302 results about "Molecular geometry" patented technology

The invention relates to the technical field of computer-aided drug design, in particular to a drug molecule screening and optimizing method based on artificial intelligence prediction, which comprises the following steps: S1, obtaining a dynamic protein conformation set and molecular multi-dimensional characterization: obtaining a dynamic conformation set of a target protein and a physicochemical property spatial distribution diagram of a binding pocket of the dynamic conformation set, a two-dimensional molecular map topological structure and three-dimensional conformation coordinates of the drug molecules are obtained; s2, multi-modal fusion prediction is carried out; s3, generating interpretable optimization guidance; and S4, automatic iterative optimization: performing batch prediction and screening on the new candidate molecular structure, taking the screened optimal molecule as a new starting point, repeatedly executing the interpretability optimization guidance generation step and the step until an iteration termination condition is met, and outputting a final optimized molecule list. Through the multi-modal fusion deep learning model, the interaction strength of the drug molecules and the target protein can be quickly and accurately predicted, and the screening efficiency of the drug molecules is greatly improved.

The invention relates to a generative molecule reverse design system based on reinforcement learning, which comprises a data set construction module, a multi-target performance prediction model establishment module, a pre-training module, a reward function construction module and an optimization module, and is characterized in that the data set construction module is used for constructing and screening to obtain a molecular structure performance data set; the multi-target performance prediction model establishment module is used for establishing a multi-target performance prediction model based on the constructed molecular structure performance data set; the pre-training module is used for pre-training a molecular generation model by using the screened molecular structure data; the reward function construction module is used for constructing a layered multi-target reward function; and the optimization module is used for rapidly evaluating key indexes by using a performance prediction model by adopting a reinforcement learning method, and carrying out optimization adjustment on the molecular generation model through a layered multi-target reward function. According to the invention, efficient and systematic reverse design of lithium metal negative electrode interface self-assembly molecules can be realized.

The invention relates to the technical field of molecular simulation, and provides a method for constructing a heavy component average molecular structure of thickened oil, which comprises the following steps: inputting parameters, generating a core aromatic ring unit, constructing an inner-layer aliphatic ring unit, constructing an outer-layer aliphatic ring unit, generating a side chain unit, and storing and outputting structural information. A plurality of molecular structures meeting experimental data are generated at a time through a traversal algorithm thought, composition characteristics of a heavy oil complex mixture are met, more available molecular structures are provided for molecular simulation, the accuracy and efficiency of molecular modeling are improved, the molecular structures completely meeting the experimental data can be rapidly generated, and the method is suitable for large-scale popularization and application. And only the mass spectrometer, the elemental analyzer, the nuclear magnetic resonance hydrogen spectrum and the infrared spectrum data need to be input, so that the experiment cost is reduced.

The invention discloses a three-dimensional molecule generation method fusing conditional flow matching and Van der Waals radius constraint, which comprises the following steps: processing a molecule training data set, and extracting a total number of atoms and a training element component histogram; based on the optimal transmission path interpolation, combining the sampling time step and the standard Gaussian noise to construct a noise coordinate and a target condition velocity field; the noise coordinates are input into a continuous flow matching prediction model, node features are extracted through affine transformation modulation, a prediction velocity field is obtained, soft atom type distribution is generated, and the expected Van der Waals radius of each atom type is calculated; calculating flow matching loss through a prediction velocity field and a target condition velocity field, calculating a geometric collision penalty term in combination with an expected Van der Waals radius and a noise coordinate, and constructing a total loss function training model parameter; and defining an ordinary differential equation by using the trained parameters for solving, and outputting a three-dimensional molecular structure file. According to the method, atom space overlapping is inhibited, and the physical rationality and chemical effectiveness of generated molecules are improved.

A molecular structure prediction method based on a multi-granularity graph neural network comprises the steps of data set preprocessing, multi-granularity level data feature construction, multi-granularity graph neural network construction, multi-granularity graph neural network training, multi-granularity graph neural network verification and multi-granularity graph neural network testing. The key connection relation between atoms, the incidence relation of substructure keys and graph-level global feature representation are determined in the molecular graph, and the problem of local information loss in graph representation learning is effectively relieved; a message passing mechanism of graph neural sub-networks with different granularities is improved, a multi-granularity molecular graph data structure is trained, unique information of each hierarchical molecular structure is fully utilized, and the modeling capability of a model for a complex chemical structure is enhanced. The method has the advantages of being high in prediction accuracy, reducing errors, relieving local information loss existing in the prediction method, being good in prediction interpretability and the like, and can be applied to the technical fields of drug discovery, molecular property prediction and the like.

The invention discloses a method and device for predicting the transmembrane capability of a substance molecule, terminal equipment and a storage medium, and the method comprises the steps: obtaining a target molecular substance, and determining a molecular structure diagram corresponding to the target molecular substance; according to a pre-trained prediction model, determining a permeation free energy prediction value of the target molecular substance; according to the prediction model, multilevel feature extraction is carried out on a molecular structure diagram of sample substance molecules by utilizing a graph neural network model, contribution values of atoms in the sample substance molecules to transmembrane behaviors are evaluated through a gradient weighted class activation mapping method, and permeation free energy prediction values corresponding to the sample substance molecules are output. The graph neural network is trained to obtain the prediction model, and the prediction model is adopted to predict the transmembrane capability of the target molecular substance, so that intelligent screening of the transmembrane capability of a large-scale drug molecular library can be realized, the prediction efficiency and flux are remarkably improved, and intelligent prediction of the transmembrane free energy of substance small molecules is realized.

The invention discloses a multi-target drug molecule generation model construction method and a multi-target drug design method, and relates to computer drug design and bioinformatics. Determining a corresponding two-dimensional molecular map based on the SMILES sequence; the fully-connected pharmacophore diagram and the two-dimensional molecular diagram are used as input of a GatedGCN module, each atomic node in the two-dimensional molecular diagram is connected with all pharmacophore nodes in the fully-connected pharmacophore diagram in message transmission, and information exchange between different nodes is achieved; the masked SMILES sequence serves as the input of an encoder, and the decoder generates an SMILES sequence conforming to pharmacophore characteristics in an autoregression mode; freezing the network parameters of the GatedGCN module and the encoder; and training the multi-target drug molecule generation model through the elite multi-target molecule set, and reversely optimizing and updating decoder network parameters according to an output result. According to the method, the model fully learns the multi-target molecular structure characteristics, and the multi-target drug molecule generation accuracy is improved.

The invention discloses an ultra-low interfacial tension surfactant screening method and system based on AI4S, and relates to the technical field of oil and gas exploitation. The method comprises the following steps: constructing a database covering molecular structures of typical surfactants; constructing a prediction model for predicting the surface tension; generating candidate molecules by using a generative artificial intelligence model comprising a variational auto-encoder and a generative adversarial network; inputting the candidate molecules and the experimental conditions into the trained prediction model to predict the interfacial tension of the prediction model under the target condition, and screening the candidate molecules with the predicted performance reaching the standard; and performing experimental determination on the screened candidate molecules until molecules with target performance are obtained, and feeding back experimental results to a database to form closed-loop optimization of data, prediction and experiments. The research and development efficiency and success rate of the chemical oil-displacing agent can be remarkably improved, the research and development period is shortened, the research and development cost is reduced, and effective technical support is provided for improving the oil and gas recovery efficiency and guaranteeing energy safety.

The invention discloses a method and system for determining the reaction activity of material molecules based on quantum computing, and relates to the technical field of material performance analysis based on quantum computing, and the method comprises the following steps: generating a secondary quantization molecular Hamiltonian based on a target material molecular structure; traversing non-zero coefficient excitation terms in the molecular Hamiltonian, and screening an excitation operator conforming to the symmetry of a molecular point group as a target excitation operator; generating symmetric unitary coupling cluster single-double excitation simulation according to the target excitation operator; constructing a quantum circuit based on the configuration; and determining the reaction activity of the target material molecule by using the quantum circuit. According to the method, traditional manual symmetry analysis is avoided, professional knowledge is not needed, and the operation process is simplified. The method is systematized, easy to program and implement, high in universality and capable of being automatically expanded to any molecule, the design cost is remarkably reduced, and the molecular energy level solving efficiency is improved.

The invention discloses a multi-mode chemical reaction yield prediction method based on adaptive data screening, and belongs to the technical field of chemical synthesis and machine learning crossing. According to the method, a multi-modal input system containing one-dimensional chemical attribute data, a two-dimensional molecular structure map and a three-dimensional spatial configuration is constructed, so that multi-dimensional analysis of molecular interaction is realized; and meanwhile, a multi-stage training process is adopted to simulate a human cognitive rule, so that the model progressively learns layer by layer from a basic reaction feature to a complex reaction mechanism. According to the method, the screening threshold is dynamically adjusted, noise interference is effectively suppressed, and the adaptability to long-tail distribution data is improved. Compared with a traditional single-mode prediction model, the method has the advantages that the prediction accuracy on a noisy data set is improved, the trial and error cost of chemical synthesis experiments is remarkably reduced, and reliable technical support is provided for efficiently screening reaction conditions and accelerating research and development of new compounds.

This invention relates to a method and apparatus for predicting molecular properties by integrating three-dimensional structure and prior features. The method includes: identifying a set of predicted properties and a set of prior properties; constructing a molecular property prediction model based on the predicted and prior property sets; constructing a model training dataset and training the molecular property prediction model based on the dataset; after model training, receiving first molecular information input by the user, preparing model input data based on the first molecular information and the prior property set, and inputting the three-dimensional molecular structure M and prior characteristic vector X obtained from this data preparation into the molecular property prediction model to predict the corresponding property prediction vector Y, which is then fed back to the current user. This invention can reduce prediction complexity, shorten prediction time, and improve prediction efficiency.

A method for a structural analysis of a sample molecule using a mass spectrometer includes: a process for obtaining a standard analysis condition for each of setting items for performing a molecular-related ion measurement, where the setting items include an ion amount setting item, a mass-to-charge-ratio range setting item, and a signal intensity setting item; a process for performing a product ion measurement under an altered analysis condition, to acquire mass spectrum data of the product ion measurement, where the altered analysis condition is prepared by changing at least one of analysis conditions of the ion amount setting item, the mass-to-charge-ratio range setting item and the signal intensity setting item in the standard analysis condition; a process for extracting peaks corresponding to the fragment ions; and a process for determining at least a portion of the structure of the sample molecule based on mass information of the extracted peaks.

A method and apparatus for the characterization of matter is described. Light from or modulated by the matter is analyzed by means of the spectral correlation. The spectral correlation can report on the composition, static and temporally dynamic characteristics of the matter. The amplitude and temporal characteristics of the spectral correlation are measurement features that report on properties of matter, or changes in local molecular forces, chemical composition, molecular structure, shape, size, charging state, mass, and more. The advantage of spectral correlation of scattered light lies in the increased time-resolution and minimized noise in the spectroscopic characterization of matter, in particular in the characterization of fluctuations. In one embodiment, an apparatus may combine an optical system to illuminate the matter, collect scattered photons, direct these photons into an optical interferometer, and the time-resolved detection of the photons. By performing temporal intensity correlation of the light after the interferometer the spectral correlation is obtained.

The invention discloses a monomolecular structure design and generation method based on a large language model, and relates to the crossing field of monomolecular electronics and artificial intelligence. In order to solve the problems of low single molecule design efficiency and insufficient data, SingMolT5 is obtained through field fine tuning on the basis of MolT5, and generation from a natural language to a molecular SMILES is realized. According to the method, data are collected from literatures in the single molecule field and an existing disclosed molecule library, a fine tuning data set containing 329 high-quality instructions is constructed, and the fine tuning data set comprises three basic data types including a molecular skeleton, an anchoring group and molecules. The model is initialized by a MolT5 check point, and a cross entropy loss function is used in the fine tuning process; reasoning, generating and using a beam search strategy; the follow-up evaluation indexes comprise BLEU, Levenshtein, MACCS, RDK, Morga and effectiveness. The method provides an innovative technical path for'low experience and data dependent type single molecule design 'in the field of single molecule electronics, and is particularly suitable for aided design of a single molecule device core function unit and a single molecule sensing probe molecule, namely, a target molecular structure can be quickly converted through a natural language, the design threshold of researchers is reduced, and the design efficiency is improved. And a traceable technical basis can be provided for screening and iteration of a monomolecular structure based on a high-quality data set and rigorous evaluation logic.

The invention discloses a liquid crystal performance prediction method and device based on molecular dynamics and a liquid crystal theory, and relates to the technical field of liquid crystal performance prediction.The method comprises the steps that the molecular structure of liquid crystal to be simulated is confirmed, and the molecular structure of the liquid crystal is obtained; generating a coarse graining molecular structure and a coarse graining force field parameter through a coarse graining scheme; constructing a nematic phase initial structure of the volume phase system; performing molecular dynamics simulation at a set temperature to obtain molecular dynamics simulation trajectory data, calculating to obtain a molecular director of liquid crystal molecules, and determining an overall director; a second-order-sequence parameter and a fourth-order-sequence parameter are obtained through calculation; constructing a correlation function and fitting to obtain the rotation relaxation time of the system; according to a Nemtsov-Zakharov formula or a Fialkowski formula, the rotary viscosity of the liquid crystal is obtained through calculation; constructing a change relation of the second-order-sequence parameter along with the temperature; and performing fitting analysis through a Maier-Saupe theory or a critical index formula to determine a phase change point. The liquid crystal performance prediction efficiency can be improved, the cost is reduced, and the prediction precision is improved.

A method for predicting a molecular structure includes: preparing a learning data set including first learning data including an eigenvector value and a quantum mechanics calculation value for a monoatomic and molecular structural model, a bulk structural model, a slab structural model, and a nanoparticle structural model of a material including a plurality of elements; learning an artificial neural network using the learning data set to obtain a potential value; and predicting a molecular structure of another material by using the potential value.

The application discloses a bismaleimide resin performance prediction model training method, a performance prediction method, a structure design method and related products, relates to the technical field of bismaleimide resin, and the performance prediction model training method comprises the following steps: taking a molecular structure as input, taking a test condition as a supplementary input feature, taking a dielectric constant as a label, taking a predicted value of the dielectric constant as output, training a supervised learning model to obtain a dielectric constant prediction model; taking the molecular structure as input, taking melting point data as a label, taking a predicted value of the melting point data as output, training the supervised learning model to obtain a melting point prediction model; and taking the molecular structure as input, taking a 5% thermal decomposition temperature as a label, taking a predicted value of the 5% thermal decomposition temperature as output, training the supervised learning model to obtain a thermal decomposition temperature prediction model. The three prediction models are obtained by training the supervised learning model, so that the performance prediction of the bismaleimide resin material can be realized based on the theoretical model.

This application provides a method and apparatus for training a molecular multiconformation prediction model. The method includes: acquiring training data, which includes molecular structure representation samples and molecular conformation samples; processing the molecular structure representation samples and molecular conformation samples using the molecular multiconformation prediction model to be trained to obtain a training output; constructing a loss function value based on the training output and the molecular conformation samples; and, if the loss function value or the number of training rounds is within a preset range, using the trained molecular multiconformation prediction model as the molecular multiconformation prediction model. This application can achieve fast and accurate prediction of multiple three-dimensional conformations of molecules without analyzing their potential energy surfaces, reducing the computational resources and time consumed in resolving multiple three-dimensional conformations of molecules.

The invention discloses a method for identifying a molecular structure and constructing a reaction network based on atomic coordinates, and belongs to the technical field of material simulation. According to the method for recognizing the molecular structure based on the atomic coordinates, by obtaining trajectory data, atoms outside a box are converted into atoms inside the box through minimum mirror agreement; grids are divided according to a space truncation radius, potential bonding atoms are accurately screened, and redundancy of distance calculation between every two atoms of the whole system is avoided; atomic clusters are quickly divided in combination with depth-first search, and molecular structure recognition is achieved. The method only depends on basic data such as atomic coordinates and does not depend on a simulation software post-processing module, and the limitation that only a single simulation system is adapted in the prior art is broken through. According to the method for constructing the reaction network, molecular structure description is compared frame by frame, and invalid reactions are offset and filtered through positive and negative reactions to obtain net reactions; counting the net reaction times to determine the reaction probability so as to quantify the tendency of the reaction; and constructing a reaction network by taking the molecular structure as a node and the reaction frequency as an edge weight.

This invention provides a system and program that facilitates the process of generating and / or visualizing molecular networks for library molecular structure content. [Solution] In system 500, the molecular network generation system 502 for library molecular structure content includes an evaluation component 512 that performs a comparison between a first molecular fingerprint including first molecular structure data of a first molecular structure and a second molecular fingerprint including second molecular structure data of a second molecular structure, and a visualization component 516 that generates display data that visualizes a representation of the structural similarity score obtained from the first molecular structure, the second molecular structure and the comparison. The representation by the visualization component includes edges corresponding to the structural similarity score that extend between pairs of nodes corresponding to the first molecular structure and the second molecular structure.

The invention discloses a rapid and accurate high-throughput drug screening system based on deep learning, and relates to the field, and the method comprises the steps: obtaining molecular structure data containing two-dimensional structural features and three-dimensional conformation features in a to-be-screened drug molecular library, and converting the molecular structure data into a numerical feature matrix; and constructing a feature association network based on the feature matrix, identifying influence nodes through node connectivity, and determining a dynamic weight coefficient. And performing weighted reconstruction on the feature matrix by using the dynamic weight coefficient to generate optimized feature data, and calculating the binding strength and spatial complementarity of the candidate drug molecules and the target sites to obtain molecular activity distribution data. Constructing a map according to the activity distribution data, determining an initial screening interval, and optimizing and adjusting the screening interval by calculating a structural skeleton difference value between candidate molecules, so as to screen out target drug molecules. According to the method, large-scale drug molecule screening work can be rapidly and accurately completed, and the drug research and development efficiency is improved.

Embodiments of the present application provide a three-dimensional molecular structure generation method, device, equipment and storage medium, at least applied to the field of artificial intelligence and the field of drug synthesis, wherein the method comprises: obtaining atomic type noise and atomic coordinate noise at a current time step; constructing a full connection adjacency matrix based on the atomic type noise; performing three-dimensional isometry processing on the atomic type noise, the atomic coordinate noise and the full connection adjacency matrix to obtain a molecular structure distribution average value at the current time step; generating a molecular structure of the molecule iteratively based on the molecular structure distribution average value at the current time step to obtain a molecular representation of the molecule; and constructing the three-dimensional molecular structure through the molecular representation. Through the present application, the efficiency of three-dimensional molecular structure generation can be accelerated and cumulative errors can be avoided, so as to accurately generate a three-dimensional molecular structure of an effective molecule.

The application discloses a screening method of a phosphogypsum flotation collector, and comprises the following steps: providing a molecular structure of the collector, and obtaining the composition of impurities in the phosphogypsum; constructing a molecular model of the collector and a unit cell model of the impurities respectively, and performing geometric optimization to obtain an optimized collector molecular model and an optimized impurity unit cell model; constructing a unit cell box of the impurities; constructing an adsorption model between the collector and the impurities based on the optimized collector molecular model and the unit cell box of the impurities; performing molecular dynamics simulation calculation on the adsorption model to obtain the adsorption energy between the collector and the impurities; and sorting the adsorption energy according to the size to screen out the collector with better collection effect. The application reveals the adsorption mechanism of the collector and the phosphogypsum impurities from the molecular and atomic scales, can more quickly select the optimal collector compounding scheme based on different regional phosphogypsum impurity phases, and provides theoretical support for phosphogypsum impurity removal.

The invention discloses a quantum chemical screening method for extracting arbutin, and relates to the technical field of natural product extraction, and the quantum chemical screening method comprises the following steps: constructing an arbutin single-molecule model and a hydrogen bond donor and hydrogen bond receptor single-molecule model by using GaussView software, firstly performing conformation search through MOPAC and molplus programs to obtain a most stable molecular structure, and then extracting the arbutin from the molecular structure; carrying out geometric optimization and single-point energy calculation by utilizing Gaussian; the optimized hydrogen bond donor and hydrogen bond acceptor establish random clusters according to different molar ratios, the random clusters with target molecules are established, a cluster model with an optimal structure is selected for geometric optimization and single-point energy calculation, meanwhile, the binding energy between the eutectic solvent and arbutin is calculated, and IGMH analysis, ESP analysis and AIM analysis are carried out. The invention provides a universal method for green and efficient extraction of active ingredients of natural products.

The invention relates to the technical field of electrical data processing, and provides an anthocyanin stabilizer screening method and system based on fusion of quantum chemistry calculation and machine learning. The method comprises the following steps: constructing a molecular structure data set; performing geometric structure optimization on the anthocyanin molecules and the candidate polyphenol molecules to obtain a ground state optimization structure; performing multi-dimensional molecule descriptor calculation on the candidate polyphenol molecules to obtain a polyphenol molecule descriptor matrix; performing structure optimization, single-point energy calculation and free energy calculation on the anthocyanin polyphenol compound system to obtain a combined free energy value set; integrating the polyphenol molecule descriptor matrix and the combination free energy value set, constructing a training data set, and performing feature screening; performing model training through a machine learning algorithm to obtain a combined free energy prediction model; and predicting the binding free energy of the anthocyanin polyphenol compound through the binding free energy prediction model to obtain candidate molecules of the anthocyanin stabilizer. The screening efficiency of the anthocyanin stabilizer is improved.

The present disclosure relates to the technical field of nanoparticle material simulation, and particularly relates to a method and device for generating active nanoparticles and an electronic device. The method comprises: generating a base of active nanoparticles according to a third quantity and first configuration parameters; generating a modification structure containing a second quantity of molecular structures according to the second quantity; performing a preset operation on the modification structure to obtain a processed modification structure; connecting a connection site in the processed modification structure with a modification site of the base of the active nanoparticles to obtain a molecular model of the active nanoparticles; determining a charge carried by each atom in the molecular model based on force field parameters of a chemical environment corresponding to the molecular model; and performing relaxation processing on the molecular model based on atomic coordinates of each atom in the molecular model, the charge carried by each atom, and the force field parameters to obtain the active nanoparticles.

The invention relates to a characterization method of a thermal decomposition mechanism of an azo initiator. The characterization method comprises the following steps: (1) carrying out thermal stability analysis on the azo initiator; (2) acquiring microscopic decomposition characteristic data of the azo initiator by using a thermogravimetric-Fourier transform infrared spectroscopy combined technology; (3) constructing a three-dimensional molecular structure model of the azo initiator; (4) calculating the bond dissociation energy of the key covalent bond based on the density functional theory; and (5) the thermal decomposition characteristic temperature, the heat release behavior, the decomposition product and the evolution path of the azo initiator are determined by integrating the experimental data measured by the combination technology and the bond dissociation energy obtained by theoretical calculation, and the complete thermal decomposition mechanism of the azo initiator is clarified. According to the method, the thermal decomposition rule of the azo initiator is disclosed comprehensively and accurately through a technical path of combining experimental analysis and theoretical simulation and verifying macroscopic characteristics and micromechanism, and a key theoretical basis and technical support are provided for preventing and controlling the thermal runaway risk of such substances.

The invention belongs to the technical field of computational chemistry and artificial intelligence crossing, and provides a molecular potential energy surface prediction method based on deep learning. The method comprises the following steps: inputting a structure file of an unoptimized molecule, and reading and predicting energy and force of a current molecular structure by utilizing a machine learning potential model; and then, judging whether the structure is converged or not according to a prediction result by adopting a GeomeTRIC algorithm, and if not, adjusting coordinates to generate a new structure and performing circular prediction until the structure is converged. According to the method, the quantum chemistry and the deep learning technology are combined, the calculation efficiency can be greatly improved, complex inter-atomic interaction in molecules can be fully captured, and the prediction accuracy is ensured; and meanwhile, batch processing and automatic generation of a quantum chemistry software interface are supported, and the method is particularly suitable for high-throughput molecular structure optimization in the fields of drug discovery, material science and chemical product design.

The application discloses a molecular property prediction method based on spectral position coding and bionic side inhibition gating, and relates to the field of olfactory perception. The method comprises the following steps: constructing a molecular structured input comprising atomic types, a Coulomb matrix and graph Laplacian eigenvector; fusing atomic chemical attributes and spectral domain topological positions through a feature extraction module to generate initial atomic features; using a multi-scale aggregation module, hierarchical neighborhood aggregation is carried out based on preset scale constraints to extract multi-level structure features covering chemical bonds, functional groups and molecular skeletons; with the help of a global interaction module of bionic side inhibition gating, local features are dynamically modulated by a global query signal to realize adaptive denoising and semantic sharpening; finally, through mask pooling and a decoupled multi-label prediction head, the scores of each odor attribute of the molecule are output. The application significantly improves the accuracy, structure perception ability and cross-task generalization performance of molecular property prediction.

The challenge is to reduce the computational cost when performing distributed learning with parallel data generation. [Solution] An information processing device that performs machine learning on a machine learning model that extracts molecular structure features using different molecules as multiple training data obtains the number of atoms contained in each of the multiple training data. If there is a first training data with a number of atoms greater than or equal to a threshold among the obtained number of atoms in each of the multiple training data, the first training data is replaced with another number of atoms that is less than the threshold, based on the other training data among the multiple training data.