have been proposed for biopolymers such as proteins and nucleic acid molecules. Patent Document 1 describes one such technique: "A sequencing method for determining the sequence of multiple monomers constituting a biopolymer by measuring a tunneling current flowing between a pair of electrodes, the sequencing method comprising: a) measuring a current value between the electrodes at predetermined time intervals to obtain current value data; b) selecting an analysis region including a signal-dense region from the current value data; c) determining, in the analysis region, a minimum representative value having the smallest current value among multiple representative values ​​and a maximum representative value among the multiple representative values; and d) identifying the type of monomer corresponding to the signal based on the minimum representative value and the maximum representative value."

[0003] International Publication No. 2019 / 208225

[0004] As described above, although sequence analysis methods for biopolymers and the like have been established, sequence analysis methods have not yet been established for many industrially used polymers. Therefore, an object of the present invention is to provide a simple method for analyzing the sequence of polymers. Another object of the present invention is to provide a sequence analysis device, a polymerization condition proposal device, and an automated synthesis device.

[0005] As a result of extensive research into solving the above problems, the present inventors have found that the above problems can be solved by the following configuration.

[0006] [1] A polymer sequence analysis method for estimating the content of multipliers, which are formed by arranging a plurality of units derived from the monomer, in a polymer obtained by polymerizing a monomer selected from a monomer set including two or more types of monomers, the method comprising the steps of: determining a number of variants K of the multipliers according to the number of types of monomers included in the monomer set and the number of the units constituting the multipliers; heating a reference specimen, which is a polymer formed from the monomers, and a specimen to be estimated, sequentially ionizing the gas components produced, to obtain a data matrix including a two-dimensional mass spectrum of m / z versus heating temperature; and performing nonnegative matrix factorization on the data matrix, thereby decomposing the data matrix into a product of a matrix representing a normalized basis spectrum and its intensity distribution matrix. a second NMF process for decomposing the intensity distribution matrix of each of the specimens into a product of a matrix representing the mass fraction of a model polymer composed only of the multiplicands in the specimen and a matrix representing a feature vector of the model polymer; defining a K-1-dimensional simplex containing all of the feature vectors of the specimen, with the feature vectors of the model polymer being endmembers; defining distances between the K endmembers and the feature vectors of the specimen to be estimated by a Riemannian metric distance that takes into account non-orthogonality of the basis spectra of the first NMF process; and estimating the content ratio of each of the multiplicands in the specimen to be estimated from the ratio of the distances. [2] The sequence analysis method according to [1], wherein, when the number of variants K is 3 or more, at least one of the feature vectors of the reference specimen is located in each of the outer regions of a hypersphere inscribing the K-1-dimensional simplex, or the reference specimen includes at least one of the endmembers. [3] The sequence analysis method according to [1] or [2], further comprising, after the second NMF processing, reconstructing the spectrum of the model polymer by a matrix product of a matrix representing the feature vector of the model polymer and a matrix representing the basis spectrum, and identifying the polyadduct to which the spectrum of the model polymer belongs. [4] The sequence analysis method according to [3], wherein the identification is performed by comparing the sum of the mass numbers of the units constituting the polyadduct with the m / z of a peak in the spectrum of the model polymer.[5] The sequence analysis method according to [3] or [4], wherein, if the identification results in a spectrum of the model polymer that cannot be assigned to any of the multi-adducts, the number of variants K is changed and the first NMF process, the second NMF process, and the identification are repeated. [6] The sequence analysis method according to [5], wherein the change is to subtract a predetermined number from the number of variants K. [7] The sequence analysis method according to [3] or [4], wherein, if the identification results in a spectrum of the model polymer that cannot be assigned to any of the multi-adducts, the reference sample is added and the acquisition of the data matrix, the first NMF process, the second NMF process, and the identification are repeated. [8] When the number of types is j and j is 3 or greater, the number of variants K is expressed by the formula: K= ｊ C ３ +3 ｊ C ２ + ｊ C １[9] The sequence analysis method according to any one of [1] to [7], wherein the polymer comprises a resist resin.

[10] The sequence analysis method according to any one of [1] to [9], wherein the reference sample and the suspected target sample are polymerized by different methods.

[11] A polymer sequence analysis device for estimating the content of multiplets formed by arranging a plurality of units derived from the monomer in a polymer obtained by polymerizing a monomer selected from a monomer set including two or more types of monomers, the device comprising: a mass spectrometer for sequentially ionizing gas components generated by heating a reference specimen, which is a polymer composed of the monomers, and a specimen composed of an estimation target specimen, and continuously observing mass spectra; and an information processing device for processing the observed mass spectra, the information processing device comprising: a data matrix creation unit for obtaining a data matrix including a two-dimensional mass spectrum of m / z versus heating temperature; a variant number determination unit for determining the number of variants K of the multiplets in accordance with the number of types of monomers included in the monomer set and the number of the units constituting the multiplets; and a matrix factorization unit for performing non-negative matrix factorization of the data matrix to obtain a matrix representing a normalized basis spectrum. a first NMF processing unit that performs NMF processing to decompose the intensity distribution matrix of each of the specimens into a product of a matrix representing a mass fraction of a model polymer composed only of the multiplicands in the specimen and a matrix representing a feature vector of the model polymer, thereby obtaining a feature vector of the model polymer; a vector projection unit that sets a K-1-dimensional simplex containing all of the feature vectors of the specimen, with the feature vectors of the model polymer as end members; and a composition estimation unit that defines distances between the K end members and the feature vector of the specimen to be estimated by a Riemannian metric distance that takes into account non-orthogonality of the basis spectra of the first NMF processing, and estimates a content ratio of each of the multiplicands in the specimen to be estimated from the ratio of the distances.

[12] The sequence analysis device according to

[11] , further comprising a model polymer spectrum identification unit that reconstructs the spectrum of the model polymer by a matrix product of a matrix representing the feature vector of the model polymer and a matrix representing the basis spectrum, and identifies the polyadduct to which the spectrum of the model polymer belongs.

[13] The sequence analysis device according to

[12] , wherein the identification is performed by comparing the sum of the mass numbers of the units constituting the polyadduct with the m / z of a peak in the spectrum of the model polymer.

[14] If, as a result of the identification, there is a spectrum of the model polymer that cannot be assigned to any of the polyadducts, the information processing device according to

[12] or

[13] changes the number of variants K, and repeats the NMF processing by the first NMF processing unit, the NMF processing by the second NMF processing unit, and the identification by the model polymer spectrum identification unit.

[15] A polymerization condition proposal device further comprising the sequence analysis device according to any one of

[11] to

[14] , and a policy proposal unit that has been machine-learned using the sequence analysis results from the sequence analysis device and the polymerization conditions of the suspected target specimen as training data, wherein the policy proposal unit compares the sequence analysis results with a predetermined target sequence and proposes new polymerization conditions for obtaining a polymer of the target sequence.

[16] An automated synthesis device comprising the polymerization condition proposal device according to

[15] , and an apparatus for synthesizing the polymer, wherein the apparatus has a monomer supply mechanism, a reaction vessel that receives the monomer supplied from the supply mechanism and reacts the monomer, and a control device, and the control device controls at least one selected from the group consisting of the supply mechanism and the reaction vessel based on the polymerization conditions proposed by the polymerization condition proposal device, to synthesize a new polymer.

[0007] According to the present invention, a simple method for analyzing the sequence of a polymer can be provided. Furthermore, according to the present invention, a sequence analysis device, a polymerization condition proposing device, and an automatic synthesizer can be provided.

[0008] 1 is a flowchart of an embodiment of a sequence analysis method of the present invention. A data matrix was acquired from a sample synthesized using methyl methacrylate (M) and styrene (S) as monomers, and an M spectrum was calculated from each matrix obtained by the first NMF process and the second NMF process. A conceptual diagram showing a K-1-dimensional simplex (two-dimensional simplex, triangle) when K is 3. A conceptual diagram showing a K-1-dimensional simplex (two-dimensional simplex, triangle) when K is 3. A hardware configuration diagram of an embodiment of a sequence analysis device of the present invention. A functional block diagram of a second embodiment of a sequence analysis device. A functional block diagram of an embodiment of a polymerization condition proposal device of the present invention. A functional block diagram of an embodiment of an automated synthesis device of the present invention. A model polymer spectrum for each multi-unit obtained by calculation. A diagram showing the calculation results of a model polymer spectrum when the number of types of monomers is 2 and the multi-unit length is 5. A diagram showing the relationship between polymerization time and conversion rate. These are sequence analysis results, showing the relationship between the mass-based content of BBB (A) and the mass-based content of BBS (B) versus conversion rate. FIG. 1 is a diagram showing the results of sequence analysis and the relationship between the mass content of BBS and the conversion rate.

[0009] The present invention will be described in detail below. The following description of the constituent elements may be based on a representative embodiment of the present invention, but the present invention is not limited to such an embodiment. In this specification, a numerical range expressed using "to" means a range that includes the numerical values ​​before and after "to" as the lower and upper limits.

[0010] The embodiment shown below is an example that embodies the technical concept of the present invention, and the technical concept of the present invention does not limit the materials, shapes, structures, and arrangements of the components to the following embodiment. The drawings are schematic. Therefore, the relationship and ratio between thickness and planar dimensions may differ from the actual ones, and the relationship and ratio between dimensions may also differ between the drawings.

[0011] [Definition of Terms] The terms used in this specification are explained below. Terms not explained below are used in the sense that they are commonly understood by those skilled in the art.

[0012] As used herein, the term "monomer" refers to a compound (monomer) used in the synthesis of a polymer, which is a specimen. The "reference specimen" and "presumed specimen" described below are both polymers. A polymer is synthesized from one or more monomers selected from a monomer set consisting of a predetermined number of monomers.

[0013] In this specification, the term "unit" refers to a part of the structure of a polymer that is derived from a monomer. For example, vinyl chloride (CH ２ Polyvinyl chloride (CH =CHCl) is synthesized by polymerization ２ Regarding CHCl), vinyl chloride corresponds to the "monomer", polyvinyl chloride corresponds to the "polymer", and ２ "CHCl" corresponds to the "unit".

[0014] As used herein, "polyads" refers to a partial structure of a polymer composed of a finite number of units arranged in one direction. For example, possible combinations of polyads in a polymer synthesized from monomers A and B include diads such as AA, BB, and AB (or BA); triads such as AAB; and the like. Polyads are the units of sequence analysis, and sequence analysis in this specification refers to estimating the type of polyads contained in a target sample and their mass-based content.

[0015] In the above examples, the "number of units constituting the multi-adduct" is 2 for AA, BB, and AB(BA), etc., and 3 for AAA, AAB, BBA, BBB, and ABA(BAB), etc. In the following explanation, the "number of units constituting the multi-adduct" may be simply referred to as the "length of the multi-adduct."

[0016] "Number of variants of a multiply unit" refers to the combination variations of units in a multiply unit. For example, if the number of types of monomers included in a monomer set is three (monomers A, B, and C) and the number of units constituting the multiply unit (length of the multiply unit) is three (a triad), the number of variants of the triad is 13: AAA, BBB, CCC, AAC, AC{AC}, CCA, BBA, AB{AB}, AAB, BBC, BC{BC}, BCC, and ABC. Note that "AC{AC}" means a repeat of "AC." A repeat of "AC" can be expressed as "ACA" or "CAC" as a triad, but is expressed as "AC{AC}" to distinguish it from the different arrangements "AAC{AAC}" and "CCA{CCA}." The same applies to others.

[0017] The number of polyadrenaline variants is uniquely determined as the number of possible combinations depending on the length of the polyadrenaline and the number of types of monomers contained in the monomer set. However, depending on the type of monomer and the polymerization form, there are polyadrenaline combinations that exist in theory but cannot actually occur. For example, if monomers A and B do not form an alternating copolymer, "AB{AB}" exists in theory but is a polyadrenaline that cannot actually occur. Therefore, the number of polyadrenaline variants is the number of combinations uniquely determined depending on the length of the polyadrenaline and the number of types of monomers contained in the monomer set, or a number less than that. In this specification, the number of polyadrenaline variants is sometimes represented as "K (a number greater than or equal to 1)."

[0018] A "suspected sample" is a sample whose sequence is to be estimated. The suspected sample is composed of one or more of two or more monomers contained in the sample set. The type and amount of the monomer used in the synthesis may be unknown. The suspected sample may be a so-called homopolymer, which is composed of one type of monomer. Note that the term "homopolymer" used herein includes both a polymer that is actually composed of only one type of unit and a polymer that is suspected (or appears) to be composed of only one type of unit based on mass spectrometry. In other words, even if a polymer is suspected to be composed of only one type of unit based on mass spectrometry but actually contains other units below the detection limit, it is treated as a "homopolymer" in this specification. The same treatment applies to reference samples.

[0019] The "reference sample" refers to a sample required for determining the K end members of a K-1-dimensional simplex, as described below, and is a polymer synthesized from one or more monomers selected from the "sample set," similar to the sample to be estimated. The reference sample contains at least one or more types of multiplies selected from K types of multiplies. The types of multiplies contained are not particularly limited and may range from 1 to K types. The reference sample may contain one having the same composition as the sample to be estimated. In other words, the sample to be estimated and the reference sample may be identical, but the reference samples are different from each other. The reference samples being "different" from each other means that they differ in at least one type selected from the group consisting of the type of unit contained and the sequence of the units.

[0020] An "end member" refers to a vector corresponding to a vertex of a K-1 dimensional simplex, and an end member corresponds to a feature vector of a polymer (model polymer) consisting of only one type of polyadduct out of K types.

[0021] [Sequence Analysis Method] The sequence analysis method of the present invention is a sequence analysis method that takes as input the number of variants K, which is determined based on the number of types of monomers contained in a sample set and the number of units constituting a multiply-linked molecule, as well as two-dimensional mass spectra obtained from a reference sample and a suspected target sample, and outputs the mass-based content ratio of multiply-linked molecules in the suspected target sample.

[0022] The sequence analysis method of the present invention will be described in detail with reference to the drawings. FIG. 1 is a flowchart of an embodiment of the present invention. First, in step S1, the number of multi-adjuvant variants K (K is an integer of 2 or greater) is determined based on the number of types of monomers contained in the monomer set and the number of units constituting the multi-adjuvant. In one embodiment, the number of multi-adjuvant variants is a number that can be uniquely determined as the number of possible combinations based on the length of the multi-adjuvant and the number of types of monomers contained in the monomer set.

[0023] The number of types of monomers is an integer of two or more, and the upper limit is not particularly limited, but in one embodiment, 10 or less is preferable. For example, if the number of types of monomers included in the monomer set is 10, the reference sample and the suspected sample are synthesized using one or more of the 10 types of monomers. That is, the reference sample may be a (co)polymer obtained by adding one or more monomers selected from the sample set to a reaction vessel and polymerizing them under various conditions (temperature and time). Furthermore, the suspected sample may be synthesized using any one or more monomers included in the monomer set.

[0024] The number of units constituting the multiply (length of the multiply) is 2 or more, and although there is no particular upper limit, it is preferably 10 or less. In particular, the length of the multiply is preferably 3 or more, more preferably 9 or less, more preferably 7 or less, and even more preferably 5 or less.

[0025] In this sequence analysis method, the polymer sequence is estimated as the ratio of multi-unit contents, and the longer the multi-units, the closer it is to uniquely defining the entire polymer chain. On the other hand, when the number of multi-units is 10 or less, the increase in the number of multi-unit variants and end members in combinatorial calculations is kept to a certain level, and the number of reference samples required is unlikely to increase. When the length of the multi-units is 3 or more, it is easier to predict the physical properties of the polymer based on the analysis results of this sequence analysis method. When the length is 5 or less, the variation in reference samples is likely to be reduced, making analysis easier.

[0026] The number of variants K of the multiad can be uniquely determined once the number of types of monomers included in the monomer set and the length of the multiad are determined. For example, if the number of monomers is 3 or more and the length of the multiad is 3 (a triad), K = ｊ C ３ +3 ｊ C ２ + ｊ C １ Furthermore, when the number of monomers is 2 and the lengths of the multiadducts are 2, 3, 4, 5, ..., K will be 3, 5, 6, 9, .... Although the length of the multiadduct is uniquely determined as described above from the number of theoretically possible combinations, as will be described later, the realizable number may be smaller than this, and therefore it is preferable that K is the number of theoretically possible combinations or a number less than this.

[0027] Next, in step S2, the reference specimen, which is a polymer composed of monomers, and the suspected specimen are heated, and the resulting gas components are sequentially ionized, and a data matrix containing a two-dimensional mass spectrum of m / z (originally written in italics; defined as a dimensionless quantity obtained by dividing the mass of an ion by the absolute value of the unified atomic mass unit and the number of charges on the ion) versus heating temperature is obtained.

[0028] The method for observing (obtaining) a mass spectrum is not particularly limited, but a method of performing mass analysis on a sample under ambient conditions without pretreatment is preferred. As an example of such an ionization method and mass analysis method, a mass spectrometer called "DART-MS" is known, which combines an ion source called a "DART" (registered trademark, Direct Analysis in Real Time) ion source with a mass spectrometer. The mass spectrometer is not particularly limited, and is preferably one capable of accurate mass analysis, and may be either a quadrupole type or a time-of-flight (TOF) type.

[0029] There are no particular limitations on the specific conditions for obtaining a mass spectrum, but one non-limiting example is a procedure in which a specimen is successively heated at a temperature increase rate of 50°C / min, and helium ions are injected at intervals of 50 shots / min into pyrolysis gases generated in the temperature range of 50 to 550°C to ionize the gases, thereby obtaining a two-dimensional mass spectrum with m / z on the horizontal axis and temperature on the vertical axis.

[0030] In one form, the obtained two-dimensional mass spectra may be stored for each specimen and heating temperature, and at least two or more of these two-dimensional mass spectra may be converted together into a data matrix.

[0031] In this step, mass spectra are continuously acquired at predetermined intervals of temperature rise. These mass spectra may be used directly to create a data matrix, or may be averaged over a predetermined temperature rise range. By averaging the mass spectra over a predetermined temperature rise range into a single mass spectrum, the amount of data can be compressed. An example of such a temperature rise range is approximately 10 to 30°C.

[0032] Furthermore, the peak intensities in each spectrum may be normalized, for example, so that the sum of squares of the peak intensities becomes 1.

[0033] In this way, a single measurement of a given sample will yield a predetermined number of mass spectra (e.g., 20, depending on how they are grouped) for each heating temperature (or for each heating temperature grouped into a predetermined range). These mass spectra are stored in rows and the heating temperatures are stored in columns, resulting in a two-dimensional mass spectrum for each sample.

[0034] Once two-dimensional mass spectra for each sample are obtained in this manner, at least two of these spectra are collectively converted into a data matrix X. The number of two-dimensional mass spectra used to create the data matrix X is not particularly limited as long as it is two or more, but it is preferable to use two-dimensional mass spectra for all samples (all samples included in the sample set). Note that when measurements are performed two or more times for one sample, some or all of the two-dimensional mass spectra obtained from the two or more measurements may be used to create the data matrix X.

[0035] Next, in step S3, a first NMF process is performed in which the data matrix is ​​subjected to non-negative matrix factorization and decomposed into a matrix consisting of normalized basis spectra and the product of the intensity distribution matrix thereof.

[0036] In the first NMF process, the data matrix X is decomposed into the product of an intensity distribution matrix A and a matrix S consisting of basis spectra.

[0037] Here, the input data matrix X is expressed by the following formula:

[0038] The output intensity distribution matrix A and the matrix representing the basis spectrum (basis spectrum matrix) S are expressed by the following equations.

[0039] Here, N is the number of samples, T is the number of temperature band divisions, D is the number of channels, and M is the number of basis spectra.

[0040]

[0041] The above represents a non-negative or real NxM matrix.

[0042] In contrast, represent the n-th row vector, the m-th row vector, and the (n, m)-element, respectively. Ｔ represents the transpose of X, represents the Frobenius norm. is the nth row of X Tr(X) represents the trace, and diag(X) represents the diagonal matrix consisting of diagonal elements. Ｎ and 11 Ｎ represents an N-dimensional vector or an (N,N) matrix with all elements being 1. Ｎ represents an N-dimensional unit matrix.

[0043] Returning to the explanation of the first NMF process, the first NMF process in this analysis method was developed to better suit the interpretation of MS (Mass spectrometry) data by adding the following three main changes to the ARD-SO-NMF proposed by Shiga et al.

[0044] ・Change 1: Variance-covariance matrix of Gaussian noise for each channel - Change 2: Apply soft orthogonal constraints between base fragment spectra - Change 3: Merging fragment spectra with similar intensity distributions (extension of merging conditions)

[0045] Regarding change 1, for the time being, the variance σ ２ We assume independent and equal variance (i.i.d) Gaussian noise, and introduce a variance-covariance matrix R along the way. Assuming i.i.d noise, the probability generation model for the data matrix X is Since the number of basis spectra M is unknown, automatic relevance determination (ARD) is introduced based on the sparsity of the distribution matrix A so that it can be estimated automatically. First, λ ｍ Assume an exponential distribution parametrized by as the prior distribution of A.

[0046] That is, In contrast, The overall probability model is Here, p(S) can be written as The uniform distribution on the hypersphere of p(λ|a, b) is the inverse gamma distribution parameterized by (a, b), i.e. a is a hyperparameter that adjusts sparsity, and a = 1 + 10 －１６ , b is empirically A ｉｍ The expected value E(A ｉｍ ) can be estimated from This is further related to E(X ｉｄ )and There is a relationship such that E(X ｉｄ ) is the average μ of X Ｘ Approximating it with Therefore, b is Therefore, the negative log-likelihood function is Since this is a downward convex function with respect to λ, That is, The update formula for λ is obtained as follows. Up to this point, we have completely followed the derivation of ARD-SO-NMF reported by Shiga et al.

[0047] It is known that the assumption of an i.i.d Gaussian distribution for noise does not hold for MS data, and larger signals tend to have larger noise. was calculated as the residual component of the linear regression using natural isotope peaks. where are the isotope peaks at m / z ±3 of channel d in the channels [d-30, d-20, d-10, d+10, d+20, d+30] centered on channel d (note that the channel spacing is 0.1 m / z). The variance-covariance matrix for each channel teeth Using this, we can rewrite the likelihood function as Note that R is a constant matrix. The overall negative log-likelihood function is Substituting the update equation for λ (Eq. S3) and dropping the constant term to simplify, we get: The minimization of this function with respect to A and S is performed by hierarchical alternating least squares (HALS). For convenience, the following vector notation is used: In HALS, the residual X-AS is where This allows L(X, A, S, λ) to be written separately for each component m, The soft orthogonal constraint on S is the penalty term can be incorporated into the objective function L as This term represents the non-orthogonality between the m-component and other components, and ξ ｍ represents the Lagrange multiplier when orthogonality is strictly satisfied. This term is further divided into Ｏ ∈[0,1]. The objective function to be minimized is therefore and a ｍ and ｍ The gradient with respect to is If we set these to zero, the update formula becomes It becomes. ｍ Since is normalized for each update, the constant coefficient is omitted. The non-negativity constraint is satisfied by projecting to the non-negative quadrant for each update. Specifically, for example, where means a vector with the absolute value of each element. ｍ teeth Multiplying Eq. S6 from the left side, the strict orthogonal condition, And lol Ｏ By applying =1, Using Eqs. S1 to S7 above, we propose the following Algorithm 1.

[0048]

[0049] Here, similar components are merged each time A and S are updated. Shiga et al. proposed merging components with similar spectra, but here we also decided to merge components with similar intensity distributions. This allows isotope peaks, fragment series ionized by different ion additions, and oligomer peak series with different unit numbers to be combined into a single component, providing results that are easier to interpret.

[0050] Furthermore, noise components contained in the intensity distribution matrix may be extracted by canonical correlation analysis between the obtained basis spectrum matrix and the data matrix, and the intensity distribution matrix may be corrected to reduce the influence of the noise components, thereby obtaining a corrected intensity distribution matrix.

[0051] Since NMF is a low-rank approximation of a data matrix, even if a component k does not actually exist in the i-th spectrum, if the approximation improves in the least squares sense by assuming that it exists, then C ｉｋ In many cases, such a C ｉｋ is very small and is often not a problem in NMF analysis.

[0052] However, in the detection of trace components, C exists in trace amounts in the j-th spectrum. ｊｋ >0 and the NMF artifact C ｉｋ >0 and distinguish ghost peaks C ｉｋ It is preferable to substitute 0 for , because by removing the false peaks resulting from the artifacts of the NMF algorithm as one of the noises, more accurate estimation results can be obtained.

[0053] One method for solving the above problem is to use canonical correlation analysis, which the inventors have named the method the Canonical Correlation Analysis (CCA) Filter.

[0054] Conceptually, the CCA filter scans sample-wise to determine whether each component of the basis spectrum output from NMF was actually included in the original data, and if no similar peak pattern is found in the original data, it is removed from the spectrum of that sample. The CCA filter will be described in detail below.

[0055] The input is the basis spectral matrix of the output in the first NMF process. , and the background spectrum,

[0056] The output is a list of multiplicities determined to be from the background.

[0057] The M-components obtained by the first NMF process include components derived from background and contaminants, which may distort the sequence analysis results, and are therefore preferably removed from A and S. If the CCA-filter determines that the M'-component is a multi-adduct derived from the background, and are respectively, and For simplicity, the number of components after applying the CCA-filter is M-M', but we will use M consistently. In order to use the CCA-filter, It is necessary to include in the data set and perform the first NMF together with the sample spectrum. If some contaminants are expected, the spectrum measured for the contaminant should be added to X ＢＧ The CCA-filter can be used as follows: ｍ： is X ＢＧ The first step is to check whether the M-spectrum is included in the The S is then ｍ： A set of spectra similar to and dissimilar spectral sets This division is This is done so that the following is satisfied. １ ∈[0, 1] is a threshold value, and in the present invention,１ = 0.2 was used. ｍ： is always included in Y, so S is in the first column of Y. ｍ： Z is X ＢＧ Combined with So that Y and Z have an average of zero, CCA is performed on these two spectrum sets. CCA creates a pair of spectra that are as similar as possible by linearly combining the two spectrum sets.

[0058] and The coefficients of the linear combination of and Therefore, each spectrum pair is stored as and The similarity is evaluated by the correlation coefficient ρ, where: The problem setting of CCA is as follows: The solution can be written as is given as the solution to the generalized eigenvalue problem, where is the eigenvector is a matrix obtained by arranging the eigenvalues ​​as a column vector in descending order. is the corresponding (u ＊ , v ＊ ) is the correlation coefficient of y and z, which is created by linear combination with the coefficient ρ>t ２ u corresponding to all eigenvalues ​​that satisfy ＊ Take out S ｍ： The first element corresponding to the coefficient of u ＊ If it is a large component that contributes significantly to If t, the component m is determined to be a component derived from the background and is removed from the system. ２ ∈[0.9,0.99], t ３ ∈[0, 1]. This is summarized below as Algorithm 2.

[0059] After the background components were identified, they were removed from the system by deleting the corresponding column vectors of A and row vectors of S. (In the following, for simplicity, M-M' is replaced with M, ) corrects the assigned intensity with the sample amount and the internal standard peak. Then, the submatrix part A of A for sample n is （ｎ） is converted into a one-dimensional vector may be used as the feature vector of sample n and input to the second NMF. However, since the temperature distribution of M-fragments is unnecessary information for composition analysis, all temperature bands are summed for each sample to represent the M-fragment abundance (FA) for each sample. can be used as input. For simplicity, We will conduct a second NMF on this.

[0060] In the following sections, we frequently use fitting by Non-negative least squares (NNLS). This problem is solved by fitting optimal non-negative coefficients. Using the constant matrix By linear combination of the column vectors It approximates To solve this problem, many optimization methods can be used, including alternating direction multiplier methods (ADMM) (1). Here, we use the ADMM-NNLS method developed by Fu et al. (2) to solve this problem and obtain the solution. The approximate vector set Y = [y １ ,... ,y Ｌ ], the non-negative coefficient vector x ｌ ＊ (l=1,...,L) can be calculated individually, or in matrix form, Here, A similar problem, a problem with constraints such that the sum of the coefficient vectors is 1, is called fully constrained least squares (FCLS), and can also be solved using ADMM, as follows: or in matrix notation, I will write it as follows.

[0061] Next, the second NMF process in step S4 will be described in detail. In the second NMF process, the FA for each sample is decomposed into a product of a matrix C representing the mass fraction of the model polymer in the specimen and a matrix B representing the feature vector of the model polymer, as expressed by the following equation:

[0062] The input is the FA for each specimen. , spectrum for each base M-fragment , and the number of variants is K, and the output is a matrix representing the mass fraction of the model polymer in the sample.

[0063] , a matrix representing the feature vector of the model polymer, is.

[0064] Second NMF: is calculated as the approximate residual using the Riemannian metric. Evaluate, i.e., where: A lower triangular matrix Cholesky decomposition G = LL Ｔ If we obtain where and This is equivalent to matrix decomposition as follows: The volume term of the simplex spanned by the row vector of The non-orthogonality terms between the row vectors of Let α>0 and β∈[0,1] be weights, then By Here, Fu et al. (3) add robustness to outliers by where p∈(0,2] and ε is a small regularization parameter. In the present invention, we consistently use p=1.5 and ε=10 －８ The optimization problem is C and is optimized alternately according to the block coordinate descent (BCD) theory. Now, t optimizations are performed, For initialization, we use vertex component analysis (VCA) (4) to obtain Among the N-row vectors of the matrix, K row vectors closest to the K-vertices of the simplex are selected, and these are It was decided. The update of C based on This can be done simply by （ｔ） Based on from Consider updating to. in, is an undetermined multiplier. where: is a small regularization parameter, and is consistently τ=10 －８ where: By the tangent inequality, we can see that the Based on where teeth H （ｔ） The gradient at twist, where: and const is a constant term. of Substituting this and writing all penalty terms together gives where Therefore The update of is obtained by solving To organize complex constraints, The constraint is expressed as Lagrange multipliers is incorporated into the objective function and this problem is solved within the framework of ADMM. where: and μ is a hyperparameter of the ADMM (here, μ = 1 is used consistently). Z′ represents Z one period ago. The objective function is maximized with respect to Z when This gives It is sufficient to optimize for We cyclically optimized the algorithm and summarized it in Algorithm3.

[0065] Objective function teeth Since this is an optimization of a quadratic function without constraints, is further minimized. can be solved by NNLS, From the above, the original problem Eq. S11 can be solved as follows: Finally, the update formula V = αF （ｔ） +βΛ （ｔ） Think about updating F （ｔ） About This can be easily calculated as follows. （ｔ） is a strict orthogonality condition, i.e. and By combining Applying this to the right side of Eq. S19, From Eq. S18 and Eq. S20, Using the above, the algorithm for solving problem Eq. S11 is as follows:

[0066] (Projection of test data onto the hyperplane spanned by S and B) After estimating S and B from the dataset, the data that was not used to estimate S and B (here called test data) We will now discuss how to project onto the hyperplane spanned by S and B. First, by projecting onto the S-hyperplane, That is, Take the sum over the temperature range, After transforming it into, by projection onto the B-hyperplane and normalization, That is,

[0067] The cited references are as follows:

[0068] Next, the calculation of the model polymer spectrum (hereinafter sometimes referred to as "M spectrum") in step S5 will be described in detail. In step S5, the spectrum (M spectrum) of a model polymer composed of only one type of polyadduct is reconstructed from the calculation results up to this point. The method for this is to calculate the matrix product of a matrix representing the basis spectrum obtained by the first NMF process and a matrix representing the feature vector of the model polymer obtained by the second NMF process.

[0069] The spectrum of a model polymer (M spectrum) is calculated by multiplying the basis spectrum by a matrix (similar to coefficients) that represents a feature vector specific to each model polymer. That is, the M spectrum is calculated as the matrix product of the matrix that represents the basis spectrum and the matrix that represents the feature vector of the model polymer.

[0070] That is, when the first NMF treatment yields X = AS, the second NMF treatment decomposes this into X = AS = (CB)S, which can then be interpreted as C(BS), thereby yielding the mass ratio C of the model compound and the spectrum BS of the model polymer. This BS corresponds to the M spectrum.

[0071] The M spectrum calculated in this step can be said to be an estimate of the mass spectrum of a sample containing only the model polymer, and in step S6, identification is performed to determine to which multiplicities each of these M spectra can be assigned. As already explained, the number of multiplicities is determined based on the number of theoretically possible combinations, and in principle, all M spectra can be assigned to each multiplicity. However, as will be explained in detail later, there are cases where one or more M spectra cannot be assigned to a multiplicity.

[0072] In step S6, it is determined whether there is an M spectrum that cannot be attributed to a multi-adjuvant. If there is an M spectrum that cannot be attributed to a multi-adjuvant (step S6: YES), the number of multi-adjuvant variants K is changed, and the first NMF process and the second NMF process are repeated. If there is an M spectrum that cannot be attributed to a multi-adjuvant, one of the reasons for this is that at least one of the multi-adjuvant variants does not actually exist or is not sufficiently contained in the reference sample.

[0073] As an example of the former case, consider a case where A and B are not alternating copolymerizable in a triad of a polymer obtained from monomers A, B, and C. In this case, if the number of variants K is determined and an M spectrum is obtained on the assumption that "AB{AB}" exists as one of the triad variants, it is not possible to obtain an M spectrum derived from (attributed to) "AB{AB}". This is because the sample does not contain a polymer having such a triad. As a result, one or more M spectra will not be attributable to a multi-unit.

[0074] In such cases, correction can be made by subtracting 1 from K and then performing the first NMF process and the second NMF process again. That is, if an M spectrum that cannot be assigned to a multi-unit is present, more accurate analysis results can be obtained by mechanically subtracting 1 from K and re-determining the M spectrum, even if the fact that A and B do not have alternating copolymerization is unknown. As described above, when the present sequence analysis method includes steps S5 and S6, even if a non-existent multi-unit variant is incorporated into K due to the combination of individual monomers or the like and the cause is unknown, the validity of the analysis can be evaluated and corrected simply by confirming whether the M spectrum can be assigned to the set multi-unit.

[0075] Furthermore, even if the "AB" multiplicity actually exists, if the sample (typically, the reference sample) does not contain enough of that multiplicity, the same situation as above may occur. In this case, correction can be made by adding more samples, acquiring the data matrix, and performing the first NMF process and the second NMF process again. Note that even in this case, the reference sample may remain the same, and the first NMF process and the second NMF process may be performed with K decreased by 1.

[0076] On the other hand, if an M spectrum is assigned to each of the multiplets (step S6: NO), the next step (S7) is performed. As will be described in detail later, the estimation of the multiplet content ratio in this sequence analysis method is performed by projecting the feature vector of the sample onto a K-1 dimensional simplex, and "endmembers" that correspond to the vertices of this K-1 dimensional simplex are required. These "endmembers" are defined by the feature vector of a polymer (model polymer) composed only of multiplets. However, it is often difficult to actually prepare a model polymer as a reference sample.

[0077] For example, in triad analysis when a sample set consists of three types of monomers, A, B, and C, the number of variants, K, is 13, and feature vectors of 13 types of "model polymers," including three types of homopolymers, are required as "end members." However, it is often not practical to precisely control the arrangement of units, synthesize these 13 types of model (co)polymers, and obtain mass spectra.

[0078] In contrast, one of the features of the present sequence analysis method is that it combines the first NMF process and the second NMF process as described above to estimate the feature vector of an "end member," eliminating the need to actually measure the mass spectrum of a model polymer.

[0079] An example of the M spectrum analysis in steps S5 and S6 will now be described based on experimental results. Figure 2 shows the M spectrum calculated from the matrices obtained by the first NMF process and the second NMF process after acquiring a data matrix from a specimen synthesized using methyl methacrylate (M) and styrene (S) as monomers. The experimental procedure will now be described.

[0080] First, the synthesis of the sample was carried out by adding methyl methacrylate, styrene, and a polymerization initiator (dimethyl 2,2'-azobis(isobutyrate)) to a reaction vessel and heating for a predetermined period of time. The amounts of methyl methacrylate and styrene added, the heating temperature, and the reaction time were adjusted in various ways to prepare polymers of different compositions, each of which was used as a sample. Note that the detailed procedures were the same as those in the examples described below, and therefore will not be explained here. The polymerization conditions were as shown in the table below.

[0081]

[0082] In the table, "M initial fraction" and "S initial fraction" represent the ratio of the amounts of methyl methacrylate and styrene added, respectively, "Time (h)" represents the polymerization time, and "Temp (°C)" represents the heating temperature. In this way, 31 different specimens were prepared. As described above, the number of samples was 31, the temperature range for mass analysis was 200 to 450°C, and the m / z range was 50 to 410. Other hyperparameters are as follows:

[0083]

[0084] In this experiment, the monomer set contained two types of monomers: methyl methacrylate and styrene. The length of the multi-unit was set to 3 (tri-unit analysis), and the variants of the multi-unit were all five types assumed as theoretical combinations: MMM, MMS, MS{MS}, SSM, and SSS (number of variants K = 5).

[0085] Figure 2 shows the calculated M spectrum. In Figure 2, spectra (1) to (5) are the calculated M spectrum. Here, labels indicating which multi-adduct each M spectrum originates from are not attached to the first NMF process and the second NMF process. Even if labels are not attached, this does not affect the analysis itself in the subsequent stages. On the other hand, by attributing the M spectrum to a multi-adduct, it is possible to check the validity of the analysis conditions, etc., resulting in more accurate analysis.

[0086] As already explained, step S6 is a step of identifying the multiplicity to which the M spectrum belongs. In principle, the M spectrum will belong to any multiplicity, and therefore, in step S6, the multiplicity is identified. Although the method of identification is not particularly limited, it is preferable that the identification be performed by comparing the sum of the mass numbers of the units constituting the multiplicity with the m / z at the peak of the M spectrum.

[0087] In the case of Figure 2, the sum of the mass numbers of the units constituting the multi-unit is 300 for MMM, 304 for MMS, 304 / 308 for MS(MS), 308 for SSM, and 312 for SSS. In contrast, the peaks of the M spectrum are 300+1 for spectrum (1), 304+1 for (2), 304+1 and 308+1 for (3), 308+1 for (4), and 312+1 for (5), which coincide with the protonated peaks of the multi-unit. Based on the above, the M spectra of (1) to (5) are identified as MMM, MMS, MS(MS), SSM, and SSS, respectively. The labels of each spectrum in Figure 2 are the results of this step.

[0088] The spectrum (6) in Figure 2 represents the mass spectrum obtained by actually polymerizing an M / S alternating copolymer, and it can be seen that it almost coincides with the calculated spectrum (3). Since the M spectra obtained as a result of the first NMF treatment and the second NMF treatment in this sequence analysis method are almost the same as the actually measured mass spectra as described above, the identification process in this step can be easily performed based on the structure and mass number of the multi-unit units.

[0089] 1 , next, in step S7, the feature vectors of the model polymer are set as endmembers, and a K-1 dimensional simplex that includes all of the feature vectors (each corresponding to a specimen) is set. In this sequence analysis method, the K-1 dimensional simplex is set by the endmembers estimated by the second NMF process, regardless of whether the reference specimen contains endmembers or not.

[0090] When the number of variants K is 3 or more, it is preferable that at least one of the feature vectors of the reference analyte is located in each of the outer regions of the hypersphere inscribed in the K-1-dimensional simplex, or that the reference analyte includes at least one of the end members. This allows for more accurate analysis results.

[0091] Here, each position within the region of the K-1-dimensional simplex represents the mass-based content ratio of the endmembers. Therefore, the region outside the hypersphere inscribed in the K-1-dimensional simplex roughly represents an arbitrary endmember, i.e., a region where the content of arbitrary multiplies is equal to or greater than a predetermined amount. The location of the feature vector of the reference analyte at such a position means that the analyte set includes a reference analyte where the content of arbitrary multiplies is equal to or greater than a predetermined amount. In one embodiment, the reference analyte includes an endmember.

[0092] 3 is an image diagram showing a K-1-dimensional simplex (two-dimensional simplex, triangle) when K is 3. A K-1-dimensional simplex 10 is a triangle with end members 13, 14, and 15 determined by reference specimens 16, 17, and 18 as vertices.

[0093] As shown in FIG. 3, when the reference specimens 16, 17, and 18 are located in the outer regions 19 (hatched) of the inscribed hypersphere 12 (in this case, a circle) of the K-1 dimensional simplex, highly accurate quantitative analysis results can be obtained.

[0094] Next, the distances between the K endmembers and the feature vectors of the target specimen are calculated, and the content ratio of multi-units in the target specimen is estimated (step S8). Note that the distances are defined by the Riemannian metric distance, which takes into account the non-orthogonality of the basis spectra obtained by the first NMF process.

[0095] When a reference analyte contains at least one endmember (at least one of the reference analytes is an endmember), the feature vectors of other reference analytes may be located in the inner region of a hypersphere inscribed in the K-1-dimensional simplex. Similar to FIG. 3, FIG. 4 is an image diagram showing a K-1-dimensional simplex (a two-dimensional simplex, a triangle) when K is 3. The K-1-dimensional simplex 20 is a triangle with vertices at endmembers 24 and 25 determined by reference analyte 21, which is an endmember, other reference analyte 22, and reference analyte 23. The difference from FIG. 3 is that reference analytes 22 and 23 are located in the inner region of a hypersphere inscribed in the K-1-dimensional simplex.

[0096] Even when at least one of the reference analytes is an endmember, the other reference analytes may be located on a hypersphere inscribed in the K-1-dimensional simplex, or may be located in an outer region. That is, in this case, the other reference analytes may be located at any position. In particular, in terms of obtaining a more excellent effect of the present invention, the other reference analytes preferably contain 20% by mass or more, and more preferably 40% by mass or more, of components of endmembers (components of other endmembers) different from the reference analyte that is an endmember.

[0097] This sequence analysis method enables accurate sequence analysis using a simple procedure without requiring any special pretreatment, even for polymers for which it is difficult to prepare reference samples. This sequence analysis method can significantly reduce the time required to obtain results, particularly when applied to quality control of polymers synthesized from multiple monomers and investigation of the cause of defects.

[0098] The polymers to which this sequence analysis method can be applied are not particularly limited and may be either synthetic or natural polymers. In the examples described below, (co)polymers synthesized based on monomers having ethylenically unsaturated bonds are described, but the (main chain) structure of the polymer is not limited to this. The polymerization method is also not particularly limited and may be synthesised by any method, such as addition polymerization, open-chain polymerization, polycondensation, polyaddition, or addition polymerization. Furthermore, the reference sample and the suspected target sample may be synthesised by different synthesis methods (polymerization methods) as long as they are synthesized from the same set of monomers. For example, the reference sample may be prepared by conventional radical polymerization under variously modified polymerization conditions, while the suspected target sample may be polymerized under precise control by other methods (e.g., living polymerization such as atom transfer radical polymerization or reversible addition-fragmentation chain transfer (RAFT) polymerization).

[0099] One application of this sequence analysis method is to apply it to the sequence analysis of (photo)resist resins. It is known that there is a correlation between the developability of resist resins and their multi-unit sequences. By using this sequence analysis method to analyze the sequence of resist resins, it becomes easier to develop resist resins with better developability and to investigate the causes of development problems with resist resins. There are no particular limitations on the resist resins to which this sequence analysis method can be applied. Examples of resist resins include resins synthesized from the following monomers:

[0100] 3-hydroxy-1-adamantyl methacrylate, 1-adamantyl acrylate, 1-adamantyl methacrylate, 2-methyl-2-adamantyl methacrylate, 2-methyladamantan-2-yl acrylate, 2-ethyl-2-adamantyl methacrylate, 2-ethyl-2-adamantane acrylate, dicyclopentanyl acrylate, 2-isopropyladamantan-2-yl acrylate, tetrahydrodicyclopentadienyl methacrylate, 5-methacryloyloxy-2,6-norbornane Benzyl acrylate, β-hydroxy-γ-butyrolactone methacrylate, 1-ethylcyclopentyl methacrylate, α-methacryloxy-γ-butyrolactone, 1-ethylcyclohexyl methacrylate, 1-methylcyclopentyl methacrylate, 4-acetoxystyrene, 2-oxo-2-(2,2,3,3,3-pentafluoropropoxy)ethyl methacrylate, 3-hydroxy-1-adamantyl acrylate, (adamantan-1-yloxy)methyl methacrylate, 2-isopropyl-2-adamantyl Methacrylate, 1,3-adamantanediol diacrylate, 1,3-adamantanediol dimethacrylate, 1-methyl-1-ethyl-1-adamantylmethanol methacrylate, 1,1-diethyl-1-adamantylmethanol methacrylate, 5,7-dimethyl-1,3-adamantanediol diacrylate, 5,7-dimethyl-1,3-adamantanediol dimethacrylate, 5-ethyl-1,3-diadamantanediol diacrylate, 5-ethyl-1,3-diadamantanediol dimethacrylate, 2-methyl-2-propenoic acid 2-oxo-2-[(5-oxo-4-oxatricyclo[4.3.1.13,8]undec-2-yl)oxy]ethyl ester, 2-propenoic acid, 2-methyl-,2-[(hexahydro-2-oxo-3,5-methano-2H-cyclopenta[b]furan-6-yl)oxy]-2-oxoethyl ester, 2-(2,2-difluoroethenyl)bicyclo[2.2.1]heptane, 6-methacryloyl-6-azabicyclo[3.2.0]heptan-7-one, 2-propenoic acid, (3R,3aS,6R,7R,8aS)-octahydro-3,6,8,8-tetramethyl-1H-3a,7-Methanoazulen-6-ol ester, 2-propenoic acid, (3R,3aS,6R,7R,8aS)-octahydro-3,6,8,8-tetramethyl-1H-3a,7-methanoazulen-6-yl ester, 2-cyclohexylpropan-2-yl methacrylate, 1-isopropylcyclohexyl methacrylate, 1-methylcyclohexyl methacrylate, 1-ethylcyclopentyl acrylate, 1-methylcyclohexyl acrylate, tetrahydropyranyl methacrylate, tetrahydro-2-furanyl methacrylate, (3-methyl-5-oxooxolan-3-yl)2-methylprop-2-enoate, 2-oxotetrahydrofuran-3-yl Acrylates, (5-oxotetrahydrofuran-2-yl)methyl methacrylate, (2-oxo-1,3-dioxolan-4-yl)methyl methacrylate, 1-ethoxyethyl methacrylate, N-(methoxymethyl)methacrylamide, N-isopropylmethacrylamide, 2-(bromomethyl)ethyl acrylate, 2-(bromomethyl)methyl acrylate, N-butyl-2-(bromobutyl)acrylate, 2,2,3,3,4,4,4-heptafluorobutyl methacrylate, 2,5-dimethylhexane-2,5-diyl Bis(2-methyl methacrylate), 2-(trifluoromethanesulfamido)ethyl methacrylate, 3-[dimethoxy(methyl)silyl]propyl acrylate, 2,3-dihydroxypropyl acrylate, 9H-fluorene-9,9-dimethanol dimethacrylate, 9,9-bis[(acryloyloxy)methyl]fluorene, 9-anthrylmethyl methacrylate, 4-hydroxyphenyl methacrylate, 4-(4-acryloyl Oxybutoxy)benzoic acid, 3-(4-hydroxy-phenoxy)propyl acrylate, 10-([1,1'-biphenyl]-2-yloxy)decyl acrylate, 2-vinylnaphthalene, 4-tert-butoxystyrene, 4-isopropenylphenol, 4-([1-ethoxyethoxy)styrene, 3,4-diacetoxystyrene, 4-allyloxystyrene, 3-tert-butoxystyrene, 2-acetoxystyrene, 4-ethenyl-1,2-bis(1-ethoxyethoxy)benzene, tetrahydro-2-[4-(1-methylethenyl)phenoxy]furan, 2-[(4-vinylphenoxy)methyl]oxirane, 3,5-diacetoxystyrene, 2,3-difluoro-4-vinylphenol, 3-fluoro-4-vinylphenol, 1,1,2,2-tetramethylpropyl acrylate, 1-ethenyl-4-propan-2-yloxybenzene, 4-vinylphenyl benzoate, 1-ethylhexyl methacrylate, 2-isopropyl-2-adamantyl methacrylate, 3-hydroxy-1-adamantyl methacrylate, 1,1-dimethylpentyl methacrylate, 1,1-dimethylhexyl methacrylate, neopentyl methacrylate, and 2,2,2-trifluoroethyl methacrylate.

[0101] Specific examples of the monomer combination include γ-butyrolactone (meth)acrylate / 2-methyl-2-adamantyl (meth)acrylate / 3-hydroxy-1-adamantyl (meth)acrylate, and 4-hydroxystyrene / 2-methyl-2-adamantyl (meth)acrylate / styrene.

[0102] [Sequence Analysis Apparatus (First Embodiment)] Next, a sequence analysis apparatus according to an embodiment of the present invention will be described with reference to the drawings. Figure 5 is a hardware configuration diagram of one embodiment of the sequence analysis apparatus of the present invention. The sequence analysis apparatus 30 includes a mass spectrometer 31 and an information processing device 32. The information processing device 32 has a processor 33, a storage device 34, a display device (not shown), and an input / output interface (I / F) 35 for connecting an input device. The mass spectrometer 31 and the information processing device 32 are configured to be able to send and receive data to and from each other.

[0103] The processor 33 is, for example, a microprocessor, a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a general-purpose computing on graphics processing unit (GPGPU).

[0104] The storage device 34 has a function of temporarily and / or non-temporarily storing various programs and data, and provides a working area for the processor 33. The storage device 34 is, for example, a read-only memory (ROM), a random access memory (RAM), a hard disk drive (HDD), a flash memory, or a solid state drive (SSD).

[0105] An input device connected to the input / output I / F 35 can accept input of various information and also accept input of instructions to the sequence analysis device 30. The input device may be a keyboard, a mouse, a scanner, a touch panel, or the like.

[0106] Furthermore, a display device connected to the input / output I / F 35 can display the status of the sequence analysis apparatus 30, the progress of the analysis, the sequence analysis results, etc. The display device may be a liquid crystal display, an organic electroluminescence (EL) display, etc. The display device may also be configured as an integrated part with an input device. In this case, the display device may be a touch panel display that provides a GUI (Graphical User Interface), for example.

[0107] The information processing device 32, which includes a processor 33, a storage device 34, and an input / output I / F 35 that can communicate data with each other via a data bus, is typically a computer.

[0108] The mass spectrometer 31 is typically a mass spectrometer equipped with a "DART" ion source, a sample heating device, and a time-of-flight mass spectrometer. The ion source and mass spectrometer equipped in the mass spectrometer 31 are non-limiting examples, and the configuration of the mass spectrometer equipped in the sequence analysis device is not limited to the above.

[0109] 6 is a functional block diagram of the sequence analysis device 30. The sequence analysis device 30 includes a mass spectrometer 31 that performs mass analysis of a sample, and an information processing device 32 that processes the mass spectrum obtained by the mass spectrometer 31.

[0110] The mass spectrometer 31 is controlled by a processor 33 executing a program stored in a storage device 34 of an information processing device 32. The mass spectrometer 31 heats the loaded sample and thermally desorbs and / or pyrolyzes the sample, ionizes the gas components produced, sequentially performs mass analysis, and outputs a mass spectrum.

[0111] The mass spectrum acquired by the mass spectrometer 31 is passed to a data matrix creation unit 41 of the information processing device 32. The data matrix creation unit 41 is a function realized by the processor 33 executing a program stored in the storage device 34. The data matrix creation unit 41 creates a data matrix from the two-dimensional mass spectrum in which a mass spectrum is stored in each row, and passes the data matrix to a first NMF processing unit 43, which will be described later. Details of the data matrix have already been explained.

[0112] The variant number determination unit 42 is a function realized by the processor 33 executing a program stored in the storage device 34. The variant number determination unit 42 determines the variant number K according to the number of types of monomers and the length of the multi-adducts, which are acquired from the outside 47 via the input / output I / F 35, and passes the determined number of variants K to the first NMF processing unit 43. The storage device 34 stores a predetermined maximum value of the variant number K according to the number of types of monomers and the length of the multi-adducts, and the variant number determination unit 42 determines the variant number K by referring to this maximum value. Note that in this embodiment, the number of types of monomers and the length of the multi-adducts are acquired from the outside 47 via the input / output I / F 35, but the number of types of monomers and the length of the multi-adducts may also be acquired from an external network using communications other than the input / output I / F 35. Furthermore, the maximum value of the number of variants K may be stored in the storage device 34 as a table of the number of types of monomers and the length of the multi-adduct, or may be stored as a function of the number of types of monomers and the length of the multi-adduct.

[0113] The first NMF processing unit 43 is a function realized by the processor 33 executing a program stored in the storage device 34. The first NMF processing unit 43 performs nonnegative matrix factorization (NMF) on the data matrix based on the data matrix provided by the data matrix creation unit 41 and the number of variants K provided by the number of variants determination unit 42, and decomposes the data matrix into a product of an intensity distribution matrix and a matrix representing the base spectrum. The matrix decomposition method has already been described. The result is passed to the second NMF processing unit 44.

[0114] The second NMF processing unit 44 is a function realized by the processor 33 executing a program stored in the storage device 34. The second NMF processing unit 44 performs nonnegative matrix factorization on the intensity distribution matrix provided by the first NMF processing unit 43, and decomposes it into the product of a matrix representing the mass fraction of a model polymer composed only of polyadducts in the specimen and a matrix representing the feature vector of the model polymer. The method of this matrix factorization has already been explained. The result is passed to the vector projection unit 45.

[0115] The vector projection unit 45 is a function realized by the processor 33 executing a program stored in the storage device 34. The vector projection unit 45 sets a K-1 dimensional simplex that includes all of the feature vectors of the specimen, using the feature vectors of the model polymer provided by the second NMF processing unit 44 as end members. The method for setting the K-1 dimensional simplex has already been described.

[0116] The composition estimation unit 46 is a function realized by the processor 33 executing a program stored in the storage device 34. The composition estimation unit 46 calculates the distances between the K endmembers in the K-1-dimensional simplex set by the vector projection unit 45 and the feature vector of the estimation target specimen, and estimates the content ratio of each of the multi-units in the estimation target specimen from the ratio of the distances. The composition estimation unit 46 then outputs the sequence analysis results to the outside 48 via the input / output I / F 35. Note that, in this embodiment, the sequence analysis results are output to the outside 48 via the input / output I / F 35, but the sequence analysis results may also be transmitted to an external network using communications other than the input / output I / F 35.

[0117] The sequence analysis device 30 enables accurate sequence analysis with a simple procedure, without requiring any special pretreatment, even for polymers for which it is difficult to prepare reference samples. The sequence analysis device 30 can significantly reduce the time required to obtain results, particularly when applied to quality control of polymers synthesized from multiple monomers and investigation of the cause of defects.

[0118] [Sequence Analysis Apparatus (Second Embodiment)] Fig. 7 is a functional block diagram of a second embodiment of a sequence analysis apparatus. The sequence analysis apparatus 50 is similar to the sequence analysis apparatus 30, except that the information processing device 32 has a model polymer spectrum identification unit 51 (denoted as "M spectrum identification unit" in the figure). Differences from the sequence analysis apparatus 30 will be described below.

[0119] The model polymer spectrum identification unit 51 of the sequence analysis apparatus 50 is a function realized by the processor 33 executing a program stored in the storage device 34. The model polymer spectrum identification unit 51 identifies the multi-atom to which the M spectrum belongs based on the basis spectrum provided by the first NMF processing unit 43, the feature vector of the model polymer provided by the second NMF processing unit 44, and the mass numbers of the units constituting the multi-atom obtained from the outside 52. The details of the identification method have already been described. The mass numbers of the units constituting the multi-atom may be stored in advance in the storage device 34.

[0120] If there are any M spectra that cannot be assigned to multi-adducts, the model polymer spectrum identification unit 51 provides the first NMF processing unit 43 with a new number of variants K (typically, the initial number of variants K minus a predetermined number), and causes the first NMF processing unit 43 to perform the first NMF process and the second NMF processing unit 44 to perform the second NMF process. This process is repeated until all of the M spectra can be assigned to multi-adducts. The method for reducing the number of variants is not particularly limited, but a typical example is to subtract one.

[0121] When all of the M spectra have been assigned to polyadrenergic molecules by the M spectrum identification unit 51, the matrix representing the mass fraction of the model polymer in the sample calculated by the second NMF processing unit 44 and the matrix representing the feature vector of the model polymer are provided to the vector projection unit 45. The subsequent processing is the same as that of the sequence analysis device 30.

[0122] Since the sequence analysis device 50 has the model polymer spectrum identification unit 51, even if the value of K initially set as a theoretically possible value does not correspond to reality and the reason for this is unknown, the sequence analysis device 50 can reset K to an appropriate value by a predetermined process, thereby providing more accurate sequence analysis results.

[0123] [Polymerization Condition Proposal Apparatus] Next, a polymerization condition proposal apparatus according to an embodiment of the present invention will be described with reference to the drawings. Fig. 8 is a functional block diagram of one embodiment of the polymerization condition proposal apparatus according to the present invention.

[0124] The polymerization condition proposal device 60 has an information processing device 61. In addition to the functions of the sequence analysis device 30, the information processing device 61 further has a policy proposal unit 62. The hardware of the polymerization condition proposal device 60 is similar to that of the sequence analysis device 30, and the policy proposal unit 62 is a function realized by the processor 33 executing a program stored in the storage device 34. The functions of the polymerization condition proposal device 60 are similar to those of the sequence analysis device 30 except for the inclusion of the policy proposal unit 62, and therefore, the functions of the policy proposal unit 62 will be described below.

[0125] The policy proposal unit 62 is a learning model generated by machine learning using, as training data, a plurality of actual measurement data associating polymerization conditions with the sequence analysis results of the resulting polymer. The policy proposal unit 62 generates a polymerization condition dataset including a plurality of polymerization conditions under which the sequence of the resulting polymer is unknown, and calculates a prediction result (polymer sequence) for each polymerization condition. Furthermore, the policy proposal unit 62 creates a prediction dataset associating polymerization conditions with the prediction results, identifies a prediction result that is closest to the target sequence from the obtained prediction results, and extracts the polymerization conditions associated with the identified prediction result.

[0126] The policy proposal unit 62 receives data relating the sequence analysis results of the estimation target specimen and the polymerization conditions from the composition estimation unit 46, and further receives data on the target sequence from the outside 63 via the input / output I / F 35. The policy proposal unit 62 generates multiple polymerization condition data sets and predicts the sequence. From these, the policy proposal unit 62 extracts conditions that will result in a sequence closer to the target sequence than the sequence analysis results obtained from the composition estimation unit 46, and proposes these as "polymerization conditions" to the outside 64 via the input / output I / F 35.

[0127] The learning model may be, for example, a trained neural network trained using each parameter of the polymerization conditions as an explanatory variable and the sequence analysis result of the obtained polymer as a target variable. Known methods can be used to construct such learning models, for example, the methods described in WO 2020 / 054183, WO 2020 / 066309, and JP 2008-501837 A can be used.

[0128] The polymerization condition proposing device 60 has a policy proposing unit 62, and therefore can compare the sequence analysis results with the target sequence and propose polymerization conditions that are expected to produce a polymer with a sequence closer to the target sequence. As a result of the above, more efficient material design becomes possible even for complex systems with a large number of types of monomers and / or long multi-unit lengths.

[0129] Although the polymerization condition proposing device 60 does not have the model polymer spectrum identifying unit 51, it is preferable that the polymerization condition proposing device of the present invention has the model polymer spectrum identifying unit 51.

[0130] [Automated Synthesizer] Next, an automated synthesizer according to an embodiment of the present invention will be described with reference to the drawings. FIG. 9 is a functional block diagram of one embodiment of the automated synthesizer of the present invention. The automated synthesizer 70 includes a polymer synthesizer 71 in addition to a polymerization condition proposal device 60. The polymer synthesizer 71 includes a monomer supply mechanism 75, a reaction vessel 76, and a control device 72 that controls these. Note that in FIG. 8, some of the functions of the polymerization condition proposal device 60 are omitted, and only parts necessary for explanation are shown, but the polymerization condition proposal device 60 has the same functions as the polymerization condition proposal device 60 described above.

[0131] The polymer synthesis apparatus 71 may typically be a flow reactor. A monomer and / or a monomer solution in which a monomer is dissolved in a solvent is supplied to the reaction vessel 76 from a monomer supply mechanism 75. The synthesis apparatus 71 may have a plurality of monomer supply mechanisms 75, each of which is independently controlled by the control device 72. The monomer supply mechanism 75 typically has a container for storing the monomer (or solution), a conduit from the container to the reaction vessel 76, and a pump. The type and amount of monomer supplied to the reaction vessel 76 are adjusted by the output of the pump.

[0132] The reaction tank 76 is a hollow portion provided in a pipeline connected to the supply mechanism 75, and is typically a vessel-shaped reaction site. The reaction tank 76 includes a heater, a gas pipeline for adjusting the atmosphere, a valve, a pump, a stirring blade, and the like.

[0133] The monomer supply mechanism 75 and the reaction tank 76 are controlled by the processor 73 executing a program stored in the storage device 74. Specifically, when the polymerization conditions provided by the polymerization condition proposal device 60, i.e., the polymerization conditions predicted to produce a polymer closer to the target sequence, are received, the supply mechanism 75 is controlled in accordance with the conditions to adjust the type of monomers to be supplied to the reaction tank 76 and the amount of each monomer to be supplied. The reaction tank 76 is also controlled to adjust the reaction temperature, reaction time, stirring speed, etc.

[0134] Furthermore, after the reaction for the predetermined reaction time has ended, the control device 72 controls the pump in the reaction tank 76 to send the resulting polymer from the reaction tank 76 to the mass spectrometer 31. In the automated synthesis system 70, the reaction tank 76 and the mass spectrometer 31 are connected by a pipeline, and the synthesized polymer is subjected to sequence analysis again. With the automated synthesis system 70 configured in this manner, polymerization is automatically carried out under the polymerization conditions proposed by the polymerization condition proposal device 60, and the polymer is subjected to sequence analysis again, and the evaluation of the results is repeated again. In this way, a polymer is automatically synthesized along the target sequence.

[0135] The present invention will be described below with reference to examples, but the present invention is not limited to these examples.

[0136] [Example 1: Triad analysis of MMA / St / BA] Triad analysis was performed using methyl methacrylate (M), styrene (S), and butyl acrylate as a monomer set. Methyl methacrylate, styrene, and butyl acrylate were each manufactured by Tokyo Chemical Industry Co., Ltd. A predetermined amount of these monomers was poured into a vial, and dimethyl 2,2'-azobis(isobutyrate) was added as a polymerization initiator. After replacing the atmosphere with nitrogen gas, polymerization was carried out at a predetermined temperature for a predetermined time while stirring, and then the reaction was terminated with methanol. The obtained polymer was dried and then subjected to mass spectrometry using "DART-MS."

[0137] The procedure for mass spectrometry using "DART-MS" is as follows. The polymer was decomposed by heating from 50°C to 500°C at a heating rate of 50°C / min on a heater (trade name "ionRocket" manufactured by Biochromato). Measurements were performed for 11 minutes per sample, including a 2-minute preheating time from room temperature to 50°C. The pyrolysis gas was continuously ionized with excited He gas using a "DART" ion source (trade name "DART-OS" manufactured by IonSense). MS ("LCMS-2020" manufactured by Shimadzu) spectra were recorded in positive ion mode at 50 scans / min, obtaining 550 spectra per sample. The mass range was 50-1500 m / z, the interval scale was 0.05 m / z, and the mass resolution was 2000.

[0138] Spectra were output in CDF file format and converted to Numpy format using the Python module netCDF4. All data processing was performed using Python 3.7 on a Windows 11 laptop equipped with an AMD Ryzen 9 4900HS without the assistance of an external GPU. Total processing time was 2-3 hours.

[0139] The following table summarizes the polymerization conditions for the M / S / B ternary system. In the table, "mass (mg)" represents the mass of the obtained polymer, "M initial fraction," "S initial fraction," and "B initial fraction" represent the charge ratios (by mass) of M, S, and B, respectively, "polymer. Time (h)" represents the reaction time (h), and "polymer. Temp (C)" represents the reaction temperature (°C). As shown in the table below, 85 different polymers were synthesized under different reaction conditions.

[0140]

[0141] As described above, the number of samples was 85, the temperature range of mass analysis was 200 to 450°C, and the m / z range was 50 to 410. Other hyperparameters were as follows:

[0142]

[0143] Figure 10 shows the model polymer spectrum for each multi-unit unit calculated by calculation. Since the number of types of monomers is three and the analysis is a triad, K is 13, and each spectrum has a peak at a reasonable position when compared with the sum of the mass numbers of the monomers. In the figure, the "(XXX)" written next to each spectrum indicates the ｌ " indicates the type of multi-unit, and "XXX" written at the peak position indicates the peak position of the identified multi-unit. Each model polymer spectrum in Figure 10 was reasonably assigned to a triad, and it became clear that the calculations were carried out as intended.

[0144] Example 2: Pentamer Analysis of St / BA Sequence analysis was carried out in the same manner as in Example 1, except that the monomer set was changed from MMA / St / BA to St / BA and pentamer analysis was performed. The following table shows the polymerization conditions. In the table, "mass (mg)" represents the mass of the obtained polymer, "S initial fraction" and "B initial fraction" represent the charge ratios of S and B (based on mass), respectively, "polymer. Time (h)" represents the reaction time (h), and "polymer. Temp (C)" represents the reaction temperature (°C). As shown in the table below, 81 different polymers were synthesized under different reaction conditions.

[0145]

[0146]

[0147] As described above, the number of samples was 81, the temperature range of mass analysis was 200 to 450°C, and the m / z range was 100 to 700. Other hyperparameters were as follows:

[0148]

[0149] The analysis was performed with K=9. When the number of types of monomers is 2 and the length of the multi-unit is 5, the theoretical number of combinations is 9. The results of the model polymer spectrum are shown in Figure 11. When compared with the sum of the mass numbers of the monomers, each spectrum had peaks at reasonable positions. In the figure, the "(XXXXXX)" written next to each spectrum indicates the ｌ " indicates the type of multi-unit, and "XXXXX" written at the peak position indicates the peak position of the identified multi-unit. Each model polymer spectrum in Figure 11 was reasonably assigned to a triad, and it became clear that the calculations were carried out as intended.

[0150] [Example 3: Comparison with NMR Measurement Results] Using the pentad basis spectrum learned in Example 2, a polymer was grown in a living radical polymerization system using two types of monomers, styrene (S) and butyl acrylate (B), and samples were collected over time and analyzed. The same samples were further analyzed by NMR, and the results were compared with the theoretical curve calculated using the Alfrey-Mayo equation.

[0151] In comparison with the NMR data, because NMR only provides information on the triad centered on "B," the basis spectrum of the pentads learned in Example 2 was downgraded (aggregated) to a triad composition to create a new basis spectrum. The method for this is described in detail below.

[0152] First, in accordance with the method already explained, the pentad composition of the target specimen is obtained by projecting it onto the hyperplane spanned by S and B. ｔｅｓｔ (Here, let's call it K-9). is multiplied from the right by the transformation matrix from pentad to triad. The transformation matrix is ​​shown in the table below.

[0153]

[0154] In the above table, "Sequence-defined copolymers" refers to the matrix representing the base spectrum of the pentads calculated in Example 2, and "B-centered triad matrix, T Ｂ" represents a transformation matrix to a triplet with B at the center (e.g., BBS), and "B-centered triad matrix, T Ｓ " represents the transformation matrix to a triplet with S at the center. In this verification, we used T, which is centered on B and for which NMR data can be obtained. Ｂ According to the above transformation matrix,

[0155]

[0156] The three-dimensional vector is resolved into the mass ratios of BBB, BBS, and SBS.

[0157] Living radical polymerization using two monomers, styrene (S) and butyl acrylate (B), was carried out according to the following procedure. 72.9 mg of 2-(dodecylthiocarbonothioylthio)-2-methylpropionic acid (DDMAT) and 9.9 mg of azobis(isobutylnitrile) (AIBN) were placed in a reaction vessel and the atmosphere was replaced with nitrogen. 2.1 mL of styrene, 2 mL of n-butyl acrylate, and 2 mL of 1,4-dioxane were placed in a separate vessel, and oxygen was removed by bubbling nitrogen gas for 30 minutes. This mixture was then added to the reaction vessel. The reaction vessel was heated to 70°C with stirring, and the polymerization solution was sampled periodically to determine conversion and sequence analysis. Figure 12 shows the relationship between polymerization time and conversion.

[0158] Figures 13 and 14 show the results of sequence analysis. Figure 13(A) shows the change in the content of BBB triads in the obtained copolymer. The horizontal axis represents the conversion rate (%), and the vertical axis represents the mass fraction of BBB triads. Similarly, Figure 13(B) shows the change in the content of BBS triads in the obtained copolymer, and Figure 14 shows the change in the content of SBS triads. In both triad analysis results, the NMR analysis results and the analysis results obtained in the examples of the analytical method of the present invention, "RQPMS; 'reference-free' quantitative pyrolysis MS," were in good agreement, and further, the results were in agreement with the theoretical curve calculated using the Alfrey-Mayo equation.

[0159] 10, 20 K-1-dimensional simplex 12 Inscribed hypersphere 13, 14, 24, 25 Endmember 16, 17, 21-23 Reference specimen 19 Region 30, 50 Sequence analysis device 31 Mass spectrometer 32 Information processing device 33, 73 Processor 34, 74 Storage device 35 Input / output interface (I / F) 41 Data matrix creation unit 42 Variant number determination unit 43 First NMF processing unit 44 Second NMF processing unit 45 Vector projection unit 46 Composition estimation unit 51 Model polymer spectrum identification unit 60 Polymerization condition proposal unit 62 Policy proposal unit 70 Automated synthesis device 71 Synthesis device 72 Control device 75 Supply mechanism 76 Reaction vessel

Claims

1. A method for analyzing a polymer sequence, comprising: estimating a content of a multi-unit formed by arranging a plurality of units derived from a monomer selected from a monomer set including two or more kinds of monomers in a polymer obtained by polymerizing the monomer, the method comprising the steps of: determining a number K of variants of the multi-unit according to the number of types of monomers included in the monomer set and the number of the units constituting the multi-unit; a reference specimen, which is a polymer composed of the monomer, and a suspected specimen are heated, and the gas components generated are sequentially ionized to obtain a data matrix including a two-dimensional mass spectrum of m / z versus heating temperature; performing a first NMF process of decomposing the data matrix into a non-negative matrix factorization and a product of a matrix representing a normalized basis spectrum and its intensity distribution matrix; performing a second NMF process in which the intensity distribution matrix of each of the specimens is decomposed into a non-negative matrix factorization, and is decomposed into a product of a matrix representing a mass fraction of a model polymer composed only of the multiplicities in the specimen and a matrix representing a feature vector of the model polymer; setting a K-1-dimensional simplex including all of the feature vectors of the specimen, with the feature vectors of the model polymer as end members; defining distances between the K end members and the feature vector of the estimated target specimen by a Riemannian metric distance that takes into account non-orthogonality of the basis spectrum of the first NMF processing, and estimating a content ratio of each of the multi-links in the estimated target specimen from a ratio of the distances.

2. When the number of variants K is 3 or more, at least one of the feature vectors of the reference sample is located in each of the outer regions of a hypersphere inscribed in the K-1 dimensional simplex, or the reference sample includes at least one of the end members. The sequence analysis method of claim 1.

3. After the second NMF treatment, The sequence analysis method of claim 1, further comprising: reconstructing the spectrum of the model polymer by a matrix product of a matrix representing the feature vector of the model polymer and a matrix representing the basis spectrum, and identifying the polyadduct to which the spectrum of the model polymer belongs.

4. 4. The method of claim 3, wherein the identification is performed by comparing the sum of the mass numbers of the units constituting the polyadduct with the m / z of a peak in the spectrum of the model polymer.

5. 4. The method for sequence analysis according to claim 3, wherein, when the identification results in the existence of a spectrum of the model polymer that is not assigned to any of the polyadducts, the number of variants K is changed, and the first NMF process, the second NMF process, and the identification are repeated.

6. The method for analyzing sequences according to claim 5 , wherein the modification is to reduce the number of variants K by a predetermined number.

7. 4. The method of claim 3, wherein, when the identification results in a spectrum of the model polymer that cannot be assigned to the polyadduct, the reference sample is added, and the acquisition of the data matrix, the first NMF processing, the second NMF processing, and the identification are repeated.

8. The number of types is j, and when j is 3 or more, the number of variants K is expressed by the formula: K= ｊ C ３ +3 ｊ C ２ + ｊ C １ The method of claim 1, wherein the sequence is determined by:

9. The sequence analysis method according to any one of claims 1 to 8, wherein the polymer comprises a resist resin.

10. The sequence analysis method according to claim 1 , wherein the polymerization methods of the reference sample and the suspected target sample are different.

11. A polymer sequence analysis device that estimates a content of multiunits formed by arranging a plurality of units derived from a monomer in a polymer obtained by polymerizing a monomer selected from a monomer set including two or more kinds of monomers, the device comprising: a mass spectrometer for heating a reference specimen, which is a polymer composed of the monomer, and a specimen, which is a suspected target specimen, and successively ionizing the gas components generated by the heating and continuously observing the mass spectra; and an information processing device for processing the observed mass spectrum, The information processing device includes: a data matrix creation unit for obtaining a data matrix including a two-dimensional mass spectrum of m / z versus heating temperature; a variant number determination unit that determines a variant number K of the multi-unit according to the number of types of monomers included in the monomer set and the number of the units constituting the multi-unit; a first NMF processor that performs NMF processing to decompose the data matrix into a matrix representing a normalized basis spectrum and a product of the matrix and an intensity distribution matrix; a second NMF processing unit that performs NMF processing to decompose the intensity distribution matrix of each of the specimens into a non-negative matrix factorization, and decomposes the matrix into a product of a matrix representing a mass ratio of a model polymer composed only of the multiplicities in the specimen and a matrix representing a feature vector of the model polymer, thereby obtaining a feature vector of the model polymer; a vector projection unit that sets a K-1-dimensional simplex including all of the feature vectors of the specimen, the feature vectors of the model polymer being end members; a composition estimation unit that defines a distance between the K end members and the feature vector of the estimated target specimen by a Riemannian metric distance that takes into account non-orthogonality of the basis spectrum of the first NMF processing, and estimates a content ratio of each of the multi-links in the estimated target specimen from a ratio of the distances.

12. The sequence analysis apparatus according to claim 11, further comprising a model polymer spectrum identification unit that reconstructs the spectrum of the model polymer by a matrix product of a matrix representing the feature vector of the model polymer and a matrix representing the base spectrum, and identifies the polyadduct to which the spectrum of the model polymer belongs.

13. 13. The sequence analysis apparatus according to claim 12, wherein the identification is performed by comparing the sum of the mass numbers of the units constituting the polyadduct with the m / z of a peak in the spectrum of the model polymer.

14. 13. The sequence analysis device according to claim 12, wherein, when the identification result indicates that there is a spectrum of the model polymer that is not assigned to any of the polyadrenals, the information processing device changes the number of variants K and repeats the NMF processing by the first NMF processing unit, the NMF processing by the second NMF processing unit, and the identification by the model polymer spectrum identification unit.

15. A sequence analysis device according to any one of claims 11 to 14, A policy proposal unit that is machine-learned using the sequence analysis result by the sequence analysis device and the polymerization conditions of the suspected target specimen as training data, The policy proposal unit is a polymerization condition proposal device that compares a sequence analysis result with a predetermined target sequence and proposes new polymerization conditions for obtaining a polymer having the target sequence.

16. The polymerization condition proposal device according to claim 15, and a polymer synthesis apparatus, The synthesis apparatus comprises: a monomer supply mechanism, a reaction tank that receives the monomer from the monomer supply mechanism and reacts the monomer, and a control device; The control device controls at least one selected from the group consisting of the supply mechanism and the reaction tank, based on the polymerization conditions proposed by the polymerization condition proposal device, to synthesize a new polymer.