Physical and chemical knowledge guided multi-granularity molecular property prediction method and system

Abstract

The invention belongs to the technical field of molecular representation learning and molecular property prediction, and provides a multi-granularity molecular property prediction method and system guided by physical and chemical knowledge. Edge granularity (chemical bond), point granularity (atom), sub-graph granularity (functional group) and graph granularity (molecular graph) information are considered at the same time to comprehensively encode various hierarchical structures of molecules, molecular potential energy serves as physical information to be spread in the edge granularity and the point granularity, and the functional group serves as chemical information to be fully embedded in the sub-graph granularity. Finally, the outputs of the granularities will be summarized in the graph granularities, producing high quality and interpretable molecular characterization, and applying them to downstream molecular property prediction tasks.

Description

Technical Field

[0001] The present invention belongs to the technical field of molecular representation learning and molecular property prediction, and in particular to a multi-granularity molecular property prediction method and system guided by knowledge in the fields of physics and chemistry. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The purpose of molecular property prediction is to accurately predict the properties of small molecules in screening libraries to facilitate the selection of potential drug candidates and reduce the time and cost of drug development.

[0004] The inventors have found that in the current molecular property prediction technology, since this type of molecular data itself contains very rich physical and chemical information. From the perspective of molecular dynamics, physical information is crucial to ensure molecular stability. In addition, certain chemical information has a significant impact on molecular properties. However, the current molecular property prediction technology lacks consideration of this information, resulting in insufficient physical and chemical interpretability of the model. In addition, the prior art only focuses on encoding atoms within molecules, ignoring other granularities and various interaction forces within molecules. Moreover, the small-scale message transmission mechanism of the prior art cannot fully embed the local structural information of molecules, and there will be a problem of global structural information loss. The above problems will lead to poor prediction performance of molecules. Summary of the invention

[0005] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a multi-granularity molecular property prediction method and system guided by knowledge in the fields of physics and chemistry, which can be guided by physical and chemical knowledge, and simultaneously consider edge granularity (chemical bonds), point granularity (atoms), subgraph granularity (functional groups) and graph granularity (molecular graph) information to comprehensively encode various hierarchical structures of molecules. Among them, molecular potential energy is propagated as physical information at edge granularity and point granularity, and functional groups are fully embedded as chemical information at subgraph granularity. Finally, the outputs of the above granularities will be summarized at the graph granularity to produce high-quality and interpretable molecular representations, and they will be applied to downstream molecular property prediction tasks.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a multi-particle size molecular property prediction method guided by knowledge in the field of physics and chemistry, comprising the following steps:

[0008] Obtain multi-granularity information of the molecule to be predicted;

[0009] Combining the physical and chemical knowledge of the molecule to be predicted, according to the multi-granularity information of the molecule to be predicted and the constructed molecular property prediction model, the molecular representation is obtained for molecular property prediction;

[0010] The process of constructing the molecular property prediction model includes:

[0011] The chemical bonds in the molecule are regarded as the edges of the molecular graph, and the atoms in the molecule are regarded as the points in the molecular graph. The molecular potential energy is introduced. Based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, the relative importance of each hop neighbor of each edge or point in the molecular graph is calculated to update the embedded representation of the edges and points in the molecular graph.

[0012] Each functional group is regarded as a subgraph in the molecular graph and compressed into subgraph nodes. The 3D distance between subgraph nodes is used as a bias, and the relationship between subgraph nodes is considered to update the representation of subgraph nodes.

[0013] Molecular representation is obtained based on the embedded representation of edges and nodes in the molecular graph, the representation of subgraph nodes and the original molecular features.

[0014] Furthermore, the introduction of molecular potential energy, based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, calculates the relative importance of each hop neighbor of each edge or point in the molecular graph to them, and updates the embedded representation of the edges and points in the molecular graph, specifically including:

[0015] Calculate the bond lengths, bond angles, dihedral angles, and non-bonded distances of a molecule based on the coordinates of the atoms within the molecule;

[0016] The bond length, bond angle, dihedral angle and non-bond distance of the molecule are passed through the OPLS force field to embed the corresponding stretching energy, bending energy, torsion energy and non-bonded interaction energy;

[0017] Calculate the relative importance of each edge's 1-2 hop neighbors based on bending and torsional energies;

[0018] The relative importance of each point's 1-3 hop neighbors is calculated based on stretching energy, bending energy, torsion energy, and non-bonded interaction energy;

[0019] Update the embedding representation of edges and points in the molecular graph based on the relative importance of each edge's 1-2 hop neighbors and the relative importance of each point's 1-3 hop neighbors.

[0020] Further, the relative importance of 1-2 hops of adjacent edges of each edge is calculated based on the bending energy and the torsion energy, including:

[0021] Each edge aggregates information from the directly connected point and the 1-2 hop adjacent edges to update its own representation, where the embedded representation of the directly connected point is directly passed to the target edge, the bond angle bending energy between the 1-hop adjacent edge and the target edge is used as the message passing weight of the 1-hop adjacent edge, and the dihedral angle twisting energy between the 2-hop adjacent edge and the target edge is used as the message passing weight of the 2-hop adjacent edge.

[0022] Furthermore, the weight of a 1-hop neighbor edge is inversely proportional to the bond angle, and the weight of a 2-hop neighbor edge is inversely proportional to the dihedral angle. The larger the bond angle or dihedral angle, the higher the bending energy or torsion energy, and the lower the weight assigned by the target edge to the corresponding neighbor edge.

[0023] Furthermore, in the process of assigning neighbor edge weights, if they are both 1-hop neighbor edges, the neighbor edge with a larger bond angle is assigned a lower weight by the target edge; if they are both 2-hop neighbor edges, the neighbor edge with a larger dihedral angle is assigned a lower weight by the target edge.

[0024] Furthermore, the relative importance of 1-3 hop neighboring points of each point is calculated based on the stretching energy, bending energy, torsion energy and non-bonded interaction energy, including:

[0025] Each point aggregates information from directly connected edges and 1-3 hop neighboring points to update its own representation, where the embedded representation of directly connected edges is directly passed to the target point, the bond length stretching energy between 1-hop neighboring points and the target point is used as the aggregation weight of 1-hop neighboring points, the bond angle bending energy and non-bonded interaction energy between 2-hop neighboring points and the target point are used as the aggregation weight of 2-hop neighboring points, and the dihedral angle torsion energy and non-bonded interaction energy between 3-hop neighboring points and the target point are used as the aggregation weight of 3-hop neighboring points.

[0026] Furthermore, the weight of a 1-hop neighbor to the target point is inversely proportional to the bond length, the weight of a 2-hop neighbor to the target point is inversely proportional to the bond angle and the distance between points, and the weight of a 3-hop neighbor to the target point is inversely proportional to the dihedral angle and the distance between points.

[0027] The second aspect of the present invention provides a multi-granularity molecular property prediction system guided by knowledge in the fields of physics and chemistry, comprising:

[0028] An information acquisition module, which is used to obtain multi-granularity information of the molecule to be predicted;

[0029] The molecular property prediction module is used to combine the physical and chemical knowledge of the molecule to be predicted, obtain the molecular representation for molecular property prediction according to the multi-granularity information of the molecule to be predicted and the constructed molecular property prediction model;

[0030] The process of constructing the molecular property prediction model includes:

[0031] The chemical bonds in the molecule are regarded as the edges of the molecular graph, and the atoms in the molecule are regarded as the points in the molecular graph. The molecular potential energy is introduced. Based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, the relative importance of each hop neighbor of each edge or point in the molecular graph is calculated to update the embedded representation of the edges and points in the molecular graph.

[0032] Each functional group is regarded as a subgraph in the molecular graph and compressed into subgraph nodes. The 3D distance between subgraph nodes is used as a bias, and the relationship between subgraph nodes is considered to update the representation of subgraph nodes.

[0033] Molecular representation is obtained based on the embedded representation of edges and nodes in the molecular graph, the representation of subgraph nodes and the original molecular features.

[0034] A third aspect of the present invention provides a computer-readable storage medium.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the multi-particle-size molecular property prediction method guided by knowledge in the fields of physics and chemistry as described above.

[0036] A fourth aspect of the present invention provides a computer device.

[0037] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the multi-particle size molecular property prediction method guided by knowledge in the physical and chemical fields as described above are implemented.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention provides a multi-granularity molecular property prediction method and system guided by knowledge in the fields of physics and chemistry, which comprehensively and integratedly aggregates information from edges, points, subgraphs and graphs in a molecular graph, encodes them through different granularity modules and enables interaction, thereby realizing the comprehensive and high-quality molecular expression capability of the model; based on the important concept of "molecular potential energy" in physics and the important concept of "functional group" in chemistry, they are introduced into the message passing network and the self-attention mechanism respectively, which achieves superior performance while also complementing the local and global structural information within the molecule, thereby increasing the physical and chemical interpretability of the model prediction results.

[0040] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0042] Figure 1 is a flow chart of a multi-particle size molecular property prediction method guided by knowledge in the field of physics and chemistry according to an embodiment of the present invention;

[0043] Figure 2 is a molecular expression flow chart guided by physical domain knowledge in an embodiment of the present invention;

[0044] Figure 3 is a molecular expression flow chart guided by chemical field knowledge in an embodiment of the present invention;

[0045] Figure 4 It is a schematic diagram of the structure of a multi-particle size molecular property prediction system guided by knowledge in the fields of physics and chemistry according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0049] The present invention provides a multi-granularity molecular property prediction method and system guided by knowledge in the fields of physics and chemistry, wherein the physical information is molecular potential energy and the chemical information is functional groups. Multi-granularity includes: chemical bonds, atoms, functional groups and molecular graphs. First, it is necessary to obtain the characteristics of all chemical bonds and atoms of the molecule whose properties are to be predicted, the coordinates of all atoms, the characteristics of all functional groups, the coordinates of all functional groups and the characteristics of the molecules. Secondly, the characteristics of chemical bonds and atoms and the coordinates of atoms are input into the edge granularity module and the point granularity module, and the molecular potential energy is used as guiding information. The characteristics and coordinates of the functional groups are input into the subgraph granularity module. Finally, the molecular characteristics and the output results of the above modules are summarized in the graph granularity module to generate molecular representations and used for property prediction.

[0050] Embodiment 1

[0051] like Figure 1-2 As shown, this embodiment provides a multi-particle size molecular property prediction method guided by knowledge in the fields of physics and chemistry, comprising the following steps:

[0052] S101: Acquire multi-granularity information of the molecule to be predicted;

[0053] The multi-granularity information of the molecule to be predicted includes the characteristics of all chemical bonds and atoms, as well as the coordinates of all atoms, the characteristics and coordinates of all functional groups, and the characteristics of the molecule;

[0054] The process of obtaining the characteristics and coordinates of all functional groups is as follows:

[0055] S1011: pre-define SMART identifiers of all functional groups, and use them to retrieve all functional groups in the molecule in a subgraph matching manner;

[0056] S1012: For the characteristics of the functional groups, initialization is performed according to their chemical characteristics;

[0057] S1013: The coordinates of the functional group are calculated based on the average value of the coordinates of all atoms included in the functional group.

[0058] S102: Introducing physical and chemical knowledge as a guide, the chemical bonds within the molecule are regarded as the edges of the molecular graph, and the atoms within the molecule are regarded as points in the molecular graph. The chemical bonds are input into the edge granularity module, and the characteristics of the atoms and the coordinates of the atoms are input into the point granularity module. Secondly, the characteristics and coordinates of the functional groups are input into the subgraph granularity module. Finally, the output results of all the above modules and the characteristics of the molecules are summarized in the graph granularity module to form a molecular representation for property prediction.

[0059] like Figure 3 As shown in the figure, the potential energy term in the molecular potential energy is used to assist the calculation of the aggregation weight in the edge and point message passing process. The specific details are as follows:

[0060] First, bond lengths, bond angles, dihedral angles, and non-bonded distances are calculated based on the coordinates of the atoms.

[0061] Secondly, these geometric information are embedded in the OPLS force field to generate the corresponding stretching energy, bending energy, torsion energy and non-bonded interaction energy. The sum of these four energies constitutes the expression of molecular potential energy under the OPLS standpoint.

[0062] Finally, this energy information is used to assist the weight calculation in the message passing process of the edge granularity module and the point granularity module. The expression of molecular potential energy and the embedded expression of the four energies it contains are as follows:

[0063] E=E bond +E angle +Etorsion +E nonbond ,

[0064] E bond =[r,r 2 ],

[0065] E angle =[θ,θ 2 ],

[0066] E torsion =[cosφ, cos2φ, cos3φ, sinφ, sin2φ, sin3φ],

[0067] E nonbond =[r -1 ,r -12 ,r -6 ],

[0068] Among them, E bond represents the stretching energy in the molecular potential energy, E angle represents the bending energy in the molecular potential energy, E torsion represents the torsional energy in the molecular potential energy, E nonbond represents the non-bonded interaction energy in the molecular potential energy, r represents the bond length (interatomic distance), θ represents the bond angle, φ represents the dihedral angle, and [·] represents the connection operation.

[0069] The edge granularity module, which takes edge and vertex features as input, is used to update the edge embedding representation. Each edge in the molecular graph aggregates information from its neighboring vertex or edge.

[0070] Specifically, each edge aggregates information from the direct connection point and the 1-2 hop adjacent edges to update its own representation. The embedded representation of the direct connection point is directly transferred to the target edge, the bond angle bending energy between the 1-hop adjacent edge and the target edge is used as the message transfer weight of the 1-hop adjacent edge, and the dihedral angle twisting energy between the 2-hop adjacent edge and the target edge is used as the message transfer weight of the 2-hop adjacent edge.

[0071] The specific formula is:

[0072]

[0073]

[0074] Among them, [·] represents the connection operation, h uv Represents the characteristics of the edge, h u The features of the representative points, u represents the starting point of the edge, v represents the end point of the edge, and h v represents the feature of the point, l represents the depth of the network, m one_hop_edge Indicates the message summary of the target edge from the 1-hop adjacent edge, m two_hop_edgeIndicates the target edge’s message aggregation from 2-hop adjacent edges, h g Indicates the characteristics of the molecule, E angle is the mapping function from bond angle to bending energy, Ε dihedral It is the mapping function from dihedral angle to torsional energy. Angle() is the function for calculating bond angle, and Dihedral() is the function for calculating dihedral angle.

[0075] The point granularity module takes the features of points and edges as input to update the embedded representation of points. Each point in the molecular graph aggregates information from its adjacent edges or points. Specifically, each point aggregates information from directly connected edges and 1-3 hops of adjacent points to update its own representation. Among them, the embedded representation of directly connected edges is directly passed to the target point, the bond length stretching energy between 1-hop adjacent points and the target point is used as the aggregation weight of 1-hop adjacent points, the bond angle bending energy and non-bonded interaction energy between 2-hop adjacent points and the target point is used as the aggregation weight of 2-hop adjacent points, and the dihedral angle torsion energy and non-bonded interaction energy between 3-hop adjacent points and the target point is used as the aggregation weight of 3-hop adjacent points.

[0076] The specific formula is:

[0077]

[0078] Among them, m one_hop_node Indicates the message summary of the target point from the 1-hop neighboring point, m two_hop_node Indicates the message summary of the target point from the 2-hop neighboring points, m three_hop_node Indicates the message summary of the target point from the 3-hop neighboring points, Bond() is the function for calculating the bond length, Nonbond() is the function for calculating the non-bond distance (distance between atoms), E bond is the mapping function from bond length to stretching energy, E nonbond It is the mapping function of non-bonded distance → non-bonded interaction energy.

[0079] In the message passing process of the above point-granularity module, the embedding representation of the directly connected edge is passed to the target point as a direct message, and the embedding representation of the 1-3 hop neighboring points is passed to the target point as an indirect message. The molecular features also provide global guidance information.

[0080] like Figure 4 As shown, the functional groups within the molecule are retrieved and encoded, which improves the model performance while increasing the chemical interpretability of the molecular characterization and prediction results. The specific details include the following:

[0081] First, a SMART identifier is predefined for each functional group, based on which the functional groups present in the molecule are retrieved using a subgraph matching method. Then, each functional group is considered as a subgraph in the molecular graph and compressed into a subgraph node.

[0082] The subgraph granularity module takes the features and coordinates of the subgraph nodes as input to update the embedding representation of the subgraph. The embedding representation of each subgraph node updates its own embedding representation based on its relationship with all other subgraph nodes. Specifically, a new multi-head self-attention mechanism is defined, and the 3D distance between subgraph nodes is used as a bias. Based on this principle, the relationship between subgraph nodes is considered and the representation of subgraph nodes is updated.

[0083] The specific formula is:

[0084]

[0085] MultiHead(Q,K,V)=Concat(head 1 ,...,head h )W O ,

[0086] head k =Attention(x s Q k_i ,x s K k_j ,x s V k_j ),

[0087] Among them, Q k_i For query, K k_j is the index, V k_j For content, x s Represents the characteristics of the functional group, W Q , W K , W V and W O represents the learnable matrix, q ij represents the 3D distance between functional groups i and j, d represents the characteristic dimension of the functional group, head h represents the hth attention head.

[0088] By visualizing the weights in the edge-granular module message passing process, we analyze the relationship between the weights and the bond angles and dihedral angles. By reasoning about the relationship between the weights and angles in the point-granular module message passing process, we further determine the relationship between the message aggregation weights and energy in the message passing process of our model.

[0089] Specifically, for the edge granularity module, the relationship between the aggregate weight of all 1-hop neighbors of each edge and the bond angle, and the relationship between the aggregate weight of all 2-hop neighbors of each edge and the dihedral angle are visualized. The study found that the weight of 1-hop neighbors is inversely proportional to the bond angle, and the weight of 2-hop neighbors is inversely proportional to the dihedral angle. This shows that the larger the bond angle or dihedral angle, the higher the bending energy or torsion energy, and the lower the weight assigned to the corresponding neighbor by the target edge. This is in line with the principles of physics; the larger the angle, the higher the energy, and the more unstable the molecule. In the process of assigning the weight of the neighboring edges of the model, if they are both 1-hop neighbors, the neighboring edge with a large bond angle is assigned a lower weight by the target edge. The reason is that the angle is large and the energy is high, which is more unstable to the target edge than other 1-hop neighbors, so the weight assigned is lower. If they are both 2-hop neighbors, the neighboring edge with a large dihedral angle is assigned a lower weight by the target edge, and the reason is the same as that of the 1-hop neighbors.

[0090] Furthermore, based on the above conclusions, the weight distribution of each hop neighboring edge in the message passing process of the point granularity module is inferred; the weight from a 1-hop neighboring point to a target point is inversely proportional to the bond length, the weight from a 2-hop neighboring point to a target point is inversely proportional to the bond angle and the distance between points, and the weight from a 3-hop neighboring point to a target point is inversely proportional to the dihedral angle and the distance between points.

[0091] Embodiment 2

[0092] This embodiment provides a multi-granularity molecular property prediction system guided by knowledge in the fields of physics and chemistry, including:

[0093] An information acquisition module, which is used to obtain multi-granularity information of the molecule to be predicted;

[0094] The molecular property prediction module is used to combine the physical and chemical knowledge of the molecule to be predicted, obtain the molecular representation for molecular property prediction according to the multi-granularity information of the molecule to be predicted and the constructed molecular property prediction model;

[0095] The process of constructing the molecular property prediction model includes:

[0096] The chemical bonds in the molecule are regarded as the edges of the molecular graph, and the atoms in the molecule are regarded as the points in the molecular graph. The molecular potential energy is introduced. Based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, the relative importance of each hop neighbor of each edge or point in the molecular graph is calculated to update the embedded representation of the edges and points in the molecular graph.

[0097] Each functional group is regarded as a subgraph in the molecular graph and compressed into subgraph nodes. The 3D distance between subgraph nodes is used as a bias, and the relationship between subgraph nodes is considered to update the representation of subgraph nodes.

[0098] Molecular representation is obtained based on the embedded representation of edges and nodes in the molecular graph, the representation of subgraph nodes and the original molecular features.

[0099] Embodiment 3

[0100] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the multi-particle size molecular property prediction method guided by physical and chemical field knowledge as described above are implemented.

[0101] Embodiment 4

[0102] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the multi-particle size molecular property prediction method guided by knowledge in the physical and chemical fields as described above are implemented.

[0103] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0104] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0105] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0107] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Multi-particle molecular property prediction method guided by knowledge in physics and chemistry, It is characterized in that The steps include: Obtain multi-granularity information of the molecule to be predicted; Combining the physical and chemical knowledge of the molecule to be predicted, according to the multi-granularity information of the molecule to be predicted and the constructed molecular property prediction model, the molecular representation is obtained for molecular property prediction; The process of constructing the molecular property prediction model includes: The chemical bonds in the molecule are regarded as the edges of the molecular graph, and the atoms in the molecule are regarded as the points in the molecular graph. The molecular potential energy is introduced. Based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, the relative importance of each hop neighbor of each edge or point in the molecular graph is calculated to update the embedded representation of the edges and points in the molecular graph. Each functional group is regarded as a subgraph in the molecular graph and compressed into subgraph nodes. The 3D distance between subgraph nodes is used as a bias, and the relationship between subgraph nodes is considered to update the representation of subgraph nodes. Molecular representation is obtained based on the embedded representation of edges and nodes in the molecular graph, the representation of subgraph nodes and the original molecular features.

2. The multi-particle molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 1, It is characterized in that The introduction of molecular potential energy, based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, calculates the relative importance of each hop neighbor of each edge or point in the molecular graph to them, and updates the embedded representation of the edge and point in the molecular graph, specifically includes: Calculate the bond lengths, bond angles, dihedral angles, and non-bonded distances of a molecule based on the coordinates of the atoms within the molecule; The bond length, bond angle, dihedral angle and non-bond distance of the molecule are passed through the OPLS force field to embed the corresponding stretching energy, bending energy, torsion energy and non-bonded interaction energy; Calculate the relative importance of each edge's 1-2 hop neighbors based on bending and torsional energies; The relative importance of each point's 1-3 hop neighbors is calculated based on stretching energy, bending energy, torsion energy, and non-bonded interaction energy; Update the embedding representation of edges and points in the molecular graph based on the relative importance of each edge's 1-2 hop neighbors and the relative importance of each point's 1-3 hop neighbors.

3. The multi-particle molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 2, It is characterized in that The relative importance of the 1-2 hop neighboring edges of each edge is calculated based on the bending energy and the torsion energy, including: Each edge aggregates information from the directly connected point and the 1-2 hop adjacent edges to update its own representation, where the embedded representation of the directly connected point is directly passed to the target edge, the bond angle bending energy between the 1-hop adjacent edge and the target edge is used as the message passing weight of the 1-hop adjacent edge, and the dihedral angle twisting energy between the 2-hop adjacent edge and the target edge is used as the message passing weight of the 2-hop adjacent edge.

4. The multi-particle molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 3, It is characterized in that The weight of a 1-hop neighbor edge is inversely proportional to the bond angle, and the weight of a 2-hop neighbor edge is inversely proportional to the dihedral angle. The larger the bond angle or dihedral angle, the higher the bending energy or torsion energy, and the lower the weight assigned to the corresponding neighbor edge by the target edge.

5. The multi-particle molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 3, It is characterized in that In the process of assigning neighbor edge weights, if they are both 1-hop neighbor edges, the neighbor edge with a larger bond angle will be assigned a lower weight by the target edge; if they are both 2-hop neighbor edges, the neighbor edge with a larger dihedral angle will be assigned a lower weight by the target edge.

6. The multi-granularity molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 2, It is characterized in that The relative importance of 1-3 hops of neighbors of each point is calculated based on stretching energy, bending energy, torsion energy and non-bonded interaction energy, including: Each point aggregates information from directly connected edges and 1-3 hop neighboring points to update its own representation, where the embedded representation of directly connected edges is directly passed to the target point, the bond length stretching energy between 1-hop neighboring points and the target point is used as the aggregation weight of 1-hop neighboring points, the bond angle bending energy and non-bonded interaction energy between 2-hop neighboring points and the target point are used as the aggregation weight of 2-hop neighboring points, and the dihedral angle torsion energy and non-bonded interaction energy between 3-hop neighboring points and the target point are used as the aggregation weight of 3-hop neighboring points.

7. The multi-granularity molecular property prediction method guided by physical and chemical field knowledge as claimed in claim 6, It is characterized in that The weight of a 1-hop neighbor to the target point is inversely proportional to the bond length, the weight of a 2-hop neighbor to the target point is inversely proportional to the bond angle and the distance between points, and the weight of a 3-hop neighbor to the target point is inversely proportional to the dihedral angle and the distance between points.

8. Multi-granularity molecular property prediction system guided by knowledge in physics and chemistry, It is characterized in that include: An information acquisition module, which is used to obtain multi-granularity information of the molecule to be predicted; The molecular property prediction module is used to combine the physical and chemical knowledge of the molecule to be predicted, obtain the molecular representation for molecular property prediction according to the multi-granularity information of the molecule to be predicted and the constructed molecular property prediction model; The process of constructing the molecular property prediction model includes: The chemical bonds in the molecule are regarded as the edges of the molecular graph, and the atoms in the molecule are regarded as the points in the molecular graph. The molecular potential energy is introduced. Based on the energy of the interaction force corresponding to each potential energy term contained in the molecular potential energy, the relative importance of each hop neighbor of each edge or point in the molecular graph is calculated to update the embedded representation of the edges and points in the molecular graph. Each functional group is regarded as a subgraph in the molecular graph and compressed into subgraph nodes. The 3D distance between subgraph nodes is used as a bias, and the relationship between subgraph nodes is considered to update the representation of subgraph nodes. Molecular representation is obtained based on the embedded representation of edges and nodes in the molecular graph, the representation of subgraph nodes and the original molecular features.

9. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the steps in the multi-particle size molecular property prediction method guided by physical and chemical field knowledge as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the steps of the multi-particle size molecular property prediction method guided by physical and chemical field knowledge are implemented as described in any one of claims 1 to 7.

Images