Molecular attribute prediction

Abstract

According to the implementation of the invention, a scheme for molecular attribute prediction is provided. According to the scheme, based on at least one target atom in a molecule, an initial atom cluster which takes the at least one target atom as a center and has a specified radius is determined. And determining an adjustment strategy corresponding to the cross-cluster attribute based on the cross-cluster attribute of the cross-cluster atoms included in each initial atom cluster of the at least one target atom. And adjusting the cross-cluster atoms contained in the initial atom cluster of the at least one target atom based on the adjustment strategy to obtain a corrected atom cluster corresponding to the at least one target atom. And determining a target molecule attribute of the molecule based on the corrected atom cluster corresponding to the at least one target atom. According to the embodiment of the invention, the space is used as the dividing basis of the atom clusters, so that the method can be suitable for different objects and different environments.

Description

Background Technology

[0001] Molecular property prediction is crucial in materials science, energy applications, biotechnology, and drug research. However, some commonly used molecular property prediction methods often face problems such as high computational costs and insufficient adaptability when dealing with complex molecular structures. How to ensure computational efficiency while adapting to different molecular property prediction scenarios remains an urgent problem to be solved. Summary of the Invention

[0002] According to the implementation of this disclosure, a scheme for predicting molecular properties is proposed. In this scheme, based on at least one target atom in the molecule, an initial atomic cluster centered on the at least one target atom and having a specified radius is determined. Based on the cross-cluster properties of the cross-cluster atoms contained in each of the initial atomic clusters of the at least one target atom, an adjustment strategy corresponding to the cross-cluster properties is determined. Adjustments are performed on the cross-cluster atoms contained in the initial atomic clusters of the at least one target atom based on the adjustment strategy to obtain a modified atomic cluster corresponding to the at least one target atom. And based on the modified atomic clusters corresponding to the at least one target atom, the target molecular properties of the molecule are determined. According to embodiments of this disclosure, space is used as the basis for atomic cluster division, thus making it applicable to different objects and different environments. By selectively adjusting the cross-cluster atoms and generating modified atomic clusters, this processing method can effectively reduce boundary effects caused by virtual cutting due to radius division. Processing the cross-cluster atoms can provide more accurate local molecular information, thereby providing more accurate target molecular property prediction results.

[0003] This section is provided to simplify the presentation of the selection of objects, which will be further described in the detailed embodiments below. This section is not intended to identify key or principal features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter. Attached Figure Description

[0004] Figure 1 A block diagram of an example environment in which multiple implementations of this disclosure can be implemented is shown;

[0005] Figure 2 A flowchart illustrating a process for predicting molecular properties according to some implementations of this disclosure is shown;

[0006] Figure 3 A schematic diagram of an initial atomic cluster according to some implementations of this disclosure is shown;

[0007] Figure 4A One of the schematic diagrams of some cross-cluster atoms according to this disclosure is shown;

[0008] Figure 4B A second schematic diagram of some cross-cluster atoms according to this disclosure is shown;

[0009] Figure 4C This is shown as a third schematic diagram of some cross-cluster atoms according to this disclosure; and

[0010] Figure 5 A schematic block diagram of an electronic device capable of implementing various implementations of the present disclosure is shown. Detailed Implementation

[0011] This disclosure will now be discussed with reference to several example implementations. It should be understood that these implementations are discussed only to enable those skilled in the art to better understand and thus implement this disclosure, and not to imply any limitation on the scope of this disclosure.

[0012] As used herein, the term "comprising" and its variations are to be interpreted as open-ended terms meaning "including but not limited to". The term "based on" is to be interpreted as "at least partially based on". The terms "an implementation" and "an implementation" are to be interpreted as "at least one implementation". The term "another implementation" is to be interpreted as "at least one other implementation". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0013] It should be noted that the headings of any section / subsection provided herein are not restrictive. Various implementations are described throughout this document, and any type of implementation may be included under any section / subsection. Furthermore, an implementation described in any section / subsection may be combined in any way with any other implementation described in the same section / subsection and / or different sections / subsections.

[0014] In this document, unless explicitly stated otherwise, performing a step in response to A does not mean that the step is performed immediately after A, but may include one or more intermediate steps.

[0015] As used herein, a set of elements, group of elements, or similar expression may include zero or one such element. The set of elements may be ordered or unordered. For example, "a set of separators" may include zero or one separators. As used in the text, a sequence of elements or similar expression may include one or more such elements, and the elements in the sequence are ordered.

[0016] As used in this paper, the term "model" refers to a model that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning (DL) is a machine learning algorithm that processes inputs and provides corresponding outputs using multiple layers of processing units. A neural network model is an example of a deep learning-based model. In this paper, "model" may also be referred to as a "machine learning model," "learning model," "machine learning network," or "learning network," and these terms are used interchangeably.

[0017] Machine learning typically comprises three phases: training, testing, and inference. In the training phase, a given model is trained using a large amount of training data, iterating until the model can consistently generate inferences that meet the expected goals from the training data. Through training, the model can be considered to have learned the relationship between inputs and outputs (also known as the input-output mapping) from the training data. The parameter values ​​of the trained model are determined. In the testing phase, test inputs are applied to the trained model to test whether it can provide the correct output, thus determining the model's performance. In the inference phase, the model can be used to process actual inputs based on the trained parameter values ​​to determine the corresponding output.

[0018] Example environment and basic principles

[0019] Figure 1 A schematic diagram of an example environment 100 in which an implementation of this disclosure can be carried out is shown. (See diagram for example.) Figure 1 As shown, environment 100 includes electronic device 110. It is desirable to use such electronic device 110 to perform molecular property prediction. To this end, in some implementations, an atomic cluster partitioning module 130 and a machine learning model 150 can be deployed in electronic device 110 for molecular modeling and property prediction. The purpose of molecular modeling is to improve the efficiency of molecular property prediction while ensuring its accuracy.

[0020] like Figure 1As shown, electronic device 110 can take information related to the molecular structure of molecule 102 as input and output the result 120 of property prediction. In some implementations of this disclosure, any suitable molecular property can be predicted. Examples of molecular properties include, but are not limited to, the total energy corresponding to electronic and nuclear energies, the distribution of electrons in the molecule, molecular geometry, spectral properties, solvent effects, etc. Determining molecular properties will help in various subsequent applications. Taking the role of molecular properties in drug analysis as an example, by calculating the total energy of reactants, intermediates, and products, the metabolic pathways and reaction mechanisms of drugs in vivo can be understood. By analyzing the distribution of electrons in the molecule, the electron density distribution can be determined, which helps to understand the electrostatic interactions between the drug and the target (such as a protein) and predict binding sites and binding affinity. Information such as bond lengths and bond angles in the molecular geometry can be used to characterize drug molecules and confirm whether the synthesized products of the molecules meet expectations. By analyzing spectral effects, impurities or byproducts in drugs can be detected to ensure drug purity. By analyzing solvent effects, the behavior of drugs in water or other physiological solvents can be simulated to help predict their stability and efficacy in vivo, etc.

[0021] Molecules 102 typically include multiple atoms, and the spatial arrangement of these atoms constitutes the molecular structure. Molecular structure can influence molecular properties. Therefore, to predict the target molecular property 120 of molecule 102, molecular structure modeling is required. In some implementations of this disclosure, molecular structure modeling can be based on an atomic cluster partitioning module 130. Exemplarily, molecular structure modeling can include performing clustering on the atoms in molecule 102 to obtain multiple atomic clusters. After identifying the atomic clusters within the molecule, the relevant information of each atomic cluster can be used as input information, and calculations regarding forces and energy can be performed based on a machine learning model 150, ultimately yielding different target molecular properties based on the calculation results.

[0022] exist Figure 1 In this context, electronic device 110 can be any system with computing capabilities, such as various computing devices / systems, terminal devices, servers, etc. Terminal devices can be any type of mobile terminal, fixed terminal, or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. Servers include, but are not limited to, mainframes, edge computing nodes, computing devices in cloud environments, etc.

[0023] It should be understood that Figure 1The components and arrangements shown in the environment are merely examples, and a computing system suitable for implementing the implementations described in this disclosure may include one or more different components, other components, and / or different arrangements.

[0024] It should be understood that the structure and function of the various elements in environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of this disclosure.

[0025] As mentioned above, the molecular structure can first be modeled, and then molecular property predictions can be performed based on the modeling results. Taking molecular modeling as an example, some schemes use pairwise distances between atoms within a molecule as relative position encoding. However, molecular structures are usually complex, and the information provided by relative distances between atoms is very limited. Therefore, representations like relative position encoding are often insufficient to capture the complex interactions within molecules. Furthermore, for protein molecules, some schemes use virtual segmentation of the protein's polypeptide chain to obtain multiple protein fragments. However, since proteins usually exist in folded peptide form, the above virtual segmentation methods only consider the protein's sequence on the polypeptide chain and fail to account for spatial relationships. This leads to a decrease in the accuracy of protein molecular property predictions based on multiple protein fragments. In addition, the above methods are only applicable to protein molecules, and cannot perform predictions for other molecules in the solution environment in which the protein resides.

[0026] According to the implementation of this disclosure, a scheme for molecular modeling and molecular property prediction based on atomic spatial positional relationships is proposed. In this scheme, based on at least one target atom in the molecule, an initial atomic cluster centered on the at least one target atom and having a specified radius is determined. Based on the cross-cluster properties of cross-cluster atoms contained in each of the initial atomic clusters of the at least one target atom, an adjustment strategy corresponding to the cross-cluster properties is determined. Adjustments are performed on the cross-cluster atoms contained in the initial atomic clusters of the at least one target atom based on the adjustment strategy to obtain a modified atomic cluster corresponding to the at least one target atom. Based on the modified atomic clusters corresponding to the at least one target atom, the target molecular properties of the molecule are determined. According to the embodiments of this disclosure, the spatial positional relationships between atoms are considered in molecular modeling. Furthermore, for cross-cluster atoms appearing during the construction of atomic clusters, adjustments can be performed on the atomic clusters in a corresponding manner based on the atomic properties, so that the adjusted atomic clusters can effectively reduce the influence of boundary effects. Finally, the above processing method can be adapted to different scenarios and environments, such as performing molecular property prediction on proteins, macromolecular materials, and complex biological systems.

[0027] An example implementation of this disclosure is described below with reference to the accompanying drawings.

[0028] To perform molecular modeling, it is necessary to describe the environment in which the molecule to be predicted exists; hereinafter, the molecule to be predicted will be referred to simply as a molecule. Taking a protein molecule as an example, in real biological environments, proteins typically exist in complex solution environments, such as within the human body. Proteins are usually surrounded by water molecules, ions, and other small molecules. Taking a drug molecule as an example, drug molecules are typically organic compounds that can function in different biological environments, such as within cell membranes, in blood, or at target protein binding sites. The interactions of drug molecules with water molecules, lipid molecules, or other macromolecules can affect their efficacy.

[0029] Each atom in a molecule, its position within the molecule, and the chemical bonds surrounding it constitute its atomic properties. These properties include, at a minimum, the atoms it is connected to, the type of chemical bond (single, double, etc.), and the local electron density. Furthermore, atoms interact not only with their immediate neighbors but also with distant atoms, influenced by forces such as van der Waals and Coulomb forces. These long-range interactions play a crucial role in calculating the overall energy and forces of molecules. Accurate calculations of molecular energy and forces are particularly critical in predicting properties involving macromolecules. However, due to the complexity and scale of macromolecules (such as proteins), such calculations become extremely challenging.

[0030] Figure 2 A flowchart of a method 200 for molecular modeling according to some implementations of this disclosure is shown. Method 200 can be implemented in... Figure 1 There are 110 electronic devices. This will be discussed in conjunction with... Figure 1 The environment 100 is used to describe method 200.

[0031] In box 201, electronic device 110 determines an initial cluster of atoms centered on at least one target atom and having a specified radius, based on at least one target atom in molecule 102.

[0032] Generally speaking, a molecule contains multiple atoms. Taking proteins as an example, proteins are composed of many amino acids, each containing an average of 10-20 atoms. Protein molecules typically consist of tens to thousands of amino acids, so a single protein molecule may contain hundreds to tens of thousands of atoms.

[0033] For the multiple atoms in molecule 102, at least one target atom can be identified first. The specific process for identifying the target atom will be described later. For example, the target atom can be identified based on the atomic mass of each atom in molecule 102.

[0034] For each target atom within molecule 102, it is necessary to identify the atom cluster centered on the target atom. Therefore, the processing procedure for each target atom is the same or similar. Figure 3A schematic diagram of some implementations of the initial atom cluster determination principle 300 according to this disclosure is shown. For each target atom, a virtual range is drawn based on a specified radius. All atoms within the virtual range form an initial atom cluster centered on the target atom. As an example, Figure 3 The specified radius shown is It should be understood that in practical application scenarios, the specified radius can be flexibly adjusted based on requirements.

[0035] For a relatively small given radius, the advantage is low computational complexity, making it suitable for handling very large systems. That is, for a relatively small given radius, each cluster contains fewer atoms, thus requiring relatively less computation. Since the complexity of quantum mechanical calculations increases exponentially with the number of atoms, a relatively small given radius can reduce computational complexity.

[0036] For a relatively large specified radius, the advantage is that it can capture local interactions more comprehensively, improving computational accuracy. That is, for a relatively large specified radius, each atomic cluster contains more atoms. Therefore, it can describe the interactions between atoms in more detail. Because it considers intramolecular and intermolecular local interactions more comprehensively, the calculation results are closer to reality, thus improving computational accuracy.

[0037] In box 202, electronic device 110 determines an adjustment strategy corresponding to the cross-cluster properties based on the cross-cluster properties of cross-cluster atoms contained in each initial atom cluster of at least one target atom.

[0038] Typically, each molecule 102 contains multiple target atoms, and each target atom corresponds to an initial atomic cluster. Taking one target atom as an example, the process of adjusting the atomic clusters is explained. The initial atomic clusters determined with a specified radius centered on the target atom usually involve virtual cuts to the atoms. That is, atoms located at the boundaries of the initial atomic clusters are usually virtually cut off. For a virtually cut atom, part of it is in one initial atomic cluster, and the other part is in another. Therefore, a virtually cut atom can be called a cross-cluster atom.

[0039] Cross-cluster attributes indicate the cross-cluster status of atoms. For example, the cross-cluster attribute of atoms can be functional group cross-cluster, such as double bond cross-cluster or aromatic ring cross-cluster. It can also be single bond cross-cluster, and so on. For each cross-cluster attribute, there exists a corresponding adjustment strategy to adaptively adjust the atoms in the initial cluster. The purpose of the adjustment is to minimize changes to the chemical environment. For example, if the adjustment strategies are categorized, they can include redistribution adjustment strategies and recombination adjustment strategies. The specific execution process of the adjustment strategies will be discussed in detail later.

[0040] In box 203, electronic device 110 performs adjustments on cross-cluster atoms contained in the initial atomic cluster of at least one target atom based on an adjustment strategy, to obtain a modified atomic cluster corresponding to at least one target atom.

[0041] Once the adjustment strategy is determined, adjustments can be performed on cross-cluster atoms in the initial atomic cluster based on the adjustment strategy to obtain the adjusted modified atomic cluster. For example, the adjustment strategy may include a reallocation adjustment strategy and a recombination adjustment strategy.

[0042] A redistribution adjustment strategy can refer to the redistribution of inter-cluster atoms within the initial clusters without disrupting their original structure. For example, if the inter-cluster atom is an aromatic ring inter-cluster, meaning one part of the aromatic ring is assigned to the first initial cluster and the other part to the second initial cluster, the first and second initial clusters can be adjacent clusters within molecule 102, partially overlapping clusters, etc. A redistribution adjustment strategy could then involve completely redistributing the aromatic ring to both the first and second initial clusters.

[0043] For the recombination adjustment strategy, a specified atom can be added at the cross-cluster position corresponding to the cross-cluster atom. The addition of a specified atom is to maintain the chemical integrity of the atomic cluster and minimize the impact on the original initial atomic cluster. Therefore, hydrogen atoms can be added at the cross-cluster position to pair the free electrons. Alternatively, the recombination adjustment strategy can also be based on the actual situation, adding halogen atoms, etc., at the virtual broken single bonds. For example, the processes in boxes 201 to 203 can be executed by the atomic cluster partitioning module 130. The process in box 204 can be executed by the machine learning model 150.

[0044] In box 204, electronic device 110 determines the target molecular properties of the molecule based on the modified atom cluster corresponding to at least one target atom.

[0045] Based on the modified atomic cluster corresponding to at least one target atom within molecule 102, the target molecular property 120 of molecule 102 can be determined. For example, using the three-dimensional positional information of the target atom in the modified atomic cluster, quantum mechanical calculations can yield results such as electronic structure, molecular energy, reaction pathways, and molecular vibrational frequencies. These calculation results can further predict the target molecular properties of molecule 102. For instance, the electronic structure of the molecule can help understand reactive sites, charge transfer processes, and intermolecular interactions. Molecular energy, involving ground-state and excited-state energies, is crucial for assessing molecular stability and reactivity. Reaction pathways can help analyze energy changes between reactants and products, thereby predicting reaction kinetics. Molecular vibrational frequencies can interpret and predict experimentally observed infrared and Raman spectra, thus aiding in the analysis of molecular vibrational modes and the strength of chemical bonds.

[0046] The above scheme, using space as the basis for cluster division, is applicable to different objects and environments. It is applicable to proteins in folded peptide form as well as drug molecules in compound form. It is applicable to the complex environment within the human body, such as the environment of water molecules, ions, and other small molecules surrounding proteins, and the environment of water molecules, lipid molecules, or other large molecules surrounding drug molecules. Furthermore, by specifically adjusting cross-cluster atoms and generating modified clusters, this approach effectively reduces boundary effects caused by virtual cutting performed through radius division. Boundary effects refer to the properties of atoms near the cluster boundary that differ from those of other atoms within it. By adjusting the strategy, boundary effects can be eliminated or mitigated, thus providing more accurate target molecule property results. Finally, this treatment of cross-cluster atoms provides more accurate local molecular information, resulting in more precise predictions of target molecule properties.

[0047] In some implementations of this disclosure, when determining the adjustment strategy corresponding to the cross-cluster attribute, for a target atom in at least one target atom, the electronic device 110 can determine the cross-cluster attribute corresponding to the cross-cluster atom contained in the initial atomic cluster of the target atom, and determine the adjustment strategy based on the cross-cluster attribute.

[0048] In some implementations, cross-cluster attributes can be categorized into first cross-cluster attributes and second cross-cluster attributes. First cross-cluster attributes indicate at least one of two cross-cluster types: cross-clusters where the cross-cluster atom is a functional group or a cross-cluster where the cross-cluster atom is a bonding atom. Second cross-cluster attributes indicate single-bond cross-clusters. The first cross-cluster attribute will be described in detail below.

[0049] First, we will introduce the case where the cross-cluster atom is a functional group crossing the cluster. Figure 4AA schematic diagram of some implementations of functional group cross-cluster 400A according to this disclosure is shown. Figure 4A The functional group shown is phenol. Figure 4A The dashed line 401 can be used to represent the local boundary of the initial atomic cluster. Figure 4A In the example, the local boundary of the initial atomic cluster precisely bisects the phenol, making the phenol a cross-cluster atom. Since the cross-cluster atom is a functional group (phenol) cross-cluster, the cross-cluster property of the cross-cluster atom can be the first cross-cluster property.

[0050] Figure 4B A schematic diagram of some implementations of functional group cross-cluster 400B according to the present disclosure is shown. Figure 4B The dashed line 402 can be used to represent the local boundary of the initial atomic cluster. Figure 4B In the example, the local boundary of the initial atomic cluster precisely cuts the double bond of the atom, making the atom a cross-cluster atom. Since the cross-cluster atom is a functional group (double bond) cross-cluster, the cross-cluster property of the cross-cluster atom can be the first cross-cluster property.

[0051] Figure 4A and Figure 4B In the example, the functional group corresponds to phenol and a double bond. In real-world scenarios, the functional group can also correspond to triple bonds, aromatic rings, oxygen-containing functional groups such as hydroxyl or carbonyl groups, nitrogen-containing functional groups such as amino and amide groups, and so on.

[0052] Figure 4C A schematic diagram is shown of designated atomic segments across cluster 400C that realize bonded atoms according to this disclosure. Figure 4C The dashed line 403 in the diagram can be used to represent the local boundary of the initial atomic cluster. Figure 4C In the example, the local boundary of the initial atomic cluster precisely virtually cuts the bonded atom. A fragment 403D of this bonded atom is separated outside the boundary, and this fragment is simultaneously bonded to three atoms (atoms 403A, 403B, and 403C) inside the boundary, making this atom a cross-cluster atom. The cross-cluster property of this cross-cluster atom is that of a specified atomic fragment of the bonded atom cross-cluster; therefore, the cross-cluster property of a cross-cluster atom can be the first cross-cluster property. For a specified atomic fragment of a bonded atom to cross-cluster, it can also be simply referred to as a bonded atom cross-cluster.

[0053] In some implementations, if it is determined that the cross-cluster atoms contained in the initial atomic cluster of a target atom have a first cross-cluster attribute, then the electronic device 110 can determine that the adjustment strategy is a reallocation adjustment strategy if the target atom is the first cross-cluster attribute. That is, for a cross-cluster attribute that is the first cross-cluster attribute, an adjustment strategy corresponding to the first cross-cluster attribute can be used to adjust the cross-cluster atoms in the initial atomic cluster. For example, the reallocation adjustment strategy redistributes the affiliation of cross-cluster atoms rather than changing their structure.

[0054] Through the above process, the redistribution adjustment strategy determined based on the first cross-cluster attribute can adjust the atom assignment of cross-cluster atoms without changing their structure. Therefore, functional groups (such as double bonds and aromatic rings) possess specific chemical properties and reactivity within a molecule; separating them would alter their chemical characteristics. The redistribution strategy, by containing the entire functional group within a single cluster, ensures that its chemical properties remain unaffected. Furthermore, for bonded atoms crossing clusters—that is, bonds connecting two or more atoms within a cluster—splitting them and replacing them with designated atoms such as hydrogen atoms could lead to inaccurate structures. For example, if... Figure 4C Process 423, shown in the diagram, involves using hydrogen atoms 433 to fill in the virtually broken bonds, resulting in two or more hydrogen atoms spatially close to or overlapping each other. Due to the size of the hydrogen atoms and the repulsive effect of their electron clouds, steric hindrance occurs. The redistribution adjustment strategy is as follows... Figure 4C In process 413, these bonded atoms are redistributed into a cluster. This preserves the chemical integrity of the molecule. By avoiding unrealistic structural changes, the interactions and positions of each atom are ensured to be more accurate.

[0055] In some implementations of this disclosure, to determine whether a cross-cluster atom is a functional group cross-cluster, electronic device 110 can determine whether the cross-cluster atom includes a functional group and whether different functional group segments within the functional group cross-cluster. If electronic device 110 determines that the cross-cluster atom includes a functional group, and the first functional group segment of the functional group is outside the initial atomic cluster, and the second functional group segment of the functional group is inside the initial atomic cluster, then the cross-cluster atom can be determined to be a functional group cross-cluster corresponding to the first cross-cluster attribute.

[0056] Still combined Figure 4A and Figure 4B The situation shown. Figure 4A The case shown illustrates a double bond crossing a cluster within a functional group, meaning one end of the double bond is outside the initial atomic cluster, while the other end is inside the initial atomic cluster. In this case, the electronic device 110 can determine that the crossing atom is a functional group crossing a cluster corresponding to the first crossing cluster attribute. Figure 4B The case shown is that of phenol cross-cluster in a functional group, where one segment of phenol is outside the initial atomic cluster and the other segment is inside the initial atomic cluster. In this case, the electronic device 110 can also determine that the cross-cluster atom is a functional group cross-cluster corresponding to the first cross-cluster attribute. Furthermore, functional group cross-clusters also include triple bond cross-clusters, aromatic ring cross-clusters, and so on.

[0057] The main significance of identifying functional groups across clusters lies in determining and preserving the integrity of key reaction sites in molecules. Functional groups are crucial structures that determine the chemical reactivity of molecules. By identifying and employing appropriate strategies to handle functional groups across clusters, we can ensure that the functional properties of molecules are accurately preserved during molecular modeling and the determination of target molecule properties. For example, functional groups determine the chemical reactivity of molecules; accurately identifying and handling these structures ensures that these important reactive properties are not lost during molecular modeling. Furthermore, the identification of functional groups across clusters is applicable to various complex molecular structures, enhancing the applicability of molecular modeling techniques in fields such as materials science and drug design.

[0058] In some implementations of this disclosure, if the cross-cluster attribute of the target atom is determined to be a functional group cross-cluster attribute, the adjustment performed by the electronic device 110 on the cross-cluster atoms in the initial atomic cluster includes, based on a redistribution adjustment strategy, allocating the cross-cluster atoms corresponding to the functional group cross-cluster to the initial atomic cluster to obtain the modified atomic cluster corresponding to the target atom.

[0059] For cross-clusters with different functional group attributes, such as double bonds, triple bonds, aromatic rings, oxygen-containing functional groups, and nitrogen-containing functional groups, the corresponding adjustment strategy is a redistribution adjustment strategy. Taking the cross-cluster atom corresponding to the functional group cross-cluster as spanning two initial atomic clusters as an example, the first functional group atom fragment of the cross-cluster atom is assigned to the first initial atomic cluster, while the second functional group atom fragment of the cross-cluster atom is assigned to the second initial atomic cluster. In this case, the cross-cluster atom corresponding to the functional group cross-cluster is completely redistributed to the first initial atomic cluster and the second initial atomic cluster, respectively, resulting in the first modified atomic cluster and the second modified atomic cluster.

[0060] The first modified atomic cluster serves as a modified atomic cluster of the first initial atomic cluster. The second modified atomic cluster serves as a modified atomic cluster of the second initial atomic cluster. That is, if the trans-cluster atom corresponding to the functional group is an aromatic ring, then both the first and second modified atomic clusters contain the complete aromatic ring.

[0061] Through the above process, adjusting the initial atomic clusters using a redistribution strategy for cross-cluster functional groups ensures that the corrected atomic clusters fully contain the functional groups, avoiding changes in the chemical environment caused by virtual truncation, thereby reducing errors in subsequent molecular property analysis. Furthermore, the redistribution strategy is applicable to various complex molecular structures, ensuring accurate simulation results in different molecular environments.

[0062] In some implementations of this disclosure, to determine whether a cross-cluster atom is a designated atomic segment of a bonding atom across a cluster, electronic device 110 can determine whether the cross-cluster atom is a bonding atom and whether a specific cross-cluster condition is satisfied. If electronic device 110 determines that the first atomic segment of the bonding atom is outside the initial atomic cluster and that the first atomic segment is bonded to at least two atoms within the initial atomic cluster, then the specific cross-cluster condition can be determined to be satisfied. In this case, the first atomic segment corresponds to the designated atomic segment.

[0063] If the initial atomic cluster excludes one atomic segment (the first atomic segment) of a bonded atom from the cluster, and this atomic segment is bonded to at least two atoms within the initial atomic cluster, then the cross-cluster property of this bonded atom can be determined as the first cross-cluster property. The first atomic segment can contain only one atom or can be an atomic segment composed of multiple bondsed atoms.

[0064] In the case of a specified atomic fragment of a bonded atom spanning multiple clusters, an atomic fragment outside the cluster is simultaneously bonded to at least two atoms within the cluster. Considering the complexity of this multiple bonding relationship, which is not present in other types of cross-cluster cases, its cross-cluster property can be defined as the first cross-cluster property. Therefore, in subsequent processing, a redistribution strategy can be used to ensure the integrity and correctness of the cross-cluster atomic structure.

[0065] In some implementations of this disclosure, for a first cross-cluster attribute including a specified atomic fragment of a bonded atom, the electronic device 110 performs adjustments on the cross-cluster atoms in the initial atomic cluster, including, based on a redistribution adjustment strategy, assigning the specified atomic fragment of the bonded atom to the initial atomic cluster to obtain the modified atomic cluster corresponding to the target atom.

[0066] As previously mentioned, a designated atomic segment of a bonded atom crosses a cluster and simultaneously bonds to at least two atoms within the cluster. Based on these multiple bonding relationships, simply adding a designated atom, such as a hydrogen atom, to the cross-cluster position would result in multiple hydrogen atoms being added to the same external atomic segment. In this case, adding hydrogen atoms would disrupt the correct structure of the atomic cluster, affecting the physical and chemical properties of the molecule, thus causing steric hindrance conflicts, and consequently impacting subsequent calculations and predictions.

[0067] Based on this, for cases where a specific atomic fragment of a bonding atom crosses a cluster, the first atomic fragment can be reassigned to the initial cluster, resulting in a modified cluster. This redistribution strategy does not cause structural changes to the cross-cluster atoms, thus ensuring the integrity and correctness of the cross-cluster atomic structure. Therefore, accurate atomic and electron distributions are crucial in the quantum mechanical calculations corresponding to the determination of target molecule properties. The redistribution of cross-cluster atoms can more accurately reflect the actual bonding relationships and electron distribution of the molecule, thereby improving the accuracy of the calculation results.

[0068] The previous section detailed the example and processing procedure for determining the first type of cross-cluster property. The second type of cross-cluster property will now be discussed in detail. As mentioned earlier, the cross-cluster atoms contained in the initial atomic cluster may also include the second type of cross-cluster property, namely, the single-bond cross-cluster property. The second type of cross-cluster property typically indicates single-bond cross-clustering, meaning one atom is inside the initial atomic cluster and another atom is outside, connected by a single bond. If this virtually severed single bond is not addressed, it will cause the initial atomic cluster to generate virtual free radicals. These virtual free radicals have high chemical reactivity and may cause unexpected chemical reactions during quantum computing, leading to inaccurate calculation results.

[0069] To achieve a more complete and stable molecular structure and reduce the degrees of freedom and uncertainties that need to be handled in the calculations, a recombination adjustment strategy needs to be applied to the second cross-cluster property. This strategy can restore bond losses caused by single bond breakage, ensuring that each atom meets its bonding requirements and maintaining the chemical integrity and stability of the molecule.

[0070] In some implementations of this disclosure, for at least one target atom, if it is determined that the cross-cluster atom contained in the initial atomic cluster of the target atom is a second cross-cluster attribute, the electronic device 110 determines that the adjustment strategy for the initial atomic cluster is a recombination adjustment strategy. In some implementations of this disclosure, the process of the electronic device 110 executing the recombination adjustment strategy may include supplementing a specified atom at the cross-cluster position corresponding to the single-bond cross-cluster in the initial atomic cluster.

[0071] For single-bond cross-clusters corresponding to the second cross-cluster attribute, a recombination adjustment strategy can be employed to supplement specified atoms at the cross-cluster positions corresponding to the single-bond cross-clusters. When supplementing specified atoms, the orientation of the virtually severed single bonds is first obtained. That is, the angles and relative positions of the single bonds between atoms within the initially severed atomic cluster and those outside the initial atomic cluster are obtained. When supplementing specified atoms, the process maintains consistency with the original molecular structure or ensures that the error is within a specified error range. This is to preserve the original geometric structure and chemical properties of the molecule as much as possible.

[0072] By adding specified atoms, the direction of single bonds can be maintained while the bond length can be adjusted according to specific circumstances to ensure that the newly formed structure meets the physical and chemical requirements of the molecule. This method ensures that the directionality of the original structure is preserved while adjusting bond lengths during cross-cluster processing, resulting in a reasonable new structure that meets the requirements of chemical and physical stability. This improves the accuracy of calculations and ensures that the generated modified atomic clusters conform to the physicochemical properties of the actual molecule.

[0073] In some implementations of this disclosure, after the cluster partitioning module 130 in the electronic device 110 determines the modified clusters, the machine learning model 150 can obtain the three-dimensional position information of the target atoms in each modified cluster. Finally, based on the three-dimensional position information of the target atoms and the specified atoms, the machine learning model 150 can use a molecular target attribute prediction model to determine the predicted results of the target molecular attributes of the molecule.

[0074] For modified atomic clusters, quantum mechanical (QM) calculations can be performed on the interior of the modified atomic cluster, while mechanical and / or electronic embedding can be combined for the exterior. Molecular target property prediction models are based on modified atomic clusters, performing calculations on forces and energy, thereby yielding different target molecular properties based on the calculation results.

[0075] The 3D position information of the target atom in each modified atom cluster, as well as the 3D position information of the specified atom, can be used as input to the molecular target attribute prediction model. Here, the subsequent process is illustrated using hydrogen atoms as an example of the specified atom.

[0076] Equations (1) to (5) can be used to calculate the force. Based on molecular mechanics, the total energy of each modified atomic cluster can be expressed as follows:

[0077]

[0078] E in formula (1) QM (frag i E can be used to represent the quantum mechanical energy of the i-th modified atomic cluster. The quantum mechanical energy of the i-th modified atomic cluster includes the following components: short-range kinetic energy: that is, the kinetic energy of the electron's movement in the local environment (modified atomic cluster). Electron exchange interaction: the exchange energy due to the indistinguishability of fermions. Correlation interaction: the dynamic correlation between electrons due to the Coulomb interaction. long-range (out i ) can be used to represent the energy of the long-range interaction experienced by the i-th modified atom cluster, which corresponds to the long-range interaction effect exerted by the external environment on the modified atom cluster.

[0079] The force acting on the target atom within the modified atom cluster is calculated using the gradient of energy with respect to the target atom's position information. That is, the force acting on the target atom within the modified atom cluster can be expressed as follows:

[0080]

[0081] In formula (2) It can be used to represent the gradient calculation result of the three-dimensional position information of the target atom in the i-th modified formula (2) positive atom cluster. The underlying logic of formula (2) is that force is the rate of change of energy with respect to position.

[0082] The calculation of the energy of long-range interactions involved in Equation (1) can be combined with molecular mechanics (E MM (or electron intercalation) methods. The molecular mechanics representation is as follows:

[0083]

[0084] In formula (3) It can be used to represent van der Waals force, ε ij It can be used to represent the potential energy σ between target atom i (within the corrected atom cluster) and target atom j (outside the corrected atom cluster). ij It can be used to represent the effective distance parameter r between target atom i and target atom j. ij It can be used to represent the distance between target atom i and target atom j. It can be used to represent Coulomb interactions. q i and q j εi and εj can be used to represent the charges of target atom i and target atom j, respectively, and ε0 can be used to represent the electrical constant in vacuum. Van der Waals forces describe the attractive and repulsive forces caused by interatomic interactions. Coulomb forces describe the electrostatic interactions between charged atoms. Therefore, equation (3) represents the long-range interaction energy between an atom (target atom i) within the modified atom cluster and an atom (target atom j) outside the modified atom cluster. Target atom i and target atom j can be atoms within the modified atom cluster.

[0085] The method of electronic embedding is represented as follows:

[0086]

[0087] Formula (4) calculates the electrostatic potential at position R. Formula (4) consists of two parts: one part is the potential generated by the electron density ρ(r), and the other part is the potential generated by the nuclear charge Zm. |Rr| can represent the distance between position R and position r. The integral term can represent the electrostatic potential generated at position R due to the electron distribution ρ(r). The summation term can represent the electrostatic potential generated at position R by M target atoms, Zm can represent the charge of the m-th target atom, and Rm can represent the position of the m-th target atom.

[0088] The total energy corresponding to the electron intercalation method is expressed as follows:

[0089]

[0090] Formula (5) takes into account the electrostatic interactions of all target atoms, that is, the total energy E of the interaction between the l-th target atom and all other target atoms. emb ql can represent the charge of the l-th target atom. For example, the total number of target atoms can be L.

[0091] The above formulas integrate short-range interactions, long-range interactions, and polarization effects. This integration method enables accurate calculations of forces. For example, short-range interactions are reflected in the quantum mechanical energy E of each modified atomic cluster. QM (frag i In the calculation of ), long-range interactions are manifested in the van der Waals forces, Coulomb forces, and polarization effects exerted on each modified atom cluster by other external modified atom clusters. The polarization effect is described by the electrostatic field calculation results, that is, by the integral term and summation term in formula (4).

[0092] For energy calculations, the derivative of force with respect to displacement can be utilized. By integrating the force along the displacement, the change in relative energy can be determined. Equations (6) and (7) correspond to the energy calculation process.

[0093]

[0094] E in formula (6) cur E can represent the energy that corrects the current displacement of an atomic cluster. pre This can represent the energy that corrects a previously determined cluster of atoms, x. cur It can represent the position of the current displacement of the corrected atomic cluster, x pre It can represent the position of the previous displacement of the atom cluster, and F is the force on the atom cluster that corrects the displacement.

[0095] Since integrals are difficult to solve directly in practical calculations, numerical methods can be used for approximation. For example, the trapezoidal rule is a commonly used method in numerical integration. The trapezoidal rule is expressed as follows:

[0096]

[0097] In formula (7) F pre F can represent the force that corrects the previous displacement of the atomic cluster. cur It can represent the force that corrects the current displacement of the atomic cluster, x cur and x pre These can still represent the current displacement of the atomic cluster and the previous displacement of the atomic cluster, respectively. In other words, although it's impossible to integrate the exact functional form of the force F in numerical calculations, discrete force values ​​F can be used. pre and F cur To perform an approximate continuous integral, we can obtain the energy change between adjacent displacements.

[0098] The underlying logic of the above calculation is based on the complex positional changes and possible overlapping relationships between the modified atomic clusters within the molecule. Therefore, instead of directly calculating the energy, the energy is solved using the derivative of force with respect to displacement.

[0099] For the recombination and adjustment strategy, the atoms to be added are as small and simple as possible to avoid introducing excessive computational complexity in quantum mechanical calculations, while preserving the original local chemical environment as much as possible. In some implementations of this disclosure, the designated atom may include a hydrogen atom.

[0100] Depending on the type of molecular system and the required computational precision, other atoms or other types of structures can be selected for addition in other implementations. For example, in some organic molecules, methyl groups can be added to saturate dangling bonds. Furthermore, depending on the chemical environment of the molecule and the chemical properties to be preserved, other small groups such as fluorine and chlorine can be added.

[0101] Designating a hydrogen atom as the primary atom offers several advantages over adding methyl groups or other small groups. Firstly, hydrogen atoms are simple, easy to manipulate, and in most cases, effectively saturate broken single bonds. Secondly, adding hydrogen atoms helps to preserve the original chemical environment of the system as much as possible.

[0102] In some implementations of this disclosure, the electronic device 110 determines a target atom in a molecule based on the atomic characteristics of a set of atoms in the molecule, identifying at least one target atom from the set of atoms, wherein the atomic characteristics of the atom indicate at least one of the atom's atomic number or atomic mass.

[0103] For a set of atoms in a molecule, at least one target atom can be identified first. The target atom can be determined based on at least one of the atomic numbers and atomic masses of the atoms in the molecule. For example, the atomic number is the position of an element in the periodic table, representing the number of protons in an atom's nucleus. A larger atomic number generally means that the atom has more protons and neutrons. Therefore, atoms with atomic numbers not less than a specified number can be identified as target atoms. As another example, atomic mass, usually expressed in atomic mass units, is the total mass of protons and neutrons in an atom. Therefore, atoms with atomic masses not less than a specified atomic mass threshold can also be identified as target atoms.

[0104] The significance of identifying the target atom lies in its ability to focus on determining molecular properties. It allows for calculations based on the local region surrounding the target atom, reducing overall computational complexity and time, and improving efficiency. Furthermore, selecting the target atom and modifying its surrounding environment (modifying the atomic cluster) preserves the local structure and chemical environment of the molecule, ensuring the accuracy of the calculation results. Finally, since the target atom is often the location of reaction centers or functional groups, precise calculations of these regions can better predict molecular properties such as reactivity and mechanisms.

[0105] It should be understood that the architecture of the model described above is merely exemplary and is not intended to impose any limitations.

[0106] Example device

[0107] Figure 5 A schematic block diagram of an electronic device capable of implementing various implementations of this disclosure is shown. It should be understood that... Figure 5 The electronic device 500 shown is merely exemplary and should not be construed as limiting the functionality and scope of the implementation described in this disclosure.

[0108] like Figure 5 As shown, electronic device 500 includes electronic device 500 in the form of general-purpose computing device. Components of electronic device 500 may include, but are not limited to, one or more processors or processing devices 510, memory 520, storage device 530, one or more communication units 540, one or more input devices 550, and one or more output devices 560.

[0109] In some implementations of this disclosure, the electronic device 500 can be implemented as a computing device, computing system, server, mainframe, or other device with computing capabilities.

[0110] Processing device 510 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 520. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 500. Processing device 510 may include a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor, a controller, and / or a microcontroller, etc.

[0111] Electronic device 500 typically includes multiple computer storage media. Such media can be any available media accessible to electronic device 500, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 520 may include volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 530 may include removable or non-removable media and may include computer-readable media, such as memory, flash drives, disks, or any other media capable of storing information and / or data and accessible within electronic device 500.

[0112] Electronic device 500 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 5 As shown, disk drives for reading from or writing to removable, non-volatile disks and optical disc drives for reading from or writing to removable, non-volatile optical discs can be provided. In these cases, each drive can be connected to a bus (not shown) via one or more data media interfaces.

[0113] The communication unit 540 enables communication with other computing devices via a communication medium. Additionally, the functionality of the components of the electronic device 500 can be implemented as a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, the electronic device 500 can operate in a networked environment using logical connections to one or more other servers, personal computers (PCs), or other general network nodes.

[0114] Input device 550 can be one or more various input devices, such as a mouse, keyboard, data import device, etc. Output device 560 can be one or more output devices, such as a monitor, data export device, etc. Electronic device 500 can also communicate with one or more external devices (not shown) via communication unit 540 as needed. External devices include storage devices, display devices, etc., and can communicate with one or more devices that enable user interaction with electronic device 500, or with any device that enables electronic device 500 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication can be performed via input / output (I / O) interface (not shown).

[0115] In some implementations of this disclosure, in addition to being integrated into a single device, some or all of the various components of electronic device 500 may be configured in the form of a cloud computing architecture. In a cloud computing architecture, these components can be remotely deployed and can work together to achieve the functions described in this disclosure. In some implementations, cloud computing provides computing, software, data access, and storage services without requiring end users to know the physical location or configuration of the systems or hardware providing these services. In various implementations, cloud computing provides services over a wide area network (such as the Internet) using appropriate protocols. For example, cloud computing providers offer applications over a wide area network, and these applications can be accessed through a web browser or any other computing component. The software or components of the cloud computing architecture, along with the corresponding data, may be stored on servers at remote locations. Computing resources in a cloud computing environment may be consolidated at remote data center locations or they may be distributed. Cloud computing infrastructure can provide services through shared data centers, even if they appear as a single access point for users. Therefore, the components and functions described herein can be provided from service providers at remote locations using a cloud computing architecture. Alternatively, they may be provided from conventional servers, or they may be installed directly or otherwise on client devices.

[0116] Electronic device 500 can be used to implement molecular modeling in various implementations of this disclosure. Memory 520 may include one or more modules having one or more program instructions that can be accessed and executed by processing unit 510 to implement the functionality of the various implementations described herein. For example, memory 520 may include molecular modeling module 522 for performing molecular modeling in one or more of the above implementations. Figure 5 As shown, the electronic device 500 can acquire the input required for molecular modeling through the input device 550 and provide the output of molecular modeling, such as the determined target molecular properties, through the output device 560. In some implementations, the electronic device 500 can also receive input from other devices (not shown) via the communication unit 540.

[0117] Example implementation

[0118] The following are some example implementations of this disclosure.

[0119] In one aspect, this disclosure provides an electronic device. The electronic device includes: a processing unit; and

[0120] A memory, coupled to a processing unit and containing instructions stored thereon, which, when executed by the processing unit, cause the device to perform the following actions: Based on at least one target atom in the molecule, determine an initial atomic cluster centered on the at least one target atom and having a specified radius; based on the cross-cluster properties of cross-cluster atoms contained in each of the initial atomic clusters of the at least one target atom, determine an adjustment strategy corresponding to the cross-cluster properties; perform adjustments on the cross-cluster atoms contained in the initial atomic clusters of the at least one target atom based on the adjustment strategy, to obtain a modified atomic cluster corresponding to the at least one target atom; and determine a target molecular property of the molecule based on the modified atomic cluster corresponding to the at least one target atom.

[0121] In some implementations of this disclosure, determining the adjustment strategy corresponding to the cross-cluster attribute may include: for a target atom in at least one target atom, in response to the cross-cluster atom contained in the initial atomic cluster of the target atom being a first cross-cluster attribute, determining the adjustment strategy as a reallocation adjustment strategy, wherein the first cross-cluster attribute indicates at least one of functional group cross-cluster or a specified atomic fragment of a bonded atom cross-cluster.

[0122] In some implementations of this disclosure, the action may further include: in response to the cross-cluster atom including a functional group, and the first functional group segment of the functional group being outside the initial atomic cluster and the second functional group segment of the functional group being inside the initial atomic cluster, then determining the cross-cluster atom as the functional group cross-cluster corresponding to the first cross-cluster attribute.

[0123] In some implementations of this disclosure, the first cross-cluster attribute may include functional group cross-cluster, and adjusting the cross-cluster atoms contained in the initial atomic cluster of at least one target atom based on the adjustment strategy includes allocating the cross-cluster atoms corresponding to the functional group cross-cluster to the initial atomic cluster based on the redistribution adjustment strategy to obtain the modified atomic cluster.

[0124] In some implementations of this disclosure, the action may further include: in response to the cross-cluster atom including a bonding atom, the first atomic fragment of the bonding atom being outside the initial atomic cluster, and the first atomic fragment being bonded to at least two atoms within the initial atomic cluster, then determining the cross-cluster atom as a specified atomic fragment cross-cluster of the bonding atom corresponding to the first cross-cluster attribute, and the first atomic fragment corresponding to the specified atomic fragment.

[0125] In some implementations of this disclosure, the first cross-cluster attribute includes a specified atomic fragment of the bonded atom across the cluster, and adjusting the cross-cluster atoms contained in the initial atomic cluster of at least one target atom based on an adjustment strategy may include assigning the specified atomic fragment of the bonded atom to the initial atomic cluster based on a redistribution adjustment strategy to obtain a modified atomic cluster.

[0126] In some implementations of this disclosure, determining the adjustment strategy corresponding to the cross-cluster attribute includes: for a target atom in at least one target atom, in response to the cross-cluster atom contained in the initial atomic cluster of the target atom being a second cross-cluster attribute, determining the adjustment strategy as a recombination adjustment strategy, wherein the second cross-cluster attribute indicates single-bond cross-cluster.

[0127] In some implementations of this disclosure, adjusting cross-cluster atoms in the initial atomic cluster of at least one target atom based on an adjustment strategy may include supplementing specified atoms at the cross-cluster positions corresponding to single-bond cross-clusters in the initial atomic cluster based on a recombination adjustment strategy.

[0128] In some implementations of this disclosure, determining the target molecular properties of a molecule may include: obtaining the three-dimensional position information of target atoms in each modified atom cluster; obtaining the three-dimensional position information of specified atoms in each modified atom cluster; and determining the target molecular properties of the molecule using a molecular target property prediction model based on the three-dimensional position information of the target atoms and the specified atoms.

[0129] In some implementations of this disclosure, the specified atom may include a hydrogen atom.

[0130] In some implementations of this disclosure, the target atom in the molecule may be determined by identifying at least one target atom from the set of atoms based on the atomic characteristics of the set of atoms in the molecule, wherein the atomic characteristics of the atom indicate at least one of the atomic number or atomic mass of the atom.

[0131] In another aspect, this disclosure provides a computer-readable medium having stored computer-executable instructions thereon, which, when executed by a device, cause the device to perform one or more example implementations of the methods described above.

[0132] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, without limitation, example types of hardware logic components that can be used include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Load Programmable Logic Devices (CPLDs), and so on.

[0133] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0134] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0135] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of a single implementation may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0136] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A computer-implemented method comprising: determining, based on at least one target atom in a molecule, an initial atomic cluster centered at the at least one target atom and having a specified radius; determining, based on a cross-cluster property of a cross-cluster atom included in each initial atomic cluster of the at least one target atom, an adjustment policy corresponding to the cross-cluster property; performing, based on the adjustment policy, an adjustment on the cross-cluster atom included in the initial atomic cluster of the at least one target atom, to obtain a revised atomic cluster corresponding to the at least one target atom; and determining, based on the revised atomic cluster corresponding to the at least one target atom, a target molecular property of the molecule.

2. The method of claim 1, wherein determining the adjustment policy corresponding to the cross-cluster property comprises: for a target atom of the at least one target atom, in response to a cross-cluster atom included in the initial atomic cluster of the target atom being a first cross-cluster property, determining the adjustment policy to be a redistribution adjustment policy, the first cross-cluster property indicating at least one of a functional group cross-cluster or a specified atomic fragment of a bonded atom cross-cluster.

3. The method of claim 2, further comprising: in response to the cross-cluster atom including a functional group, and a first functional group fragment of the functional group being outside the initial atomic cluster, and a second functional group fragment of the functional group being inside the initial atomic cluster, determining the cross-cluster atom to be the functional group cross-cluster corresponding to the first cross-cluster property.

4. The method of claim 2 or 3, wherein the first cross-cluster property comprises a functional group cross-cluster, and performing, based on the adjustment policy, the adjustment on the cross-cluster atom included in the initial atomic cluster of the at least one target atom comprises: based on the redistribution adjustment policy, distributing the cross-cluster atom corresponding to the functional group cross-cluster into the initial atomic cluster, to obtain the revised atomic cluster.

5. The method of claim 2, further comprising: in response to the cross-cluster atom including a bonded atom, a first atomic fragment of the bonded atom being outside the initial atomic cluster, and the first atomic fragment being bonded to at least two atoms inside the initial atomic cluster, determining the cross-cluster atom to be the specified atomic fragment of the bonded atom cross-cluster corresponding to the first cross-cluster property, and the first atomic fragment corresponding the specified atomic fragment of the bonded atom.

6. The method of claim 2 or 5, wherein the first cross-cluster property comprises a specified atomic fragment of a bonded atom cross-cluster, and performing, based on the adjustment policy, the adjustment on the cross-cluster atom included in the initial atomic cluster of the at least one target atom comprises: based on the redistribution adjustment policy, distributing the specified atomic fragment of the bonded atom into the initial atomic cluster, to obtain the revised atomic cluster.

7. The method of claim 1, wherein determining the adjustment policy corresponding to the cross-cluster property comprises: for a target atom of the at least one target atom, in response to a cross-cluster atom included in the initial atomic cluster of the target atom being a second cross-cluster property, determining the adjustment policy to be a recombination adjustment policy, the second cross-cluster property indicating a single bond cross-cluster. ​ 8. The method of claim 7, wherein performing adjustment on the cross-cluster atoms in the initial atomic cluster of the at least one target atom based on the adjustment policy comprises: supplementing a cross-cluster position corresponding to a single-bond cross-cluster in the initial atomic cluster with a designated atom based on the recombination adjustment policy.

9. The method of claim 1, wherein determining the target molecular property of the molecule comprises: obtaining three-dimensional position information of a target atom in each of the revised atomic clusters; obtaining three-dimensional position information of a designated atom in each of the revised atomic clusters; and determining the target molecular property of the molecule based on the three-dimensional position information of the target atom and the three-dimensional position information of the designated atom using a molecular target property prediction model.

10. The method of claim 8 or 9, wherein the designated atom comprises a hydrogen atom.

11. The method of claim 1, wherein the target atom in the molecule is determined by: determining at least one target atom from a set of atoms in the molecule based on atomic features of the atoms in the set, the atomic feature of an atom indicating at least one of an atomic number or an atomic mass of the atom.

12. An electronic device, comprising: a processing unit; and a memory coupled to the processing unit and containing instructions stored thereon that, when executed by the processing unit, cause the device to perform actions comprising: determining an initial atomic cluster centered at at least one target atom in a molecule and having a designated radius based on the at least one target atom; determining an adjustment policy corresponding to a cross-cluster property of a cross-cluster atom included in a respective initial atomic cluster of the at least one target atom based on the cross-cluster property; performing adjustment on the cross-cluster atom included in the initial atomic cluster of the at least one target atom based on the adjustment policy to obtain a revised atomic cluster corresponding to the at least one target atom; and determining a target molecular property of the molecule based on the revised atomic cluster corresponding to the at least one target atom.

13. The device of claim 12, wherein determining the adjustment policy corresponding to the cross-cluster property comprises: for a target atom of the at least one target atom, in response to a cross-cluster atom included in the initial atomic cluster of the target atom being a first cross-cluster property, determining the adjustment policy to be a redistribution adjustment policy, the first cross-cluster property indicating at least one of a functional group cross-cluster or a designated atom fragment of a bonded atom cross-cluster.

14. The device of claim 13, the actions further comprising: in response to the cross-cluster atom comprising a functional group, and a first functional group fragment of the functional group being outside the initial atomic cluster, a second functional group fragment of the functional group being inside the initial atomic cluster, determining the cross-cluster atom to be a functional group cross-cluster corresponding to the first cross-cluster property.

15. The device of claim 13 or 14, wherein the first cross-cluster property comprises a functional group cross-cluster, and performing adjustment on the cross-cluster atom included in the initial atomic cluster of the at least one target atom based on the adjustment policy comprises: ​ ​ based on the redistribution adjustment policy, assigning the functional group to a cross-cluster atom corresponding to the cross-cluster pair into the initial atomic cluster to obtain the modified atomic cluster.

16. The device of claim 13, wherein the acts further comprise: in response to the cross-cluster atom including a bonded atom, a first atomic fragment of the bonded atom being outside the initial atomic cluster, and the first atomic fragment being bonded to at least two atoms within the initial atomic cluster, determining the cross-cluster atom as a designated atomic fragment cross-cluster of the bonded atom corresponding to the first cross-cluster property, and the first atomic fragment corresponding the designated atomic fragment of the bonded atom.

17. The device of claim 13 or 16, wherein the first cross-cluster property includes a designated atomic fragment cross-cluster of a bonded atom, and performing adjustment on the cross-cluster atom contained in the initial atomic cluster of the at least one target atom based on the adjustment policy comprises: based on the redistribution adjustment policy, assigning the designated atomic fragment of the bonded atom to the initial atomic cluster to obtain the modified atomic cluster.

18. The device of claim 12, wherein determining an adjustment policy corresponding to the cross-cluster property comprises: for a target atom of the at least one target atom, in response to a cross-cluster atom contained in the initial atomic cluster of the target atom being a second cross-cluster property, determining the adjustment policy as a reorganization adjustment policy, the second cross-cluster property indicating a single bond cross-cluster.

19. The device of claim 18, wherein performing adjustment on the cross-cluster atom in the initial atomic cluster of the at least one target atom based on the adjustment policy comprises: based on the reorganization adjustment policy, supplementing a designated atom at a cross-cluster position corresponding to the single bond cross-cluster in the initial atomic cluster.

20. A computer program product tangibly stored in a computer storage medium and comprising computer executable instructions that, when executed by a device, cause the device to perform acts comprising: based on at least one target atom in a molecule, determining an initial atomic cluster centered at the at least one target atom and having a designated radius; based on a cross-cluster property of a cross-cluster atom contained in each initial atomic cluster of the at least one target atom, determining an adjustment policy corresponding to the cross-cluster property; performing adjustment on the cross-cluster atom contained in the initial atomic cluster of the at least one target atom based on the adjustment policy to obtain a modified atomic cluster corresponding to the at least one target atom; and based on the modified atomic cluster corresponding to the at least one target atom, determining a target molecular property of the molecule.

21. A device comprising: a memory that stores computer executable instructions; and a processor, wherein the processor, when executing the computer executable instructions, causes the device to perform acts comprising: based on at least one target atom in a molecule, determining an initial atomic cluster centered at the at least one target atom and having a designated radius; based on a cross-cluster property of a cross-cluster atom contained in each initial atomic cluster of the at least one target atom, determining an adjustment policy corresponding to the cross-cluster property; performing adjustment on the cross-cluster atom contained in the initial atomic cluster of the at least one target atom based on the adjustment policy to obtain a modified atomic cluster corresponding to the at least one target atom; and based on the modified atomic cluster corresponding to the at least one target atom, determining a target molecular property of the molecule.

Images