Structure analysis method, structure analysis device, and program therefor

Abstract

To predict a physical property attributable to a microstructure in the inside of a molecule with high accuracy while suppressing a calculation load.SOLUTION: A structure analysis method pertaining to the present invention comprises: a step of dividing a space of an analysis object into a plurality of spaces; a step of calculating an atom occupancy rate of each of the divided spaces; a step of identifying one or more voids on the basis of the space(s) whose prescribed occupancy rate is low; and a step of outputting the identified void(s). A structure analysis device pertaining to the present invention comprises: a space division part for dividing the space to be analyzed into the plurality of spaces; an atom occupancy rate calculation part for calculating the atom occupancy rate of each of the spaces divided by the space division part; a void identification part for identifying the void(s) on the basis of the space(s) whose prescribed occupancy rage is low; and a void output part for outputting the void(s) identified by the void identification part.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a structure analysis method and a structure analysis apparatus.

Background Art

[0002] In recent years, the accuracy of identifying molecular structures using X-rays, neutron rays, NMR, etc. has been improving, and the identified molecular structures are registered in a database, making a lot of data available.

[0003] As one of the methods for extracting feature quantities from molecular structures, there is a quantitative structure-activity relationship analysis method (QSAR). QSAR analysis is a technique that calculates structural, quantum chemical, and physicochemical feature quantities called structure descriptors from a molecular structure, and statistically predicts the biological activity of a compound using these as explanatory variables. In the case of this technique, mainly in the field of drug discovery, a method has been proposed in which deep learning technology is used to perform information reduction using a molecular image file as input information and perform QSAR with a lower-dimensional data representation. In such a method, in order to construct an information reduction model, it is necessary to obtain conformational information of various molecules and create teacher data using a huge number of molecular image files from various angles. Constructing such an information reduction model is costly, and it is difficult to verify whether the obtained reduction model functions effectively.

[0004] Also, as one of the feature quantities of a molecule, the concept of free volume is used. For example, as an experimental measurement method of free volume, there is a positron annihilation lifetime spectroscopy (PALS). In this method, positrons are used as probes, and the size of the free volume is measured from the lifetime of the positrons. Thus, the measured value of the free volume obtained experimentally is generally useful when specifying the relevance with the macroscopic physical properties of the molecule. However, it is difficult to specify the relevance with the physical properties caused by the microscopic structure inside the molecule, such as what is near a specific atom.

[0005] Here, as one approach to actively investigate the relationship between the microscopic structure of molecules and free volume, for example, there is a method that utilizes molecular simulations such as molecular dynamics as described in Non-Patent Document 1. In such an approach, the definition of free volume is defined as the volume in which one atom focused on within the molecule can move around.

Prior Art Documents

Non-Patent Documents

[0006]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] In the approach to the microscopic structure of molecules based on free volume, since it is necessary to execute simulations for moving atoms one by one, the computational load increases as the molecule becomes larger, and there is a possibility that the simulation becomes difficult or costly. Also, free volume is defined independently in each field, and for example, the volume measured by the above-mentioned PALS and the volume calculated by molecular simulation do not necessarily match.

[0008] An object of the present invention is to accurately predict the physical properties resulting from the microscopic structure inside the molecule while suppressing the computational load.

Means for Solving the Problems

[0009] If an example of the structural analysis method of the present invention is given, it includes steps of dividing the space to be analyzed into a plurality of parts, calculating the occupancy rate of atoms in each divided space, identifying vacancies based on spaces with an occupancy rate lower than a predetermined occupancy rate, and outputting the identified vacancies.

[0010] Also, if an example of the structural analysis apparatus of the present invention is given, it includes a space division unit that divides the space to be analyzed into a plurality of parts, an atomic occupancy rate calculation unit that calculates the occupancy rate of atoms in each space divided by the space division unit, a vacancy identification unit that identifies vacancies based on spaces with an occupancy rate lower than a predetermined occupancy rate, and a vacancy output unit that outputs the vacancies identified by the vacancy identification unit.

Effect of the Invention

[0011] According to the present invention, for physical properties caused by the microscopic structure inside a molecule, it is possible to accurately predict with a suppressed computational load.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Best Mode for Carrying Out the Invention

[0013] Embodiments of the present invention will be described with reference to FIGS. 1 and 2.

[0014] FIG. 1 is a configuration diagram showing a structure analysis system in which a data input / output device and an analysis device that performs calculations and recordings are separated into different computers (computers). As shown in FIG. 1, a data input / output device 402 and an analysis device 410 are connected by a network 420. Therefore, in the data input / output device 402 operated by the user, there is no need to perform calculations or recordings for structure analysis.

[0015] The data input / output device 402 is composed of an input device 413, a screen display unit 414, an input / output control unit 412, and a network I / O unit 404. On the other hand, the analysis device 410 is composed of a CPU 408, a memory 406, a network I / O unit 404, a storage 418, and a bus unit 409.

[0016] The structure analysis device can be realized in the computer of the analysis device 410 by storing a predetermined program on the memory 406 and the CPU 408 executing the program on the memory 406. This predetermined program is stored directly on the memory 406 from a storage medium in which the program is stored via a reading device (not shown), or from the network 420 via the network I / O unit 404, or is once stored in the storage 418 and then stored in the memory 406.

[0017] FIG. 2 is a configuration diagram showing a structure analysis system in which a data input / output device and an analysis device that performs calculation and recording are integrated. That is, in this system, the analysis device 410 includes an input device 413, a screen display unit 414, an input / output control unit 412, a network I / O unit 404, a CPU 408, a memory 406, a storage 418, and a bus unit 409.

Example

[0018] FIG. 3 is a block diagram showing the functions executed by the structure analysis apparatus according to Example 1. As shown in FIG. 3, the structure analysis apparatus of this example includes a setting input unit 501, a space division unit 502, an atomic occupancy calculation unit 503, a vacancy identification unit 504, and a vacancy output unit 505. The setting input unit 501 performs input of information regarding atoms and settings such as thresholds necessary for analysis. The space division unit 502 divides the space to be analyzed into a plurality of parts. The atomic occupancy calculation unit 503 calculates the occupancy of atoms in each space divided by the space division unit 502. The vacancy identification unit 504 identifies vacancies based on spaces with a lower occupancy than a predetermined occupancy. The vacancy output unit 505 outputs the vacancies identified by the vacancy identification unit 504. Each function shown in FIG. 3 like this is executed by operating a predetermined program stored in the memory 406 of the computer of the analysis apparatus 410 by the CPU 408. Note that in this example, a case where a polymer including resin, fiber, protein, etc. is taken as an analysis target will be described as an example.

[0019] FIG. 4 is a flowchart showing a method of extracting vacancies as characteristic quantities of a molecular structure by the structure analysis apparatus according to Example 1.

[0020] [Input of Data (Step S101)] First, the user uses the input device 413 to input information regarding the atomic positions and elements of the polymer to be analyzed. Here, the information on atomic positions is obtained by measurement using X-rays, neutron rays, NMR, etc. or calculation using simulation. In addition, the information on elements includes the atomic radius of the elements. The atomic radius can be arbitrarily specified by the user, and suitable examples are the covalent bond radius and the van der Waals radius corresponding to each element. Further, the AV threshold used for identifying vacancies and the m_coef coefficient representing the magnitude of the vibrational motion of atoms inside the substance are also input via the input device 413. Note that the information input via the input device 413 is read by the input setting unit of the CPU 408 and held in the memory 406.

[0021] [Polyhedral Division of Analysis Space (Step S102)] Next, the space division unit 502 calculates a Voronoi polyhedron centered on the coordinate point of each atom in the molecule using the data of the atomic positions held in the memory 406. FIG. 5 is a diagram showing an example of the positional relationship between the coordinate points of the atoms and the Voronoi polyhedron. As shown in FIG. 5, the entire analysis target space surrounded by the analysis boundary 1000 is divided by the perpendicular bisecting planes between the coordinate points of five respective atoms (coordinate point 1021 of atom A, coordinate point 1022 of atom B, coordinate point 1023 of atom C, coordinate point 1024 of atom D, coordinate point 1025 of atom E) into Voronoi polyhedrons (Voronoi polyhedron 1101 of atom A, Voronoi polyhedron 1102 of atom B, Voronoi polyhedron 1103 of atom C, Voronoi polyhedron 1104 of atom D, Voronoi polyhedron 1105 of atom E) centered on the coordinate points of the respective atoms.

[0022] FIG. 6 is a diagram showing an example of the data structure of the Voronoi polyhedron information. As shown in FIG. 6, the Voronoi polyhedron information 5000 includes geometric information regarding the Voronoi polyhedron (vertex coordinate list 5100, edge list 5200, face list 5300, designation of the faces constituting the volume 5400), as well as the element symbol 5600 of the atom, the atomic radius 5700, the atomic position 5500, and the atomic occupancy 5800. Here, the atomic occupancy 5800 is defined as Va / Vv, which is the ratio of the effective atomic sphere volume Va based on the radius of the atom included in the Voronoi polyhedron and the volume Vv of the Voronoi polyhedron.

[0023] The vertex coordinate list 5100 stores the vertex label 5101 and the coordinate value 5102 of the vertex. The edge list 5200 stores the edge label 5201 and the value 5202 of the edge (the label of the starting vertex and the label of the ending vertex). The face list 5300 stores the face label 5301 and the value 5302 of the face (the labels of the surrounding edges).

[0024] In addition, the information 5000 of the Voronoi polyhedron may include chemical properties of atoms, such as atomic number, mass, number of electrons, electronegativity, electron configuration, etc. Note that even when the analysis target is other than polymers and the space has a periodic structure like a crystal, by setting the crystal lattice as the analysis boundary 1000, it can be similarly divided into Voronoi polyhedra.

[0025] In this way, since the entire space is divided into a plurality of Voronoi polyhedra, the analysis results are retained in the memory 406 as a set 2000 of information of Voronoi polyhedra. The set 2000 of information of Voronoi polyhedra is preferably stored in the memory 406 as data in a format such as a network type, a list type, or a graph type.

[0026] FIG. 7 is a diagram showing the set of information of the five Voronoi polyhedra shown in FIG. 5 in a network-type data format. In FIG. 7, five nodes (2001 to 2005) from A to E are arranged corresponding to the five atomic positions, and two Voronoi polyhedra including the atomic positions that are adjacent to each other are connected between the nodes, and this connection state is recorded as edges 2011 to 2019. For example, node A (2001) is connected to node C (2003), node B (2002), and node D (2004), and node C (2003) is connected to node A (2001), node B (2002), node D (2004), and node E (2005). In addition, the information 5001 to 5005 of each Voronoi polyhedron is associated with each node. As another method of representing the set of information of Voronoi polyhedra in a network-type data format, a Delaunay diagram with atomic positions as coordinate points may be created.

[0027] FIG. 8 is a diagram showing a set of information on the five Voronoi polyhedra shown in FIG. 5 in a list-type data format. When storing information in the memory 406 in a list-type data format, a list vector of reference numbers of the list (array) is used. For example, in node A (2001), an edge list 2021 corresponding to the connection state between nodes and information 5001 of the Voronoi polyhedron are stored. Note that in the edge list 2021, the names of nodes C, B, and D, which are the connection destinations of node A, are stored.

[0028] When storing the set 2000 of information on the Voronoi polyhedron in a data format such as that in FIG. 7 or FIG. 8, it is possible to easily remove or add a Voronoi polyhedron from the set according to various conditions. For example, when removing the information on the Voronoi polyhedron belonging to node B shown in FIG. 7 from the set, it is sufficient to delete node B and also delete edge 2016, edge 2012, and edge 2017. When removing the information on the Voronoi polyhedron belonging to node B shown in FIG. 8 from the set, it is sufficient to delete node B and also delete the name of node B from edge list 2021, edge list 2023, and edge list 2025.

[0029] Also, in this step S102, when the space to be analyzed exists in isolation outside the boundary 1000 of the analysis and the periodic boundary condition is not used, the Voronoi polyhedron including the coordinate points belonging to the vertices, edges, or faces of the boundary 1000 of the analysis in the vertex coordinate list 5100 is deleted from the set 2000 of information on the Voronoi polyhedron. For example, in the example of FIG. 5, Voronoi polyhedra other than the Voronoi polyhedron 1103 at the coordinate point 1023 of atom C are excluded from the analysis target.

[0030] [Calculation of atomic occupancy (step S103)] Next, the atomic occupancy calculation unit 503 refers to the atomic radius 5700 from the element information included in the information 5000 of the Voronoi polyhedron and calculates the effective atomic spherical volume Va. At this time, assuming the atomic radius is r_a, the effective atomic spherical volume Va can be expressed by the following (Equation 1).

[0031] Va = 4 / 3 * pi * (r_a)^3 (Equation 1) Also, the effective atomic sphere volume Va can also be expressed by the following (Equation 2) when using the m_coef coefficient.

[0032] Va = 4 / 3 * pi * (r_a * m_coef)^3 (Equation 2) Next, the atomic occupancy calculation unit 503 calculates the ratio Va / Vv of the effective atomic sphere volume Va and the volume Vv of the Voronoi polyhedron included in the Voronoi polyhedron information 5000 as the atomic occupancy 5800, and holds it in the memory 406 as one of the Voronoi polyhedron information 5000. The calculation and holding of the value of Va / Vv are performed for all Voronoi polyhedra included in the set 2000 of Voronoi polyhedron information.

[0033] [Vacancy determination (Step S104)] When Va / Vv is large, since the atomic occupancy of the corresponding Voronoi polyhedron is large and it is in a dense state, it is considered inappropriate as a part constituting a vacancy. Therefore, in step S104, the vacancy identification unit 504 compares the atomic occupancy (Va / Vv) with a preset AV threshold value. If it is larger than the AV threshold value, the corresponding Voronoi polyhedron does not meet the conditions for constituting a vacancy, and thus it is deleted from the set 2000 of Voronoi polyhedron information. The vacancy identification unit 504 performs such determination for all Voronoi polyhedra included in the set 2000 of Voronoi polyhedron information. As a result, only the sparse-state Voronoi polyhedra with an atomic occupancy below the AV threshold value remain in the set 2000 of Voronoi polyhedron information.

[0034] [Vacancy generation (Step S105)] Next, the vacancy identification unit 504 checks the connectivity between each of the remaining Voronoi polyhedra in the set 2000 of vacancy Voronoi polyhedron information, and connects the adjacent ones to form one group (vacancy). When this operation is completed, the number of groups (vacancies) of connected Voronoi polyhedra becomes three states: 0, 1, 2 or more. Note that the information regarding the vacancies identified by grouping is stored in the storage 418 or the like.

[0035] [Output of Data (Step S106)] Next, the pore output unit 505 outputs the identified pores. FIG. 9 is a diagram showing an example of visualizing the input coordinate information and the region including the formed pores. As shown in FIG. 9, in the analysis result display unit 3000, in addition to the atoms 3001 constituting the molecule and the bonds 3002 between the atoms, a region 3003 corresponding to the pores formed by a plurality of Voronoi polyhedra and a region 3004 corresponding to the pores formed by a single Voronoi polyhedron are superimposed and displayed. In the example of FIG. 9, the number of pores is two. By the pore output unit 505 outputting the pores as shown in FIG. 9, the user can easily recognize the atoms included in the region determined to be pores. Also, in FIG. 9, although the boundaries of the pores after bonding are output, the edge lines of each Voronoi polyhedron are not output, and it is a simplified display, so it is effective especially when confirming a complicated structure such as a polymer.

[0036] FIG. 10 is a diagram showing an example of visualizing not only the input coordinate information and the region including the pores but also the edge lines 3005 of each Voronoi polyhedron forming the pores. For example, it is effective when the user wants to confirm the details inside the pores when the molecule is not large, such as in a crystal structure. Also, even for a polymer, it is effective when the user wants to confirm the details partially. In FIG. 10, the edge lines 3006 of the Voronoi polyhedra that do not form pores are also output, but these edge lines 3006 may not be output.

[0037] FIG. 11 is a diagram showing an example of visualizing the molecular structure and the pore structure side by side. On the left side of FIG. 11, only the coordinate information (and the bonds between atoms) input in step S101 is displayed as the molecular structure, and on the right side of FIG. 11, the information corresponding to FIG. 9 is displayed as the pore structure. By visualizing as shown in FIG. 11, it becomes easy for the user to compare the molecular structure and the pore structure.

[0038] As another output mode for this other data, it may be possible to display, in association with the AV threshold value which is the input value, the number of pores, the volume of the pores, the number of Voronoi polyhedra forming the pores, the (geometric) centroid position of the pores, the (geometric) centroid position of the Voronoi polyhedra, and so on. As yet another output mode, it may be possible to display, in association with the AV threshold value which is the input value and the m_coef coefficient, the number of pores, the volume of the pores, the number of Voronoi polyhedra forming the pores, the centroid position of the pores, the centroid position of the Voronoi polyhedra, and so on.

[0039] As shown in Non-Patent Document 1, when the volume in which one atom can move around is defined as the free volume, particularly in the case of polymers, the processing of calculations by simulation is difficult. However, as in this Example, by capturing the free volume as pores inside the molecule, it is possible to suppress the amount of calculation while predicting the properties of the molecule with high accuracy. Further, since the information regarding the pores of the molecule is associated with the information of the atoms belonging to the pores, chemometric analysis using chemical information becomes possible in the analysis of the governing factors of the macroscopic properties of the molecule.

[0040] Figure 12 is a diagram showing the crystal structure of emerald (Be3Al2Si6O 18 )), and Figure 13 is a diagram showing the result of visualizing the pores extracted by the analysis of this Example. According to Figures 12 and 13, it can be seen that it is also possible to extract cylindrical pores centered on oxygen atoms for the crystal structure of emerald. That is, the pore analysis according to this Example can also be applied to inorganic molecules.

Example

[0041] In Example 2, the user specifies in advance the ranges of the AV threshold value and the m_coef coefficient, and sequentially executes a plurality of analyses (pore identification) using a plurality of candidates for the AV threshold value and the m_coef coefficient within the said ranges, and selects, from among the execution results, the candidate that meets the determination conditions as the final AV threshold value and m_coef coefficient.

[0042] FIG. 14 is a block diagram showing functions executed by the structure analysis apparatus according to Example 2. As shown in FIG. 14, unlike Example 1, the structure analysis apparatus of this example further includes a determination unit 506. The determination unit 506 determines candidates for the AV threshold value and the m_coef coefficient that meet the conditions, specifically, the AV threshold value or the m_coef coefficient such that the number of vacancies is maximized. Each function shown in FIG. 14 like this is executed by the CPU 408 operating a predetermined program stored in the memory 406 of the computer of the analysis apparatus 410.

[0043] FIG. 15 is a flowchart showing a method of sequentially executing a plurality of analyses within a set range in the structure analysis apparatus according to Example 2.

[0044] [Setting of input value range (Step S201)] Before executing the analysis, the user uses the input device 413 to input, in addition to information on the atomic positions and elements to be analyzed, the upper and lower limit values of the AV threshold value used for specifying vacancies, the upper and lower limit values of the m_coef coefficient, and the number of division steps.

[0045] [Generation of input value list (Step S202)] The information input via the input device 413 is read by the input setting unit of the CPU 408, and a list storing a plurality of AV threshold values and a list storing a plurality of m_coef coefficients are generated and held in the memory 406. Note that the list generated in this step S202 may be either only the AV threshold value or only the m_coef coefficient.

[0046] [Execution of vacancy analysis (Step S203)] Next, the atomic occupancy calculation unit 503 and the vacancy identification unit 504 sequentially acquire the AV threshold value and the m_coef coefficient from the generated input value list, and execute continuous analysis. At this time, the atomic occupancy calculation unit 503 and the vacancy identification unit 504 may create the direct product (Cartesian product) of the AV threshold value list and the m_coef coefficient list, and use this sequentially to execute continuous analysis. Further, the analysis result is recorded in the memory 406 or the storage 418.

[0047] [Determination (Step S204)] Next, the determination unit 506 determines whether the analysis result has reached a predetermined condition. If not, it returns to step S203 and repeats the analysis. The predetermined condition is, for example, the maximum number of vacancies.

[0048] Here, the case where the number of vacancies in the molecule is small is roughly classified into a situation where the interatomic distance is short, atoms are densely present, and there is no gap in the molecule, and a situation where the connection of the Voronoi polyhedra forming the vacancies extends over the entire molecule. The latter situation may be due to the fact that the AV threshold value and the m_coef coefficient, which are input values, are not suitable values for expressing the characteristics of the molecular structure. Therefore, in this embodiment, by using a plurality of AV threshold values and m_coef coefficients within the set range and executing a plurality of analyses, an optimal analysis result is obtained.

[0049] [Display (Step S205)] As a result of the determination, the AV threshold value and the m_coef coefficient that satisfy the predetermined condition are displayed on the screen display unit 414 and recorded in the memory 406 or the storage 418.

Example

[0050] In Example 3, when sequentially executing a plurality of analyses (vacancy identification), the candidate for the AV threshold value that can extract characteristic time from the molecular dynamics trajectory is used as the final AV threshold value.

[0051] FIG. 16 is a block diagram showing functions executed by the structure analysis apparatus according to Example 3. As shown in FIG. 16, different from Example 2, the structure analysis apparatus of this example further includes a survival time calculation unit 507. The survival time calculation unit 507 sequentially calculates the survival time from the generation to the disappearance of vacancies by using a plurality of candidates for the AV threshold within the range set by the setting input unit 501. Each function shown in FIG. 16 like this is executed by operating a predetermined program stored in the memory 406 of the computer of the analysis apparatus 410 by the CPU 408.

[0052] FIG. 17 is a flowchart showing a method of sequentially executing a plurality of analyses within a set range in the structure analysis apparatus according to Example 3. Here, the time when the molecular dynamics calculation starts is set to zero, and the time counted in time steps until the calculation ends is defined as the analysis time. The trajectory obtained by the molecular dynamics calculation is time-series data that can specify the molecular structure (atomic positions within the molecule) at the analysis time.

[0053] [Setting of input value range (Step S301)] Before executing the analysis, the user uses the input device 413 to input, in addition to information regarding the atomic positions and elements to be analyzed, the upper and lower limit values of the AV threshold used for specifying vacancies and the number of division steps.

[0054] [Generation of input value list (Step S302)] Information regarding the range of the AV threshold input via the input device 413 is read by the input setting unit of the CPU 408, a list storing a plurality of AV thresholds is generated, and is held in the memory 406. One AV threshold is selected from the generated list and used as the input value.

[0055] [Initialization of analysis time (Step S303)] The analysis time (t) is set to zero.

[0056] [Reading of molecular structure at analysis time (t) (Step S304)] The molecular structure at the analysis time (t) is read.

[0057] [Execution of void analysis (step S305)] The atomic occupancy calculation unit 503 and the void identification unit 504 execute the analysis. The analysis results are stored in the memory 406 in association with the AV threshold values, and the voids included in each result are numbered.

[0058] [Change of analysis time (step S306)] The analysis time is updated. The update is performed by t = t + Δt.

[0059] [Judgment 1 (step S307)] It is determined whether the analysis time has reached the end of the simulation. If not, the process returns to step S304.

[0060] [Judgment 2 (step S308)] It is determined whether all the analyses using the plurality of AV threshold values stored in the list have been executed. If not, the process returns to step S303 with the next AV threshold value as the input value.

[0061] [Calculation of survival time (step S309)] When all the analyses using the plurality of AV threshold values have been executed, first, the survival time calculation unit 507 compares the void analysis result at the analysis time (t) with the void analysis result at the analysis time (t + Δt) using the same AV threshold value. Here, Δt is the time step of the molecular dynamics simulation. As a result of the comparison, changes in voids may be extracted between the result at the analysis time (t) and the result at the analysis time (t + Δt). For example, a change in which a void that existed at the analysis time (t) disappears at the analysis time (t + Δt), or conversely, a change in which a void that did not exist at the analysis time (t) is generated at the analysis time (t + Δt), is extracted. In this embodiment, the time from when a void is generated until it disappears is defined as the survival time. Also, for a certain AV threshold value, when there is a repetition of generation and disappearance in voids including the same atomic position, the sum of the survival times of those voids is defined as the integrated survival time.

[0062] Next, the survival time calculation unit 507 calculates the survival time of each vacancy based on the above definition. When the vacancy analysis with the analysis time sequentially changed is executed for all the AV thresholds stored in the list by the survival time calculation unit 507, it can be seen that changes in the generation and disappearance of vacancies occur for each AV threshold, and the survival time and the integrated survival time change depending on the AV threshold.

[0063] [Determination of AV Threshold (Step S310)] FIG. 18 is a diagram showing how the survival time of a vacancy changes depending on the AV threshold. The horizontal axis represents the analysis time, and the vertical axis represents the AV threshold. For a certain AV threshold, the times when a vacancy is generated (6002) and disappears (6004) are indicated by black circles, and the survival time during that period is indicated by a solid line (6006). In the example of FIG. 18, it can be seen that the survival time of the AV threshold AV(3) is longer than that of AV(2) and AV(1).

[0064] Here, in physical phenomena and chemical reactions, the kinetic relaxation time and the diffusion time are characteristic times that govern the phenomena and reactions. Therefore, vacancies that exist longer than the relaxation time and the diffusion time are considered to have a high correlation with the properties of molecules. Thus, in this embodiment, when the relaxation time and the diffusion time in the target physical phenomenon or chemical reaction are known, the AV threshold at which the survival time and the integrated survival time become longer than these times is set as the AV threshold to be obtained.

Example

[0065] Example 4 constructs a model for predicting molecular properties using information about pores as a characteristic quantity of the molecular structure. Specific examples of characteristic quantities include the total volume of pores, the volume of each pore and their frequency distributions, the pore centroid, the pore radius and their frequency distributions, etc. In regression analysis and the construction of a property prediction model, the characteristic quantities are included as explanatory variables, and the properties of the molecule (water absorption rate, permeability, refractive index, magnetic properties, Young's modulus, etc.) are included as the objective variable. In the construction of the property prediction model, the explanatory variables and the objective variable are each divided into learning and test uses, and only the learning data is treated as the input to the property prediction model. Then, the performance of the property prediction model is evaluated using new test data that has not been used for learning. When modeling the relationship between the information of the manufacturing process and the characteristic quantities of the molecular structure, the process conditions for actually creating sample molecules are set as explanatory variables, and the characteristic quantities are set as the objective variable to perform model learning.

[0066] In particular, the properties of polymers often show non-linear and complex changes with respect to chemical structure and process conditions. That is, the atomic positions of polymers and their time-series data are very high-dimensional data, and using all of this data to determine the relationship between process conditions and properties means constructing a high-dimensional prediction model, resulting in an enormous amount of learning data and learning costs. On the other hand, in prediction models using the average properties of molecules (such as density, molecular weight, etc.), although the learning data and learning costs can be reduced, the prediction accuracy obtained is often not high. However, by adopting the concept of pores as appropriate characteristic quantities according to the state of the polymer as in this example, it is possible to construct a model with high prediction accuracy while reducing the learning data and learning costs.

Explanation of Symbols

[0067] 402: Data input / output device, 404: Network I / O unit, 406: Memory, 408: CPU, 409: Bus unit, 410: Analysis device, 412: Input / output control unit, 413: Input device, 414: Screen display unit, 418: Storage, 420: Network, 501: Setting input unit, 502: Space division unit, 503: Atomic occupancy calculation unit, 504: Vacancy identification unit, 505: Vacancy output unit, 506: Judgment unit, 507: Lifetime calculation unit

Claims

A structural analysis method executed by a computer having a memory and a storage, comprising: dividing a space to be analyzed into Voronoi polyhedra centered on the coordinate points of each atom in a molecule by using data of atomic positions held in the memory; calculating, as Va / Vv which is a ratio of an effective atomic sphere volume Va based on the radii of atoms included in the Voronoi polyhedron and a volume Vv of the Voronoi polyhedron, the occupancy of atoms in each divided Voronoi polyhedron, and holding the result in the memory; leaving the Voronoi polyhedra for which Va / Vv is equal to or less than a preset threshold, and connecting adjacent ones among the remaining Voronoi polyhedra to identify voids and store them in the storage; outputting by overlapping a region corresponding to the identified voids in addition to atoms constituting the molecule and bonds between the atoms; A structural analysis method comprising the above steps.

2. In the structural analysis method according to claim 1, setting a range of the threshold or a coefficient used when calculating the effective atomic sphere volume; sequentially identifying the voids by using a plurality of candidates of the threshold or the coefficient within the set range; determining, as the final threshold or coefficient, a candidate of the threshold or the coefficient for which the number of voids becomes maximum; A structural analysis method further comprising the above steps.

3. In the structural analysis method according to claim 1, setting a range of the threshold; sequentially calculating the lifetime from when the voids are generated until they disappear by using a plurality of candidates of the threshold within the set range; determining, as the final threshold, a candidate of the threshold for which the lifetime is longer than a predetermined time; A structural analysis method further comprising the above steps.

4. In the structural analysis method according to claim 1, As an output mode of the voids, a mode of outputting the boundary of the connected voids and also outputting the edges of each Voronoi polyhedron; a mode of outputting the boundary of the connected voids and not outputting the edges of each Voronoi polyhedron; A structural analysis method, characterized in that either mode can be selected.

5. In the structural analysis method according to claim 1, The object to be analyzed is a polymer. A structural analysis method characterized by this.

6. A space division unit that divides the space to be analyzed into Voronoi polyhedra centered on the coordinate points of each atom in the molecule, using the data of the atomic positions held in the memory; An atomic occupancy calculation unit that calculates, as Va / Vv, which is the ratio of the effective atomic spherical volume Va based on the radii of the atoms included in the Voronoi polyhedron and the volume Vv of the Voronoi polyhedron, the occupancy of the atoms in each Voronoi polyhedron divided by the space division unit, and holds it in the memory; A void identification unit that leaves the Voronoi polyhedra for which Va / Vv is equal to or less than a preset threshold value, and connects adjacent ones among the remaining Voronoi polyhedra to identify voids and store them in storage; A void output unit that outputs, in addition to the atoms constituting the molecule and the bonds between the atoms, the region corresponding to the voids identified by the void identification unit; A structure analysis apparatus comprising the above.

7. In the structure analysis apparatus according to claim 6, further comprising a setting input unit and a determination unit, the setting input unit is for setting the threshold value or the range of coefficients used when calculating the effective atomic spherical volume, the void identification unit sequentially identifies the voids using a plurality of candidates for the threshold value or the coefficient within the range set by the setting input unit, the determination unit determines, as the final threshold value or coefficient, the candidate for the threshold value or the coefficient for which the number of voids becomes maximum. A structure analysis apparatus characterized by this.

8. In the structure analysis apparatus according to claim 6, a setting input unit for setting the range of the threshold value, a survival time calculation unit that sequentially calculates the survival time from when the voids are generated until they disappear, using a plurality of candidates for the threshold value within the range set by the setting input unit, a determination unit that determines, as the final threshold value, the candidate for the threshold value for which the survival time is longer than a predetermined time. A structure analysis apparatus further comprising the above.

9. In the structure analysis apparatus according to claim 6, the void output unit outputs the boundaries of the connected voids and also outputs the edge lines of each Voronoi polyhedron; outputs the boundaries of the connected voids without outputting the edge lines of each Voronoi polyhedron; A structure analysis apparatus characterized by selectively outputting the above.

10. In the structure analysis apparatus according to claim 6, the object to be analyzed is a polymer. A structure analysis apparatus characterized by this.

11. A program that causes a computer to function as a structural analysis device, a space division unit that divides a space to be analyzed into Voronoi polyhedra centered on the coordinate points of each atom in a molecule, using data on the atomic positions held in a memory, an atomic occupancy rate calculation unit that calculates, as Va / Vv, which is the ratio of the effective atomic sphere volume Va based on the radii of the atoms included in the Voronoi polyhedron and the volume Vv of the Voronoi polyhedron, the occupancy rate of the atoms in each Voronoi polyhedron divided by the space division unit, and holds it in the memory, a void identification unit that leaves the Voronoi polyhedra for which Va / Vv is equal to or less than a preset threshold value, and connects adjacent ones among the remaining Voronoi polyhedra to identify voids and store them in storage, and a void output unit that outputs, in an overlapping manner, the atoms constituting the molecule, the bonds between the atoms, and the regions corresponding to the voids identified by the void identification unit, a program for causing the above to function.

Images