Method, apparatus, medium and program product for drug screening

Abstract

The invention relates to a drug screening method, apparatus, medium and program product. The method comprises the following steps: screening a plurality of chemical molecules in a chemical molecule library through a first deep learning model so as to obtain a plurality of to-be-selected molecules meeting preset conditions; determining binding characterization data of each to-be-selected molecule in the plurality of to-be-selected molecules and a specified target so as to select candidate molecules from the plurality of to-be-selected molecules based on the calculated binding characterization data; constructing a molecular feature map for the candidate molecules, the molecular feature map at least indicating features of molecules, functional groups, and atoms about the candidate molecules; and determining molecular characterization data of the candidate molecules based on the constructed molecular feature map, so as to determine target molecules for drugs in the candidate molecules based on the molecular characterization data. According to the method, high screening efficiency and good universality and accuracy can be taken into account.

Description

Technical Field

[0001] The present invention relates generally to biological information processing, and in particular, to methods, computing devices, computer storage media, and computer program products for drug screening. Background Art

[0002] Existing methods for drug screening include, for example: drug screening methods based on high-throughput cell or molecular experiments, virtual drug screening methods based on molecular docking, and molecular screening methods based on deep learning.

[0003] In drug screening methods based on high-throughput cell or molecular experiments, they are mainly based on high-throughput molecular and cell-level experimental methods, combined with large-scale data analysis and processing procedures. Therefore, they face the problems of high investment, long cycle and low efficiency.

[0004] In the virtual drug screening method, special molecules targeting specific targets are screened mainly by calculating the degree of spatial binding between drug molecules and targets. However, molecular docking is too dependent on target selection. However, in most cases, accurate targets are not easy to obtain. Therefore, in the case of unclear or unknown targets, molecular docking technology is difficult to play its advantages. Therefore, the versatility and accuracy of virtual drug screening methods are poor.

[0005] As for the molecular screening method based on deep learning, although it has the technical advantages of high efficiency and good transferability, the existing deep learning-based methods cannot characterize the molecular properties well, which leads to the unsatisfactory accuracy of molecular screening. In addition, the existing deep learning-based methods have poor interpretability and it is difficult to provide an intuitive explanation for the results.

[0006] In summary, the shortcomings of traditional methods for drug screening are that it is difficult to achieve both high screening efficiency and good versatility and accuracy. Summary of the invention

[0007] The present invention provides a method, computing device, computer storage medium and computer program product for drug screening, which can achieve high screening efficiency, good versatility and accuracy.

[0008] According to a first aspect of the present invention, a method for drug screening is provided. The method comprises: for obtaining a plurality of chemical molecules from a chemical molecule library, screening is performed via a first deep learning model to obtain a plurality of candidate molecules that meet predetermined conditions; determining the binding characterization data of each candidate molecule in the plurality of candidate molecules with a specified target, so as to select a candidate molecule from the plurality of candidates based on the calculated binding characterization data; for the candidate molecules, constructing a molecular feature graph, the molecular feature graph at least indicating the characteristics of the molecules, functional groups, and atoms of the candidate molecules; and determining the molecular characterization data of the candidate molecules based on the constructed molecular feature graph, so as to determine the target molecule for the drug in the candidate molecules based on the molecular characterization data.

[0009] According to a second aspect of the present invention, a computing device is also provided, comprising: a memory configured to store one or more computer programs; and a processor coupled to the memory and configured to execute the one or more programs so that the device performs the method of the first aspect of the present invention.

[0010] According to a third aspect of the present invention, a non-transitory computer-readable storage medium is further provided, wherein machine-executable instructions are stored on the non-transitory computer-readable storage medium, and when the machine-executable instructions are executed, the machine executes the method of the first aspect of the present invention.

[0011] According to a fourth aspect of the present invention, a computer program product is further provided, wherein a computer program is stored thereon, and when the computer program is executed by a machine, the method according to the first aspect of the present invention is implemented.

[0012] In some embodiments, based on the constructed molecular feature graph, determining the molecular characterization data of the candidate molecule includes: based on the constructed molecular feature graph, extracting the relationship information of the nodes and edges of the molecular feature graph via a heterogeneous graph attention network model, so as to aggregate the extracted relationship information of the nodes and edges, thereby generating molecular characterization data about the candidate molecule.

[0013] In some embodiments, aggregating the extracted relationship information of nodes and edges to generate molecular characterization data about candidate molecules includes: aggregating the relationship information of nodes and edges about molecular nodes, two-dimensional graph atomic nodes, and three-dimensional graph atomic nodes to functional group nodes for generating molecular characterization data about candidate molecules.

[0014] In some embodiments, constructing a molecular feature graph for a candidate molecule includes: using the molecule of the candidate molecule, each functional group in the candidate molecule, each atom in the two-dimensional chemical structure of the candidate molecule, and each atom in the three-dimensional atomic space conformation of the candidate molecule as nodes in the molecular feature graph, respectively, to construct a node set; making each node in the node set associated with a mapping function of a corresponding node type in the set of node types; constructing a connection relationship between molecular nodes, functional group nodes, two-dimensional graph atomic nodes, and three-dimensional graph atomic nodes in the form of edges, to form an edge set; and making each edge in the edge set associated with a mapping function of a corresponding edge type in the set of edge types.

[0015] In some embodiments, based on the constructed molecular feature graph, determining the molecular characterization data of the candidate molecule includes: initializing each node in the molecular feature graph into a vector to generate an initial feature of the node, the dimensions of the initial features of different types of nodes being different; mapping the initial features of different nodes into the same dimension; and performing message passing and aggregation on the initial features of the mapped nodes to obtain the molecular characterization data.

[0016] In some embodiments, message passing and aggregation are performed on initial features of nodes via mapping to obtain molecular characterization data, including: calculating attention coefficients of a current node in a molecular feature graph and other nodes connected to the current node; aggregating features of neighboring nodes of the current node for updating features of the current node; and determining molecular characterization data about candidate molecules based on the features of the updated nodes.

[0017] In some embodiments, determining molecular characterization data about the candidate molecule based on the features of the updated nodes includes: selecting features of the clustered and updated functional group nodes; and generating molecular characterization data about the candidate molecule via a maximum pooling operation for the features of the clustered and updated functional group nodes.

[0018] In some embodiments, based on the calculated binding characterization data, selecting candidate molecules from the multiple candidates includes: determining the binding characterization data of each candidate molecule with a specified target so as to rank the candidate molecules based on the calculated binding characterization data; and based on the ranking result, selecting a predetermined number of candidate molecules with the highest ranking as candidate molecules.

[0019] In some embodiments, the method for drug screening also includes: in response to determining that the ratio of the depth between mitochondrial fusion breakpoints to the depth of other regions is greater than a predetermined ratio threshold, determining the presence of variant sites and regions; and screening the determined variant sites and regions based on a predetermined variant database to obtain pathogenic variant sites.

[0020] In some embodiments, the target molecule identified is a target molecule for anti-aging

[0021] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the invention, nor is it intended to limit the scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of a computing device for implementing a method for drug screening according to an embodiment of the present invention is shown.

[0023] Figure 2 A flow chart for implementing a method for drug screening according to an embodiment of the present invention is shown.

[0024] Figure 3 A schematic diagram of a molecular signature map according to an embodiment of the present invention is shown.

[0025] Figure 4 A schematic diagram showing a lateral comparison between the method for drug screening according to an embodiment of the present invention and other methods is shown.

[0026] Figure 5 A flow chart of a method for constructing a molecular feature map according to an embodiment of the present invention is shown.

[0027] Figure 6 A flow chart of a method for determining molecular characterization data of a candidate molecule according to an embodiment of the present invention is shown.

[0028] Figure 7 A schematic diagram of a method for determining molecular characterization data of a candidate molecule according to an embodiment of the present invention is shown.

[0029] Figure 8 A flow chart of a method for obtaining molecular characterization data via node messaging and aggregation according to an embodiment of the present invention is shown.

[0030] Fig. 9 A schematic diagram of a method for determining molecular characterization data of a candidate molecule according to some other embodiments of the present invention is shown.

[0031] Fig.10 The block diagram of an electronic device suitable for implementing the embodiments of the present invention is schematically shown.

[0032] Fig.11 A molecular schematic diagram of target molecules for drugs determined according to some embodiments of the present invention is shown.

[0033] Fig.12Molecular schematic diagrams of target molecules for drugs determined according to other embodiments of the present invention are shown. Fig.13 Molecular schematic diagrams of target molecules for drugs determined according to further embodiments of the present invention are shown.

[0034] Fig.14 A schematic diagram of a molecular signature map constructed according to some embodiments of the present invention is shown.

[0035] Fig.15 Schematic diagrams of molecular feature maps constructed according to other embodiments of the present invention are shown.

[0036] In the various drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0037] The preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0038] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects.

[0039] As described above, the disadvantages of traditional methods for drug screening are that it is difficult to achieve both high screening efficiency and good versatility and accuracy.

[0040] In order to at least partially solve one or more of the above problems and other potential problems, an exemplary embodiment of the present invention proposes a method for drug screening. In this method, the candidate molecules screened out by the deep learning model are screened again by the method of molecular docking. On the one hand, the scope of candidate molecules is further narrowed, and the efficiency of screening is significantly improved. On the other hand, more interpretable basis is provided for the entire screening process; and the present invention obtains more modal information of candidate molecules (molecular global information, molecular pharmacophore information, and atomic local information of molecules) through the multi-view structure of the molecular feature map, so that the molecular characterization data determined based on the molecular feature map can better characterize the information of the candidate molecules, improve the generalization ability of the method and the accuracy of screening the properties of candidate molecules. Therefore, the present invention can take into account higher screening efficiency, as well as good versatility and accuracy.

[0041] Figure 1 A schematic diagram of a computing device for a method for drug screening according to an embodiment of the present invention is shown.

[0042] Regarding the computing device 100, it is used for drug screening, for example. Specifically, the computing device 100 can obtain a plurality of chemical molecules from a chemical molecule library, and screen them via a first deep learning model to obtain a plurality of candidate molecules that meet predetermined conditions; and determine the binding characterization data of each candidate molecule in the plurality of candidate molecules with a specified target, so as to select a candidate molecule from the plurality of candidates based on the calculated binding characterization data. In addition, the computing device 100 can construct a molecular feature map for the candidate molecules, the molecular feature map at least indicating the characteristics of the molecules, functional groups, and atoms of the candidate molecules; and determine the molecular characterization data of the candidate molecules based on the constructed molecular feature map, so as to determine the target molecule for the drug in the candidate molecules based on the molecular characterization data.

[0043] In some embodiments, the computing device 100 may have one or more processing units, including dedicated processing units such as GPU, FPGA and ASIC, and general processing units such as CPU. In addition, one or more virtual machines may also be running on each computing device. The computing device 100 includes, for example: a candidate molecule screening unit 112, a combined characterization data calculation and candidate molecule selection unit 114, a molecular feature map construction unit 116, and a target molecule determination unit 118. The above-mentioned candidate molecule screening unit 112, the combined characterization data calculation and candidate molecule selection unit 114, the molecular feature map construction unit 116, and the target molecule determination unit 118 can be configured on one or more computing devices 100.

[0044] The candidate molecule screening unit 112 is used to screen multiple chemical molecules obtained from the chemical molecule library through a first deep learning model so as to obtain multiple candidate molecules that meet predetermined conditions.

[0045] The binding characterization data calculation and candidate molecule selection unit 114 is used to determine the binding characterization data of each candidate molecule in a plurality of candidate molecules and a designated target, so as to select a candidate molecule from the plurality of candidate molecules based on the calculated binding characterization data.

[0046] The molecular feature map constructing unit 116 is used to construct a molecular feature map for a candidate molecule, wherein the molecular feature map at least indicates features of molecules, functional groups, and atoms of the candidate molecule.

[0047] The target molecule determination unit 118 is used to determine the molecular characterization data of the candidate molecules based on the constructed molecular feature graph, so as to determine the target molecule for the drug among the candidate molecules based on the molecular characterization data.

[0048] The following will be combined Figure 2 , Figure 3 A method for drug screening according to an embodiment of the present invention is described. Figure 2 A flow chart of a method 200 for drug screening according to an embodiment of the present invention is shown. Figure 3 Schematic diagram of a molecular feature map 300 according to an embodiment of the present invention is shown. The method 200 may be performed, for example, in Fig.10 The electronic device 1000 described above may also be used. Figure 1 The method 200 is executed at the described computing device 100. It should be understood that the method 200 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0049] At step 202 , the computing device 100 obtains a plurality of chemical molecules from a chemical molecule library and screens them using a first deep learning model to obtain a plurality of candidate molecules that meet predetermined conditions.

[0050] For example, the computing device 100 first inputs all or part of the chemical molecules in the chemical molecule library into the trained first deep learning model for preliminary screening, so as to screen out the candidate molecules that meet the predetermined conditions. The first deep learning model is trained, for example, by a large number of samples of candidate molecules that are labeled as having effective ingredients (for example, having therapeutic effects on a certain disease or being effective for anti-aging).

[0051] The candidate molecules satisfying the predetermined conditions are, for example, chemical molecules having effective ingredients, for example, chemical molecules containing effective ingredients effective for delaying aging.

[0052] At step 204 , the computing device 100 determines binding characterization data of each candidate molecule in a plurality of candidate molecules with a designated target, so as to select a candidate molecule from the plurality of candidate molecules based on the calculated binding characterization data.

[0053] Regarding the method for calculating the binding characterization data (e.g., binding score) of the candidate molecule and the predetermined target, it includes, for example: the computing device 100 obtains the target in the specific pathway in the known literature; retrieves the spatial structure and possible binding site information of the corresponding target in the Protein DataBank (PDB), which is a database that specifically collects the three-dimensional structure data of proteins and nucleic acids; based on the spatial structure and possible binding site information, calculates the binding characterization data of the candidate molecule and the specified target. In some embodiments, for example, MGLTools and AutoDock Vina software for molecular docking can be used to calculate the binding characterization data of the candidate molecule and the specified target. It should be understood that AutoDock Vina is an existing molecular docking software, which is used based on the MGLTools toolkit. The MGLTools toolkit includes AutoDock Tools and Python Molecular Viewer. AutoDock Tools is used to generate input files for Vina, and Python Molecular Viewer is used to view the results.

[0054] Regarding the method for selecting candidate molecules, it includes, for example: determining the binding characterization data of each candidate molecule with a specified target, so as to rank the candidate molecules based on the calculated binding characterization data; and based on the ranking result, selecting a predetermined number of candidate molecules with top rankings as candidate molecules.

[0055] At step 206, the computing device 100 constructs a molecular feature map for the candidate molecule, wherein the molecular feature map at least indicates the characteristics of the molecules, functional groups, and atoms of the candidate molecule. It should be understood that the molecular feature map constructed by the present invention indicates more information about the candidate molecule, and involves information of multiple types (rather than a single type) of molecules, functional groups, and atoms, thereby facilitating a more accurate and comprehensive identification of the properties of the candidate molecule.

[0056] About the characteristics of the molecule of the candidate molecule, it is, for example, a molecular view of the candidate molecule. In some embodiments, the molecular view is generated based on the SMILES sequence of the candidate molecule via a pre-trained model (for example, but not limited to: MOLBERT model). MOLBERT uses the Bert model architecture and utilizes the simplified molecular input line entry specification (SMILES) sequence information of the molecule to learn the global representation of the molecule by predicting the masked atoms or other characters in the molecular sequence. It should be understood that the SMILES sequence information represents the molecular structure in the form of a string. The SMILES encoding corresponds to the chemical structure. It uses letters such as C, H, O, N, S to represent the corresponding atoms. Two adjacent atoms represent a single bond connection. Double bonds between atoms are represented by "=". Triple bonds between atoms are represented by "#". Brackets are used to indicate that atoms contain side chains, etc.

[0057] About the feature of the functional group of the candidate molecule, it is, for example, the functional group view of the candidate molecule, which is used to indicate the property of the functional group of the candidate molecule. It should be understood that RDKit has built-in multiple molecular fingerprint calculation methods. For example, the computing device 100 uses, for example, the built-in "rdkit.Chem.MACCSkeys.GenMACCSKeys()" function of RDKit to calculate the MACCS fingerprint of the functional group, and the built-in "factory.GetFeaturesForMol(m)" function of RDKit can obtain the characteristics of the functional group. The features of the functional group of the candidate molecule can be obtained based on the calculated MACCS fingerprint of the functional group and the characteristics of the functional group.

[0058] The features of the atoms of the candidate molecule include, for example, at least one of a two-dimensional atomic view of the candidate molecule and a three-dimensional atomic view of the candidate molecule.

[0059] The two-dimensional atomic view of the candidate molecule is constructed, for example, according to the chemical structure of the candidate molecule. For example, the atoms of the candidate molecule are set as points, and the chemical bonds of the candidate molecule are set as edges, so as to construct the two-dimensional atomic view of the candidate molecule.

[0060] Regarding the three-dimensional atomic view of the candidate molecule, it is, for example, constructed by a molecular space conformation generation model (for example, but not limited to the known GEODIFF model). For example, first, the computing device 100 uses the GEODIFF model to use a diffusion model to simulate the process of a molecule from a chaotic spatial geometric structure to a stable spatial conformation, so as to obtain a three-dimensional molecular space conformation, which contains the three-dimensional coordinates of each atom in the molecule. Then, the atoms in the three-dimensional molecular space conformation are configured as nodes, and the Euclidean distance of the atoms in the three-dimensional molecular space conformation generated by the GEODIFF model is calculated. Determine whether the calculated Euclidean distance is less than a predetermined value; if the calculated Euclidean distance is less than a predetermined value (for example, a predetermined value of 4), an edge connection is configured between the atoms. Afterwards, the three-dimensional atomic view of the candidate molecule is generated based on the configured nodes and edge connections. Regarding the atomic representation in the three-dimensional atomic view of the candidate molecule, it is, for example, generated by Uni-Mol model pre-training. For example, the computing device 100 uses the Uni-Mol model to add fixed spatial position encoding, atom pair information, and molecular smiles sequence information to the Transformer architecture to output the spatial structure and atomic representation of the molecule. Since the spatial structure of the molecule is used in the training process, the atomic representation (512 dimensions) output by the Uni-Mol model contains the spatial conformation information of the molecule.

[0061] In some embodiments, the three-dimensional atomic view of the candidate molecule indicates the three-dimensional atomic spatial conformation of the candidate molecule.

[0062] Regarding the molecular feature graph, in some embodiments, it includes: a node set, an edge set, a node type set, and an edge type set, wherein the node set includes a molecule node, a functional group node, a two-dimensional graph atom node, and a three-dimensional graph atom node. Figure 3As shown, the molecular feature graph 300 is a heterogeneous topological structure graph about nodes and edges. The topmost node of the molecular feature graph 300 is a molecular node 310. The molecular node 310 represents the entire molecule of the candidate molecule. The second layer of the molecular feature graph 300 includes one or more functional group nodes 312. It should be understood that the functional group is an atom or atomic group that determines the chemical properties of the organic compound. Functional groups include, for example, carbon-carbon double bonds, carbon-carbon triple bonds, hydroxyl groups, carboxyl groups, ether bonds, aldehyde groups, benzene rings, etc. The present invention projects each functional group in all functional groups in the candidate molecule into a functional group node. Each molecular node is connected to each functional group node in the form of an edge (for example, as shown in mark 314). If there is an atomic connection between the functional groups, the functional group nodes with atomic connections are connected in the form of edges (for example, as shown in mark 316). Atomic nodes are divided into two types. For example, the left half of the third layer of the molecular feature graph 300 indicates the two-dimensional chemical structure of the candidate molecule. The two-dimensional graph atom nodes 318 indicate atoms in the two-dimensional chemical structure of the candidate molecule, and the edges connected by the two-dimensional graph atom nodes (e.g., as shown by the mark 320) indicate the chemical bonds of the atoms. The third layer right half of the molecular feature graph 300 indicates the spatial conformation of the atoms of the candidate molecule. The atoms in the spatial conformation are the three-dimensional graph atom nodes 324.

[0063] The following formula (1) schematically shows the functional expression of the molecular feature map.

[0064]

[0065] In the above formula (1), V represents the node set of the molecular feature graph, and E represents the edge set of the molecular feature graph. A collection of node types representing molecular feature graphs. A collection of edge types representing a molecular feature graph.

[0066] It should be understood that in the molecular feature graph, there are four types of nodes, namely, M, P, A1, and A2. Fig.14 Schematic diagrams of molecular feature maps constructed according to some embodiments of the present invention are shown. Fig.14 The molecular feature graph constructed for the molecule "CC=CC(=O)N(CC)c1ccccc1C" is used as an example to specifically illustrate the structure of the molecular feature graph and its construction method. M represents the molecular node of the candidate molecule (such as Fig.14 There is only one molecule node in the molecular feature graph of a candidate chemical molecule. P represents a functional group node (e.g. Fig.14 Each functional group is represented as a node. A1 represents an atomic node in the two-dimensional graph (e.g. Fig.14A2 represents the atomic nodes of the three-dimensional graph, and each atom represents a node (such as Fig.14 1408 in FIG. 14B ).

[0067] In some embodiments, in the molecular feature graph, there are six types of edges, which are used to indicate the connection relationship between the objects (i.e., nodes) connected to the two ends of the edge. For example, the edges are MP, PP, P-A1, P-A2, A1-A1, and A2-A2. Among them, MP represents the connection relationship between the molecule node and the functional group node (e.g. Fig.14 In some embodiments, the computing device 100 connects all functional group nodes (i.e., P) to the molecule node (i.e., M). PP represents the connection relationship between the functional group nodes (e.g., Fig.14 P-A1 represents the connection relationship between the atomic nodes in the two-dimensional graph and the functional group nodes to which they belong (e.g. Fig.14 P-A2 represents the connection relationship between the atomic nodes of the three-dimensional graph and the functional group nodes to which they belong (eg Fig.14 A1-A1 represents the chemical bonds in the candidate molecule (e.g. Fig.14 ), A2-A2 represents the distance relationship of the atoms of the candidate molecule in space (as indicated by the mark 1420 in FIG. Fig.14 1422 in FIG. 14).

[0068] Regarding the method for constructing a molecular feature graph, it includes, for example: the computing device 100 uses the molecule of the candidate molecule, each functional group in the candidate molecule, each atom in the two-dimensional chemical structure of the candidate molecule, and each atom in the three-dimensional atomic space conformation of the candidate molecule as nodes in the molecular feature graph, so as to construct a node set; so that each node in the node set is associated with a mapping function of a corresponding node type in the set of node types; through edges, the connection relationship between the molecular nodes, the functional group nodes, the two-dimensional graph atomic nodes, and the three-dimensional graph atomic nodes is constructed to form an edge set; and so that each edge in the edge set is associated with a mapping function of a corresponding edge type in the set of edge types. The following will be combined with Figure 4 The specific implementation of the method 400 for constructing a molecular feature map is described in detail and will not be repeated here.

[0069] At step 208 , the computing device 100 determines molecular characterization data of the candidate molecules based on the constructed molecular feature graph, so as to determine a target molecule for a drug among the candidate molecules based on the molecular characterization data.

[0070] In some embodiments, the candidate molecule determined by method 200 is, for example, O=C1C=2C=CC=CC2N(C=3OC(=C(C)CC4C5=C(OC(C)(C)C4)C=6C=CC=CC6N(C5=O)C)CC13)C; the target molecule determined for drug use is, for example, Fig.11 shown.

[0071] In other embodiments, the candidate molecules determined by method 200 are, for example, O=C(C=CC1=CC=C(OC)C(=C1)COC=2C=CC=CC2C3CCCCC3)C4=CC=C(O)C=C4; the target molecules for drugs determined are, for example, Fig.12 In some other embodiments, the candidate molecules determined by method 200 are, for example, FC(F)(F)C1=CC=C(C=C1)CN2CCC(C3=NC=4C=C(C=CC4N3)C)C2; the target molecules for drugs determined are, for example, Fig.13 shown.

[0072] Regarding the method of generating molecular characterization data, in some embodiments, for example, it includes: the computing device 100 extracts the relationship information of nodes and edges of the molecular characteristic graph based on the constructed molecular characteristic graph through the heterogeneous graph attention network model, so as to aggregate the extracted relationship information of nodes and edges, thereby generating molecular characterization data about the candidate molecule. Specifically, for example, Figure 7 As shown, the computing device 100 aggregates the node and edge relationship information about the molecule node 720, the two-dimensional graph atom node 726, and the three-dimensional graph atom node 728 to the functional group node 724, so as to generate the molecular characterization data 740 about the candidate molecule. Afterwards, based on the molecular characterization data 740, the properties of the candidate molecule (e.g., determining whether it is effective in delaying aging) are determined by the trained second neural network model 742 (e.g., MLP network), so as to screen out the target molecule from the candidate molecules.

[0073] The following will combine Figure 5 The specific implementation of the method for generating molecular characterization data is specifically described and will not be repeated here.

[0074] In order to verify the effect of the present invention, the present invention selects data from an anti-aging data set, which is, for example but not limited to, an experimentally verified anti-aging molecular data set provided by Wong F et al. The data of the anti-aging data set used contains a total of 2353 different small molecule drugs, of which 45 molecules are positive. The 45 positive molecules represent that they can clear senescent cells. In order to make the test results more representative, the present invention uses the Scaffold data set division method to divide the anti-aging data set, wherein the ratio of the training set, the validation set, and the test set is 7:1:2 respectively. In order to accurately compare and evaluate the drug screening effect of the method of the present invention with other methods, the present invention uses a five-fold cross validation, that is, the data is divided into 5 equal parts, and one part is taken for testing each time, and the rest is used for training. The average value is calculated for 5 experiments. In addition, the present invention uses the calculated average AUC (Area under Curve, i.e., the area under the Roc curve) as the criterion. The methods used as comparative examples are, for example, the known D-MPNN-based method, the FraGAT model-based method, the CoMPT model-based method, the ABT-MPNN model-based method, and the PharmHGT model-based method. Figure 4 A schematic diagram of a horizontal comparison between the method for drug screening according to an embodiment of the present invention and other methods is shown. Mark 410 indicates the average AUC of the method according to the present invention for the test set. Mark 412 indicates the average AUC of the method based on the D-MPNN model for the test set. Mark 414 indicates the average AUC of the method based on the FraGAT model for the test set. Mark 416 indicates the average AUC of the method based on the CoMPT model for the test set. Mark 418 indicates the average AUC of the method based on the ABT-MPNN model for the test set. Mark 420 indicates the average AUC of the method based on the PharmHGT model for the test set. Figure 4 It can be seen that the method of the present invention shows better performance than the comparative method, and even when the training set is only 70%, the average AUC on the test set still reaches 80%. Compared with the method based on other models (for example, D-MPNN), the effect is improved by 15%.

[0075] In the above method, the present invention selects candidate molecules based on the combined characterization data of the candidate molecules screened by the first deep learning model and the specified target, and constructs a molecular feature map including atoms, functional group views, and molecular characterization for the candidate molecules, and determines the molecular characterization data of the candidate molecules based on the molecular feature map, so as to determine the target molecules for the drug, and the present invention can screen the candidate molecules screened by the deep learning model again by the method of molecular docking, on the one hand, further narrowing the scope of the candidate molecules, significantly improving the efficiency of screening, and on the other hand, providing more interpretable basis for the entire screening process; and the present invention obtains more modal information of the candidate molecules (molecular global information, molecular pharmacophore information, and atomic local information of the molecule) through the multi-view structure of the molecular feature map, so that the molecular characterization data determined based on the molecular feature map can better characterize the information of the candidate molecules, improve the generalization ability of the method and the accuracy of screening the properties of the candidate molecules. Therefore, the present invention can take into account higher screening efficiency, as well as good versatility and accuracy.

[0076] The following will be combined Figure 5 A method for constructing a molecular signature map according to an embodiment of the present invention is described. Figure 5 FIG. 5 shows a flow chart of a method 500 for constructing a molecular feature map according to an embodiment of the present invention. It should be understood that the method 500 can be implemented, for example, in Fig.10 The electronic device 1000 described above may also be used. Figure 1 The method 500 is executed at the described computing device 100. It should be understood that the method 500 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0077] At step 502, the computing device 100 uses the molecules of the candidate molecule, each functional group in the candidate molecule, each atom in the two-dimensional chemical structure of the candidate molecule, and each atom in the three-dimensional atomic spatial conformation of the candidate molecule as nodes in the molecular feature graph, so as to construct a node set. It should be understood that by simultaneously constructing the atoms of the two-dimensional view and the atomic nodes of the three-dimensional view, the present invention can more accurately and stably identify the properties of the candidate molecule.

[0078] For example, Figure 3 As shown, the computing device 100 constructs a molecular node with the entire molecule of each candidate molecule (for example, as shown by label 310); constructs each of all functional groups in the candidate molecule into a functional group node (for example, as shown by label 312); constructs the atoms of the two-dimensional chemical structure of the candidate molecule into two-dimensional graph atomic nodes (for example, as shown by label 318); and constructs the atoms of the atomic spatial conformation of the candidate molecule into three-dimensional graph atomic nodes (for example, as shown by label 324).

[0079] At step 504 , the computing device 100 associates each node in the set of nodes with a mapping function of a corresponding node type in the set of node types.

[0080] It should be understood that each node in the molecular feature graph belongs to a node set. That is, v∈V (where v represents a node and V represents a node set). Each node in the molecular feature graph is associated with its corresponding node type mapping function φ(v). That is, (Where V represents the node set, Represents a collection of node types).

[0081] At step 506 , the computing device 100 constructs connection relationships between molecular nodes, functional group nodes, two-dimensional graph atom nodes, and three-dimensional graph atom nodes by means of edges to form an edge set.

[0082] Regarding the connection method between the functional group nodes (ie, P) and the molecular nodes (ie, M), it includes, for example: the computing device 100 connects all the functional group nodes (ie, P) and the molecular nodes (ie, M) in the form of edges.

[0083] Regarding the connection mode between functional group nodes, for example, it includes: if the computing device 100 determines that there are atoms connected by chemical bonds between any two functional groups, then the two functional group nodes are connected by an edge.

[0084] Regarding the connection method between the two-dimensional graph atomic nodes and the functional group nodes, it includes, for example: the computing device 100 connects the two-dimensional graph atomic nodes and the functional group nodes to which they belong via edges.

[0085] Regarding the connection method between the two-dimensional graph atomic nodes, it includes, for example: if there are chemical bonds between the atoms corresponding to the two-dimensional graph atomic nodes of the computing device 100, the two-dimensional graph atomic nodes are connected through edges (for example, as shown by mark 320).

[0086] Regarding the connection method of the three-dimensional graph atomic nodes and functional group nodes, for example, it includes: the computing device 100 connects the three-dimensional graph atomic nodes and the corresponding functional group nodes through edges. Figure 3 As shown, if the functional group is connected to the atom, the corresponding functional group node and the two-dimensional graph atom node or the three-dimensional graph atom node will be connected by an edge (for example, as shown by marks 322 and 328).

[0087] Regarding the connection mode between the three-dimensional graph atomic nodes, for example, it includes: if the computing device 100 determines that the spatial distance between the atoms corresponding to the three-dimensional graph atomic nodes is less than a set threshold, then the corresponding three-dimensional graph atomic nodes are connected through edges. Figure 3 As shown, if the distance between corresponding atoms of the three-dimensional graph atomic nodes is less than a set threshold (for example but not limited to 4), the three-dimensional graph atomic nodes of the corresponding atoms will be connected in the form of edges (for example, as shown by mark 326).

[0088] At step 508 , the computing device 100 associates each edge in the set of edges with a mapping function of a corresponding edge type in the set of edge types.

[0089] Each edge in the molecular feature graph belongs to an edge set. That is, e∈E (where e represents an edge and E represents an edge set). Each edge in the molecular feature graph is also associated with the edge type mapping function ψ(e) to which it belongs. That is, (Where E represents the edge set, represents a set of edge types). Fig.15 Schematic diagrams of molecular feature maps constructed according to other embodiments of the present invention are shown. Fig.15 Shown is a schematic diagram of the constructed molecular feature map constructed for the candidate molecule "COc1cc(O)cc(O)c1C(=O)c1ccccc1".

[0090] In the above scheme, the molecular characteristic map constructed by the present invention includes a multi-view model, which not only contains the global information of the molecule, the pharmacophore information of the molecule, and the atomic information of the molecule, but also the atomic information includes both the structural atomic view and the spatial atomic view. Therefore, the constructed molecular characteristic map can obtain more levels of information about the molecule and can better characterize the molecular information.

[0091] The following will be combined Figure 6 and Figure 7 A method for determining molecular characterization data of a candidate molecule according to an embodiment of the present invention is described. Figure 6 A flow chart of a method 600 for determining molecular characterization data of a candidate molecule according to an embodiment of the present invention is shown. Figure 7 Schematic diagram of a method for determining molecular characterization data of a candidate molecule according to an embodiment of the present invention is shown. It should be understood that method 600 can be used, for example, in Fig.10 The electronic device 1000 described above may also be used. Figure 1 The method 600 is executed at the described computing device 100. It should be understood that the method 600 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0092] At step 602, the computing device 100 initializes each node in the molecular feature graph into a vector to generate an initial feature of the node, and the dimensions of the initial features of different types of nodes are different.

[0093] For example, the computing device 100 performs a multi-step process for each node v in the candidate molecular feature graph. i (1≤i≤N n ), initialized to a corresponding node vector x i , in order to generate the initial features of the nodes. Represents the initial features of the node. N n Represents the number of all types of nodes in the candidate molecule. k Represents the initialization dimensions of different types of nodes. The following Table 1 schematically shows the initial feature dimensions of different types of nodes.

[0094] Table 1

[0095]

[0096]

[0097] Regarding the method of generating initial features of a node, for example, it includes: the computing device 100 constructs corresponding node vectors using different patterns for different types of nodes.

[0098] Specifically, in some embodiments, for a molecule node (M), the computing device 100 is initialized using, for example, a pre-trained model MOLBERT to form a corresponding node vector. It should be understood that the MOLBERT pre-trained model uses the Bert model architecture, utilizes the SMILES sequence information of the molecule, and learns the global representation of the molecule by predicting the masked atoms or other characters in the molecule sequence.

[0099] For the functional group node (P), the computing device 100 is initialized, for example, using the RDKit software, and is represented by molecular fingerprints and one-hot encoding. It should be understood that a molecular fragment is a group of connected atoms and may contain related functional groups. A series of tools for analyzing and manipulating molecular fragments are provided in the RDKit software. For example, to obtain the functional group library: "FunctionalGroups.txt" in the RDConfig.RDDataDir directory, a parameterizer can be instantiated according to the functional group library: FragmentCatalog.FragCatParams().

[0100] For the two-dimensional graph atomic node (A1), the computing device 100 constructs a corresponding node vector based on, for example, multiple atomic information (atom type, atomic degree, atomic number, atomic formal charge, chirality, hybridization, whether there is aromatic substance, mass, number of connected H atoms, etc.), and all types of atomic information are encoded using one-hot encoding, for example. The following Table 2 schematically shows the feature dimensions corresponding to different atomic information.

[0101] Table 2

[0102]

[0103]

[0104] For the three-dimensional graph atomic node (A2), the computing device 100 generates the corresponding node vector using the Uni-Mol model, for example. It should be understood that the node vector corresponding to the three-dimensional graph atomic node generated by the Uni-Mo model contains not only the information of the atom in the molecular sequence, but also the distribution information of the atom in the molecular space.

[0105] At step 604 , the computing device 100 maps initial features of different nodes into the same dimension.

[0106] For example, the computing device 100 uses different multilayer perceptron (MLP) structures to Figure 7 The dimensions of the corresponding node initial features of different types of nodes (molecule nodes, functional group nodes, two-dimensional graph atom nodes and three-dimensional graph atom nodes) in the molecular feature graph 710 of the candidate molecule are mapped to the same dimension.

[0107] At step 606 , the computing device 100 performs message passing and aggregation on the initial features of the mapped nodes to obtain molecular characterization data.

[0108] In some embodiments, the computing device 100 uses, for example, a Heterogeneous Graph Transformer (HGT) model to perform message passing and aggregation on the initial features of the mapped nodes to obtain molecular characterization data. Figure 7 The mark 720 in schematically shows the message passing and aggregation process for the initial features of each node. It should be understood that the molecular feature graph is a heterogeneous graph. The heterogeneous mutual attention of the HGT model is defined by decomposing each edge e=(s, t) based on the meta-relation triple. The following formula (2) schematically shows the function expression of the edge. Formula (3) schematically shows the expression of the meta-relation of the edge.

[0109] e=(s,t) (2)

[0110] <φ(s),ψ(e),φ(t)> (3)

[0111] In the above formulas (2) and (3), s and t represent two nodes in the heterogeneous graph, respectively.

[0112] φ(s) and φ(t) represent the node type mappings of two nodes s and t, respectively. ψ(e) represents the edge type mapping of edge e. As described above, there are 4 node type mappings, corresponding to M, P, A1, and A2, respectively. There are 6 edge type mappings, corresponding to MP, PP,

[0113] P-A1, P-A2, A1-A1, A2-A2.

[0114] It should be understood that the Heterogeneous Graph Transformer (HGT) model includes three processes, namely: a heterogeneous mutual attention process, a heterogeneous message passing process, and an aggregation (or "aggregation") update process. In some embodiments, obtaining molecular characterization data includes: the computing device 100 calculates the attention coefficient of the current node in the molecular feature graph and other nodes connected to the current node; aggregating the features of the neighboring nodes of the current node to update the features of the current node; and determining the molecular characterization data about the candidate molecule based on the updated node features.

[0115] Table 3 below schematically shows the main parameters of the model.

[0116] Table 3

[0117]

[0118] The following will combine Figure 8 A specific embodiment of the method for obtaining molecular characterization data via node message transmission and aggregation is specifically described, which will not be repeated here.

[0119] It should be understood that in addition to using the Transformer (HGT) model to perform message passing and aggregation on the initial features of the mapped nodes in order to obtain the implementation method of molecular characterization data. The present invention can also replace the method of the Heterogeneous Graph Transformer (HGT) model by algorithms such as HEAT and HAN. It should be understood that the algorithm HEAT uses a multi-channel trajectory prediction structure of a three-channel framework, taking into account the individual dynamics, interactions, and edge structures and features of each edge of the nodes. The algorithm HAN uses different meta-paths to convert heterogeneous graphs into multiple isomorphic sub-graphs, learns node-level and semantic-level features based on the attention mechanism in different word graphs, and finally weights each meta-path sub-graph vector to obtain a global representation.

[0120] Fig. 9 Schematic diagram of a method 900 for determining molecular characterization data of a candidate molecule according to some other embodiments of the present invention is shown. In some other embodiments, Fig. 9 As shown, the present invention can also use a homogeneous graph algorithm for the atomic layer in the heterogeneous graph of the molecular feature graph, for example, the graph convolutional network (Graph Convolutional Networks, GCN) model 910 de-aggregates the two-dimensional graph atomic information and the three-dimensional graph atomic information, and then, the atomic information 912 after aggregation is input into the heterogeneous graph network, and the edges (A1-A1) between the two-dimensional graph atomic nodes and the edges (A2-A2) between the three-dimensional graph atomic nodes in the heterogeneous graph network are removed. Afterwards, the initial features after removing the atomic edges are message-transmitted and aggregated (as indicated by mark 920) to obtain molecular characterization data. It should be understood that if the local fragment information of the molecule is more useful, then by using GCN to aggregate the atomic view in advance, it is possible to obtain local fragment information that is different from the functional group feature, and this information is then transmitted to the functional group via the heterogeneous graph, which may more effectively characterize the molecular characteristics.

[0121] The following will be combined Figure 7 and Figure 8 A method for obtaining molecular characterization data via node messaging and aggregation according to an embodiment of the present invention is described. Figure 8 800 is a flowchart of a method 800 for obtaining molecular characterization data via node message transmission and aggregation according to an embodiment of the present invention. It should be understood that the method 800 can be implemented, for example, in Fig.10 The electronic device 1000 described above may also be used. Figure 1 The method 800 is executed at the described computing device 100. It should be understood that the method 800 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0122] At step 802, the computing device 100 calculates the attention coefficients of the current node and other nodes connected to the current node in the molecular feature graph. The following formulas (4) to (7) schematically illustrate an algorithm for calculating heterogeneous mutual attention coefficients.

[0123]

[0124] In the above formulas (4) to (7), Attention(s,e,t) represents the heterogeneous attention coefficient for (s,e,t). Where s and t represent two nodes in the heterogeneous graph. e represents the edge connecting the two nodes s and t. N(t) represents all other nodes connected to node t. i represents the i-th head in the multi-head attention, and the value range of i is [1,h]. Hl-1 Represents the characteristics of the node at the l-1 layer. l-1 [s] represents the feature of node s at layer l-1. l-1 [t] represents the feature of node t at layer l-1. i K(t) represents the feature of node t after being mapped by the linear mapping parameters corresponding to the node type of the i-th head. i (s) represents the features of node s after being mapped by the linear mapping parameters corresponding to the node type of the i-th head. Indicates the linear mapping parameters corresponding to the node type s of the i-th head. φ(s) represents the linear mapping corresponding to the node type of node s. φ(t) represents the linear mapping corresponding to the node type of node t. ψ(e) represents the mapping of the edge type of edge e. μ represents the coefficient corresponding to different edge types, and m is a constant. ATT-head i (s, e, t) represents the attention parameters of the i-th head. The above formula (6) indicates the use of parameters The feature H of node s at layer l-1 is l-1 [s] is mapped to K i (s). Formula (7) indicates the use of parameters The feature H of node t at layer l-1 is l-1 [t] is mapped to Q i (t). Represents the edge mapping parameter of edge e. For the same type of edge, different attention heads share this parameter.

[0125] At step 804, the computing device 100 aggregates features of neighboring nodes of the current node to update the features of the current node. Figure 7 As shown, the features of the neighboring nodes around the functional group node 724, i.e., the features of the molecule node 722, the two-dimensional graph atom node 726, and the three-dimensional graph atom node 728, are aggregated to update the features of the aggregated functional group node 724. It should be understood that the computing device 100 updates the features of each node using the above-mentioned aggregation and updating method.

[0126] Regarding the method of aggregating the features of the neighboring nodes of the current node for updating the current features, the following formulas (8) to (9) schematically illustrate the algorithm for aggregating the features of the neighboring nodes of any node (i.e., the message passing process). It should be understood that the message passing includes a multi-head linear mapping process.

[0127]

[0128] In the above formulas (8) to (9), and They represent linear mappings respectively. S,t represents nodes, e represents edges. h represents the number of multi-head attention heads. l-1 [s] represents the feature of node s at layer l-1.

[0129] The following formulas (10) to (11) schematically illustrate the algorithm of the method for updating the characteristics of any node (i.e., the aggregation process). Formula (10) represents the algorithm for aggregating the information of different neighboring nodes of node t. Formula (11) represents the algorithm for updating the characteristics of node t using the aggregated information.

[0130]

[0131] In the above formulas (10) to (11), WA φ(t) represents the linear mapping parameter, σ represents the activation function, s, t represents the node, and e represents the edge. l-1 [t] represents the feature of node t at layer l-1. Represents the features of node t at layer l.

[0132] At step 806 , the computing device 100 determines molecular characterization data about the candidate molecule based on the updated features of the node.

[0133] For example, after the computing device 100 obtains the features of each node after the update, the features of the functional group node (i.e., P) are selected to generate molecular characterization data about the candidate molecule. This is because the functional group node, as an intermediate node connecting the molecular node (i.e., M) and different atomic nodes (e.g., the two-dimensional graph atomic node A1, the three-dimensional graph atomic node (i.e., A2), can better represent the molecular characterization of the candidate molecule, and at the same time, a large part of the candidate molecule is played by the pharmacophore. Therefore, the present invention selects the features of all functional group nodes to characterize the molecular characterization data of the final candidate molecule (e.g., obtain Figure 7 Molecular characterization data of the candidate molecule indicated by label 740 in FIG.

[0134] Specifically, the method for generating molecular characterization data about a candidate molecule based on the features of a functional group node (ie, P) includes, for example: the computing device 100 uses a maximum pooling operation to obtain the molecular characterization data of the candidate molecule.

[0135] The following formula (12) schematically shows an algorithm for generating molecular characterization data about a candidate molecule.

[0136]

[0137] In the above formula (12), c i Represents molecular characterization data C = {c1, c2, ..., c d}.p Represents the number of functional group nodes (ie, P).

[0138] By adopting the above method, the present invention can not only better learn the local features of nodes in the molecular feature graph, but also better learn the global features of molecules, so that the global representation of candidate molecules can be well obtained.

[0139] Fig.10 The block diagram of the electronic device 1000 suitable for implementing the embodiment of the present invention is schematically shown. The electronic device 1000 may be used to implement the execution Figure 2 , Figure 5 , Figure 6 , Figure 8 The methods 200, 500, 600 and 800 shown in FIG. Fig.10 As shown, the electronic device 1000 includes a central processing unit (i.e., CPU 1001), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (i.e., ROM 1002) or loaded from a storage unit 1008 to a random access memory (i.e., RAM 1003). In RAM 1003, various programs and data required for the operation of the electronic device 1000 can also be stored. CPU 1001, ROM 1002, and RAM 1003 are connected to each other via a bus 1004. An input / output interface (i.e., I / O interface 1008) is also connected to the bus 1004.

[0140] Multiple components in the electronic device 1000 are connected to the I / O interface 1008, including: an input unit 1006, an output unit 1008, and a storage unit 1008. The CPU 1001 performs the various methods and processes described above, such as executing methods 200, 500, 600, and 800. For example, in some embodiments, methods 200, 500, 600, and 800 may be implemented as a computer software program, which is stored in a machine-readable medium, such as the storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the CPU 1001, one or more operations of the methods 200, 500, 600, and 800 described above may be performed. Alternatively, in other embodiments, the CPU 1001 may be configured to perform one or more actions of methods 200 , 500 , 600 , and 800 in any other appropriate manner (eg, by means of firmware).

[0141] It should be further noted that the present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0142] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: 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), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0143] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0144] The computer program instructions for performing the operation of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present invention.

[0145] Here, various aspects of the present invention are described with reference to the flow chart and / or block diagram of the method, device (system), and computer program product according to an embodiment of the present invention. It should be understood that each square frame of the flow chart and / or block diagram and the combination of each square frame in the flow chart and / or block diagram can be implemented by computer-readable program instructions.

[0146] These computer-readable program instructions can be provided to a processor in a voice interaction device, a general-purpose computer, a special-purpose computer, or a processing unit of other programmable data processing devices, thereby producing a machine, so that when these instructions are executed by a processing unit of a computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions enable the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0147] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0148] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the equipment, method and computer program product according to multiple embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of this module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0149] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

[0150] The above are only optional embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for drug screening, characterized in that: include: For obtaining a plurality of chemical molecules in the chemical molecule library, screening is performed using a first deep learning model to obtain a plurality of candidate molecules that meet predetermined conditions; determining binding characterization data of each candidate molecule in a plurality of candidate molecules to a designated target, so as to select a candidate molecule in the plurality of candidate molecules based on the calculated binding characterization data; For the candidate molecule, construct a molecular feature map, wherein the molecular feature map at least indicates the features of the molecule, the functional group, and the atom of the candidate molecule; as well as Based on the constructed molecular feature graph, molecular characterization data of the candidate molecules are determined, so as to determine target molecules for drugs among the candidate molecules based on the molecular characterization data.

2. The method according to claim 1, characterized in that The molecular feature graph includes a node set, an edge set, a node type set and an edge type set. The node set includes: at least one of two-dimensional graph atomic nodes and three-dimensional graph atomic nodes, a molecular node, and a functional group node.

3. The method according to claim 2, characterized in that Based on the constructed molecular feature graph, the molecular characterization data of the candidate molecule is determined to include: Based on the constructed molecular feature graph, the relationship information of the nodes and edges of the molecular feature graph is extracted through a heterogeneous graph attention network model, so as to aggregate the extracted relationship information of the nodes and edges, thereby generating molecular characterization data about the candidate molecules.

4. The method according to claim 3, characterized in that Aggregating the extracted node and edge relationship information to generate molecular characterization data about the candidate molecules includes: The node and edge relationship information about the molecule node, the two-dimensional graph atom node, and the three-dimensional graph atom node is aggregated to the functional group node to generate the molecular characterization data about the candidate molecule.

5. The method according to claim 2, characterized in that: Building a molecular feature map for candidate molecules includes: The molecule of the candidate molecule, each functional group in the candidate molecule, each atom in the two-dimensional chemical structure of the candidate molecule, and each atom in the three-dimensional atomic spatial conformation of the candidate molecule are respectively used as nodes in the molecular feature graph to construct a node set; Associating each node in the node set with a mapping function of a corresponding node type in the node type set; By means of edges, connection relationships between molecular nodes, functional group nodes, two-dimensional graph atom nodes, and three-dimensional graph atom nodes are constructed to form an edge set; and Each edge in the edge set is associated with a mapping function of a corresponding edge type in the set of edge types.

6. The method according to claim 4, characterized in that Based on the constructed molecular feature graph, the molecular characterization data of the candidate molecule is determined to include: Initialize each node in the molecular feature graph into a vector to generate the initial features of the node. The dimensions of the initial features of different types of nodes are different. Mapping the initial features of different nodes into the same dimension; and Message passing and aggregation are performed on the initial features of the mapped nodes to obtain molecular characterization data.

7. The method according to claim 4, characterized in that Message passing and aggregation are performed on the initial features of the mapped nodes to obtain molecular characterization data including: Calculate the attention coefficients of the current node in the molecular feature graph and other nodes connected to the current node; Aggregating features of neighboring nodes of the current node to update the feature of the current node; and Based on the updated features of the nodes, molecular characterization data about the candidate molecules is determined.

8. The method according to claim 7, characterized in that Determining molecular characterization data about the candidate molecule based on the updated node features includes: selecting features of the clustered and updated functional group nodes; and The molecular characterization data about the candidate molecules are generated through a maximum pooling operation for the features of the clustered and updated functional group nodes.

9. The method according to claim 1, characterized in that: Selecting a candidate molecule from the plurality of candidates based on the calculated binding characterization data comprises: determining binding characterization data of each candidate molecule to a designated target, so as to rank the candidate molecules based on the calculated binding characterization data; and Based on the ranking results, a predetermined number of molecules to be selected that are ranked high are selected as candidate molecules.

10. The method according to claim 3, characterized in that: The identified target molecules are target molecules for anti-aging.

11. A computing device comprising: at least one processing unit; At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the device to perform the steps of the method according to any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 10 when executed by a machine.

13. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a machine, performs the method according to any one of claims 1 to 10.

Images