Method for predicting physical properties of polymers and polymer

Abstract

To provide a highly accurate method for predicting a physical property of a polymer and a polymer having excellent physical properties such as an ability to selectively capture specific cells and biocompatibility.SOLUTION: A method for predicting a physical property of a polymer includes: step 1 of acquiring a monomer structure from a structure of a polymer; step 2 of converting the acquired monomer structure or a converted monomer structure into an expression format recognizable by a computer; step 3 of computing a descriptor from the expression format recognizable by the computer; step 4 of acquiring a physical property of the polymer, using it as an objective variable and the descriptor as an explanatory variable, performing regression calculation to generate a regression model; step 5 of computing an expression format of a new monomer from the expression format recognizable by the computer; step 6 of computing the descriptor from the expression format of the new monomer or a converted expression format of the new monomer; and step 7 of applying the descriptor of the new monomer to the regression model to predict a physical property of a polymer obtained by polymerizing the new monomer.SELECTED DRAWING: None

Description

[Technical field]

[0001] The present invention relates to a method for predicting physical properties of a polymer and the polymer. [Background technology]

[0002] It has been proposed to create a regression model that predicts the physical properties of compounds using machine learning, and predict the physical properties of new compounds based on the regression model. Furthermore, new compounds are synthesized, their physical properties are measured, and the predicted values ​​are verified. At the same time, the new compounds and their actually measured physical properties are added to improve the accuracy of the regression model. In addition, useful compounds are discovered from new compounds and their physical properties, and this is often done in the pharmaceutical field.

[0003] However, in the field of polymers, it cannot be said that such techniques are being used sufficiently, and in particular, they have not been fully implemented when it comes to predicting specific physical properties of polymers (e.g., water content, intermediate water content, amount of unfrozen water, adhesiveness to specific cells (cancer cells and stem cells), and biocompatibility).

[0004] Meanwhile, substrates coated with special polymers have been proposed as devices for capturing specific cells in blood and body fluids (blood cells, cancer cells present in blood and body fluids, etc.) (see Patent Document 1, etc.).

[0005] However, such a substrate may capture not only specific cells such as cancer cells but also blood cells at the same time, and therefore it is desired to provide a polymer that can selectively capture specific cells such as cancer cells and suppress the capture of normal cells such as blood cells. It would also be desirable to provide polymers that are more biocompatible. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Special Publication No. 2005-523981 Summary of the Invention [Problem to be solved by the invention]

[0007] An object of the present invention is to solve the above problems and to provide a method for predicting the physical properties of a polymer with high accuracy, and to provide a polymer having excellent physical properties such as the ability to selectively capture specific cells and biocompatibility. [Means for solving the problem]

[0008] The present invention includes the steps of: obtaining a monomer structure from a polymer structure; Step 2 of converting the monomer structure directly or in a converted form and then converting it into a representation format that can be recognized by a computer; A step 3 of computing a descriptor from the computer-recognized representation; Step 4: Obtain the physical properties of the polymer as the objective variables, use the descriptors as explanatory variables, perform regression calculations, and create a regression model; Step 5 of calculating the representation of the new monomer from the representations that the computer can recognize; Step 6 of calculating the descriptors of the new monomers, either directly or in a converted form; Step 7: applying the descriptors of the new monomer to the regression model to predict the properties of a polymer polymerized from the new monomer; The present invention relates to a method for predicting the physical properties of a polymer comprising the steps of: Effect of the Invention

[0009] According to the present invention, since the method for predicting the physical properties of a polymer includes the above steps 1 to 7, it is possible to provide a highly accurate method for predicting the physical properties of a polymer. Therefore, it is expected that the method will enable rapid discovery of many polymers (candidate polymers) with excellent physical properties. It is also expected that the present invention will provide polymers that have excellent physical properties, such as the ability to selectively capture specific cells, such as cancer cells, and biocompatibility. [Brief description of the drawings]

[0010] [Figure 1] FIG. 13 is a diagram showing the relationship between the predicted values ​​of the intermediate water contents of the monomer structures of 33 new monomers and the predicted values ​​of the intermediate water contents of the compounds obtained by converting the monomer structures, using the regression model created in step 4. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] <Method of predicting polymer properties> The method for predicting physical properties of a polymer of the present invention includes the steps of: step 1 acquiring a monomer structure from a polymer structure; step 2 acquiring the monomer structure as is or after conversion, and further converting it into an expression format recognizable by a computer; step 3 calculating a descriptor from the expression format recognizable by a computer; step 4 acquiring the physical properties of the polymer as a response variable and performing regression calculations using the descriptors as explanatory variables to create a regression model; step 5 calculating an expression format of a new monomer from the expression format recognizable by a computer; step 6 acquiring the expression format of the new monomer as is or after conversion, and calculating a descriptor; and step 7 applying the descriptor of the new monomer to the regression model to predict the physical properties of a polymer polymerized from the new monomer.

[0012] The polymer property prediction method includes step 1 of acquiring a monomer structure from a polymer structure. In step 1, for example, a monomer structure having a carbon-carbon double bond constituting the polymer is acquired from the polymer structure, and specifically, when the polymer structure is polyacrylic acid, acrylic acid, which is the structure of a monomer constituting polyacrylic acid, is acquired.

[0013] In the above step 1, the polymer is not particularly limited, and any known polymer can be used as appropriate. Examples of the polymer include a homopolymer of one type of monomer and a copolymer of two or more types of monomers. The polymer may be used alone or in combination of two or more types.

[0014] The polymer can be produced by a known method, for example, by using a solution of the monomers constituting the polymer and polymerizing the monomers by a known method. The solvent of the monomer solution is not particularly limited, and for example, the solvents described below can be used. Among them, toluene and methanol are preferable.

[0015] From the viewpoints of the selective capture of specific cells such as cancer cells and biocompatibility, the above polymer is desirably a polymer having hydrophilic properties (hydrophilic polymer).

[0016] Examples of the hydrophilic polymer include homopolymers and copolymers of one or more hydrophilic monomers, copolymers of one or more hydrophilic monomers and one or more other monomers, etc. These may be used alone or in combination of two or more.

[0017] The hydrophilic monomer is not particularly limited, and for example, various monomers having a hydrophilic group can be used. Examples of the hydrophilic group include known hydrophilic groups such as an amide group, a sulfate group, a sulfonic acid group, a carboxyl group, a hydroxyl group, an amino group, an amide group, and an oxyethylene group.

[0018] Specific examples of the hydrophilic monomer include (meth)acrylic acid, (meth)acrylic acid esters, (alkoxyalkyl (meth)acrylates such as methoxyethyl (meth)acrylate, hydroxyalkyl (meth)acrylates such as hydroxyethyl (meth)acrylate), (meth)acrylamide, and (meth)acrylamide derivatives having a cyclic group ((meth)acryloylmorpholine, etc.). From the viewpoint of selective capture of specific cells such as cancer cells and biocompatibility, (meth)acrylic acid, (meth)acrylic acid esters, and alkoxyalkyl (meth)acrylates are preferred, alkoxyalkyl (meth)acrylates are more preferred, and 2-methoxyethyl acrylate is particularly preferred. These may be used alone or in combination of two or more.

[0019] The other monomers may be appropriately selected within a range that does not inhibit the action and effect of the hydrophilic polymer. Specific examples of the other monomers include aromatic monomers such as styrene, vinyl acetate, and N-isopropylacrylamide that can impart temperature responsiveness. These may be used alone or in combination of two or more.

[0020] Specific examples of the above homopolymers and copolymers include homopolymers composed of one hydrophilic monomer such as polyacrylic acid, polyacrylic acid esters, polymethacrylic acid, polymethacrylic acid esters, polyacryloylmorpholine, polymethacryloylmorpholine, polyacrylamide, polymethacrylamide, polyalkoxyalkyl acrylates, and polyalkoxyalkyl methacrylates; copolymers composed of two or more of the above-listed hydrophilic monomers; copolymers composed of one or more of the above-listed hydrophilic monomers and one or more of the above-listed other monomers; etc. The above hydrophilic polymers may be used alone or in combination.

[0021] Specific examples of homopolymers and copolymers of the above hydrophilic monomers include polyacrylic acid, polyacrylic acid esters, polymethacrylic acid, polymethacrylic acid esters, polyacryloylmorpholine, polymethacryloylmorpholine, polyacrylamide, polymethacrylamide, polyalkoxyalkyl acrylates, and polyalkoxyalkyl methacrylates.

[0022] From the viewpoints of selective capture of specific cells such as cancer cells and biocompatibility, the hydrophilic polymer is desirably a polymer represented by the following formula (I). These may be used alone or in combination of two or more kinds. [ka] (In the formula, R 51 is a hydrogen atom or a methyl group, R 52 represents an alkyl group. p is 1 to 8, m is 1 to 5, and n is the number of repetitions.

[0023] From the viewpoints of selective capture of specific cells such as cancer cells and biocompatibility, the polymer represented by the above formula (I) is preferably a polymer represented by the following formula (I-1), which may be used alone or in combination of two or more kinds. [ka] (In the formula, R 51 is a hydrogen atom or a methyl group, R 52 represents an alkyl group, m is 1 to 5, and n is the number of repetitions.

[0024] R 52 The number of carbon atoms in the alkyl group of R is preferably 1 to 10, and more preferably 1 to 5. 52 is particularly preferably a methyl group or an ethyl group. p is preferably an integer of 1 to 5, and more preferably an integer of 1 to 3. m is preferably an integer of 1 to 3. n (the number of repeating units) is preferably an integer of 15 to 1,500, and more preferably an integer of 40 to 1,200.

[0025] From the viewpoints of selective capture of specific cells such as cancer cells and biocompatibility, the hydrophilic polymer is desirably a copolymer of a compound represented by the following formula (II) and other monomers. These may be used alone or in combination of two or more kinds.

[0026] [ka] (In the formula, R 51 , R 52 , p, m are the same as above.)

[0027] From the viewpoint of selective capture of specific cells such as cancer cells and biocompatibility, the compound represented by the above formula (II) is preferably a compound represented by the following formula (II-1), which may be used alone or in combination of two or more kinds. [ka] (In the formula, R 51 , R 52 , m is the same as above.)

[0028] Among the above hydrophilic polymers, from the viewpoints of selective capture of specific cells such as cancer cells and biocompatibility, the hydrophilic polymer represented by the above formula (I) is preferred, the hydrophilic polymer represented by the above formula (I-1) is more preferred, and polyalkoxyalkyl acrylates and polyalkoxyalkyl methacrylates are even more preferred.

[0029] From the viewpoint of obtaining a better effect, the number average molecular weight (Mn) of the polymer is preferably 8000 to 150000, more preferably 10000 to 60000, and further preferably 10000 to 39000. When the polymer is a hydrophilic polymer, the same thickness (film thickness) is also desirable. In this specification, the number average molecular weight (Mn) and weight average molecular weight (Mw) can be determined by standard polystyrene conversion based on the measured values ​​obtained by gel permeation chromatography (GPC) (GPC-8000 series manufactured by Tosoh Corporation, detector: differential refractometer, column: TSKGEL SUPERMALTPORE HZ-M manufactured by Tosoh Corporation).

[0030] In the above-mentioned method for predicting physical properties of a polymer, in the above-mentioned step 1, for example, when the polymer structure is poly(2-methoxyethyl acrylate), 2-methoxyethyl acrylate (monomer structure) is obtained.

[0031] In the polymer property prediction method, after step 1, step 2 is carried out in which the monomer structure acquired in step 1 is converted or used as is, and then converted into an expression format that can be recognized by a computer. That is, in step 2, the compound of the monomer structure itself acquired in step 1 (the acquired monomer structure as is) or the compound obtained by converting the monomer structure is converted into an expression format that can be recognized by a computer.

[0032] The method for converting the monomer structure obtained in step 1 is not particularly limited as long as it can convert the monomer structure, and examples thereof include (1) a method for converting the monomer structure obtained in step 1 (monomer structure as is (with carbon-carbon double bond)) into a compound in which the carbon-carbon double bond is opened and saturated hydrocarbon groups such as hydrogen and methyl groups are introduced at both ends (conversion to a monomer (without carbon-carbon double bond)), and (2) a method for converting the monomer structure obtained in step 1 (monomer structure as is) into a multimer in which the carbon-carbon double bond is opened and bonded with a carbon-carbon single bond, and saturated hydrocarbon groups such as hydrogen and methyl groups are introduced at both ends (conversion to a multimer without carbon-carbon double bonds such as a dimer (without carbon-carbon double bond) or trimer (without carbon-carbon double bond)). When converting into the multimer, a method for converting into a compound in which the carbon-carbon double bond of the monomer structure obtained in step 1 (monomer structure as is) is opened and bonded with a carbon-carbon single bond, and methyl groups are introduced at both ends is preferable.

[0033] In the above step 2, the compound of the monomer structure itself obtained in the above step 1 (the obtained monomer structure as it is) and the polymer having no carbon-carbon double bond obtained by converting it are converted into an expression format that can be recognized by a computer, and examples of the computer include general personal computers, servers, workstations, calculation nodes, microcomputers, etc. The hardware configuration of the computer can be a known one.

[0034] Examples of expression formats that can be recognized by the above computer include well-known ones such as the Simplified Molecular Input Line Entry System (SMILES) expression format and the MOL file expression format.

[0035] In the above-mentioned method for predicting physical properties of a polymer, for example, when the polymer structure is poly(2-methoxyethyl acrylate), first, the 2-methoxyethyl acrylate (monomer structure as it is (with carbon-carbon double bond)) obtained in the above-mentioned step 1 or the 2-methoxyethyl acrylate is converted into the following compounds (A) to (C), etc., in the above-mentioned step 2. (A) Monomer (no carbon-carbon double bond): A compound in which the carbon-carbon double bond of 2-methoxyethyl acrylate is opened and methyl groups are introduced at both ends. (B) Dimer (no carbon-carbon double bond): A compound in which the carbon-carbon double bond of 2-methoxyethyl acrylate is opened to form a dimer with a carbon-carbon single bond, and methyl groups are introduced at both ends of the dimer. (C) Trimer (no carbon-carbon double bond): A compound in which the carbon-carbon double bond of 2-methoxyethyl acrylate is opened to form a trimer with a carbon-carbon single bond, and methyl groups are introduced at both ends of the trimer.

[0036] In step 2, 2-methoxyethyl acrylate and the above compounds (A) to (C) are further converted into, for example, the following SMILES representation format (SMILES notation), which is a representation format that can be recognized by a computer. 2-Methoxyethyl acrylate: C=C(C(=O)OCCOC) (Hydrogens are not shown. Branches are in parentheses.) (A) Monomer (no carbon-carbon double bond): CCC(C(=O)OCCOC)C (B) Dimer (no carbon-carbon double bond): CCC(C(=O)OCCOC)CC(C(=O)OCCOC)C (C) Trimer (no carbon-carbon double bond): CCC(C(=O)OCCOC)CC(C(=O)OCCOC)CC(C(=O)OCCOC)C

[0037] The method for predicting physical properties of a polymer includes step 3, which is a step of calculating a descriptor from an expression format that can be recognized by a computer, in addition to step 2. That is, for a polymer whose physical properties, structure, etc. are known, a monomer structure having a carbon-carbon double bond that constitutes the polymer (monomer structure as is (with carbon-carbon double bond)) or a compound obtained by converting the monomer structure (multimers without carbon-carbon double bonds, such as the above-mentioned monomer (without carbon-carbon double bond), the above-mentioned dimer (without carbon-carbon double bond), and the above-mentioned trimer (without carbon-carbon double bond)) is converted into an expression format that can be recognized by a computer, and then a descriptor is calculated from the expression format.

[0038] The physical properties of the polymer are not particularly limited, and include glass transition temperature (Tg), melting point, refractive index, contact angle, surface free energy, density, water absorption rate, intermediate water content, unfrozen water content, selective adhesion ability to specific cells such as cancer cells, biocompatibility, etc. Among these, intermediate water content, selective adhesion ability to specific cells, and biocompatibility can be suitably predicted.

[0039] In the above step 3, examples of the expression format when the monomer structure itself constituting the polymer or the compound obtained by converting the monomer structure is converted into an expression format that can be recognized by a computer include the above-mentioned SMILES expression format and MOL file expression format. Then, from these expression formats, a descriptor (numerical data) that digitizes the chemical structure is calculated. For example, in Python, 208 or 400 descriptors are calculated from SMILES using RDkit. Descriptors can also be calculated using mordred, PaDEL, alvaDesc, etc. Another method for digitizing a chemical structure is fingerprinting.

[0040] In the above-mentioned method for predicting physical properties of polymers, the above-mentioned step 3, for example, creates a SMILES representation format of the monomer structure constituting each of 11 types of polymers whose intermediate water content properties are known (the monomer structure as is (with carbon-carbon double bonds)) and compounds into which the monomer structure is converted (the monomer (without carbon-carbon double bonds), the dimer (without carbon-carbon double bonds), and the trimer (without carbon-carbon double bonds)), and calculates 208 numerical data descriptors reflecting the chemical structure from all of the SMILES representation formats.

[0041] It should be noted that step 3 may be performed separately from steps 1 and 2, and the order of steps 1 and 2 does not matter. For example, step 3 may be performed after steps 1 and 2, step 3 may be performed separately and simultaneously with steps 1 and 2, or step 1 and 2 may be performed after step 3.

[0042] In the method for predicting physical properties of a polymer, after steps 1, 2, and 3, step 4 is carried out in which the physical properties of the polymer are acquired and used as response variables, and the descriptors are used as explanatory variables, regression calculations are performed, and a regression equation (regression model) is created by using the physical properties of the polymer as response variables and the descriptors (for example, the above 208 descriptors) as explanatory variables in step 4.

[0043] The algorithm for constructing the model can be a known one such as an algorithm used in machine learning. When performing regression calculation to create a regression formula (regression model), known methods such as multiple regression calculation, support vector regression calculation, decision tree calculation, random forest regression calculation, XGBoost regression calculation, LightGBM regression calculation, CatBoost regression calculation, Gaussian process regression calculation, neural network, and deep learning can be used as the regression calculation. The regression calculation may be one using regression calculation or a combination of two or more.

[0044] In the above-mentioned method for predicting physical properties of a polymer, in step 4, for example, four models for predicting the intermediate water content can be created using the monomer structure constituting the polymer (the monomer structure as is (with carbon-carbon double bond)) and compounds obtained by converting the monomer structure (the monomer (without carbon-carbon double bond), the dimer (without carbon-carbon double bond), and the trimer (without carbon-carbon double bond)) with the intermediate water content as the objective variable and 208 descriptors as explanatory variables, using ensemble learning methods such as random forest, XGBoost regression calculation, LightGBM regression calculation, and CatBoost regression calculation. In addition, in the case of Boost regression calculations, neural networks, and deep learning, hyperparameters are adjusted to reduce overfitting and to enable the entire data set to be used for training. There are methods for adjusting hyperparameters, such as grid research, random research, and Bayesian optimization. Hyperparameters are parameters that are determined by humans before learning. Hyperparameters are values ​​required to control the behavior of various algorithms, and typical examples include the maximum depth of a decision tree, the maximum number of leaves of a decision tree, the minimum number of samples contained in one leaf, the proportion of samples randomly extracted in each decision tree, the proportion of columns randomly extracted in each decision tree, the L1 regularization term for the weight of the leaves of a decision tree, the L2 regularization term for the weight of the leaves of a decision tree, the lower limit of loss reduction by adding leaves to a decision tree, the method of optimizing the loss function, the number of boosting rounds, the number of epochs, the learning rate, the threshold, the mini-batch size, the number of layers, and the number of neurons per layer.

[0045] In the method for predicting physical properties of a polymer, after the above steps 1, 2, 3, and 4, step 5 is performed in which the expression format of a new monomer is calculated from an expression format that can be recognized by a computer. That is, in the above step 5, the expression format of the new monomer is calculated (generated) from the expression format (SMILES expression format, MOL file expression format, etc.) of the monomer structure (with carbon-carbon double bonds) that constitutes the polymer.

[0046] The calculation (generation) of a representation format of a new monomer that can be recognized by a computer can be performed, for example, by dividing the chemical structure of the compound in the representation format into fragments and then mechanically combining them using the BRICS (Breaking of Retrosynthetically Interesting Chemical Substructures) algorithm, the RECAP (Retrosynthetic Combinatiorial Analysis Procedure) algorithm, the DOCS (Design Of Genuine Structures) algorithm, etc. Examples of the representation format that can be recognized by a computer include the above-mentioned SMILES representation format and MOL file representation format.

[0047] For example, using RECAP, organic molecules are broken down into fragments at points in the molecule where bonds are easily broken, focusing on amide, ester, amine, NC in urea, ether, C=C, ammonium, NS in sulfanamide, aromatic-aromatic, N(in aromatic)-C(sp3) and N(in lactam)-C(sp3) bonds. Using BRICS, 16 types of bonds are broken down into fragments in a similar manner to RECAP.

[0048] When a new monomer is calculated (generated) in a form that can be recognized by a computer, a compound that does not contain a carbon-carbon double bond may be generated. In that case, in order to predict the physical properties of the polymer, the compound that does not contain a carbon-carbon double bond is removed from the generated monomers.

[0049] In the above-mentioned method for predicting physical properties of polymers, in step 5, 74 compounds are generated using BRICS or the like from the SMILES representation of 11 monomer structures (monomer structures as they are (with carbon-carbon double bonds)) that make up the 11 types of polymers whose intermediate water content properties are known as described above, and from these, 33 compounds that can become monomers are selected (41 are removed), and the SMILES representation of 33 new monomers is calculated (generated).

[0050] The method for predicting physical properties of a polymer includes step 6, in which the representation format (a representation format recognizable by a computer) of the new monomer calculated (generated) in step 5 is used as is or converted to calculate a descriptor. That is, in step 6, the representation format (a representation format recognizable by a computer) of the new monomer calculated (generated) in step 5 is first converted into a representation format of a compound of the monomer structure itself (monomer structure as is) or a compound obtained by converting the monomer structure, in the same manner as in step 2, and then a descriptor (numerical data) is calculated from the representation format in the same manner as in step 3.

[0051] In the above-mentioned method for predicting physical properties of a polymer, in the above-mentioned step 6, for example, for the SMILES representation formats of the 33 new monomers calculated (generated) above, SMILES representation formats of the monomer structures of the 33 new monomers (monomer structures as they are (with carbon-carbon double bonds)) and compounds obtained by converting the monomer structures (monomers (without carbon-carbon double bonds), dimers (without carbon-carbon double bonds), and trimers (without carbon-carbon double bonds)) are created, and then, descriptors (numerical data) are calculated for each.

[0052] The method for predicting physical properties of a polymer includes step 7 of applying the descriptors of the new monomer to the regression model to predict the physical properties of a polymer polymerized from the new monomer. That is, in step 7, the descriptors (numerical data) of the new monomer calculated in step 6 are applied to the regression model created in step 4 to predict the physical properties of a polymer of the new monomer.

[0053] In the above step 7, for example, the intermediate water content of the polymer obtained by polymerizing the above 33 new monomers is predicted by applying the descriptors (numerical data) of the monomer structures (monomer structures as they are (with carbon-carbon double bonds)) of the above 33 new monomers calculated (generated) in the above step 6 and the compounds obtained by converting the monomer structures (the above monomers (without carbon-carbon double bonds), the above dimers (without carbon-carbon double bonds), and the above trimers (without carbon-carbon double bonds)) to the regression model created in the above step 4.

[0054] Specifically, Figure 1 shows the relationship between the predicted intermediate water content of polymers polymerized from the monomer structures of 33 new monomers (monomer structures as they are (with carbon-carbon double bonds)) and the predicted intermediate water content of polymers polymerized from compounds obtained by converting the monomer structures (monomers (without carbon-carbon double bonds), dimers (without carbon-carbon double bonds), and trimers (without carbon-carbon double bonds)), using the regression model created in step 4 above. It shows that there is no difference in the predicted intermediate water content between the presence and absence of carbon-carbon double bonds, and that there is no difference even when the number of monomers is increased.

[0055] Furthermore, when compounds are generated using the BRICS algorithm from the monomer structures of the above 11 types of polymers (monomer structures as is (with carbon-carbon double bonds)) with the upper limit set to 2000, over 1600 compounds (SMILES notation) are generated. Of these, compounds with no carbon-carbon double bonds and compounds with two or more carbon-carbon double bonds are removed, and 625 compounds with only one carbon-carbon double bond are selected. Descriptors for these 625 compounds are calculated, and the intermediate water content of these 625 compounds is predicted from the prediction model for compounds with only one carbon-carbon double bond.

[0056] Of these, those with intermediate water volumes of 0.01 to 0.15 are shown in Table 1-1, Table 1-2 (continued from Table 1-1), and Table 1-3 (continued from Table 1-1). In addition, those with intermediate water volumes of 0.22 to 0.5 are shown in Table 2-1, Table 2-2 (continued from Table 2-1), Table 2-3 (continued from Table 2-1), Table 2-4 (continued from Table 2-1), Table 2-5 (continued from Table 2-1), Table 2-6 (continued from Table 2-1), and Table 2-7 (continued from Table 2-1).

[0057] [Table 1-1]

[0058] [Table 1-2]

[0059] [Table 1-3]

[0060] [Table 2-1]

[0061] [Table 2-2]

[0062] [Table 2-3]

[0063] [Table 2-4]

[0064] [Table 2-5]

[0065] [Table 2-6]

[0066] [Table 2-7]

[0067] By using the above machine learning method, it was possible to create a regression equation (regression model) to predict the physical properties of a polymer. It was also found that when the intermediate water content is 0.01 to 0.15 (Tables 1-1 to 1-3), the polymer has good adhesiveness to specific cells such as cancer cells. Therefore, polymers calculated to have an intermediate water content of 0.01 to 0.15 using the above polymer property prediction method have excellent selective adhesiveness to specific cells, and are particularly excellent in selective adhesiveness to cancer cells. For example, polymers polymerized from monomers expressed in the SMILES format of Tables 1-1 to 1-3 had good adhesiveness to cancer cells when the intermediate water content was 0.01 to 0.15.

[0068] Furthermore, it was found that when the intermediate water content is 0.22 to 0.5 (Tables 2-1 to 2-7), the polymer has good biocompatibility. Therefore, polymers whose intermediate water content is calculated to be 0.22 to 0.5 by the above polymer property prediction method have excellent biocompatibility. For example, polymers polymerized from monomers expressed in the SMILES format of Tables 2-1 to 2-7 had good biocompatibility when the intermediate water content was 0.22 to 0.5.

[0069] The above is the result of random forest regression calculation with double bonds. In contrast, a prediction model was created for a monomer without double bonds using XGBoost regression calculation and CatBoost regression calculation (with the high parameters manually adjusted to reduce RMSE). In addition, a prediction model was created using the Optuna framework in XGBoost regression calculation, with the high parameters adjusted by Bayesian optimization. The monomer in SMILES notation in Table 3 was converted to a monomer without double bonds (with double bonds opened and methyl groups attached to both ends), and the descriptors were calculated. The intermediate water amounts predicted by the three prediction models are shown in Table 3. The predicted values ​​fell within a certain range, indicating that the predicted values ​​were fairly accurate.

[0070] [Table 3]

[0071] The present invention (1) comprises a step 1 of obtaining a monomer structure from a polymer structure; Step 2 of converting the monomer structure directly or in a converted form and then converting it into a representation format that can be recognized by a computer; A step 3 of computing a descriptor from the computer-recognized representation; Step 4: Obtain the physical properties of the polymer as the objective variables, use the descriptors as explanatory variables, perform regression calculations, and create a regression model; Step 5 of calculating the representation of the new monomer from the representations that the computer can recognize; Step 6 of calculating the descriptors of the new monomers, either directly or in a converted form; Step 7: applying the descriptors of the new monomer to the regression model to predict the properties of a polymer polymerized from the new monomer; The present invention relates to a method for predicting the physical properties of a polymer comprising the steps of:

[0072] The present invention (2) relates to the method for predicting physical properties of a polymer according to the present invention (1), wherein the expression format recognizable by the computer is a SMILES expression format or a MOL file expression format.

[0073] The present invention (3) relates to the method for predicting physical properties of a polymer according to the present invention (1) or (2), wherein the regression calculation is at least one selected from the group consisting of multiple regression calculation, support vector regression calculation, decision tree calculation, random forest regression calculation, XGBoost regression calculation, LightGBM regression calculation, CatBoost regression calculation, Gaussian process regression calculation, neural network, and deep learning.

[0074] The present invention (4) is a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (3), wherein the algorithm for calculating the new monomer expression form is at least one selected from the group consisting of BRICS, RECAP, and DOCS.

[0075] The present invention (5) is a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (4), wherein the physical property of the polymer is at least one selected from the group consisting of glass transition temperature, melting point, refractive index, contact angle, surface free energy, density, water absorption rate, intermediate water content, non-freezing water content, selective adhesion ability to specific cells, and biocompatibility.

[0076] The present invention (6) is a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (5), wherein the method for converting the monomer structure is a method for opening the carbon-carbon double bond of the monomer structure to form a polymer bonded with a carbon-carbon single bond, and for producing a compound having methyl groups introduced at both ends thereof.

[0077] The present invention (7) is a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (6), in which the step 2 converts the monomer structure directly or after converting it into a polymer of the monomer structure into an expression format that can be recognized by a computer.

[0078] The present invention (8) is a polymer having an intermediate water content calculated to be 0.01 to 0.15 by a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (7).

[0079] The present invention (9) is the polymer according to the present invention (8), which is a polymer obtained by polymerizing monomers represented in the SMILES format of Tables 1-1 to 1-3.

[0080] The present invention (10) relates to the polymer according to the present invention (8) or (9), which has selective adhesion ability to specific cells.

[0081] The present invention (11) relates to the polymer according to the present invention (10), wherein the specific cells are cancer cells.

[0082] The present invention (12) is a polymer having an intermediate water content calculated to be 0.22 to 0.5 by a method for predicting physical properties of a polymer in any combination with any of the present inventions (1) to (7).

[0083] The present invention (13) is a polymer according to the present invention (12), which is a polymer obtained by polymerizing monomers represented in the SMILES format of Tables 2-1 to 2-7.

[0084] The present invention (14) relates to the biocompatible polymer according to the present invention (12) or (13).

Claims

1. Step 1, obtaining a monomer structure from the structure of a polymer; Step 2: converting the monomer structure directly or in a converted form, and then converting it into a representation format that can be recognized by a computer; A step 3 of computing a descriptor from a computer-recognizable representation; Step 4: obtaining the physical properties of the polymer as objective variables and using the descriptors as explanatory variables to perform regression calculations and create a regression model; Step 5 of calculating a representation of a new monomer from a representation that can be recognized by a computer; Step 6 of calculating the descriptors of the new monomers, either directly or in a converted form; Step 7: applying the descriptors of the new monomer to the regression model to predict the properties of a polymer polymerized from the new monomer; A method for predicting the physical properties of a polymer comprising the steps of:

2. 2. The method for predicting physical properties of a polymer according to claim 1, wherein the expression format that can be recognized by a computer is the SMILES expression format or the MOL file expression format.

3. 2. The method for predicting physical properties of a polymer according to claim 1, wherein the regression calculation is at least one selected from the group consisting of multiple regression calculation, support vector regression calculation, decision tree calculation, random forest regression calculation, XGBoost regression calculation, LightGBM regression calculation, CatBoost regression calculation, Gaussian process regression calculation, neural network, and deep learning.

4. 2. The method for predicting physical properties of a polymer according to claim 1, wherein the algorithm for calculating the expression form of the new monomer is at least one selected from the group consisting of BRICS, RECAP, and DOCS.

5. The method for predicting physical properties of a polymer according to claim 1, wherein the physical property of the polymer is at least one selected from the group consisting of glass transition temperature, melting point, refractive index, contact angle, surface free energy, density, water absorption rate, intermediate water content, non-frozen water content, selective adhesion ability to specific cells, and biocompatibility.

6. 2. The method for predicting physical properties of a polymer according to claim 1, wherein the method for converting the monomer structure is a method for opening the carbon-carbon double bond of the monomer structure to form a polymer bonded with a carbon-carbon single bond, and for introducing methyl groups into both ends of the polymer to form a compound.

7. 2. The method for predicting physical properties of a polymer according to claim 1, wherein said step 2 converts said monomer structure directly or after converting it into a polymer of said monomer structure into an expression format that can be recognized by a computer.

8. A polymer having an intermediate water content of 0.01 to 0.15 calculated by the method for predicting physical properties of a polymer according to claim 1.

9. The polymer according to claim 8, which is a polymer obtained by polymerizing monomers represented by the SMILES expressions in Tables 1-1 to 1-3.

10. The polymer according to claim 8, which has selective adhesion to specific cells.

11. The polymer according to claim 10 , wherein the specific cells are cancer cells.

12. A polymer having an intermediate water content of 0.22 to 0.5 calculated by the method for predicting physical properties of a polymer according to claim 1.

13. The polymer according to claim 12, which is a polymer obtained by polymerizing monomers represented by the SMILES expression formats in Tables 2-1 to 2-7.

14. The polymer of claim 12 which is biocompatible.

Images