Design device, design method, program, and design system

Abstract

A design device for designing a classical force field, comprising: an inference means for inferring a physical property value of a chemical structure necessary for creating a classical force field, using a deep learning force field obtained by learning a first principle calculation result of an arbitrary chemical structure; and a determination means for determining a parameter of a function system of the classical force field using the inferred physical property value of the chemical structure.

Description

Design device, design method, program, and design system

[0001] The present disclosure relates to a design apparatus, a design method, a program, and a design system.

[0002] With the recent development of computational technology, various force fields (sets of simulation functions and parameters) have been used to predict a wide variety of physical properties, such as chemical reactions or chemical adsorption. Although force fields have various characteristics, including the target chemical species and their effects, each is specialized for a specific system, and therefore, there is a demand for the rapid development of a force field that can be used for general purposes.

[0003] For example, Patent Document 1 describes NNP (Neural Network Potential) as a method for realizing inference based on results obtained by first-principles calculations.

[0004] Japanese Patent Application Laid-Open No. 2022-189626

[0005] For example, first-principles calculations or quantum chemical calculations are used to analyze the behavior of chemical reactions or chemical adsorption. First-principles calculations or quantum chemical calculations have high accuracy and versatility, but have the problem of extremely high calculation costs. For example, first-principles calculations or quantum chemical calculations have sometimes reached a practical limit of calculations on the order of 100 atoms.

[0006] Simulations using a classical force field, which is one example of a force field, have very low computational costs and can be used to analyze polymers or proteins with around one million atoms. However, classical force fields have problems such as high learning costs and low versatility because they require multiple first-principles calculations to determine parameters for a function system specialized for the target system.

[0007] Furthermore, simulations using deep learning force fields, which are an example of force fields, reproduce first-principles calculation results with high accuracy and versatility through deep learning. Trained deep learning force fields have high accuracy and versatility. However, deep learning force fields have the problem of being inferior to classical force fields in terms of computational cost and memory utilization.

[0008] An object of the present disclosure is to provide a design device, a design method, a program, and a design system that support the design of classical force fields.

[0009] The present disclosure has the following configuration.

[0010] [1] A design device for designing a classical force field, comprising: an inference means for inferring physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; and a determination means for determining parameters of a function system of the classical force field using the inferred physical property values ​​of the chemical structure.

[0011] [2] The design device according to [1], wherein the determining means performs fitting by applying the inferred physical property values ​​of the chemical structure to a function system of the classical force field to determine the parameters.

[0012] [3] The design device according to [1] or [2], wherein the deep learning force field is a neural network used for NNP (Neural Network Potential), and has been trained to output a first-principles calculation result of the chemical structure when the chemical structure is input.

[0013] [4] The design device according to any one of [1] to [3], wherein the physical property values ​​of the chemical structure are energies when each of a bond length, a bond angle, and a dihedral angle is changed.

[0014] [5] A design method performed by a design device that designs a classical force field, the design method comprising: an inference procedure that infers physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; and a determination procedure that determines parameters of a function system of the classical force field using the inferred physical property values ​​of the chemical structure.

[0015] [6] A program for causing a design device that designs a classical force field to execute the following steps: an inference step of inferring physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; and a determination step of determining parameters of a function system of the classical force field using the inferred physical property values ​​of the chemical structure.

[0016] [7] A design system for designing a classical force field, comprising: first-principles calculation means for performing first-principles calculation; learning means for learning a deep learning force field using the first-principles calculation results of an arbitrary chemical structure performed by the first-principles calculation means; inference means for inferring physical property values ​​of the chemical structure required for creating a classical force field using the deep learning force field that has learned the first-principles calculation results; and determination means for determining parameters of a function system of the classical force field using the inferred physical property values ​​of the chemical structure.

[0017] According to the present disclosure, it is possible to provide a design device, a design method, a program, and a design system that support the design of a classical force field.

[0018] FIG. 1 is a configuration diagram of an example of a design system according to the present embodiment. FIG. 2 is a hardware configuration diagram of an example of a computer according to the present embodiment. FIG. 3 is a functional configuration diagram of an example of a design system according to the present embodiment. FIG. 4 is a diagram explaining an example of a classical force field creation method. FIG. 5 is a diagram explaining an example of a classical force field creation method. FIG. 6 is a flowchart of an example of a process for creating a deep learning force field according to the present embodiment. FIG. 7 is a flowchart of an example of a process for creating a classical force field according to the present embodiment.

[0019] Next, an embodiment of the present invention will be described in detail, but the present invention is not limited to the following embodiment.

[0020] [First Embodiment] <System Configuration> Fig. 1 is a configuration diagram of an example of a design system 1 according to this embodiment. The design system 1 in Fig. 1 includes a design device 10 and a user terminal 12. The design device 10 and the user terminal 12 of the design system 1 are connected to each other so as to be able to communicate data via a communication network 18 such as a local area network (LAN) or the Internet.

[0021] The user terminal 12 is an information processing terminal operated by an operator, such as a PC, tablet terminal, or smartphone. The user terminal 12 displays a screen for accepting information input from the operator on a display device and accepts the information input from the operator. The user terminal 12 also transmits the information accepted as input from the operator to the design device 10, causing the design device 10 to execute processing to support the design of a classical force field. The user terminal 12 receives information on the execution results of the processing by the design device 10 and displays it on a display device for the operator to confirm.

[0022] The design device 10 is an information processing device such as a PC or workstation that supports an operator in designing a classical force field. The design device 10 speeds up the data sampling calculations required for parameter fitting of a function system of a classical force field by using a deep learning force field that has learned the results of the first-principles calculations, rather than directly performing the calculations using first-principles calculations. The design device 10 receives information input by the operator to a user terminal 12 and executes processing to support the design of a classical force field. The design device 10 transmits processing result information to the user terminal 12 and causes the user terminal 12 to display the processing result information.

[0023] The design system 1 in Fig. 1 may be realized by a design device 10 having a web server function and a user terminal 12 that executes a web application using a web browser function. Alternatively, the design system 1 in Fig. 1 may be realized by an application installed in the user terminal 12 performing processing in cooperation with a program installed in the design device 10.

[0024] 1 is merely an example, and it goes without saying that there are various system configuration examples depending on the application and purpose. For example, the design device 10 may be realized by multiple computers or may be realized as a cloud computing service. The design system 1 may also be realized by a standalone computer.

[0025] <Hardware Configuration> The design device 10 and the user terminal 12 in FIG. 1 are realized by, for example, a computer 500 having the hardware configuration shown in FIG.

[0026] Fig. 2 is a hardware configuration diagram of an example of a computer 500 according to this embodiment. The computer 500 in Fig. 2 includes an input device 501, a display device 502, an external I / F 503, a RAM 504, a ROM 505, a CPU 506, a communication I / F 507, and an HDD 508, all of which are interconnected by a bus B. The input device 501 and the display device 502 may be connected to each other for use.

[0027] The input device 501 is a touch panel, operation keys, buttons, keyboard, mouse, etc. that are used by the operator to input various signals. The display device 502 is composed of a display such as a liquid crystal or organic EL display for displaying a screen, a speaker for outputting sound data such as voice and sound, etc. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0028] The HDD 508 is an example of a non-volatile storage device that stores programs and data. The stored programs and data include an OS, which is basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that the computer 500 may use a drive device that uses flash memory as a storage medium (e.g., a solid-state drive (SSD)) instead of the HDD 508.

[0029] The external I / F 503 is an interface with an external device. The external device may be a recording medium 503a. This allows the computer 500 to read and / or write data from and to the recording medium 503a via the external I / F 503. The recording medium 503a may be a flexible disk, a CD, a DVD, an SD memory card, a USB memory, or the like.

[0030] The ROM 505 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The ROM 505 stores programs and data such as the BIOS, OS settings, and network settings that are executed when the computer 500 starts up. The RAM 504 is an example of a volatile semiconductor memory (storage device) that temporarily retains programs and data.

[0031] The CPU 506 is a computing device that reads programs and data from storage devices such as the ROM 505 and HDD 508 onto the RAM 504 and executes processing to realize overall control and functions of the computer 500. The computer 500 according to this embodiment can realize various functions of the design device 10 and user terminal 12, which will be described later, by executing programs.

[0032] The program to be installed in the RAM 504 is installed, for example, by reading the program recorded on the recording medium 503a via the external I / F 503. Alternatively, the program may be installed by being downloaded from the communication network 18 via the communication I / F 507.

[0033] <Functional Configuration> The functional configuration of the design system 1 according to this embodiment will be described. Fig. 3 is a functional configuration diagram of an example of the design system 1 according to this embodiment. Note that in the configuration diagram of Fig. 3, portions that are not necessary for the description of this embodiment are omitted as appropriate.

[0034] 3 includes a request receiving unit 20, a response transmitting unit 22, a calculation system creating unit 24, a first-principles calculation unit 26, a learning unit 28, an inference unit 30, a determination unit 32, a verification unit 34, a display control unit 36, a deep learning force field storage unit 40, and a classical force field storage unit 42. The user terminal 12 of the design system 1 includes an information display unit 50, an operation accepting unit 52, a request transmitting unit 54, and a response receiving unit 56.

[0035] The information display unit 50 displays a screen for accepting information input from a worker and information on the execution results of processing in the design device 10 on the display device 502. The operation accepting unit 52 accepts operations by the worker, such as input of information. The request sending unit 54 sends a processing request in response to the information input from the worker to the design device 10. In addition, the response receiving unit 56 receives a response to the processing request sent by the request sending unit 54 from the design device 10.

[0036] The request receiving unit 20 receives a processing request from the user terminal 12. The response sending unit 22 responds with the execution result of the processing in response to the processing request. The calculation system creation unit 24 creates a calculation system using an arbitrary chemical structure based on information about the arbitrary chemical structure input by the operator. The first-principles calculation unit 26 performs first-principles calculations on a structural model in which the bond length, bond angle, and dihedral angle of the arbitrary chemical structure are changed, and calculates the energy of the structural model as a result of the first-principles calculation. The bond length, bond angle, and dihedral angle are examples of physical property values ​​of a chemical structure. The first-principles calculation is a calculation based on, for example, DFT (Density Function Theory) or QM (Quantum Mechanics), etc.

[0037] The learning unit 28 causes the deep learning force field to learn the relationship between a structural model in which the bond length, bond angle, and dihedral angle of an arbitrary chemical structure are changed and the energy of the structural model calculated as a result of first-principles calculation, based on an appropriate machine learning method. In this way, the machine learning of the deep learning force field may be performed using an arbitrary chemical structure and the physical property values ​​of the chemical structure calculated by first-principles calculation for the arbitrary chemical structure as training data.

[0038] Furthermore, the deep learning force field is a neural network used in NNP (Neural Network Potential), and has been trained to output the results of first-principles calculations of a chemical structure when the chemical structure is input. The trained deep learning force field is capable of inferring the results of first-principles calculations of the input chemical structure. The learning unit 28 stores the trained deep learning force field in the deep learning force field storage unit 40. The deep learning force field storage unit 40 stores the trained deep learning force field. The deep learning force field storage unit 40 may store trained deep learning force fields that are publicly available on the Internet, etc.

[0039] The inference unit 30 uses the learned deep learning force field stored in the deep learning force field memory unit 40 to infer the energy of a structural model as a result of first-principles calculation of a structural model in which the bond length, bond angle, and dihedral angle of an arbitrary chemical structure are changed.

[0040] The determination unit 32 performs fitting by applying the energy of the structural model inferred by the inference unit 30 to a function system of a classical force field, and determines the parameters of the classical force field. The determination unit 32 creates a classical force field by determining the parameters of the function system of the classical force field. Details of the processing of the determination unit 32 will be described later. The classical force field storage unit 42 stores the created classical force field.

[0041] The verification unit 34 verifies whether the parameters of the classical force field function system determined by the determination unit 32 are appropriate. The display control unit 36 ​​controls the display on the user terminal 12. Note that the functional configuration diagram in FIG. 3 is an example. The design system 1 according to this embodiment can be realized by various configurations. For example, the deep learning force field storage unit 40 and the classical force field storage unit 42 may be included in a storage device, computer, cloud storage, or the like that is capable of data communication with the design device 10.

[0042] <Processing> Figures 4A to 4C are diagrams for explaining an example of a classical force field creation method. Figure 4A is a schematic diagram of an example of a calculation system created using an arbitrary chemical structure. "l" in Figure 4A represents the bond length between two particles. "θ" in Figure 4A represents the bond angle between three particles. "ω" in Figure 4A represents the dihedral angle between four particles. Figure 4B is an example of a function system of a classical force field for the chemical structure shown in Figure 4A. The function system of the classical force field is based on the parameter k ｂ , k ａ , and V ｎ Includes.

[0043] In Figure 4C, the energy is plotted when the bond length l is changed in increments of 0.5 angstroms. The plot is made by calculating data sampling. In creating a classical force field, the parameters of the function system of the classical force field are determined by individually optimizing the parameters in Figure 4B so that the function system approximates the plot in Figure 4C.

[0044] The individual optimization of the parameters in FIG. 4B may be performed by a program that determines the parameters using, for example, the least squares method so as to minimize the difference with the plot in FIG. 4C, or may be performed by an operator adjusting the parameters.

[0045] Incidentally, when performing first-principles calculations for the plots of FIG. 4C , it is necessary to perform first-principles calculations for the number of plots. First-principles calculations have high accuracy and versatility, but require very high computational costs. Therefore, when performing first-principles calculations for the plots of FIG. 4C , the computational costs are high. Therefore, in this embodiment, a simulation is performed using a deep learning force field that has been trained on the first-principles calculation results for the plots of FIG. 4C . A simulation using a deep learning force field reproduces the first-principles calculation results with high accuracy. The trained deep learning force field has a lower computational cost than first-principles calculations. Therefore, when performing a simulation using a trained deep learning force field for the plots of FIG. 4C , the computational cost is lower than when performing first-principles calculations. Furthermore, when performing a simulation using a trained deep learning force field for the plots of FIG. 4C , memory utilization is superior to that of first-principles calculations. With better memory utilization, an increase in the number of atoms that can be calculated can be expected.

[0046] Furthermore, a deep learning force field that has already been published can be used, for example, a deep learning force field that has already been published in the Open Catalyst Project can be used. By using a published deep learning force field, the learning cost of the deep learning force field can be reduced in this embodiment.

[0047] If there is no published deep learning force field, in this embodiment, a deep learning force field is created, for example, according to the procedure shown in Fig. 5. Fig. 5 is a flowchart of an example of a process for creating a deep learning force field according to this embodiment.

[0048] In step S10, the calculation system creation unit 24 creates a calculation system using an arbitrary chemical structure such as that shown in FIG. 4A, based on information about the arbitrary chemical structure input by the operator.

[0049] In step S12, the first-principles calculation unit 26 performs first-principles calculations on a structural model in which the bond length, bond angle, and dihedral angle of an arbitrary chemical structure, for example, as shown in FIG. 4A , are changed, and calculates the energy of the structural model as a result of the first-principles calculations, for example, as shown by the plot in FIG. 4C .

[0050] In step S14, the learning unit 28 uses an appropriate machine learning method to train the deep learning force field to learn the relationship between a structural model in which the bond length, bond angle, and dihedral angle of an arbitrary chemical structure are changed and the energy of the structural model calculated as a result of a first-principles calculation.

[0051] FIG. 6 is a flowchart of an example of a process for creating a classical force field according to this embodiment.

[0052] In step S20, the calculation system creation unit 24 creates a calculation system using an arbitrary chemical structure such as that shown in FIG. 4A, based on information about the arbitrary chemical structure input by the operator.

[0053] In step S22, the inference unit 30 uses the trained deep learning force field stored in the deep learning force field memory unit 40 to infer the energy of a structural model in which the bond length, bond angle, and dihedral angle of an arbitrary chemical structure are changed, and calculates the energy of the structural model, for example, as shown by the plot in Figure 4C.

[0054] In step S24, the determination unit 32 performs fitting by applying the energy of the structural model inferred by the inference unit 30 in step S22 to a function system of a classical force field, thereby determining parameters of the classical force field. For example, the determination unit 32 determines parameters of the function system of the classical force field by individually optimizing the parameters of FIG. 4B so as to obtain a function system that approximates the plot of FIG. 4C.

[0055] In step S26, the verification unit 34 verifies whether the parameters determined by the determination unit 32 in step S24 are appropriate. For example, the verification unit 34 performs a molecular dynamics calculation using the parameters determined in step S24 in the function system of the classical force field in Fig. 4B, and verifies whether the parameters are appropriate based on the results.

[0056] If the parameters determined in step S24 are not appropriate, the design device 10 returns to the process of step S22 and continues the process. On the other hand, if the parameters determined in step S24 are appropriate, the design device 10 ends the process of the flowchart in Fig. 6. Note that the processes of the flowcharts in Fig. 5 and Fig. 6 may be performed after separating the arbitrary chemical structure when the arbitrary chemical structure is, for example, a polymer or a protein.

[0057] [Other Embodiments] The design device 10 according to this embodiment may be configured such that a plurality of computers cooperate to perform the above processing. For example, the design device 10 of the design system 1 according to this embodiment may be realized by being divided into a device that performs first-principles calculations, a device that creates a deep learning force field, and a device that designs a classical force field.

[0058] As described above, the design system 1 according to this embodiment can speed up the calculation of data sampling required for parameter fitting of the function system of a classical force field, and can efficiently determine the parameters of the function system of a classical force field. Therefore, this embodiment can provide a design device, design method, program, and design system that support the efficient design of a classical force field.

[0059] Although the present embodiment has been described above, it will be understood that various modifications in form and detail are possible without departing from the spirit and scope of the claims. While the present invention has been described above based on examples, the present invention is not limited to the above examples and various modifications are possible within the scope of the claims. This application claims priority from basic application No. 2023-080836, filed with the Japan Patent Office on May 16, 2023, the entire contents of which are incorporated herein by reference.

[0060] REFERENCE SIGNS LIST 1 design system 10 design device 12 user terminal 18 communication network 24 calculation system creation unit 26 first-principles calculation unit 28 learning unit 30 inference unit 32 determination unit 34 verification unit

Claims

1. A design device for designing a classical force field, comprising: an inference means for inferring physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; a determining means for determining parameters of a function system of a classical force field using the physical property values ​​of the inferred chemical structure; A design device comprising:

2. The determining means performs fitting by applying the inferred physical property values ​​of the chemical structure to a function system of the classical force field, and determines the parameters. The design device according to claim 1.

3. The deep learning force field is a neural network used for NNP (Neural Network Potential), and has been trained to output a first-principles calculation result of the chemical structure when the chemical structure is input.

3. The design device according to claim 1 or 2.

4. The physical properties of the chemical structure are the energies when the bond length, bond angle, and dihedral angle are changed.

3. The design device according to claim 1 or 2.

5. A design method performed by a design device that designs a classical force field, comprising: an inference procedure for inferring physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; a determination procedure for determining parameters of a function system of a classical force field using the physical property values ​​of the inferred chemical structure; A design method comprising:

6. A design device for designing classical force fields. an inference step of inferring physical property values ​​of an arbitrary chemical structure required for creating a classical force field using a deep learning force field that has learned the results of first-principles calculations of the chemical structure; a determining step of determining parameters of a function system of a classical force field using the physical property values ​​of the inferred chemical structure; A program to execute.

7. A design system for designing a classical force field, comprising: a first-principles calculation means for performing first-principles calculation; a learning means for learning a deep learning force field using the first-principles calculation result of an arbitrary chemical structure performed by the first-principles calculation means; an inference means for inferring physical property values ​​of the chemical structure required for creating a classical force field using a deep learning force field that has learned the results of the first-principles calculation; a determining means for determining parameters of a function system of a classical force field using the physical property values ​​of the inferred chemical structure; A design system comprising: