Molecular property prediction

Abstract

According to implementations of the present disclosure, a solution for molecular property prediction is provided. In the solution, an initial atom cluster centered on the at least one target atom and with a specified radius is determined based on at least one target atom in a molecule. An adjustment strategy corresponding to the cross-cluster property is determined based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom. Adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom is performed, based on the adjustment strategy, to obtain a modified atom cluster corresponding to the at least one target atom. A target molecular property of the molecule is determined based on the modified atom cluster corresponding to the at least one target atom.

Description

MOLECULAR PROPERTY PREDICTIONBACKGROUND

[0001] Molecular property prediction is critical in material science, energy application, biotechnology and drug research. However, some typical molecular property prediction methods often face problems such as high computational costs and insufficient scene adaptability when dealing with complex molecular structures. How to adapt to different scenarios of predicting molecular property while ensuring the computational efficiency remains an urgent problem to be solved.SUMMARY

[0002] According to implementations of the present disclosure, a solution for molecular property prediction is provided. In the solution, an initial atom cluster centered on the at least one target atom and with a specified radius is determined based on at least one target atom in a molecule. An adjustment strategy corresponding to the cross-cluster property is determined based on a crosscluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom. Adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom is performed, based on the adjustment strategy, to obtain a modified atom cluster corresponding to the at least one target atom. A target molecular property of the molecule is determined based on the modified atom cluster corresponding to the at least one target atom. According to the embodiments of the present disclosure, space is used as a basis for dividing atom clusters, thereby being applicable to different objects and different environments. By performing targeted adjustments on the cross-cluster atoms and generating a modified atom cluster, this processing method may effectively reduce the boundary effect caused by performing virtual cutting due to radius division. The processing for the cross-cluster atoms may provide more accurate molecular local information, thereby enabling to provide more precise prediction results of target molecular property.

[0003] This section is provided in a simplified form to introduce a selection of objects that are further described below in the Detailed Description. This section is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter.BRIEF DESCRIPTION OF THE DRAWINGS[0004JFIG. 1 illustrates a block diagram of an example environment in which various implementations of the present disclosure can be implemented;

[0005] FIG. 2 illustrates a flow diagram of a process of molecular property prediction according to some implementations of the present disclosure;[0006JFIG. 3 illustrates a schematic diagram of an initial atom cluster according to some implementations of the present disclosure;[0007JFIG. 4A illustrates the first one of schematic diagrams of some cross-cluster atoms according to the present disclosure;

[0008] FIG. 4B illustrates the second one of schematic diagrams of some cross-cluster atoms according to the present disclosure;[0009JFIG. 4C illustrates the third one of schematic diagrams of some cross-cluster atoms according to the present disclosure; and[0010JFIG. 5 illustrates a schematic block diagram of an electronic device capable of performing various implementations of the present disclosure.DETAILED DESCRIPTION[OOllJThe present disclosure will now be discussed with reference to a number of example implementations. It should be understood that these implementations are discussed only to enable one of ordinary skills in the art to better understand and thus implement the present disclosure, and not to imply any limitation on the scope of the present disclosure.

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

[0013] It should be noted that the headline of any section / subsection provided in the specification is not limiting. Various implementations are described throughout the specification, and any type of implementation may be included in any section / subsection. Furthermore, the implementations described in any section / subsection may be combined in any manner with any other implementations described in the same section / subsection and / or different sections / subsections.

[0014] Unless explicitly stated herein, 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, element set, or the similar expressions may include zero, one or more such elements. The set of elements may be ordered or disordered. For example, “a set of separators” may include zero, one or more separators. As used in the text, an element sequence or similar expressions may include one or more such elements, and the elements in the sequence are ordered.

[0016] As used herein, the term “model” may leam. from training data, associations between respective inputs and outputs, so that a corresponding output may be generated for a given inputafter training is completed. The generation of the model may be based on a machine learning technology. Depth learning (DL) is a machine learning algorithm that processes inputs and provides corresponding outputs by using multiple layers of processing units. A neural networks model is one example of a deep learning-based model. As used herein, “model'’ may also be referred to as “machine learning model.” “learning model.” “machine learning network.” or “learning network,” and these terms may be used interchangeably herein.

[0017] Generally, machine learning may roughly include three stages: ataining stage, testing stage, and usage stage (also referred to as an inference stage). In the training stage, a given model may be trained using a large amount of training data, constantly iterating until the model is able to obtain, from the training data, consistent inferences that meet expected goals. Through training, the model may be considered to be able to leam, from training data, associations (also referred to as mappings from inputs to outputs) between inputs to outputs. In the testing stage, a test input is applied to the trained model to test whether the model can provide a correct output, thereby determining the performance of the model. In the inference stage, the model can process, based on the parameter values obtained from training, actual inputs to determine corresponding outputs. Example Environment and Basic Principles

[0018] FIG. 1 illustrates a schematic diagram of an example environment 100 in which various implementations of the present disclosure can be implemented. As shown in FIG. 1. the environment 100 includes an electronic device 110 which is desired to implement molecular property prediction. To this end, in some implementations, an atom cluster division module 130 and machine learning model 150 may be deployed in the electronic device 110 for molecule modeling and property prediction. The purpose of molecule modeling is to improve the efficiency of molecular property prediction while ensuring the accuracy of molecular property prediction.

[0019] As shown in FIG. 1 , the electronic device 110 may take information related to the molecular structure of molecule 102 as input and output a property7prediction result 120. In some implementations of the present disclosure, any suitable molecular property can be predicted. Examples of molecular properties include, but are not limited to. the total energy corresponding to electron energy and nuclear energy, the distribution of electrons in the molecule, molecular geometric structure, spectral properties, solvent effects, etc. The determination of the molecular properties may contribute to various subsequent applications. Taking a role of the molecular property in drug analysis as an example, the metabolic pathways and reaction mechanisms of drugs in the body may be understand by computing the total energy of reactants, intermediates, and products. The electron density7distribution may be determined through the distribution of the electrons in the molecule, thereby contributing to understanding the electrostatic interaction between the drug and the target (e.g., proteins) and predicting binding sites and binding affinity.Information such as bond lengths and bond angles in the molecular geometric structure may be used to characterize the drug molecule and confirm whether a synthetic product of the molecule meets expectations. With the analysis of spectral effects, impurities or by-products in drugs may be detected to ensure the purity of the drugs. With the analysis of solvent effects, the behavior of the drug in water or other physiological solvents may be simulated, helping to predict the stability and effectiveness of the drug in the body, and the like.

[0020] The molecule 102 may generally include a plurality of atoms, and the spatial arrangement of these atoms forms a molecular structure. The molecular structure, in turn, may affect molecular property. Thus, it is necessary to model the molecular structure in order to predict the target molecular property 120 of the molecule 102. In some implementations of the present disclosure, the molecular structure may be modeled based on the atom cluster division module 130. For example, modeling the molecular structure may include performing a cluster division process on the atoms in the molecule 102 to obtain a plurality of atom clusters. After determining the atom clusters within the molecule, the relevant information of each atom cluster may be used as input information to perform calculations on force and energy based on the machine learning model 150, and finally obtain different target molecular properties based on the computational result.

[0021] In FIG. 1, the electronic device 110 may be any system with computing capability, such as a variety of computing devices / sy stems, terminal devices, servers, etc. The terminal device maybe any type of mobile terminal, fixed terminal, or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, 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, and the like.

[0022] It should be appreciated that the components and arrangements shown in FIG. 1 are merely examples. A computing system suitable for performing the implementations described herein may include one or more different components, other components, and / or different arrangements.

[0023] It should be understood that the structure and function of the various elements in the environment 100 are described for example purposes only and are not intended to imply any limitation on the scope of the present disclosure.

[0024] As mentioned above, the molecular structure may be first modeled, and then prediction of molecular property may be performed based on the modeling results. Taking molecule modeling as an example, some approaches employ pairwise distances betw een atoms within a molecule as relative position encoding. However, the molecular structure is generally complex, and the information that can be provided by the relative distance between atoms is limited. Thus, representations such as relative position encoding is generally insufficient to capture complexinteractions within the molecule. In addition, for protein molecules, some approaches may perform virtual segmentation on polypeptide chains of proteins in accordance with a specified length to obtain multiple protein fragments. However, since proteins typically exist in the form of folded peptides, this leads to the described virtual segmentation method only considering the sequence of proteins along polypeptide chains and failing to consider spatial position relationships. As a result, the accuracy of protein molecular property prediction, which is performed based on multiple protein fragments, is reduced. In addition, the described method is only suitable for protein molecules, and the prediction cannot be performed on other molecules in the solution environment where the protein is located.

[0025] According to the implementations of the present disclosure, a solution for performing molecule modeling based on atomic spatial position relationships and predicting a molecular property7is provided. In the solution, an initial atom cluster centered on the at least one target atom and with a specified radius is determined based on at least one target atom in a molecule. An adjustment strategy corresponding to the cross-cluster property' is determined based on a crosscluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom. Adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom is performed, based on the adjustment strategy, to obtain a modified atom cluster corresponding to the at least one target atom. A target molecular property of the molecule is determined based on the modified atom cluster corresponding to the at least one target atom. According to the embodiment of the present disclosure, a spatial position relationship between atoms is considered in molecule modeling. In addition, for a cross-cluster atom which appears in the process of constructing an atom cluster, the adjustment on the atom cluster may be performed based on the property of the atom in a corresponding manner, so as to ensure that the adjusted atom cluster may effectively reduce the influence of the boundary7effect. Finally, the described processing method may be adapted to different scenarios and environments, such as performing molecular property7prediction on proteins, macromolecular materials and complex biological systems.

[0026] Example implementations of the present disclosure are described below with reference to the accompanying drawings.

[0027] To perform molecule modeling, it is necessary7to describe the environment where a molecule to be predicted (hereafter referred to simply as the molecule) is located. Taking protein molecules as an example, in a real biological environment, proteins generally exist in complex solution environments, such as in the human body. Proteins are generally surrounded by water molecules, ions and other small molecules. Taking molecules as drug molecules as an example, the drug molecules are typically organic compounds that can function in different biologicalenvironments, such as within a cell membrane, in blood or at a target protein binding site. The interaction between the drug molecules and water molecules, lipid molecules or other macromolecules would affect the effect thereof.

[0028] Each atom in a molecule has an atomic properly formed by its position in the molecule and chemical bonds surrounding it. The atomic property at least includes atoms to which it is connected, types (single bond, double bonds, etc.) of the chemical bonds, and the local electron density. In addition, an atom not only interacts with the directly adjacent atoms, but is also affected by distant atoms, including van der Waals forces and Coulomb forces. These long-range interactions play an important role in the computation of overall energy and force of a molecule. Accurately computing energy and force of a molecule is critical, especially in predicting the properties of macromolecules. However, due to the complexity and scale of macromolecules (such as proteins), such computation may become highly challenging.

[0029] FIG. 2 illustrates a flowchart of a method 200 for molecule modeling according to some implementations of the present disclosure. The method 200 may be implemented at the electronic device 110 of FIG. 1. The method 200 will be described below in conjunction with the environment 100 of FIG. 1.

[0030] At block 201, the electronic device 110 determines, based on at least one target atom in the molecule 102, an initial atom cluster centered on the at least one target atom and with a specified radius.

[0031] Generally speaking, the molecule 102 includes a plurality of atoms. Taking protein as an example, a protein consists of many amino acids, and each amino acid contains an average of 10- 20 atoms. Protein molecules typically consist of tens to thousands of amino acids, meaning that one protein molecule may contain hundreds to tens of thousands of atoms.

[0032] For the plurality' of atoms in the molecule 102, at least one target atom may be determined first, and the specific process of determining the target atom will be described later. For example, the target atom may be determined based on the atomic mass of each atom in the molecule 102.

[0033] For each target atom within the molecule 102, it is necessary to determine an atom cluster centered on the target atom. Therefore, the processing procedure for each target atom is the same or similar. FIG. 3 illustrates a schematic diagram of initial atom cluster determination principle 300 according to some implementations of the present disclosure. 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 illustration, the specified radius shown in FIG. 3 is 6A.It should be understood that, in a practical application scenario, the specified radius may be flexibly adjusted based on requirements.

[0034] For a relatively small specified radius, the advantage is low computational complexity,making it suitable for processing very large systems. That is, for a relatively small specified radius, the number of atoms contained in each atom cluster is relatively small, therefore computational complexity is relatively small. Since the computational complexity of quantum mechanics increases exponentially as the number of atoms increases, a relatively small specified radius may reduce the computational complexity.

[0035] For a relatively large specified radius, the advantage is that local interactions can be captured more comprehensively, improving the computational precision. That is, for a relatively large specified radius, the number of atoms contained in each cluster is greater. Therefore, the interaction between atoms can be described in more detail. Due to a more comprehensive consideration of local interactions within and between molecules, the computational result is closer to the real condition, so that the computational precision can be improved.

[0036] At block 202, the electronic device 110 determines, based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom, an adjustment strategy corresponding to the cross-cluster property.

[0037] Generally speaking, each molecule 102 contains a plurality of target atoms, and each target atom corresponds to an initial atom cluster of one target atom. Taking one of the target atoms as an example, an adjustment process of an atom cluster is described. An initial atom cluster determined by a specified radius centered on the target atom generally involves a virtual cut-off of an atom. That is, an atom located at the boundary of an initial atom cluster is generally virtually cut off. For an atom virtually cut off, a portion of the atom is in one of the initial atom clusters, and another portion is in another initial atom cluster. Therefore, the atom that is virtually cut off may be referred to as a cross-cluster atom.

[0038] The cross-cluster property may indicate a cross-cluster status of the cross-cluster atom. For example, the cross-cluster property of the cross-cluster atom may be a functional group crosscluster such as a double-bond cross-cluster, aromatic ring cross-cluster, etc. It may also be a singlebond cross-cluster, and the like. For each cross-cluster property, there is an adjustment strategy corresponding to the cross-cluster property, so as to adaptively adjust the atoms in the initial atom cluster. The purpose of the adjustment is to ensure that the chemical environment is minimally changed. For example, if the adjustment strategies are classified according to categories, they may include reallocation adjustment strategy and recombination adjustment strategy. The specific implementation process of the adjustment strategy is described in detail in the following.

[0039] At block 203, the electronic device 110 performs, based on the adjustment strategy, adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom, to obtain a modified atom cluster corresponding to the at least one target atom.

[0040] In response to the determination of the adjustment strategy, the adjustment may beperformed, based on the adjustment strategy, on the cross-cluster atom contained in the initial atom cluster to obtain the adjusted modified atom cluster. For example, the adjustment strategy may include reallocation adjustment strategy' and recombination adjustment strategy.

[0041] For the reallocation adjustment strategy, it may refer to performing reallocation of the atom cluster on the cross-cluster atoms in the initial atom cluster, and the reallocation may not break the original structure of the cross-cluster atom. For example, if the property of a cross-cluster atom is an aromatic ring cross-cluster. That is, one portion of the aromatic ring is assigned to a first initial atom cluster, and another portion of the aromatic ring is assigned to a second initial aromatic ring. The first and second initial atom cluster may be two adjacent atom clusters within the molecule 102, and may also be two atom clusters partially overlapped, and so on. Then, the reallocation adjustment strategy may be to reallocate the aromatic ring completely to the first initial atom cluster and also to the second initial atom cluster.

[0042] For the recombination adjustment strategy, it may refer to performing supplementation of a specified atom at the cross-cluster position corresponding to the cross-cluster atom. The supplementation of the specified atom is to maintain the chemical integrity of the atom clusters and minimize the impact on the original initial atom clusters. Thus, hydrogen atoms may be supplemented at the cross-cluster positions so that the free electrons are paired. In addition, the recombination adjustment strategy may also be, based on actual conditions, to supplement halogen atoms, etc. at a single bond of the virtual break. For example, the processes of the described blocks 201-203 may be performed by the atom cluster division module 130, and the processes of subsequent block 204 may be performed by the machine learning model 150.

[0043] At block 204, the electronic device 110 determines a target molecular property of the molecule based on the modified atom cluster corresponding to the at least one target atom.

[0044] Based on the modified atom clusters corresponding to the at least one target atom in the molecule 102, the target molecular property 120 of the molecule 102 may be determined. For example, computational results, such as an electron structure, molecular energy, a reaction path, and a molecule vibration frequency, may be obtained through quantum mechanics computation with the three-dimensional position information of the target atom in the modified atom cluster. These computational results may further predict the target molecular property of the molecule 102. For example, the electronic structure in the molecule may facilitate the understanding of reactive sites, charge transfer processes, and intermolecular interactions. The ground state energy and excited state energy to which the molecular energy relates are critical for assessing the stability and reactivity' of the molecule. The reaction path can facilitate analysis of energy changes between reactants and products, thereby predicting the kinetic behavior of the reaction. The molecular vibration frequencies may explain and predict the infrared and Raman spectra observed inexperiments, thereby helping to analyze molecular vibration modes and the strength of chemical bonds.

[0045] By means of the solution described above, the space is used as a basis for dividing an atom cluster, and thus the solution is applicable to different objects and environments. Applicable to different objects may include applicable to proteins that exist in the form of folded peptide states as well as drug molecules in the form of compound states. Applicable to different environments may refer to applicable to the complex environment within a human body. For example, water molecules, ions and other small molecular environment surrounding a protein, and water molecules, lipid molecules or another macromolecular environment surrounding a drug molecule. In addition, by performing targeted adjustments on the cross-cluster atom and generating a modified atom cluster, such processing method may effectively reduce the boundary effect caused by performing virtual cutting due to radius division. Since the boundary7effect means that a portion of atoms close to the boundary exhibits different characteristics from other atoms inside due to the existence of the boundary of the atom clusters. By adjusting the strategy, the boundary effect may be eliminated or mitigated, thereby providing more accurate results of target molecular property7. Finally, such processing of cross-cluster atoms may provide more accurate local molecular information, thereby providing more precise prediction results of target molecular property.

[0046] In some implementations of the present disclosure, upon determining the adjustment strategy corresponding to the cross-cluster property, for a target atom of the at least one target atom, the electronic device 110 may determine the cross-cluster property corresponding to the cross-cluster atom contained in the initial atom cluster of the target atom, and determine the adjustment strategy' based on the cross-cluster property.

[0047] In some implementations, the cross-cluster property may be categorized into a first crosscluster property7and a second cross-cluster property7. The first cross-cluster property7is used to indicate at least one of: a functional group cross cluster or a specified atom fragment cross cluster of a bonded atom. The second cross-cluster property indicates a single-bond cross cluster. The first cross-cluster property will be first introduced in detail below.

[0048] The case where the cross-cluster atom is a functional group cross cluster is first introduced. FIG. 4A illustrates a schematic diagram of a functional group cross cluster 400A according to some implementations of the present disclosure. The functional group shown in FIG. 4Ais phenol, and a dashed line 401 in FIG. 4 A may be used to represent a local boundary of an initial atom cluster. In the example of FIG. 4A, the local boundary7of the initial atom cluster virtually cuts off the phenol so that the phenol is referred to as a cross-cluster atom. Since the cross-cluster atom is the functional group (phenol) cross-cluster, the cross-cluster property of the cross-cluster atommay be the first cross-cluster property.[0049JFIG 4B illustrates a schematic diagram of a functional group cross cluster 400B according to some implementations of the present disclosure. A dashed line 402 in FIG. 4B may be used to represent a local boundary of an initial atom cluster. In the example of FIG. 4B, the local boundary of the initial atom cluster virtually cuts off the double bonds of the atom so that the atom is referred to as a cross-cluster atom. Since the cross-cluster atom is the functional group (double bonds) cross-cluster, the cross-cluster property7of the cross-cluster atom may be the first cross-cluster property.

[0050] In the examples of FIGS. 4A and 4B, the functional groups correspond to the phenol and double bonds. In actual scenarios, the functional groups may further correspond to triple bonds, aromatic rings, oxygen-containing functional groups (such as a hydroxyl group or carbonyl group), nitrogen-containing functional groups (such as an amino group or an amide), and so on.[0051JFIG. 4C illustrates a schematic diagram of a specified atomic fragment cross cluster 400C of a bonded atom according to some implementations of the present disclosure. A dashed line 403 in FIG. 4C may be used to represent a local boundary of an initial atom cluster. In the example of FIG. 4C, the local boundary7of the initial atom cluster virtually cuts off the bonded atom, with a fragment 403D of the bonded atom being located outside the boundary, and the fragment is simultaneously bonded to three atoms (an atom 403A, atom 403B and atom 403C) inside the boundary7, such that the atom is referred to as a cross-cluster atom. The cross-cluster property of the cross-cluster atom is the specified atom fragment cross cluster of the bonded atom, and therefore the cross-cluster property of the cross-cluster atom may be the first cross-cluster property7. The specified atom fragment cross cluster of the bonded atom may also be simply referred to as a bonded atom cross cluster.

[0052] In some implementations, if it is determined that a cross-cluster atom contained in an initial atom cluster of a certain target atom has the first cross-cluster property7. If the target atom is a first cross-cluster, the electronic device 110 may determine that the adjustment strategy is the reallocation adjustment strategy. That is, for the first cross-cluster property as the cross-cluster property, the reallocation adjustment strategy corresponding to the first cross-cluster property may be employed to perform the adjustment on the cross-cluster atom in the initial atom cluster. For example, the reallocation adjustment strategy7is to perform a reallocation on the assignment of the cross-cluster atoms rather than changing the structure of the cross-cluster atoms.

[0053] By means of the described process, the reallocation adjustment strategy determined based on the first cross-cluster property' may not change the structure of the cross-cluster atom, but only adjust the assignment of the cross-cluster atom. Thus, functional groups (such as double bonds and aromatic rings) have specific chemical properties and reactivities in the molecule. If they arecut off, their chemical characteristics would be changed. Through the reallocation strategy, the entire functional group is contained within one cluster, ensuring its chemical characteristics are not affected. In addition, for a bonded atom cross-cluster, i.e., a bonded atom connected to atoms in two or more clusters, if the bonded atom is cut off and replaced with a specified atom such as a hydrogen atom, it may result in an unrealistic structure. For example, if a process 423 shown in FIG. 4C is employed, i. e., a hydrogen atom 433 is used to supplement the virtually cut-off bond, two or more hydrogen atoms may be spatially close to each other or overlapping. Due to the volume of hydrogen atoms and the repulsion of the electron clouds, it may lead to steric hindrance. The reallocation adjustment strategy, such as a process 413 in FIG. 4C, reallocates these bonded atoms to one cluster. Thus, the chemical integrity of the molecule may be preserved. The interaction and position of each atom may be ensured to be more accurate by avoiding an unrealistic structural change.

[0054] In some implementations of the present disclosure, to determine whether a cross-cluster atom is the functional group cross cluster, the electronic device 110 may determine whether the cross-cluster atom includes a functional group and determine whether different functional group fragments in the functional group are cross-cluster. If the electronic device 110 determines that the cross-cluster atom includes the functional group, and the first functional group fragment of the functional group is outside the initial atom cluster, and a second functional group fragment of the functional group is inside the initial atom cluster, it may be determined that the cross-cluster atom as a functional group cross cluster corresponding to the first cross-cluster property.

[0055] Still referring to FIGS. 4A and 4B, the case shown in FIG. 4A is the double-bond cross cluster in the functional group, that is, one end of the double bonds is outside the initial atom cluster, and the other end is inside the initial atom cluster. For this case, the electronic device 110 may determine that the cross-cluster atom is with a functional group cross cluster corresponding to the first cross-cluster property. The case shown in FIG. 4B is the phenol cross cluster in the functional group, i.e., one fragment of the phenol is outside the initial atom cluster and the other fragment is inside the initial atom cluster. For this case, the electronic device 110 may further determine that the cross-cluster atom is a functional group cross cluster corresponding to the first cross-cluster property. In addition, functional group cross clusters may further include a triplebond cross cluster, aromatic ring cross cluster, and the like.

[0056] Through the identification of functional group cross clusters described above, the main significance lies in determination and preservation of the integrity of key reactive sites in the molecule. The functional group is a key structure determining the chemical reactivity of a molecule. By identifying and employing a corresponding strategy to process a cross-cluster functional group, it may be ensured that the functional characteristics of a molecule are accuratelypreserved in the process of molecule modeling and the determination of the target molecular property. For example, the functional group determines the chemical reactivity of a molecule, accurately identifying and processing these structures may ensure that these important reactive characteristics are not lost in molecule modeling. In addition, the identification of functional group cross-clusters may be applied to various complex molecular structures, enhancing the applicability of molecule modeling technology in fields such as material science and drug design.

[0057] In some implementations of the present disclosure, if it is determined that the cross-cluster property of the target atom is the functional group cross-cluster property, the adjustment performed by the electronic device 110 on the cross-cluster atom in the initial atom cluster includes: allocating, based on the reallocation adjustment strategy, the cross-cluster atom corresponding to the functional group cross cluster to the initial atom cluster, to obtain the modified atom cluster corresponding to the target atom.

[0058] For different types of functional group cross clusters with the first cross-cluster property, such as double bonds, triple bonds, aromatic rings, oxygen-containing functional groups, nitrogencontaining functional groups and so on, the corresponding adjustment strategy is the reallocation adjustment strategy7. Taking the cross-cluster atom corresponding to the functional group cross cluster crossing two initial atom clusters as an example, the first functional group atom fragment of the cross-cluster atom is assigned into the first initial atom cluster, and the second functional group atom fragment of the cross-cluster atom is assigned into the second initial atom cluster. For this case, the cross-cluster atom corresponding to the functional group cross cluster is completely reallocated to the first initial atom cluster and the second initial atom cluster respectively, to obtain the first and second modified atom clusters.

[0059] The first modified atom cluster serves as a modified atom cluster of the first initial atom cluster, and the second modified atom cluster serves as a modified atom cluster of the second initial atom cluster. That is, if the cross-cluster atom corresponding to the functional group cross cluster is an aromatic ring, the complete aromatic ring is contained in both the first and second modified atom clusters.

[0060] Through the process described above, for functional group cross-clusters, the adjustment is performed on the initial atom cluster with the reallocation adjustment strategy which may ensure that the functional group is completely included in the modified atom cluster, avoiding changes in the chemical environment caused by virtual cut-off. thereby reducing errors in subsequent molecular property analysis. In addition, the reallocation strategy is applicable to various complex molecular structures, ensuring that accurate simulation results may be obtained in different molecular environments.

[0061] In some implementations of the present disclosure, to determine whether the cross-clusteratom is the specified atom fragment cross cluster of the bonded atom, the electronic device 110 may determine whether the cross-cluster atom is the bonded atom and satisfy a particular crosscluster condition. If the electronic device 110 determines that a first atom fragment of the bonded atom is outside the initial atom cluster and the first atom fragment is bonded to at least two atoms within the initial atom cluster, it may be determined that the particular cross-cluster condition is satisfied. For this case, the first atom fragment may correspond to the specified atomic fragment.

[0062] If the initial atom cluster assigns one atom fragment (the first atom fragment) of the bonded atom outside the cluster, and the atom fragment is further bonded to at least two atoms in the initial atom cluster, it may be determined that the cross-cluster property of the bonded atom is the first cross-cluster property. The first atom fragment may have only one atom, or may be an atom fragment formed by bonding multiple atoms.

[0063] If the specified atom fragment of the bonded atom crosses clusters, one atom fragment outside the cluster is simultaneously bonded to at least two atoms within the cluster. Given that this multiple bonding relationship has complexity that is not present in the case of other types of cross clusters, its cross-cluster property may be defined as the first cross-cluster property. Thus, the integrity7and correctness of the cross-cluster atom structure may be ensured with the reallocation strategy when processing subsequently.

[0064] In some implementations of the present disclosure, in response to the first cross-cluster property including the specified atom fragment cross cluster of the bonded atom, the electronic device 110 performs the adjustment on the cross-cluster atom in the initial atom cluster includes: allocating, based on the reallocation adjustment strategy, the specified atom fragment of the bonded atom to the initial atom cluster to obtain the modified atom cluster corresponding to the target atom.

[0065] As mentioned above, the specified atom fragment cross cluster of the bonded atom corresponds that one atom fragment outside the cluster (i.e., the specified atom fragment) is simultaneously bonded to at least two atoms within the cluster. Based on these multiple bonding relationships, if a specified atom, such as a hydrogen atom, is simply supplemented at a position corresponding to the cross cluster, it would result in a plurality of hydrogen atoms being supplemented to the same atom fragment outside the cluster. For this case, supplementing the hydrogen, in contrast, may destroy the correct structure of the atom cluster, and affect the physical and chemical properties of the molecule, thereby causing a steric hindrance conflict, and further affecting subsequent computation and prediction.

[0066] Based on this, if the first cross-cluster property is the specified atom fragment cross cluster of the bonded atom, the first atom fragment of the bonded atom may be allocated to the initial atom cluster to obtain a modified atom cluster. Such reallocation strategy would not causestructural changes to the cross-cluster atom, thereby ensuring the integrity and correctness of the cross-cluster atom structure. In this regard, in the quantum mechanics computation corresponding to the target molecular property determination process, accurate atomic and electronic distribution is crucial. The reallocation of the cross-cluster atom may reflect the actual bonding relationships and electronic distribution of the molecules more accurately, thereby improving the accuracy of the computational results.

[0067] The judgment example and processing procedure for the cross-cluster property being the first cross-cluster property have been described in detail above. Next, the second cross-cluster property is introduced in detail. As mentioned above, the cross-cluster atom contained in the initial atom cluster may further have the second cross-cluster property, that is, a single-bond cross-cluster property. The second cross-cluster property may generally indicate single-bond cross-cluster, that is, one atom is inside the initial atom cluster, the other atom is outside the initial atom cluster, and the two atoms are connected through a single bond. In this regard, if the virtual cut-off single bond is unprocessed, the initial atom cluster may be caused to generate virtual free radicals. These virtual free radicals have high chemical activity and may lead to unexpected chemical reactions during the quantum computation process, causing inaccuracies in the computational results.

[0068] To make the molecular structure more complete and stabler and reduce the degree of freedom and uncertainty required to be processed in the computation, the recombination adjustment strategy needs to be performed on the second cross-cluster property. The recombination adjustment strategy7may restore the bond loss caused by the breaking of a single bond, ensuring that each atom satisfies the bonding requirements thereof and maintaining the chemical integrity and stability of the molecule.

[0069] In some implementations of the present disclosure, for a target atom of the at least one target atom, if a cross-cluster atom contained in an initial atom cluster of the target atom is determined to be with the second cross-cluster property, the electronic device 110 determines that the adjustment strategy for the initial atom cluster is the recombination adjustment strategy. In some implementations of the present disclosure, the process of the electronic device 110 performing the recombination adjustment strategy' may include supplementing the specified atom at the cross-cluster position corresponding to a single-bond cross cluster in the initial atom cluster.

[0070] With regard to the single-bond cross cluster corresponding to the second cross-cluster property, the recombination adjustment strategy may be employed to supplement the specified atom at the cross-cluster position corresponding to the single-bond cross cluster. When the specified atom is supplemented, a direction of the virtual cut-off single bond is obtained first. That is, an angle and a relative position of the virtual cut-off single bond between an atom inside the initial atom cluster and an atom outside the initial atom cluster are obtained. When the specifiedatom is supplemented, it would remain consistent with the original molecular structure or satisfy an error within a specified error range. This is to maintain the original geometry structure and chemical characteristics of the molecule as much as possible.

[0071] While the supplemented specified atom maintains the direction of the single bond, a length of the bond may also be adjusted based on specific conditions, so as to ensure that the newly formed structure may satisfy the physical and chemical requirements in the molecule. In this way, it may be ensured that during the processing of the cross-cluster, not only the directionality' of the original structure is preserved, but also the length of the bond may be adjusted, causing the newly formed structure to be reasonable and satisfy the requirements for chemical and phy sical stability. In this way, the computational accuracy may be improved, and it is ensured that the generated modified atom clusters conform to physical and chemical characteristics of actual molecules.

[0072] In some implementations of the present disclosure, after the atom cluster division module 130 in the electronic device 110 determines the modified atom clusters, the machine learning model 150 may obtain the three-dimensional position information of the target atom in each modified atom cluster, and obtain the three-dimensional position information of the specified atom in each modified atom cluster. Finally, the machine learning model 150 may determine, based on the three-dimensional position information of the target atom and the three-dimensional position information of the specified atom, the prediction result of the target molecular property of the molecule with a molecular target property prediction model.

[0073] For the modified atom cluster, the quantum mechanics (QM) computation may be performed on the interior of the modified atom cluster, and for the exterior of the modified atom cluster, mechanical and / or electronic embedding may be combined. The molecular target property’ prediction model, based on the modified atom cluster, performs computation on force and energy, so that different target molecular properties may be obtained based on the computational results.

[0074] The three-dimensional position information of the target atom and the three-dimensional position information of the specified atom in each modified atom cluster may be used as an input of the molecular target property prediction model. Here, the subsequent process is illustrated by taking the specific atom being a hydrogen atom as an example.

[0075] Equations (1) - (5) may correspond to the computational process of force. Based on molecular mechanics, the total energy of each modified atom cluster may be represented as follows:

[0076] The energy in quantum mechanics of the ithmodified atom cluster may be represented by EQMfragi') in Equation (1). The energy in quantum mechanics of the i* modified atom cluster includes the following components: a short-range kinetic energy (that is, the kinetic energy of theelectrons in the local environment (the modified atom cluster)), electron exchange interaction energy (an exchange energy caused by the indistinguishability of fermions), and correlation interaction energy (a dynamic association between electrons due to Coulomb interaction.usec| forrepresenting the energy of the long-range interaction on the i* modified atom cluster, that is, corresponding to the long-distance interaction effects applied to the modified atom cluster by the external environment.

[0077] The force acting on the target atom within the modified atom cluster is computed by the gradient of energy with respect to the position information of the target atom. That is, the force acting on the target atom within the modified atom cluster may be represented as follows:Fi= 7,Eotal(2)

[0078] In Equation (2). Vj may represent the computational result of the gradient of the three- dimensional position information of the target atom in the ithmodified atom cluster. The underlying logic of Equation (2) reflects that the force is the rate of change of energy with respect to position.

[0079] Computation of the energy of the long-range interactions involved in Equation (1) may be combined with molecular mechanics (EAIM) or electron embedding methods. Herein, the molecular mechanics is represented as follows:

[0080] In Equation (represent a potential energy between the target atom i (inside the modified atom cluster) and the target atom j (outside the modified atom cluster). <Jij may represent a effective distance parameter between the target atom i and the target atom j. rtj may represent a distance between the target atom / and the target atom / . mav represent Coulomb interactions, t^and q, mayrepresent charges of the target atom i and the target atom / , respectively. £0may represent an electrical constant in vacuum. The van der Waals forces describe the attractive and repulsive forces caused by interactions between atoms. Coulomb forces describe electrostatic interactions between charged atoms. Thus, Equation (3) represents the long-range interaction energy between the atom (the target atom z) inside the modified cluster and the atom (the target atom y) outside the modified cluster. The target atoms i and j may be atoms in the modified atom cluster.

[0081] The method of electronic embedding is represented as follows:

[0082] The electrostatic potential at the position R is computed in Equation (4), and Equation (4)is divided into two parts: one part is the potential generated by the electron density p(r), and the other part is the potential generated by the atomic nuclear charge Zm. \R-r | may represent a distance between the positions R and r. The integral term may represent the electrostatic potential generated at the position R due to the electron distribution p(r). The summation term may represent the electrostatic potential generated by M target atoms at the position / ?. Zm may represent the charge of the m&target atom, and Rm may represent the position of the mthtarget atom.

[0083] The corresponding total energy of the electronic embedding method is represented as follows:

[0084] In Equation (5), the electrostatic interactions of all target atoms are considered, i.e., the total energy Eemhof the interaction between the 7thtarget atom and all other target atoms, ql may represent the charge of the 7thtarget atom. For example, the total number of target atoms may be L.

[0085] By means of the equations described above, short-range interaction, long-range interaction and polarization effects are integrated. Such an integration method may obtain precise computational results regarding force. For example, short-range interactions are reflected in the computation of the energy in quantum mechanics EQM(Jragi) of each modified atom cluster. The long-range interaction is reflected in that each modified atom cluster is subjected to the van der Waals force, Coulomb force, and polarization effects of other external modified atom clusters. The polarization effect is described by the computational result of the electrostatic field, i.e., by the integral term and summation term in Equation (4).

[0086] For energy computation, the fact that force is the derivative of energy with respect to displacement may be utilized. By integrating the force along the displacement, the variation in the relative energy may be determined. Equations (6) and (7) may correspond to a computational process of energy.

[0087] In Equation (6), Ecurmay represent the energy of the modified atom cluster at the current displacement, Epremay represent the energy of the modified atom cluster determined previously, xcurmay represent a position of the current displacement of the modified atom cluster, xpremay represent a position of the previous displacement of the modified atom cluster, and F is the force acting on the modified atom cluster.

[0088] Since it is difficult to directly solve the integration in the actual computation, approximation may be performed using a numerical method. For example, the trapezoidal rule is a method commonly used in numerical integration. The trapezoidal rule is represented as follows:

[0089] In Equation (7), Fpremay represent the force acting on the modified atom cluster at the position of the previous displacement, F may represent the force acting on the modified atom cluster at the position of the current displacement, and xcurand xpremay still represent the positions of the current displacement of the modified atom cluster and previous displacement of the modified atom cluster, respectively. That is, although the exact functional form of the acted force F cannot be integrated in the numerical computation, an approximately continuous integration may be performed using discrete force values FpreandF”47-to obtain the energy variation between adjacent displacements.

[0090] The underlying logic of the above computation is that due to the complex positional variation and possibly overlapping positional relationships between the various modified atom clusters within the molecule, the energy is therefore not computed directly, but rather solved using the derivative of energy with respect to displacement.

[0091] For the recombination adjustment strategy, the atom added for the specified atom is as small and simple as possible, so that excessive computational complexity would not be introduced in the quantum mechanics computation, while maintaining the original local chemical environment as much as possible. In some implementations of the present disclosure, the specified atom may include a hydrogen atom.

[0092] Depending on the type of molecular system and the required computational accuracy, other atoms or other types of structures may be selected for addition in other implementations. For example, in some organic molecules, a methyl group may be added to saturate a dangling bond. In addition, other small groups such as fluorine and chlorine may also be added based on the chemical environment in the molecule and chemical properties that need to be maintained.

[0093] Determining the specified atom as a hydrogen atom may also provide more benefits compared to adding a methyl group or other small groups. On the one hand, hydrogen atoms are simple and easy to process, and in most cases, may effectively saturate the broken single bonds. On the other hand, supplementing hydrogen atoms is to maintain the original chemical environment of the system as much as possible.

[0094] In some implementations of the present disclosure, the electronic device 110 may determine the target atom in the molecule by determining, based on respective atom features of a set of atoms in the molecule, the at least one target atom from the set of atoms. An atom feature of an atom indicates at least one of: an atomic number or an atomic mass of the atom.

[0095] For the set of atoms in the molecule, at least one target atom may first be determined from the set of atoms. The target atom may be determined based on at least one of the atomic numbers or the atomic mass of each atom contained in the molecule. For example, the atomic number isthe position of the element in the periodic table, representing the number of protons in one atomic nucleus. A larger atomic number generally means that the atom has more protons and neutrons, thus an atom with an atomic number no less than the specified number may be determined as the target atom. As another example, the atomic mass is generally represented in units of atomic mass, which is the total mass of protons and neutrons in an atom. Therefore, an atom with an atomic mass not less than the specified atomic mass threshold may also be determined as the target atom.

[0096] The significance of determining the target atom lies in the fact that it can be focused on the process of determining the molecular properties, based on the local region surrounding the target atom, thereby reducing the complexity and time of the overall computation, and improving the computational efficiency. In addition, selecting the target atom and modifying the surrounding environment thereof (the modified atom cluster) may preserve the local structure and chemical environment of the molecule and ensure the accuracy of computational results. Finally, since the target atom tends to be the reaction center or the position of the functional group, the precise computation of these regions may better predict the molecular properties such as the reactivity and mechanism of the molecule.

[0097] It should be understood that the architecture of the models described above is merely an example and is not intended to be limiting.Example Devices[0098JFIG. 5 illustrates a schematic block diagram of an electronic device capable of implementing multiple implementations of the present disclosure. It should be understood that the electronic device 500 shown in FIG. 5 is merely an example and should not constitute any limitation on the functionality and scope of the implementations described in this disclosure.

[0099] As shown in FIG. 5, the electronic device 500 includes an electronic device 500 in the form of a general-purpose computing device. Components of the electronic device 500 may include, but are not limited to, one or more processors or processing devices 510, a memory 520, a storage 530, one or more communications units 540. one or more input devices 550. and one or more output devices 560.[OlOOJIn some implementations of the present disclosure, the electronic device 500 may be implemented as a computing device, a computing system, a server, a mainframe, or other computing-capable device.[OlOlJThe processing device 510 may be an actual or virtual processor and may perform various processes according to programs stored in the memory 520. In a multiprocessor system, a plurality of processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 500. The processing device 510 may include a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor, a controller,and / or a microcontroller, etc.

[0102] The electronic device 500 typically includes a plurality of computer storage media. Such media may be any available media that is accessible to the 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)), nonvolatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory’ (EEPROM), flash memory'), or some combination thereof. Storage 530 may include removable or non-removable media, and may include computer-readable media such as memory, flash drives, magnetic disks, or any other medium that can be used to store information and / or data and that can be accessed within electronic device 500.

[0103] The electronic device 500 may further comprise additional removable / non-removable, volatile / nonvolatile storage media. Although not shown in FIG. 5, a magnetic disk drive for reading from or writing to a removable, nonvolatile magnetic disk and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk may be provided. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces.

[0104] The communication unit 540 enables communication with another computing device via a communication medium. Additionally, the functionality of the components of the electronic device 500 may be implemented in a single computing cluster or multiple computing machines that are capable of communicating via a communication connection. Thus, the electronic device 500 may operate in a networked environment using logical connections to one or more other servers, personal computers (PCs), or another general network node.

[0105] Input device 550 may be one or more of a variety of input devices, such as a mouse, keyboard, data entry device, etc. The output device(s) 560 may be one or more output devices, such as a display, a data-deriving device, etc. The electronic device 500 may also communicate with one or more external devices (not shown) via the communication unit 540, as needed, an external device such as a storage device, a display device, etc., in communication with one or more devices that enable a user to interact with the electronic device 500, or any device that causes the electronic device 500 to communicate with one or more other computing devices (e.g., a network card and a modem) communicate with each other. Such communication may be performed via an input / output (I / O) interface (not shown).

[0106] In some implementations of the present disclosure, some or all of the various components of the electronic device 500, in addition to being integrated into a single device, may also be provided in the form of a cloud computing architecture. In a cloud computing architecture, these components may be arranged remotely and may w ork together to implement the functionality described in this disclosure. In some implementations, cloud computing provides computing,software, data access, and storage services that do not require end-users to know the physical location or configuration of the systems or hardware providing these services. In various implementations, cloud computing provides sendees over a wide area network, such as the Internet, using appropriate protocols. For example, cloud computing providers provide applications over a wide area network, and they may be accessed through a web browser or any other computing component. Software or components of the cloud-computing architecture, and corresponding data, may be stored on servers at a remote location. The computing resources in the cloud computing environment may be consolidated at the remote data center locations or they may be dispersed. Cloud computing infrastructures may provide sen ices through shared data centers, even though they appear as a single point of access for a user. Thus, the components and functions described herein may be provided from a service provider at a remote location using a cloud computing architecture. Alternatively, they may be provided from a conventional server, or they may be installed directly or otherwise on the client device.

[0107] Electronic device 500 may be used to implement molecule modeling in various implementations of the present disclosure. Memory 520 may include one or more modules having one or more program instructions that may be accessed and executed by processing unit 510 to implement the functionality of the various implementations described herein. For example, memory 520 may include molecule modeling module 522 for performing molecule modeling in one or more implementations described above. As shown in FIG. 5, the electronic device 500 may obtain an input required for molecule modeling via the input device 550, and may provide an output of molecule modeling via the output device 560, e.g., the determined target molecular property of the molecule. In some implementations, the electronic device 500 may also receive input from other devices (not shown) via the communication unit 540.Example Implementations

[0108] Some example implementations of the present disclosure are listed below.

[0109] In one aspect, the present disclosure provides an electronic device. The electronic device includes: a processing unit; and a memory coupled to the processing unit and containing instructions stored thereon which, when executed by the processing unit, cause the device to perform acts comprising: determining, based on at least one target atom in a molecule, an initial atom cluster centered on the at least one target atom and with a specified radius; determining, based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom, an adjustment strategy corresponding to the cross-cluster property; performing, based on the adjustment strategy, adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom, to obtain a modified atom cluster corresponding to the at least one target atom; and determining a target molecular property of the molecule basedon the modified atom cluster corresponding to the at least one target atom.[OllOJIn some implementations of the present disclosure, determining the adjustment strategy corresponding to the cross-cluster property may comprises: for a target atom of the at least one target atom, in response to a cross-cluster atom contained in an initial atom cluster of the target atom being with a first cross-cluster property, determining the adjustment strategy as a reallocation adjustment strategy, the first cross-cluster property indicating at least one of: a functional group cross cluster or a specified atom fragment cross cluster of a bonded atom.[OlllJIn some implementations of the present disclosure, the actions may further include: 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 atom cluster, and a second functional group fragment of the functional group being inside the initial atom cluster, determining the crosscluster atom as a functional group cross cluster corresponding to the first cross-cluster property.

[0112] In some implementations of the present disclosure, the first cross-cluster properly comprises the functional group cross cluster, and performing adjustment on the cross-cluster atom contained in the initial atom cluster of the at least one target atom based on the adjustment strategy comprises: allocating, based on the reallocation adjustment strategy; the cross-cluster atom corresponding to the functional group cross cluster to the initial atom cluster, to obtain the modified atom cluster.

[0113] In some implementations of the present disclosure, the actions may further include: in response to the cross-cluster atom comprising a bonded atom, a first atom fragment of the bonded atom being outside the initial atom cluster, and the first atom fragment being bonded to at least two atoms within the initial atom cluster, determining the cross-cluster atom as the specified atom fragment cross cluster of the bonded atom corresponding to the first cross-cluster property, and the first atom fragment to be corresponding to the specified atom fragment of the bonded atom.

[0114] In some implementations of the present disclosure, the first cross-cluster property comprises the specified atom fragment cross cluster of the bonded atom, and performing adjustment on the cross-cluster atom contained in the initial atom cluster of the at least one target atom based on the adjustment strategy comprises: allocating, based on the reallocation adjustment strategy, the specified atom fragment of the bonded atom to the initial atom cluster, to obtain the modified atom cluster.

[0115] In some implementations of the present disclosure, determining the adjustment strategy’ corresponding to the cross-cluster property includes: for a target atom of the at least one target atom, in response to a cross-cluster atom contained in an initial atom cluster of the target atom being with a second cross-cluster property, determining the adjustment strategy as a recombination adjustment strategy the second cross-cluster property indicating a single bond cross cluster.

[0116] In some implementations of the present disclosure, performing adjustment on the crosscluster atom in the initial atom cluster of the at least one target atom based on the adjustment strategy includes: supplementing, based on the recombination adjustment strategy, a specified atom at a cross-cluster position corresponding to the single bond cross cluster in the initial atom cluster.

[0117] In some implementations of the present disclosure, determining the target molecular property of the molecule may include: obtaining three-dimensional position information of each target atom in the modified atom cluster; obtaining three-dimensional position information of each specified atom in the modified atom cluster; and determining a target molecular property of the molecule with a molecular target property prediction model based on the three-dimensional position information of the target atom and the three-dimensional position information of the specified atom.

[0118] In some implementations of the present disclosure, the specified atom may include a hydrogen atom.

[0119] In some implementations of the present disclosure, the target atom in the molecule may be determined by: determining, based on respective atom features of a set of atoms in the molecule, the at least one target atom from the set of atoms, an atom feature of an atom indicating at least one of: an atomic number or an atomic mass of the atom.

[0120] In another aspect, the present disclosure provides a computer readable medium having computer-executable instructions stored thereon which, when executed by a device, cause the device to perform one or more example implementations of the method of the above aspect.

[0121] The functions described herein above may be performed, at least partially, 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), Applicationspecific Integrated Circuits (ASICs), Application-specific Standard Products (ASSPs), System- on-a-chip systems (SOCs). Load Programmable Logic Devices (CPLDs). etc.

[0122] Program code for implementing the methodologies of the present disclosure may be written in any combination of one or more programming languages. The 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 the program code, when executed by the processor or controller, causes the functions / operations specified in the flowchart and / or block diagrams to be performed. The program code may be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on the remote machine or entirely on the remote machine or server.

[0123] In the context of this disclosure, a machine-readable medium may be tangible media thatmay contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0124] In addition, while operations are depicted in a particular order, this should be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed to achieve desirable results. Multitasking and parallel processing may be advantageous in certain circumstances. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features that are described in the context of separate implementations can also be implemented in combination in a single implementation. Conversely, various features that are described in the context of a single implementation may also be implemented in multiple implementations separately or in any suitable sub-combination.

[0125] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

CLAIMS1. A computer-implemented method comprising: determining, based on at least one target atom in a molecule, an initial atom cluster centered on the at least one target atom and with a specified radius; determining, based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom, an adjustment strategy corresponding to the crosscluster property; performing, based on the adjustment strategy, adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom, to obtain a modified atom cluster corresponding to the at least one target atom; and determining a target molecular property of the molecule based on the modified atom cluster corresponding to the at least one target atom.

2. The method of claim 1, wherein determining the adjustment strategy 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 an initial atom cluster of the target atom being with a first cross-cluster property; determining the adjustment strategy as a reallocation adjustment strategy, the first cross-cluster property' indicating at least one of: a functional group cross cluster or a specified atom fragment cross cluster of a bonded atom.

3. The method of claim 2, 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 atom cluster, and a second functional group fragment of the functional group being inside the initial atom cluster, determining the cross-cluster atom as a functional group cross cluster corresponding to the first cross-cluster property7.

4. The method of claim 2 or 3, wherein the first cross-cluster property7comprises the functional group cross cluster, and performing adjustment on the cross-cluster atom contained in the initial atom cluster of the at least one target atom based on the adjustment strategy comprises: allocating, based on the reallocation adjustment strategy7, the cross-cluster atom corresponding to the functional group cross cluster to the initial atom cluster, to obtain the modified atom cluster.

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

6. The method of claim 2 or 5, wherein the first cross-cluster property comprises the specified atom fragment cross cluster of the bonded atom, and performing adjustment on the cross-cluster atom contained in the initial atom cluster of the at least one target atom based on the adjustment strategy comprises: allocating, based on the reallocation adjustment strategy, the specified atom fragment of the bonded atom to the initial atom cluster, to obtain the modified atom cluster.

7. The method of claim 1, wherein determining the adjustment strategy 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 an initial atom cluster of the target atom being with a second cross-cluster property7, determining the adjustment strategy7as a recombination adjustment strategy, the second cross-cluster property7indicating a single bond cross cluster.

8. The method of claim 7, wherein performing adjustment on the cross-cluster atom in the initial atom cluster of the at least one target atom based on the adjustment strategy7comprises: supplementing, based on the recombination adjustment strategy, a specified atom at a crosscluster position corresponding to the single bond cross cluster in the initial atom cluster.

9. The method of claim 1, wherein determining the target molecular property7of the molecule comprises: obtaining three-dimensional position information of each target atom in the modified atom cluster; obtaining three-dimensional position information of each specified atom in the modified atom cluster; and determining a target molecular property of the molecule with a molecular target property7prediction model based on the three-dimensional position information of the target atom and the three-dimensional position information of the specified atom.

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

11. The method of claim 1, wherein the target atom in the molecule is determined by: determining, based on respective atom features of a set of atoms in the molecule, the at least one target atom from the set of atoms, an atom 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 which,when executed by the processing unit, cause the device to perform acts comprising: determining, based on at least one target atom in a molecule, an initial atom cluster centered on the at least one target atom and with a specified radius; determining, based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom, an adjustment strategy corresponding to the crosscluster property; performing, based on the adjustment strategy, adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom, to obtain a modified atom cluster corresponding to the at least one target atom; and determining a target molecular property of the molecule based on the modified atom cluster corresponding to the at least one target atom.

13. The device of claim 12, wherein determining the adjustment strategy 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 an initial atom cluster of the target atom being with a first cross-cluster property, determining the adjustment strategy as a reallocation adjustment strategy the first cross-cluster property7indicating at least one of: a functional group cross cluster or a specified atom fragment cross cluster of a bonded atom.

14. The device of claim 13, w herein the acts further comprise: 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 atom cluster, and a second functional group fragment of the functional group being inside the initial atom cluster, determining the cross-cluster atom as a functional group cross cluster corresponding to the first cross-cluster property.

15. A computer program product being tangibly stored in a computer storage medium and comprising computer executable instructions which, when executed by a device, cause the device to perform acts comprising: determining, based on at least one target atom in a molecule, an initial atom cluster centered on the at least one target atom and with a specified radius; determining, based on a cross-cluster property of a cross-cluster atom contained in the initial atom cluster of the at least one target atom, an adjustment strategy corresponding to the crosscluster property; performing, based on the adjustment strategy; adjustment on a cross-cluster atom contained in the initial atom cluster of the at least one target atom, to obtain a modified atom cluster corresponding to the at least one target atom; anddetermining a target molecular property of the molecule based on the modified atom cluster corresponding to the at least one target atom.

Images