Molecular force field modeling method and apparatus, electronic device, and storage medium

Abstract

The present application provides a molecular force field modeling method and apparatus, an electronic device, and a storage medium. The method comprises: establishing fingerprint spectrums of atoms in a target molecule on the basis of structural feature information and energy feature information of the target molecule; clustering the fingerprint spectrums of the atoms in the target molecule to obtain a plurality of atomic categories; performing modeling on different categories of atomic clusters to obtain a potential energy equation between the atomic clusters; and establishing a molecular force field according to the potential energy equation of different categories of atomic clusters. By clustering the fingerprint spectrums of the atoms in the target molecule, the clustering process is effectively used to replace the situation where atom types are classified on the basis of human experience; a molecular network probabilistic graphical model is established by means of a Bayesian method, data mining is carried out, and multi-body energy is simulated, so that the dependence of a molecular force field on fitting data is reduced, and the accuracy of a simulated molecular force field is improved.

Description

A molecular force field modeling method, device, electronic device and storage medium Technical Field

[0001] The present application relates to the technical field of structural biology and biomolecular modeling, and in particular to a molecular force field modeling method, apparatus, electronic device, and storage medium. Background Art

[0002] Currently, atomic types in molecular force fields are largely defined empirically. In practice, it has been found that the classification of atomic types based on chemists' knowledge of molecular structure and the mathematical description of interatomic interaction energies are not objective and precise enough. This is due to the limitations of current cognition. Cognitive barriers prevent objective classification of atoms based on the electron density distribution characteristics of different molecular structures using existing knowledge and human intuition, and the establishment of mathematical models to accurately calculate the energy of complex target molecular systems within the computational power of existing computers. Based on current computing capabilities, most relevant models use classical Newtonian mechanics formulas to calculate molecular energy and describe molecular motion accordingly. This calculation method is subject to cumulative errors during the calculation process, resulting in an inaccurate description of atomic-level objects between macroscopic and microscopic scales, which means that the simulated molecular force fields are not accurate enough.

[0003] Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a molecular force field modeling method, device, electronic device and storage medium for improving the problem of insufficient accuracy of simulated molecular force fields.

[0005] The embodiment of the present application provides a molecular force field modeling method, including: establishing a fingerprint of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule; clustering the fingerprint of each atom in the target molecule to obtain multiple atomic categories; for each atomic category in the multiple atomic categories, modeling atomic clusters of different categories to obtain potential energy equations between atomic clusters; establishing a molecular force field according to the potential energy equations of atomic clusters of different categories. In the implementation process of the above scheme, by establishing a fingerprint of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule, and clustering the fingerprint of each atom in the target molecule, the clustering process is effectively used to replace the situation of classifying atomic types based on human experience, thereby improving the accuracy of the simulated molecular force field.

[0006] Optionally, in an embodiment of the present application, a fingerprint of each atom in the target molecule is established based on the structural characteristic information and energy characteristic information of the target molecule, including: obtaining the total charge distribution of all atoms in the target molecule; projecting the total charge distribution of all atoms in the target molecule onto a Fibonacci sphere lattice constructed for each atom to simulate the charge density distribution on the surface of each atom; and performing dimensionality reduction sorting on the charge density distribution on the surface of each atom in the target molecule to obtain a fingerprint of each atom in the target molecule. In the implementation process of the above scheme, the fingerprint of each atom in the target molecule is obtained by projecting the total charge distribution of all atoms in the target molecule onto a Fibonacci sphere lattice constructed for each atom and performing dimensionality reduction sorting on the charge density distribution on the surface of each atom in the target molecule. Since there is no need to consider the global configuration of the molecule, the speed of processing the molecular structure and properties is improved.

[0007] Optionally, in an embodiment of the present application, obtaining the total charge distribution of all atoms in the target molecule includes: extracting structural characteristic information and energy characteristic information of the target molecule from a standard complex molecule file; and determining the total charge distribution of each atom in the target molecule based on the structural characteristic information and energy characteristic information. In implementing the above scheme, by extracting information from the file, the structural characteristic information and energy characteristic information of the target molecule can be obtained more efficiently, avoiding the time and resource consumption of conducting experiments or computational simulations, and improving the efficiency and accuracy of determining the total charge distribution.

[0008] Optionally, in an embodiment of the present application, the fingerprint spectrum of each atom in the target molecule is clustered, including: clustering the fingerprint spectrum of each atom in the target molecule according to the charge density distribution of the atom, evaluating the clustering results using a clustering evaluation method, and finally establishing a classification standard.

[0009] Optionally, in an embodiment of the present application, the molecular force field includes: a potential energy equation between atomic clusters; modeling different types of atomic clusters, including: modeling the atomic clusters to obtain a Boltzmann probability distribution between atoms in the atomic clusters; and determining the potential energy equation between the atomic clusters based on the Boltzmann probability distribution between atoms in the atomic clusters. In the implementation of the above scheme, the Boltzmann probability distribution between atoms in the atomic clusters can more accurately describe the interactions within the molecule, thereby effectively improving the accuracy of the potential energy equation.

[0010] Optionally, in an embodiment of the present application, the atomic clusters include: atomic pairs consisting of two atoms, or atomic groups consisting of three atoms; modeling of different types of atomic clusters includes: dividing all atoms in the target molecule into core region atoms and background region atoms; and iteratively removing the influence of the background region atoms on the probability density distribution of the core region atoms to obtain the potential energy equation between the atomic groups, or the potential energy equation between the atomic pairs. During the implementation of the above scheme, the iterative method can gradually reduce the influence of the background region atoms on the core region atoms, thereby obtaining a more accurate probability density distribution, thereby effectively improving the accuracy of the potential energy equation.

[0011] Optionally, in an embodiment of the present application, a molecular force field is established based on the potential energy equations of different categories of atomic clusters, including: using atoms in the target molecule as nodes, and the potential energy equations between atomic clusters as edges between nodes, to model the different categories of atomic clusters to obtain a molecular force field. In the implementation of the above scheme, by modeling different categories of atomic clusters, it is possible to more accurately describe the interactions in molecules, thereby improving the modeling precision and accuracy of the molecular force field.

[0012] Optionally, in an embodiment of the present application, the target molecules include: biological molecules and drug-like molecules.

[0013] An embodiment of the present application also provides a molecular force field simulation device, including: a fingerprint map establishment module, which is used to establish a fingerprint map of each atom in the target molecule based on the structural feature information and energy feature information of the target molecule; an atomic category acquisition module, which is used to cluster the fingerprint map of each atom in the target molecule to obtain multiple atomic categories; a potential energy equation acquisition module, which is used to model different categories of atomic clusters for each atomic category in the multiple atomic categories to obtain the potential energy equations between the atomic clusters; and a molecular force field establishment module, which is used to establish a molecular force field based on the potential energy equations of different categories of atomic clusters.

[0014] Optionally, in an embodiment of the present application, the fingerprint map establishment module includes: a charge distribution acquisition submodule, which is used to obtain the total charge distribution of all atoms in the target molecule; a charge density simulation submodule, which is used to project the total charge distribution of all atoms in the target molecule onto a Fibonacci sphere lattice constructed for each atom to simulate the charge density distribution on the surface of each atom; and a fingerprint map acquisition submodule, which is used to perform dimensionality reduction sorting on the charge density distribution on the surface of each atom in the target molecule to obtain a fingerprint map of each atom in the target molecule.

[0015] Optionally, in an embodiment of the present application, the charge distribution acquisition submodule includes: a feature information extraction unit, used to extract the structural feature information of the target molecule and the energy feature information of the target molecule from the standard complex molecule file; and a charge distribution determination unit, used to determine the total charge distribution of each atom in the target molecule based on the structural feature information and the energy feature information.

[0016] Optionally, in an embodiment of the present application, the atom category acquisition module includes: a fingerprint clustering submodule, which is used to cluster the fingerprint of each atom in the target molecule according to the charge density distribution of the atom.

[0017] Optionally, in an embodiment of the present application, the molecular force field includes: a potential energy equation between atomic clusters; a potential energy equation acquisition module, including: a probability distribution acquisition submodule, used to model different categories of atomic clusters and obtain the Boltzmann probability distribution between atoms in the atomic clusters; a potential energy equation determination submodule, used to determine the potential energy equation between atomic clusters based on the Boltzmann probability distribution between atoms in the atomic clusters.

[0018] Optionally, in an embodiment of the present application, the atomic cluster includes: an atomic pair consisting of two atoms, or an atomic group consisting of three atoms; the probability distribution acquisition submodule includes: a molecular atom division unit, used to divide all atoms in the target molecule into core area atoms and background area atoms; a background influence removal unit, used to iteratively remove the influence of background area atoms on the probability density distribution of core area atoms, and obtain the potential energy equation between atomic groups, or the potential energy equation between atomic pairs.

[0019] Optionally, in an embodiment of the present application, the molecular force field establishment module includes: an atomic cluster modeling submodule, which is used to model different categories of atomic clusters using atoms in the target molecule as nodes and potential energy equations between atomic clusters as edges between nodes to obtain a molecular force field.

[0020] Optionally, in an embodiment of the present application, the target molecules include: biological molecules and drug-like molecules.

[0021] An embodiment of the present application further provides an electronic device, including: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the machine-readable instructions are executed by the processor, the method described above is performed.

[0022] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described above is executed. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only illustrate certain embodiments of the embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0024] FIG1 is a schematic flow chart of a molecular force field modeling method provided in an embodiment of the present application;

[0025] FIG2 is a schematic diagram showing a mapping of the molecular structure of the atom pairs provided in an embodiment of the present application as core atoms;

[0026] FIG3 shows a directed acyclic graph of an atom pair consisting of two atoms and an atom group consisting of three atoms provided in an embodiment of the present application;

[0027] FIG4 is a schematic diagram showing a process of simplifying a molecular network using BFT according to an embodiment of the present application;

[0028] FIG5 is a schematic structural diagram of a molecular force field simulation device provided in an embodiment of the present application;

[0029] FIG6 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the embodiments of the present application only serve the purpose of illustration and description and are not intended to limit the scope of protection of the embodiments of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in the embodiments of the present application illustrate the operations implemented according to some embodiments of the embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art, guided by the contents of the embodiments of the present application, can add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0031] In addition, the described embodiments are only a portion of the embodiments of the present application, rather than all embodiments. The components of the embodiments of the present application generally described and shown in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed embodiments of the present application, but rather merely represents selected embodiments of the embodiments of the present application.

[0032] It will be understood that the "first" and "second" in the embodiments of the present application are used to distinguish similar objects. Those skilled in the art will understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different. In the description of the embodiments of the present application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that there may be three relationships, such as A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the related objects before and after are in an "or" relationship. The term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups).

[0033] Before introducing the molecular force field modeling method provided in the embodiments of the present application, some concepts involved in the embodiments of the present application are first introduced:

[0034] A molecular force field is a physical mathematical model that describes the interactions within and between molecules. It is typically based on the potential function in classical physics, describing the relationship between relative position and orientation, thereby calculating the forces of interaction between molecules. Molecular force fields can be used to simulate, calculate, and predict information on molecular structure, properties, and dynamics, including conformation, vibrations, thermodynamics, thermochemistry, and spectroscopy.

[0035] Molecular structural characteristics refer to the information required to determine the structure of a molecular system. This information typically includes the positions of atoms, the types and lengths of chemical bonds, and the molecular geometry. Molecular structural characteristics are fundamental to molecular dynamics, simulation, and computation, and are crucial for understanding the structure, properties, and function of molecules.

[0036] Molecular energy signatures refer to various information describing the energy of a molecular system, including total molecular energy, kinetic energy, potential energy, and intramolecular interaction energy. Molecular energy signatures are fundamental to understanding various chemical processes, including molecular stability, reaction kinetics, photochemical reactions, and spectroscopy.

[0037] Clustering is the process of grouping character sequences into clusters of similar character sequences along one or more dimensions. In other words, a cluster is a collection of data objects that are similar to objects in the same cluster but different from objects in other clusters, which is the common saying that "birds of a feather flock together."

[0038] Bayesian networks are currently one of the most effective theoretical models in the field of uncertain knowledge expression and reasoning. Bayesian networks use directed edges to express the relationships between nodes, have stronger semantics, and provide a powerful tool for studying the conditional probability distribution of the geometric structure of target atomic pairs.

[0039] The Boltzmann probability distribution is a mathematical model used to describe the distribution of particles in a thermodynamic system. It is based on the principles of statistical physics and explains the probability of particles being distributed at different energy levels. In a thermodynamic system, particles can be at different energy levels, each with a certain energy. The Boltzmann probability distribution tells us that the probability of a particle being at energy level i is related to the energy at that level and the temperature of the system.

[0040] It should be noted that the molecular force field modeling method provided in the embodiments of this application can be executed by electronic devices. Electronic devices herein refer to device terminals or servers capable of executing computer programs, such as smartphones, personal computers, tablet computers, personal digital assistants, or mobile internet devices. A server refers to a device that provides computing services over a network, such as x86 servers and non-x86 servers. Non-x86 servers include mainframes, minicomputers, and UNIX servers.

[0041] Please refer to FIG1 , which shows a schematic flow chart of a molecular force field modeling method provided in an embodiment of the present application. The embodiment of the present application provides a molecular force field modeling method, comprising:

[0042] Step S110: establishing a fingerprint of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule.

[0043] The target molecule refers to the molecule for which the molecular force field needs to be modeled. The target molecule may include: at least one of a biological molecule and a drug-like molecule; wherein the biological molecule and the drug-like molecule here have a relative size relationship, and the biological molecule is larger than the drug-like molecule. Therefore, the biological molecule is also called a biological macromolecule, and the drug-like molecule is also called a drug-like small molecule.

[0044] The structural characteristic information of the above-mentioned target molecules can be obtained through multiple methods, such as experimental techniques such as X-ray crystallography, nuclear magnetic resonance, mass spectrometry, and computational methods based on quantum chemistry calculations, molecular dynamics simulations, etc. These methods can be used to measure and calculate the geometric configuration, bond length, bond angle, electronic structure and other structural characteristic information of the molecules.

[0045] The energy signature information of the target molecule can be obtained through a variety of computational methods, including quantum chemistry calculations, classical molecular dynamics, and molecular force fields. These methods can be used to calculate the energy signature information of the molecule under a given environment, including its total energy, vibrational energy, and intramolecular interaction energy.

[0046] Step S120: clustering the fingerprint of each atom in the target molecule to obtain multiple atomic categories.

[0047] Fingerprints are patterns used to identify each atom. Different atoms have different fingerprints, which can be used to identify and distinguish them. Fingerprints are primarily characterized by different structural and energy signatures.

[0048] It is understandable that since the target molecule may be a complex biological molecule with a great diversity of physical and chemical properties, and the traditional molecular force field mostly defines the atomic categories based on human experience (i.e., classification is performed through human experience), therefore, in order to eliminate the limitations of the traditional molecular force field in classifying atomic types based on human experience, machine learning methods (such as Bayesian networks) can be used to learn and classify the commonalities and differences in the high-dimensional charge density distribution within atoms. The charge density can be regarded as one of the characteristic variables of the fingerprint map. In the specific practice, the K-means clustering algorithm (K-means) can be used to perform cluster analysis on the fingerprint map of each atom in the target molecule (such as a biological macromolecule system) to obtain multiple atomic categories.

[0049] Step S130: for each of the multiple atomic categories, model the atomic clusters of different categories to obtain potential energy equations between the atomic clusters.

[0050] Since there are many types of atoms in the target molecule, and the probability density distribution of atoms of each atomic category affects the calculation of the potential energy equation, it is necessary to model the atomic clusters of each atomic category in order to obtain the potential energy equation between the atomic clusters of each atomic category.

[0051] Step S140: establishing a molecular force field according to potential energy equations of different types of atomic clusters.

[0052] In the implementation process of the above scheme, a fingerprint map of each atom in the target molecule is established according to the structural characteristic information and energy characteristic information of the target molecule, and the fingerprint map of each atom in the target molecule is clustered, effectively using the clustering process to replace the situation of classifying the atom types based on human experience; the Bayesian method can also be used to establish a molecular graph model, perform data mining, simulate multi-body energy, and establish a potential energy function, which effectively solves the problem of insufficient accuracy of the current force field model based on Newtonian mechanics, thereby reducing the dependence of the molecular force field on fitting data, reducing the calculation time of the simulation process, and improving the accuracy of the simulated molecular force field.

[0053] As an optional implementation of the above step S110, the implementation of establishing the fingerprint of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule may include:

[0054] Step S111: Obtain the total charge distribution of all atoms in the target molecule.

[0055] All electrons on the surface of each atom in the target molecule can be obtained, and the total charge distribution of the atom can be calculated from all the electrons on the surface of each atom.

[0056] Step S112: Projecting the total charge distribution of all atoms in the target molecule onto the Fibonacci sphere lattice constructed for each atom to simulate the charge density distribution on the surface of each atom.

[0057] In practice, the total charge distribution of each atom can be projected using a charge density projection, thereby projecting the electrons of each atom onto a sphere and obtaining a spherical point for each atom on the sphere. Then, based on the spherical points of each atom on the sphere, an equidistantly distributed spherical point lattice (such as the Fibonacci lattice of evenly distributed atoms) is constructed. Finally, a Gaussian distribution function is used to project the charge density of each spherical point, thereby constructing a charge density distribution for each atom.

[0058] It is understood that the charge density of an atom is primarily determined by the total charge of the protons at the atom's nucleus and the distribution of electron density around it, as well as the chemical environment in which the atom resides. This chemical environment includes, for example, the types and geometric positions of the atoms surrounding the atom in the molecular system, the interconnections between the atoms, and the relative geometric positions of the atoms. Therefore, the calculation of charge density is relatively complex. In practice, the charge density distribution can be calculated using quantum mechanical calculations, such as calculating the charge density information that converges to the space occupied by each atom in the molecular structure of protein molecules and drug-like compounds. This charge density information includes the charge density distribution described above.

[0059] It is understandable that since the chemical environment in which the atoms are located is a kind of contextual environment, and the text corpus of natural language processing (NLP) is also a kind of contextual environment, the chemical environment in which the atoms are located is similar to the text corpus of NLP. Therefore, in the specific practice process, the distributed representation vector framework in the natural language processing algorithm can also be used to extract the charge density information of each atom in the three-dimensional structure of biological molecules. This charge density information includes the above-mentioned charge density distribution.

[0060] Step S113: performing dimensionality reduction sorting on the charge density distribution on the surface of each atom in the target molecule to obtain a fingerprint of each atom in the target molecule.

[0061] An implementation method of the above-mentioned step S113 is, for example: using the equation form of the Gauss basis group to project the charge of each atom in the molecule onto the atomic sphere, obtaining the charge density of each atom at the spherical point on the sphere, and performing dimensionality reduction sorting on the charge density distribution on the surface of each atom, thereby obtaining the fingerprint map of each atom.

[0062] As a first optional implementation of the above step S111, the implementation of determining the total charge distribution of atoms based on the structural feature information and energy feature information extracted from the standard complex molecule file may include:

[0063] Step S111a: extracting the structural characteristic information and the energy characteristic information of the target molecule from the standard complex molecule file.

[0064] The above-mentioned standard complex molecule file may include the structure file of the biological molecule (such as protein structure) and the structure file of the drug-like small molecule; before extracting the structural characteristic information of the target molecule and the energy characteristic information of the target molecule, the biological molecule (such as protein atom) file and the drug-like small molecule structure file may also be subjected to structural correction and structural optimization, for example: the structural characteristic information, energy characteristic information and charge density information (charge density information includes charge density distribution) of the protein and the drug-like small molecule may be obtained, the 6-31G* basis set and Restrained ElectroStatic Potential (RESP) may be used to fit the atomic charge information of the protein, and the atomic charge information of the drug-like small molecule may be obtained by calculating the am1-bcc charge.

[0065] Step S111b: Determine the total charge distribution of each atom in the target molecule based on the structural characteristic information and the energy characteristic information.

[0066] The total charge distribution refers to the total amount of information represented by the charge distribution of an atom.

[0067] As a second optional implementation of the above-mentioned step S111, the total charge distribution of the atoms can be determined based on the structural characteristic information and energy characteristic information extracted from the database. For example, the structural characteristic information and energy characteristic information of the target molecule can be extracted from the protein database, and the structural characteristic information and energy characteristic information of the target molecule can also be extracted from the drug-like small molecule database, and the total charge distribution of each atom in the target molecule can be determined based on the structural characteristic information and energy characteristic information.

[0068] As an optional implementation of the above-mentioned step S120, a machine learning method can be used to cluster the fingerprint spectrum of each atom in the target molecule. For example, a K-means clustering algorithm (K-means) is used to cluster and classify the atoms of each atomic category in the different atomic categories in the target molecule according to the atomic charge density distribution, so as to obtain multiple atomic categories; wherein, the silhouette coefficient method can be used for evaluation, and the number of classifications and the centroid coordinates with the highest scores are retained to define the atomic type for each element. Optionally, after clustering and classifying the atoms of each atomic category in the different atomic categories in the target molecule according to the atomic charge density distribution, clustering results of multiple atomic categories are obtained. In the specific practice, a clustering evaluation method can also be used to evaluate the above clustering results, so as to finally obtain a classification standard for classifying atomic categories. After obtaining the classification standard, the electronic device can directly use the classification standard to classify atoms according to the atomic charge density distribution.

[0069] As an optional implementation of the above step S130, the above molecular force field may include: a potential energy equation between atomic clusters; the implementation of modeling different types of atomic clusters may include:

[0070] Step S131: Modeling the atomic cluster to obtain the Boltzmann probability distribution between atoms in the atomic cluster.

[0071] It can be understood that in the process, all atoms in the target molecule can be divided into core area atoms and background area atoms, and the influence of the background area atoms on the probability density distribution of the core area atoms can be removed in an iterative manner, and finally the atoms corresponding to the atomic type selected in the core area each time are obtained, and the Boltzmann probability distribution between the atoms of the atomic cluster is formed under the action of the potential energy within the selected atomic type.

[0072] Step S132: determining the potential energy equation between the atomic clusters according to the Boltzmann probability distribution between the atoms in the atomic cluster.

[0073] For example, the implementation of step S132 is as follows: It can be understood that the conformational potential energy of N atoms corresponding to N selected atomic types can be derived from the Boltzmann probability, so the formula E can be used. N (R N )=-RTln[P(R N )] calculate the Boltzmann probability distribution between atoms in the atomic cluster and obtain the potential energy equation between the atomic clusters; where N represents the number of atomic categories selected in the core area, R N Represents the conformational coordinate variables of N atoms, E N Indicates that the selected N atoms have the conformational coordinate variable R N When N=2, the above formula calculates the potential energy equation of the atomic pair consisting of two atoms. Similarly, when N=3, the above formula calculates the potential energy equation of the atomic group consisting of three atoms. When N can be greater than 3, the E N Indicates that the selected N atoms have the conformational coordinate variable R N The potential energy equation of the system when .

[0074] As an optional implementation of the above step S131, the above atomic cluster may include an atomic pair consisting of two atoms, or the above atomic cluster may include an atomic group consisting of three atoms.

[0075] Please refer to FIG2 for a schematic diagram of a mapping of the molecular structure of the atom pair provided in the embodiment of the present application as the core region atom; for ease of understanding and explanation, the following is a detailed description of an atom pair consisting of two atoms as the core region atom, and a traversal method can be used to select all atoms corresponding to the specific N types of atoms in the target molecule and place them in the core region (for example, the inner area of ​​the dotted ellipse in the figure), and establish a molecular force field for the N types of atoms. The above-mentioned implementation methods for modeling different types of atomic clusters may include:

[0076] Step S133: Divide all atoms in the target molecule into core region atoms and background region atoms.

[0077] It can be understood that the core atoms refer to the atoms for which the molecular force field needs to be established, such as the atoms in the inner area of ​​the dotted ellipse in the figure. i and r j Two atoms, background atoms refer to other atoms that have an impact on the probability density distribution of core atoms, such as other atoms in the dotted elliptical area outside the figure (i.e., other atoms except r i and r j atoms other than the two atoms).

[0078] Step S134: removing the influence of the background region atoms on the probability density distribution of the core region atoms in an iterative manner, and obtaining the potential energy equation between the atomic groups, or the potential energy equation between the atomic pairs.

[0079] In practice, data mining is performed on each target molecule in a structural database containing protein-ligand structures to obtain the probability density distribution of the core region atoms. Then, the influence of the background region atoms on the probability density distribution of the core region atoms is iteratively removed to obtain the potential energy equation between the atom pairs. Similarly, the potential energy equation between atom groups is also obtained in the same way. For example, data mining is performed on each target molecule in a structural database containing protein-ligand structures to obtain the probability density distribution of the core region atoms. Then, the influence of the background region atoms on the probability density distribution of the core region atoms is iteratively removed to obtain the potential energy equation between atom groups.

[0080] As an optional implementation of the above step S140, the implementation of establishing the molecular force field according to the potential energy equations of different types of atomic clusters may include:

[0081] Step S141: Using atoms in the target molecule as nodes and potential energy equations between atomic clusters as edges between nodes, different types of atomic clusters are modeled to obtain a molecular force field.

[0082] It can be understood that a Bayesian network is a directed acyclic graph (DAG), which describes a causal chain through directed edges (conditional probabilities express the strength of relationships). In any Bayesian network, a cause can produce an effect, and the effect can also serve as a prior condition for other causes in the causal chain, and the same cause cannot serve as a prior condition for itself. Within any target molecule, there is an interaction potential between every two atoms, which allows the DAG of a molecule to be represented using a fully connected network graph model. Precisely because of these properties, Bayesian networks can be used to model different types of atomic clusters, that is, to represent different molecular structures. For example, using atoms in the target molecule as nodes (i.e., using nodes in the DAG graph to represent atoms) and potential energy equations between atomic clusters as edges between nodes (representing the energy transfer relationship between atoms), Bayesian networks can be used to model different types of atomic clusters to obtain molecular force fields.

[0083] Please refer to FIG3 for a directed acyclic graph of an atom pair consisting of two atoms and an atom group consisting of three atoms provided in an embodiment of the present application; the arrowed lines between the nodes in the figure represent the interaction relationship and direction. In order to better describe the process of modeling the Boltzmann probability of an atom pair ij using DAG, the atomic coordinate node r of the atom pair ij can be i and r j A node c is added in the middle ij To describe whether atoms i and j interact with each other.

[0084] The likelihood probability can be obtained from the Bayesian probability formula: Among them, P(c ij |r i , r j ) represents the likelihood probability between atom pair ij, P(c ij ) is the prior probability of whether the atomic pair ij composed of two atoms in the atomic pair interacts with each other. When the atomic pair ij interacts with each other, the prior probability is 1, and when the atomic pair ij does not interact with each other, the prior probability is 0. ij ) represents the unconditional probability distribution of atom pair ij without constraints.

[0085] As can be seen from Figure 2, all atoms in the molecule can reach the target atom pair through a certain path. According to the chain rule chain of probability, all nodes on these chains can interact with the target atom pair through a certain path. The following formula can be used to describe this chain.

[0086] Where C is the set of all pairwise interactions in the target molecular system, D is the set of all atom pairs at different distances in the target molecular system, R is the set of all atomic coordinate nodes in the target molecular system, and v is the sequence number of the atomic distance node or interaction node closest to the target atom pair. ij |c ij ) is the Boltzmann probability of the target atom pair, that is, the radial distribution of the target atom pair affected only by the interaction between them.

[0087] In a molecular network, each atom can be in multiple interactions, and different types of atoms can be connected in different ways. Since the interaction between target atom pairs cannot be quantified individually, using such a complex structure to train the model will produce the "curse of dimensionality". The above-mentioned Bayesian network can generate the Boltzmann probability P(d ij |c ij). In molecular networks, the true potential energy of atoms is difficult to measure directly, however, the distance between atoms can be measured directly. Therefore, the true potential energy of atoms can be modeled based on the distance between atoms, and taking into account that the farther the node is from the target node, the smaller the influence. In order to correct the molecular network, Bayesian Field Theory (BFT) can be used to assume that the background interaction chain is a Markov chain, and each node is determined only by its previous node. The above-mentioned BFT introduces background nodes to represent the indirect interactions between neighboring atoms that are farther away from the target atom. In the corrected molecular graph, each background interaction chain is separated and modeled using a method based on Bayesian theory.

[0088] Please refer to FIG4 for a schematic diagram of the process of using BFT to simplify molecular networks provided by an embodiment of the present application; the figure shows the process of using BFT to simplify the target molecule, subgraph A represents the interaction relationship between the atom r and the collar atom other than the collar atom, wherein the interaction relationship is represented by node c. Subgraph B represents a schematic diagram when the interatomic distance node dij is used to replace the target atom coordinate nodes ri and rj, and subgraph C represents the use of background node b to describe the entire background area. Each background node can be represented by a Markov chain. The interaction between the nodes, for example, the Markov chain here is: first from the background node b to the collar atom a, and then from the collar atom a to the target atom pair. When the multi-body effect and the correlation between the background nodes are not considered, in the directed graph, the multi-conditional probability after BFT simplification can be extended according to the chain rule of probability, including the probability of all chain nodes tracing back along each path until they reach the end point or collision node (node ​​pointed to by multiple arrows). In the molecular network optimized by BFT, the conditional probability of the target atom pair composed of any two atoms can be expressed as follows:

[0089] Among them, m represents the sequence number of the action path, N v Indicates the total number of action paths.

[0090] Similarly, in the molecular network after BFT optimization, the conditional probability of the target atomic group consisting of any three atoms can be expressed as follows:

[0091] Among them, m represents the sequence number of the action path, N v Indicates the total number of action paths.

[0092] In practice, statistical methods can be used to calculate the conditional probability between the neighboring atoms and the target atom or target structure in each action path. The conditional probability here can be expressed as follows:

[0093] Among them, P(d m |c m )P(d ij |c m )P(d m |b m ) represents the influence of the mth action path on the target atom pair or target triatomic structure (also known as an atomic group composed of three atoms). By iteratively deriving the above formula, we can eliminate the influence of the probability density distribution of background atoms on the core atoms, thereby obtaining the interaction probability distribution of the independent target structure in each action path that is only affected by the potential energy change. The two-atom Boltzmann probability distribution of the potential energy between the above atomic pairs can be expressed as:

[0094] Among them, P(d m |c m )P(d ij |c m )P(d m |b m ) represents the effect of the mth action path on the target atomic pair or target triatomic structure (also known as an atomic group composed of three atoms). The above two formulas can be used to derive the conditional probability distribution of the triatomic structure changing with potential energy when there is a many-body effect:

[0095] Among them, P(d m |c m )P(d ij |c m )P(d m |b m ) represents the influence of the mth action path on the target atom pair or target triatomic structure (also known as an atomic group composed of three atoms).

[0096] It is understood that the specific meanings of the letter variables not explained in the above formulas can be found in the definition table of variable symbols in Table 1 below, and the specific meanings of the letter variables between formulas can also be referenced to each other.

[0097] Table 1 Variable symbol definition table

[0098] Referring to FIG. 5 , a schematic structural diagram of a molecular force field simulation device provided in an embodiment of the present application is shown. The present application embodiment provides a molecular force field simulation device 200, comprising:

[0099] The fingerprint spectrum establishment module 210 is used to establish a fingerprint spectrum of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule.

[0100] The atom category acquisition module 220 is used to cluster the fingerprint of each atom in the target molecule to obtain multiple atom categories.

[0101] The potential energy equation obtaining module 230 is used to model different types of atomic clusters for each of the multiple atomic types, and obtain potential energy equations between the atomic clusters.

[0102] The molecular force field establishing module 240 is used to establish a molecular force field according to potential energy equations of different types of atomic clusters.

[0103] Optionally, in an embodiment of the present application, the fingerprint map establishment module includes:

[0104] The charge distribution acquisition submodule is used to obtain the total charge distribution of all atoms in the target molecule.

[0105] The charge density simulation submodule is used to project the total charge distribution of all atoms in the target molecule onto the Fibonacci sphere lattice constructed for each atom to simulate the charge density distribution on the surface of each atom.

[0106] The fingerprint acquisition submodule is used to perform dimensionality reduction sorting on the charge density distribution on the surface of each atom in the target molecule to obtain the fingerprint of each atom in the target molecule.

[0107] Optionally, in an embodiment of the present application, the charge distribution acquisition submodule includes:

[0108] The feature information extraction unit is used to extract the structural feature information and energy feature information of the target molecule from the standard complex molecule file.

[0109] The charge distribution determination unit is used to determine the total charge distribution of each atom in the target molecule based on the structural characteristic information and the energy characteristic information.

[0110] Optionally, in an embodiment of the present application, the atomic category obtaining module includes:

[0111] The fingerprint clustering submodule is used to cluster the fingerprint of each atom in the target molecule according to the charge density distribution of the atom.

[0112] Optionally, in an embodiment of the present application, the molecular force field includes: a potential energy equation between atomic clusters; and a potential energy equation acquisition module including:

[0113] The probability distribution acquisition submodule is used to model different types of atomic clusters and obtain the Boltzmann probability distribution between atoms in the atomic clusters.

[0114] The potential energy equation determination submodule is used to determine the potential energy equation between atomic clusters based on the Boltzmann probability distribution between atoms in the atomic clusters.

[0115] Optionally, in the embodiment of the present application, the atomic cluster includes: an atomic pair consisting of two atoms, or an atomic group consisting of three atoms; the probability distribution acquisition submodule includes:

[0116] The molecular atom division unit is used to divide all atoms in the target molecule into core region atoms and background region atoms.

[0117] The background influence removal unit is used to iteratively remove the influence of the background region atoms on the probability density distribution of the core region atoms, and obtain the potential energy equation between the atomic groups, or the potential energy equation between the atomic pairs.

[0118] Optionally, in an embodiment of the present application, the molecular force field establishment module includes:

[0119] The atomic cluster modeling submodule is used to model different types of atomic clusters using atoms in the target molecule as nodes and potential energy equations between atomic clusters as edges between nodes to obtain molecular force fields.

[0120] Optionally, in an embodiment of the present application, the target molecules include: biological molecules and drug-like molecules.

[0121] It should be understood that the device corresponds to the above-mentioned molecular force field modeling method embodiment and can perform each step involved in the above-mentioned method embodiment. The specific functions of the device can be found in the description above, and a detailed description is omitted here. The device includes at least one software function module that can be stored in a memory in the form of software or firmware or fixed in the operating system (OS) of the device.

[0122] Please refer to Figure 6 for a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. An electronic device 300 provided in an embodiment of the present application includes: a processor 310 and a memory 320, wherein the memory 320 stores machine-readable instructions executable by the processor 310, and when the machine-readable instructions are executed by the processor 310, the above method is performed.

[0123] The embodiment of the present application further provides a computer-readable storage medium 330 , on which a computer program is stored. When the computer program is run by the processor 310 , the above method is executed.

[0124] Among them, the computer-readable storage medium 330 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0125] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similarities between the various embodiments can be referred to in conjunction with each other. For device embodiments, since they are generally similar to method embodiments, their description is relatively simple, and for relevant details, reference can be made to the description of the method embodiments.

[0126] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are only schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible implementation architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and a module, a program segment or a part of a code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also be different from the order of occurrence marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which is mainly based on the functions involved.

[0127] In addition, the functional modules of each embodiment in the embodiments of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. In addition, in the description of this specification, the description of the reference terms "one embodiment", some embodiments, "example", specific example, "some examples", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in an appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples without contradiction.

[0128] The above description is only an optional implementation method of the embodiment of the present application, but the protection scope of the embodiment of the present application is not limited to this. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in the embodiment of the present application, and they should all be covered by the protection scope of the embodiment of the present application.

Claims

1. A molecular force field modeling method, characterized in that: include: Establishing a fingerprint spectrum of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule; Clustering the fingerprints of each atom in the target molecule to obtain multiple atomic categories; For each of the multiple atomic categories, modeling atomic clusters of different categories to obtain potential energy equations between the atomic clusters; The molecular force field is established according to the potential energy equations of the different types of atomic clusters.

2. The method according to claim 1, characterized in that The step of establishing a fingerprint spectrum of each atom in the target molecule according to the structural characteristic information and the energy characteristic information of the target molecule comprises: Obtaining the total charge distribution of all atoms in the target molecule; Projecting the total charge distribution of all atoms in the target molecule onto the Fibonacci sphere lattice constructed for each atom to simulate the charge density distribution on the surface of each atom; The charge density distribution on the surface of each atom of the target molecule is sorted by dimensionality reduction to obtain a fingerprint of each atom in the target molecule.

3. The method according to claim 2, characterized in that The obtaining the total charge distribution of all atoms in the target molecule comprises: Extracting structural characteristic information of the target molecule and energy characteristic information of the target molecule from the standard complex molecule file; The total amount of charge distribution of each atom in the target molecule is determined according to the structural characteristic information and the energy characteristic information.

4. The method according to claim 1, characterized in that: The clustering of the fingerprint spectrum of each atom in the target molecule comprises: The fingerprint spectrum of each atom in the target molecule is clustered according to the charge density distribution of the atom.

5. The method according to claim 1, characterized in that The molecular force field includes: potential energy equations between atomic clusters; the modeling of different types of atomic clusters includes: Modeling the atomic cluster to obtain a Boltzmann probability distribution between atoms in the atomic cluster; The potential energy equation between the atomic clusters is determined according to the Boltzmann probability distribution between the atoms of the atomic clusters.

6. The method according to claim 1, characterized in that The atomic cluster includes: an atomic pair consisting of two atoms, or an atomic group consisting of three atoms; the modeling of different types of atomic clusters includes: Dividing all atoms in the target molecule into core region atoms and background region atoms; The influence of the background region atoms on the probability density distribution of the core region atoms is removed in an iterative manner, so as to obtain a potential energy equation between the atom groups or a potential energy equation between the atom pairs.

7. The method according to claim 1, characterized in that The step of establishing the molecular force field according to the potential energy equations of the different types of atomic clusters comprises: The atoms in the target molecule are used as nodes, the potential energy equations between the atomic clusters are used as edges between the nodes, and the atomic clusters of different categories are modeled to obtain the molecular force field.

8. The method according to any one of claims 1 to 7, characterized in that: The target molecules include: biological molecules and drug-like molecules.

9. A molecular force field simulation device, characterized in that: include: A fingerprint spectrum establishment module is used to establish a fingerprint spectrum of each atom in the target molecule according to the structural characteristic information and energy characteristic information of the target molecule; An atomic category acquisition module, used for clustering the fingerprint spectrum of each atom in the target molecule to obtain multiple atomic categories; A potential energy equation obtaining module is used to model different types of atomic clusters for each type of atomic category in the multiple types of atomic categories, and obtain potential energy equations between the atomic clusters; The molecular force field establishment module is used to establish the molecular force field according to the potential energy equations of the atomic clusters of different categories.

10. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the machine-readable instructions are executed by the processor, the method according to any one of claims 1 to 8 is performed.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is executed.

Images