High-throughput data automatic analysis system for reactive force field fitting

Abstract

The invention belongs to the technical field of data processing, and provides a high-throughput data automatic analysis system for reactive force field fitting in order to solve the problem that the automation degree of the related data processing process is low when a reactive force field is trained at present. Comprising a high-throughput calculation data automatic analysis subsystem, a reaction force field automatic fitting subsystem and a reaction force field fitting database, and a plurality of molecular models in different formats are stored in the reaction force field fitting database; the high-throughput calculation data automatic analysis subsystem performs high-throughput processing on the molecular model to obtain corresponding first principle data; and the reaction force field automatic fitting subsystem performs high-throughput processing according to the first principle data, performs fitting according to a processing result to obtain a force field, performs molecular image drawing, and performs automatic comparison to obtain an optimal force field fitting result, thereby realizing automatic creation of an input file required by reaction force field fitting. Manual data errors are effectively reduced, and the response force field fitting efficiency is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of data processing, and in particular relates to a high-throughput data automatic analysis system for reaction force field fitting. Background Art

[0002] With the rapid development of computer computing power, high-throughput computing is increasingly used in scenarios such as first-principles calculations, molecular dynamics calculations, and preparation of training sets for force field fitting. High-throughput computing refers to the concurrent calculation of multiple models and multiple tasks on supercomputers. It has the advantages of being fast and efficient, and can save R&D time and reduce R&D costs. However, the preparation work before high-throughput computing is complicated. For example, a large number of molecular models need to be converted into a specific format suitable for the calculation program, and the corresponding calculation input files need to be called before they can be used for subsequent program calculations; the generated calculation results need to be automatically extracted and analyzed in batches. Since the format conversion between different programs is complicated, it often requires the participation of multiple programs and lacks batch automation processing functions. For example, the traditional Material Studio molecular model is converted into a data file, which makes the preparation work before high-throughput computing time-consuming and labor-intensive; in addition, there is a lack of the necessary input in file template for automatic calling of LAMMPS calculations, and the corresponding data file name needs to be modified, and specific force fields need to be called.

[0003] In addition, the training of reaction force fields often requires first-principles calculations to provide geo structure files consisting of a large number of molecular structures and trainset training set files consisting of a large number of structure names and corresponding energy data. For the geo files and trainset files required for the reaction force field training process, the traditional approach is to manually build models, manually submit task calculations, manually process data, and manually create geo files and trainset files. The lack of automated construction programs has resulted in a series of problems such as easy data errors and heavy personnel burden. Summary of the invention

[0004] In order to solve at least one of the above technical problems existing in the prior art, the present invention provides a high-throughput data automatic analysis system for reaction force field fitting.

[0005] The present invention is implemented by the following technical solution: a high-throughput data automatic analysis system for reaction force field fitting, comprising: a high-throughput calculation data automatic analysis subsystem, a reaction force field automatic fitting subsystem and a reaction force field fitting database;

[0006] The reaction force field fitting database stores molecular models, force field data and molecular structure data used for reaction force field fitting, and classifies the molecular models to obtain molecular models in several different formats, and inputs the molecular models in several different formats into the high-throughput calculation data automatic analysis subsystem;

[0007] The high-throughput calculation data automatic analysis subsystem converts the received molecular models in several different formats into molecular models in the format required for high-throughput processing, and then calls the calculation task template according to the preset calculation task type, and submits it to the corresponding calculation system for high-throughput processing to obtain the first principle data, and transmits the first principle data to the reaction force field automatic fitting subsystem and the reaction force field fitting database;

[0008] The reaction force field automatic fitting subsystem constructs a molecular information processing file based on the received first-principles data, and submits a calculation task to the corresponding computing system based on the molecular information processing file, and then fits the reaction potential function according to the calculation task to obtain a number of reaction force fields and corresponding fitting data, and draws molecular images based on the reaction force fields and the corresponding fitting data, and then automatically compares the drawn molecular images to obtain the optimal force field data and structural data, and transmits them to the reaction force field fitting database for storage.

[0009] Preferably, the high-throughput computing data automatic analysis subsystem includes a model automatic conversion module, a template calling module, a first computing task submission module and a computing data processing module;

[0010] The model automatic conversion module is used to convert molecular models in different formats into molecular models in the format required for high-throughput processing; the template calling module is used to set the corresponding calculation parameters and create the corresponding input template according to the preset calculation task type of the corresponding computing system, and then call the corresponding input template; the first calculation task submission module is used to submit the calculation task with the set calculation parameters to the corresponding computing system; the calculation data processing module is used to extract molecular information according to the molecular model and calculation task in the format required for high-throughput processing, and obtain the corresponding first-principles data.

[0011] Preferably, the reaction force field automatic fitting subsystem includes: a file construction module, a second calculation task submission module and an automatic drawing module;

[0012] The file construction module constructs a molecular information processing file based on the received first-principles data; the second computing task submission module submits these constructed molecular information processing files to the computing system, performs optimization calculations, and obtains fitting force field data in combination with preset parameter files; the automatic drawing module draws a graph based on the fitting force field data, and compares the drawn images to determine the optimal reaction force field data and structural data.

[0013] Preferably, the file construction module constructs geo files and trainset files;

[0014] The trainset file includes four parts: CHARGE, GEOMETRY, CELL PARAMETERS, and ENERGY. The CHARGE part of the trainset file is constructed by extracting Mulliken charges based on the first principle data, the GEOMETRY part of the trainset file is constructed by extracting the molecular structure topology after GAUSSIAN optimization, the CELL PARAMETERS part of the trainset file is constructed by extracting the unit cell parameters after VASP optimization, the energy between each structure is extracted, and the extracted energy is converted into a specified unit to construct the ENERGY part of the trainset file.

[0015] Preferably, the geo file and the trainset file are checked for redundant structures, and if redundant structures exist, the redundant structures are deleted.

[0016] Preferably, the first computing task submission module in the high-throughput computing data automatic analysis subsystem and the second computing task submission module in the reaction force field automatic fitting subsystem can share the same computing system.

[0017] Preferably, the reaction force field fitting database includes a molecular model database, a force field database and a structure database;

[0018] The molecular model database collects molecular models used for high-throughput calculations, the structural database collects molecular structures optimized by first-principles data during the fitting process, and the force field database collects reaction force fields obtained by GPU-accelerated fitting using first-principles data as input during the fitting process.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The high-throughput data automatic analysis system for reaction force field fitting provided by the present invention realizes the automatic creation of input files required for reaction force field fitting. At the same time, various data can be automatically analyzed, and all structural data and force fields are automatically formed into a database, which effectively reduces human data errors, realizes efficient conversion between models and efficient data processing, and greatly improves the efficiency of reaction force field fitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1It is a schematic diagram of the structure of a high-throughput data automatic analysis system for reaction force field fitting according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In conjunction with the drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0024] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so that people familiar with this technology can understand and read them. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size should fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention. It should be noted that in this specification, relational terms such as first and second are only used to distinguish one entity from several other entities, and do not necessarily require or imply any actual relationship or order between these entities.

[0025] Reaction force field is a molecular force field based on bond level, which is commonly used in molecular dynamics simulation. Through molecular dynamics or Monte Carlo, chemical reactions that may occur at micro-mesoscopic scales, as well as related thermodynamic and kinetic properties, are simulated. Therefore, when simulating and training the reaction force field, a large amount of relevant data often needs to be processed. However, most of the current practices are manual operations, which not only have a low degree of automation, but also are prone to errors in data processing. In order to solve the problem of low degree of automation in the process of processing relevant data when training the reaction force field, the present invention provides a high-throughput data automatic analysis system for reaction force field fitting, which realizes the automatic creation of input files required for reaction force field fitting, and performs high-throughput fitting, and then automatically analyzes and processes the fitting data to obtain optimal force field data and structural data, and automatically forms a database, reducing manual participation, thereby effectively reducing human data errors and improving the efficiency of reaction force field fitting.

[0026] like Figure 1 As shown, an embodiment of the present invention provides a high-throughput data automatic analysis system for reaction force field fitting, including: a high-throughput calculation data automatic analysis subsystem, a reaction force field automatic fitting subsystem and a reaction force field fitting database;

[0027] The reaction force field fitting database stores molecular models, force field data and molecular structure data used for reaction force field fitting, and classifies the molecular models to obtain molecular models in several different formats, and inputs the molecular models in several different formats into the high-throughput calculation data automatic analysis subsystem;

[0028] The high-throughput calculation data automatic analysis subsystem converts the received molecular models in several different formats into molecular models in the format required for high-throughput processing, and then calls the calculation task template according to the preset calculation task type, and submits it to the corresponding calculation system for high-throughput processing to obtain the first principle data, and transmits the first principle data to the reaction force field automatic fitting subsystem and the reaction force field fitting database;

[0029] The reaction force field automatic fitting subsystem constructs a molecular information processing file based on the received first-principles data, and submits a computing task to the corresponding computing system based on the molecular information processing file, and then fits the reaction force field according to the computing task to obtain a number of reaction force fields and corresponding fitting data, draws a molecular image based on the reaction force field and the corresponding fitting data, and then automatically compares the drawn molecular images to obtain the optimal force field data and structural data, and transmits them to the reaction force field fitting database for storage.

[0030] Optionally, the reaction force field automatic fitting subsystem includes: a file construction module, a second calculation task submission module and an automatic drawing module;

[0031] The file construction module constructs a molecular information processing file based on the received first-principles data; the second computing task submission module submits these constructed molecular information processing files to the computing system, performs optimization calculations, and obtains fitting force field data in combination with preset parameter files; the automatic drawing module draws a graph based on the fitting force field data, and compares the drawn images to determine the optimal reaction force field data and structural data.

[0032] In this embodiment, the file construction module is responsible for calling the high-throughput computing data molecular system, the second computing task submission module is responsible for submitting computing tasks in batches to the remote computer, the remote computer performs high-throughput fitting according to the computing tasks, and obtains multiple sets of reaction force fields and corresponding fitting data. The automatic drawing module is responsible for drawing according to the multiple sets of reaction force fields and the corresponding fitting data, and then comparing the multiple sets of images drawn to determine the optimal reaction force field data and structural data.

[0033] In this embodiment, there are multiple different formats of molecules or crystals stored in the molecular model database. According to the high-throughput computing task, a format conversion script is set to convert the molecular models in different formats to obtain a molecular model in a specified format. In this embodiment, the specified formats are POSCAR and gjf, which are convenient for subsequent high-throughput computing.

[0034] Optionally, the file construction module constructs geo files and trainset files;

[0035] The trainset file includes four parts: CHARGE, GEOMETRY, CELL PARAMETERS, and ENERGY. The CHARGE part of the trainset file is constructed by extracting Mulliken charges based on the first principle data, the GEOMETRY part of the trainset file is constructed by extracting the molecular structure topology after GAUSSIAN optimization, the CELL PARAMETERS part of the trainset file is constructed by extracting the unit cell parameters after VASP optimization, the energy between each structure is extracted, and the extracted energy is converted into a specified unit to construct the ENERGY part of the trainset file.

[0036] In this embodiment, the file construction module automatically calls the computational data processing module in the high-throughput computational data automatic analysis subsystem, extracts the structure file composed of the structural information in the corresponding first-principles data, constructs the geo file, extracts the energy and other data from the corresponding first-principles data, and constructs the trainset file.

[0037] The trainset file contains four parts: CHARGE, GEOMETRY, CELL PARAMETERS, and ENERGY. For the CHARGE part, the Mulliken charge is extracted to construct the trainset file; for the GEOMETRY part, the molecular structure topology after GAUSSIAN optimization is extracted to construct the trainset file; for the CELL PARAMETERS part, the unit cell parameters after VASP optimization are extracted to construct the trainset file; for the ENERGY part, the energy of different structures is extracted and converted into the reaction energy of a specific reaction, the state equation of the substance based on the stable structure, the adsorption energy, etc., and converted into the specified unit to construct the trainset file.

[0038] Optionally, check whether there is redundant structure between the geo file and the trainset file, and if there is redundant structure, delete the redundant structure.

[0039] In this embodiment, the second computing task submission module creates a PBS system file, thereby automatically submitting tasks in batches and achieving automatic fitting.

[0040] In this embodiment, the automatic drawing module can automatically call the drawing software for drawing. In this embodiment of the present invention, the system automatically calls the python program for drawing, and then compares the force field fitting results with the first principles data through images to obtain the final reaction force field.

[0041] Optionally, the high-throughput computing data automatic analysis subsystem includes a model automatic conversion module, a template calling module, a first computing task submission module and a computing data processing module;

[0042] The model automatic conversion module is used to convert molecular models in different formats into molecular models in the format required for high-throughput processing; the template calling module is used to set the corresponding calculation parameters and create the corresponding input template according to the preset calculation task type of the corresponding computing system, and then call the corresponding input template; the first calculation task submission module is used to submit the calculation task with the set calculation parameters to the corresponding computing system; the calculation data processing module is used to extract molecular information according to the molecular model and calculation task in the format required for high-throughput processing, and obtain the corresponding first-principles data.

[0043] In this embodiment, the model automatic conversion module is responsible for the format conversion of multi-format molecular models; the template calling module is responsible for calling the corresponding input templates for different calculation programs, and setting the corresponding calculation parameters according to the calculation task type; the first calculation task submission module is responsible for submitting the calculation task to the PBS system; the calculation data processing module is responsible for extracting information such as molecular structure, atomic charge, and structural energy. In this embodiment, the first calculation task submission module can also submit the calculation task to the corresponding system by customizing the system template, so as to achieve high-throughput processing.

[0044] In this embodiment, the module automatic conversion model is responsible for converting the model formats such as pdb, cif, xsd into the gjf format required by the GAUSSIAN format, the POSCAR format required by VASP, and the data format required by LAMMPS; the template calling module customizes the gif template, INCAR template and in template according to the needs in the high-throughput calculation process, and sets the corresponding parameters of each template, and then calls the gjf template, calls the INCAR template, and calls the in template. In this embodiment, the parameter settings of each template include setting the B3LPY density functional, setting the scanning of bond length, bond angle, dihedral angle, etc., setting the optimization of the model structure or single-point calculation, setting the molecular simulation time, etc.; the first calculation task submission module is responsible for generating the PBS system file submitted by the task and submitting the file to the system for program calculation; the calculation data processing module is responsible for extracting the energy data and structure data completed by the calculation, and calling python to complete the drawing according to the task requirements.

[0045] Optionally, the first computing task submission module in the high-throughput computing data automatic analysis subsystem and the second computing task submission module in the reaction force field automatic fitting subsystem may share the same computing system.

[0046] In this embodiment, the first computing task submission module and the second computing task submission module can share a PBS system or other customized computing systems. The present invention is not limited thereto, and it only needs to realize the corresponding automatic computing tasks.

[0047] Optionally, the reaction force field fitting database includes a molecular model database, a force field database and a structure database;

[0048] The molecular model database collects molecular models used for high-throughput calculations, the structural database collects molecular structures optimized by first-principles data during the fitting process, and the force field database collects reaction force fields obtained by GPU-accelerated fitting using first-principles data as input during the fitting process.

[0049] In this embodiment, the reaction force field fitting database is responsible for collecting molecular models for high-throughput calculations; collecting molecular structures, atomic charges, and corresponding energy, force, and other data after first-principle optimization during the fitting process; and collecting all expected reaction force field potential functions. Through the reaction force field fitting database, it is possible to uniformly and efficiently parse multi-source heterogeneous data generated by high-throughput calculations of different computing software, fully map the material calculation metadata model, and uniformly save it. This improves the flexibility, integrity, rigor, and processing efficiency of data integration.

[0050] The high-throughput data automatic analysis system for reaction force field fitting provided by the present invention realizes the automatic creation of input files required for reaction force field fitting. At the same time, various data can be automatically analyzed, and all structural data and force fields are automatically formed into a database, which effectively reduces human data errors, realizes efficient conversion between models and efficient data processing, and greatly improves the efficiency of reaction force field fitting.

[0051] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A high-throughput data automatic analysis system for reaction force field fitting, characterized in that: include: High-throughput computational data automatic analysis subsystem, reaction force field automatic fitting subsystem, and reaction force field fitting database; The reaction force field fitting database stores molecular models, force field data and molecular structure data used for reaction force field fitting, and classifies the molecular models to obtain molecular models in several different formats, and inputs the molecular models in several different formats into the high-throughput calculation data automatic analysis subsystem; The high-throughput calculation data automatic analysis subsystem converts the received molecular models in several different formats into molecular models in the format required for high-throughput processing, and then calls the calculation task template according to the preset calculation task type, and submits it to the corresponding calculation system for high-throughput processing to obtain the first principle data, and transmits the first principle data to the reaction force field automatic fitting subsystem and the reaction force field fitting database; The reaction force field automatic fitting subsystem constructs a molecular information processing file based on the received first-principles data, and submits a computing task to the corresponding computing system based on the molecular information processing file, and then fits the reaction force field according to the computing task to obtain a number of reaction force fields and corresponding fitting data, draws a molecular image based on the reaction force field and the corresponding fitting data, and then automatically compares the drawn molecular images to obtain the optimal reaction force field data and structural data, and transmits them to the reaction force field fitting database for storage.

2. The high-throughput data automatic analysis system for reaction force field fitting according to claim 1, characterized in that: The high-throughput computing data automatic analysis subsystem includes a model automatic conversion module, a template calling module, a first computing task submission module and a computing data processing module; The model automatic conversion module is used to convert molecular models in different formats into molecular models in the format required for high-throughput processing; The template calling module is used to set corresponding computing parameters and create corresponding input templates according to the preset computing task type of the corresponding computing system, and then call the corresponding input template; The first computing task submission module is used to submit the computing task with set computing parameters to the corresponding computing system; The computational data processing module is used to extract molecular information based on the molecular model and computational tasks in the format required for high-throughput processing, and obtain the corresponding first-principles data.

3. The high-throughput data automatic analysis system for reaction force field fitting according to claim 1, characterized in that: The reaction force field automatic fitting subsystem includes: a file construction module, a second calculation task submission module and an automatic drawing module; The file construction module constructs a molecular information processing file based on the received first-principles data; the second computing task submission module submits these constructed molecular information processing files to the computing system, performs optimization calculations, and obtains fitting force field data in combination with preset parameter files; the automatic drawing module draws a graph based on the fitting force field data, and compares the drawn images to determine the optimal reaction force field data and structural data.

4. The high-throughput data automatic analysis system for reaction force field fitting according to claim 3, characterized in that: The file construction module constructs geo files and trainset files; The trainset file includes four parts: CHARGE, GEOMETRY, CELL PARAMETERS, and ENERGY. The CHARGE part of the trainset file is constructed by extracting Mulliken charges based on the first principle data, the GEOMETRY part of the trainset file is constructed by extracting the molecular structure topology after GAUSSIAN optimization, the CELL PARAMETERS part of the trainset file is constructed by extracting the unit cell parameters after VASP optimization, the energy between each structure is extracted, and the extracted energy is converted into a specified unit to construct the ENERGY part of the trainset file.

5. The high-throughput data automatic analysis system for reaction force field fitting according to claim 4, characterized in that: Check whether there are redundant structures between the geo file and the trainset file. If there are redundant structures, delete them.

6. The high-throughput data automatic analysis system for reaction force field fitting according to claim 1, characterized in that: The first computing task submission module in the high-throughput computing data automatic analysis subsystem and the second computing task submission module in the reaction force field automatic fitting subsystem can share the same computing system.

7. The high-throughput data automatic analysis system for reaction force field fitting according to claim 1, characterized in that: The reaction force field fitting database includes molecular model database, force field database and structure database; The molecular model database collects molecular models used for high-throughput calculations, the structural database collects molecular structures optimized by first-principles data during the fitting process, and the force field database collects reaction force fields obtained by GPU-accelerated fitting using first-principles data as input during the fitting process.

Images