A drug molecule screening method and device based on a coherent Ising machine

Abstract

The application provides a drug molecule screening method and device, the method comprising: first, obtaining the chemical structure diagram corresponding to each of the candidate molecule and the target molecule. Then, an association graph is generated, which includes an association vertex set and an association edge set; the association vertex set is the Cartesian product of two atomic vertex sets in the two chemical structure diagrams, wherein any pair of association vertices connected by an association edge meets: among the four atomic vertices corresponding to the pair of association vertices, there is a chemical bond edge between each pair of atomic vertices from the same chemical structure diagram, or there is no chemical bond edge. Next, the solution space represented by the association graph is searched using a coherent Ising machine system to solve the ground state of Hamiltonian H. Finally, the most similar part of the candidate molecule and the target molecule is labeled according to the ground state of the Hamiltonian H, and the drug molecule is screened accordingly. This method not only improves the accuracy of drug molecule screening, but also improves the solving speed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of drug discovery, and in particular to a drug molecule screening method and device based on a coherent Ising machine. BACKGROUND

[0002] In the field of drug discovery, identifying the Maximum Common Subgraph (MCS) between drug candidate molecules and target molecules is a key task, which is of great significance for understanding drug action mechanisms and developing new drugs. The MCS problem aims to find the largest similar substructure between drug candidate molecules and target molecules by comparing their chemical structures. The MCS problem is a typical NP-hard problem, which means that as the size of the molecules increases, the computing resources and time required to find the best or near-optimal MCS will increase dramatically.

[0003] Existing solutions to the MCS problem include:

[0004] 1. Heuristic search algorithm:

[0005] The maximum common subgraph identification method based on heuristic search algorithm usually selects node pairs based on heuristic strategies such as node degree, but these strategies cannot well adapt to complex real molecular graph structures. Traditional heuristic search algorithms can quickly find an approximate solution when dealing with the MCS problem, but they usually cannot guarantee to find the global optimal solution, and are very sensitive to parameter settings, easily falling into local optima, and are difficult to adapt to complex graph structures.

[0006] 2. Learning search model based on Graph Neural Network (GNN):

[0007] Some research has proposed a learning search model based on GNN, which is based on the branch and bound algorithm and does not use traditional heuristic methods, but uses a Deep Q-Network (DQN) network based on GNN to select node pairs. This method uses supervision in the pre-training phase and guidance in the imitation learning phase to strengthen the training of DQN. This method can select node pairs through the DQN network, but may encounter the problem of early stopping search in the training process, leading to suboptimal solutions, and for large graph pairs, the search space is huge, and DQN may not be able to effectively explore all possible solutions. In addition, the training process of these models is complex, requiring supervision signals in the pre-training phase and guidance in the imitation learning phase.

[0008] 3. Heuristic algorithm:

[0009] Heuristic algorithms quickly find approximate solutions by using empirical rules or simplifying the problem. These algorithms can significantly reduce the solution time. Common heuristic algorithms include greedy algorithms, hill climbing algorithms, simulated annealing algorithms, etc. These methods can quickly find "good enough" solutions when dealing with NP-hard problems, but cannot guarantee to find the optimal solution.

[0010] In general, these methods have limitations in solving the MCS problem, often difficult to provide accurate MCS results within a reasonable time, especially when faced with large-scale or complex molecular structures. Therefore, the method for solving the MCS problem needs further research and improvement, especially in improving the global search ability of the algorithm, reducing the dependence on parameters, avoiding local optimization, and improving the adaptability and efficiency of the algorithm. SUMMARY

[0011] To solve the above problems, the present application provides a drug molecule screening method and device based on a coherent Ising machine.

[0012] According to the first aspect, a drug molecule screening method based on a coherent Ising machine is provided, comprising the following steps:

[0013] S1, obtaining the chemical structure graph corresponding to each of the candidate molecule and the target molecule; wherein each molecule includes a plurality of atoms and chemical bonds connecting the atoms; the chemical structure graph includes an atom vertex set and a chemical bond edge set.

[0014] S2, generating an association graph according to the two obtained chemical structure graphs; the association graph includes an association vertex set and an association edge set; the association vertex set is the Cartesian product between the two atom vertex sets corresponding to the two chemical structure graphs; wherein any pair of association vertices connected by an association edge meets: among the four atom vertices corresponding to the pair of association vertices, there is a chemical bond edge between each pair of atom vertices from the same chemical structure graph, or there is no chemical bond edge.

[0015] S3, using a coherent Ising machine system to search the solution space represented by the association graph, and solving the ground state of the Hamiltonian H, wherein the Hamiltonian H is:

[0016]

[0017] wherein, the subscripts and respectively denote the vertices and in the chemical structure graph of the candidate molecule, and the subscripts and respectively denote the vertices and in the chemical structure graph of the target molecule, represents the coupling between the vertices with subscripts whether the corresponding associated vertex is selected, the weight of , representing whether the corresponding associated vertex with subscript is selected, representing whether the corresponding associated vertex with subscript and the corresponding associated vertex with subscript has an associated edge, K1 is the coefficient of

[0018] S4, according to the ground state of the Hamiltonian H, marking the maximum similar part of the candidate molecule and the target molecule, and screening the drug molecules accordingly.

[0019] In some embodiments, the Hamiltonian H is solved by searching the solution space of the associated graph representation using a coherent Ising machine system, specifically including:

[0020] According to the Hamiltonian H, the coherent Ising machine is initialized, and the coherent Ising machine system evolution process is started.

[0021] During the evolution process of the coherent Ising machine system, the phase and intensity of the optical field in the coherent Ising machine system are measured multiple times, and the parameters of the coherent Ising machine system are adjusted in real time according to the measurement results, and finally a collective oscillation mode far higher than the threshold value is obtained, and the quantum spin state under this mode corresponds to the ground state of the Hamiltonian H.

[0022] In some embodiments, the coherent Ising machine system includes:

[0023] An optical parametric oscillator network is used to simulate the spin interaction in the Ising model.

[0024] A phase-sensitive amplifier is used to amplify signals of a specific phase.

[0025] A field programmable gate array is used to control the optical parametric oscillator network.

[0026] A phase / intensity measurer is used to measure the phase and intensity of the optical field in the cavity.

[0027] An optical modulator is used to change the phase difference between light beams, thereby simulating the interaction strength between different spin states.

[0028] A beam splitter is used to generate an interference effect to simulate the connection relationship between spins in the Ising model.

[0029] An optical fiber is used as a transmission medium for optical signals.

[0030] In some embodiments, the candidate molecules are from a molecule library; the target molecule is a biological molecule acted on by a drug, which can be a protein, DNA, hormone.

[0031] In some more specific embodiments, the screening of the drug molecules specifically includes: traversing all candidate molecules in the molecule library, marking the maximum similar part of all candidate molecules and the target molecule, and screening the drug molecules with the highest similarity to the target molecule and the similar part having a drug active functional group.

[0032] According to a second aspect, the present application also provides a drug molecule screening device, comprising:

[0033] An acquisition module is configured to acquire respective chemical structure graphs of the candidate molecules and the target molecule; each molecule includes a plurality of atoms and chemical bonds connecting the atoms; and the chemical structure graph includes an atom vertex set and a chemical bond edge set.

[0034] A graph generation module is configured to generate an association graph according to the two acquired chemical structure graphs; the association graph includes an association vertex set and an association edge set; the association vertex set is a Cartesian product between the two atom vertex sets corresponding to the two chemical structure graphs; and any pair of association vertices connected by an association edge satisfies: among the four atom vertices corresponding to the pair of association vertices, there is a chemical bond edge between each pair of atom vertices from the same chemical structure graph, or there is no chemical bond edge between each pair of atom vertices from the same chemical structure graph.

[0035] A solving module is configured to search the solution space represented by the association graph using a coherent Ising machine system, and solve the ground state of a Hamiltonian H, where the Hamiltonian H is:

[0036]

[0037] wherein, the subscripts and respectively denote the vertices and in the chemical structure graph of the candidate molecule, the subscripts and respectively denote the vertices and in the chemical structure graph of the target molecule, represents whether the association vertex corresponding to the subscript is selected, is the weight of , represents whether the association vertex corresponding to the subscript is selected, represents whether there is an association edge between the association vertex corresponding to the subscript and the association vertex corresponding to the subscript , and K1 is The coefficient.

[0038] The annotation module is used to annotate the maximum similarity between the candidate molecule and the target molecule based on the ground state of the Hamiltonian H.

[0039] The screening module is configured to screen drug molecules.

[0040] In some embodiments, the solving module is configured as follows:

[0041] The coherent Ising machine is initialized based on the Hamiltonian H, and the evolution process of the coherent Ising machine system is initiated.

[0042] During the evolution of the coherent Ising machine system, the phase and intensity of the optical field within the coherent Ising machine system are measured multiple times. Based on the measurement results, the parameters of the coherent Ising machine system are adjusted in real time, ultimately obtaining a collective oscillation mode that is much higher than the threshold. The quantum spin state in this mode corresponds to the ground state of the Hamiltonian H.

[0043] In some embodiments, the coherent Ising machine system includes:

[0044] An optical parametric oscillator network is used to simulate spin interactions in the Ising model.

[0045] A phase-sensitive amplifier is used to amplify signals of a specific phase.

[0046] Field-programmable gate arrays (FPGAs) are used to control optical parametric oscillator networks.

[0047] A phase / intensity measuring device is used to measure the phase and intensity of the optical field within the cavity.

[0048] An optical modulator is used to change the phase difference between beams, thereby simulating the interaction strength between different spin states.

[0049] A beam splitter is used to generate interference effects, simulating the connection relationship between spins in the Ising model.

[0050] Optical fiber is used as a medium for transmitting optical signals.

[0051] In some embodiments, the candidate molecules are derived from a molecular library; the target molecule is a biomolecule to which the drug acts.

[0052] In some more specific embodiments, the screening module is specifically configured to: traverse all candidate molecules in the molecular library, mark the maximum similarity between all candidate molecules and the target molecule, and thereby screen out the drug molecule that has the highest similarity to the target molecule and whose similarity has pharmacologically active functional groups.

[0053] The drug molecule screening method provided by the application can systematically search all possible matching schemes, and ensures that the found solution is globally optimal. In addition, the drug molecule screening method provided by the application is based on a coherent Ising machine, and compared with a traditional drug molecule MCS algorithm, the coherent Ising machine can more quickly explore a solution space and improve solving speed. The method not only improves the accuracy of drug molecule screening, but also greatly enhances the ability to quickly identify drug molecules with expected biological activity from massive data. By focusing on the identification of MCS, the challenge of drug screening can be converted into a more specific and easier-to-compute problem, thereby opening up an efficient and innovative path for the discovery of new drugs. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0055] Figure 1 A drug molecule screening method based on a coherent Ising machine is shown.

[0056] Figure 2 A coherent Ising machine system structure is shown.

[0057] Figure 3 A protein drug molecule screening device is shown. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the embodiments of the application more clear, the technical solutions in the embodiments of the application will be described below with reference to the drawings.

[0059] A coherent Ising machine (CIM) is a computing platform based on quantum optics simulation of Ising model, which has the advantages of fast computing speed and strong scalability, and is suitable for processing various large-scale combinatorial optimization problems. The unique parallelism of the coherent Ising machine enables it to simultaneously explore multiple possible solutions in coherent evolution, providing potential advantages for solving NP-hard problems. When processing these problems, the CIM can ensure that the found solution is globally optimal while maintaining the efficiency of the calculation and the accuracy of the solution.

[0060] Therefore, the drug screening method based on the coherent Ising machine (CIM) is expected to significantly improve the computational efficiency of the MCS search, and to process larger-scale molecular data sets while maintaining high accuracy. The present application aims to take advantage of the computational advantages of CIM to provide an efficient solution to the MCS problem, with a view to achieving breakthrough progress in the field of drug discovery.

[0061] The drug screening method provided by the present application will be described below with reference to the accompanying drawings. The flowchart of the method is shown in Figure 1 The method comprises the following steps:

[0062] S1, obtaining respective chemical structure graphs of a candidate molecule and a target molecule; wherein each molecule comprises a plurality of atoms and chemical bonds connecting the atoms; the chemical structure graph comprises an atom vertex set and a chemical bond edge set.

[0063] In some embodiments, the candidate molecule is from a molecular library; and the target molecule is a biological molecule acted on by a drug.

[0064] The candidate molecule refers to a molecule considered for further development as a drug, which is from a molecular library, and can be a small molecule compound, a natural product, a biological agent, etc. The target molecule refers to the object of the drug, which is usually a biological molecule involved in the disease process, and can be a protein, DNA, hormone, etc.

[0065] The chemical structure graph provides detailed information about the spatial structure of the two molecules and the interaction between the internal atoms in the form of a graph, and after converting the complex molecular structure into a chemical structure graph, graph theory can be used to identify and compare the similarity between molecules. In the chemical structure graph, the atom vertex set includes vertices representing atoms, and the chemical bond edge set includes edges representing chemical bonds.

[0066] S2, generating an association graph according to the two obtained chemical structure graphs; the association graph comprises an association vertex set and an association edge set; the association vertex set is the Cartesian product between the two atom vertex sets corresponding to the two chemical structure graphs; wherein a pair of association vertices connected by any association edge meet: in the four atom vertices corresponding to the pair of association vertices, there is a chemical bond edge between each pair of atom vertices from the same chemical structure graph, or there is no chemical bond edge between each pair of atom vertices from the same chemical structure graph.

[0067] The generation process of the association graph G will be explained below with reference to an example:

[0068] Let the input two chemical structure graphs be G1 = (V1, E1) and G2 = (V2, E2), where V1, V2 represent the atom vertex set, and E1, E2 represent the edge chemical bond set. The association graph G = (V, E) is constructed based on G1 and G2.

[0069] The associated vertex set V of the associated graph G is the Cartesian product of V1 and V2, V1 × V2, that is, each associated vertex in V is a combination of atomic vertices in V1 and V2, and V is all possible combinations of atomic vertices in V1 and V2.

[0070] In the associated graph, if one of the following conditions is met, an associated edge is added between the two associated vertices (u1, v1) and (u2, v2):

[0071] Condition one: u1 and u2 are adjacent in G1 (i.e., (u1, u2) ∈ E1), and v1 and v2 are also adjacent in G2 (i.e., (v1, v2) ∈ E2).

[0072] Condition two: u1 and u2 are not adjacent in G1 (i.e., (u1, u2) ∉ E1), and v1 and v2 are also not adjacent in G2 (i.e., (v1, v2) ∉ E2).

[0073] Thus, the construction of the associated graph is completed, and the MCS problem is converted into a graph theory problem, and the maximum clique of the associated graph represents the largest similar part in the two molecular structures. Next, by finding the maximum clique in the associated graph, the maximum common subgraph between the two molecular structures can be determined, which can be achieved by constructing the Hamiltonian H of the associated graph G and solving the ground state of the Hamiltonian H using CIM. The specific steps are as follows:

[0074] S3, using a coherent Ising machine system to search the solution space represented by the associated graph and solve the ground state of the Hamiltonian H, wherein the Hamiltonian H is:

[0075] (1)

[0076] wherein the meanings and possible values of the parameters are as follows:

[0077] Subscripts and refer to atomic vertices and in the chemical structure graph of the candidate molecule, and subscripts and refer to atomic vertices and in the chemical structure graph of the target molecule.

[0078] represent whether the associated vertex corresponding to subscript is selected, and similarly, represent whether the associated vertex corresponding to subscript ​​Whether the corresponding associated vertex is selected, and whether being selected corresponds to the node spin state in the coherent Ising machine. For example, it can be set to +1 when the spin is selected, indicating that this associated vertex is selected.

[0079] For is a positive number, for example, it can be 1.

[0080] Representing whether there is an associated edge between the associated vertex corresponding to the subscript and the associated vertex corresponding to the subscript , for example, 1 when there is an associated edge, and 0 when there is no associated edge.

[0081] K1 is The coefficient of, which can be set to any positive number, such as 1.

[0082] In some embodiments, the solution space of the associated graph representation is searched using a coherent Ising machine system to solve the ground state of the Hamiltonian H, specifically including:

[0083] (1) According to the Hamiltonian H, initialize the coherent Ising machine, and start the coherent Ising machine system evolution process.

[0084] (2) During the evolution process of the coherent Ising machine system, the phase and intensity of the light field in the coherent Ising machine system are measured multiple times, and the parameters of the coherent Ising machine system are adjusted in real time according to the measurement results, and finally a collective oscillation mode much higher than the threshold value is obtained. The quantum spin state in this mode corresponds to the ground state of the Hamiltonian H.

[0085] From the above, the ground state of the Hamiltonian H can be obtained using the coherent Ising machine system.

[0086] S4, according to the ground state of the Hamiltonian H, mark the maximum similar part of the candidate molecule and the target molecule.

[0087] Specifically, the ground state of the Hamiltonian H, or the nodes in the maximum group, are inversely coded into the corresponding atomic vertices and chemical bond edges in the chemical structure graphs of the two molecules, that is, the maximum similar part of the two molecules, as the final result of the MCS problem.

[0088] S5, according to the maximum similar part, screening drug molecules.

[0089] In some embodiments, this step can be implemented as: according to the maximum similar part of all candidate molecules in the molecule library and the target molecule, screening out the drug molecules with the highest similarity to the target molecule and the similar part with the drug activity functional group.

[0090] The target of drug molecule screening is to identify potential candidate molecules similar to the target molecule structure from a large molecule library. The previous steps can identify the maximum common subgraph of the candidate molecule and the target molecule. Finally, the potential candidate molecules similar to the target molecule structure can be found by marking the maximum similar part of the candidate molecules in the molecule library and the target molecule one by one, and the drug molecule screening is completed.

[0091] The above is the protein side chain structure prediction method based on the coherent Ising machine provided by the application, and the coherent Ising machine system used in the method is briefly introduced below.

[0092] The coherent Ising machine system uses a doubly resonant optical parametric oscillator (DOPO) to realize artificial spins, and realizes the enhancement of light signals with specific phases by placing a phase sensitive amplifier (PSA) in the optical cavity. The PSA is an optical amplifier based on optical parametric amplification, which can effectively amplify the 0 and π phase components relative to the pump phase. Therefore, the DOPO only uses 0 or π phase above the oscillation threshold; therefore, the discrete phase state can be used to represent the Ising spin state. The interaction between DOPO pulses is realized using a measurement feedback technique, which repeatedly measures the feedback modulation process in the cavity, while increasing the pump amplitude from 0, and finally obtains a "strongest" collective oscillation mode much higher than the threshold, which corresponds to the best solution of the given Ising problem.

[0093] The schematic diagram of the coherent Ising machine system is shown in Figure 2 , which includes:

[0094] (1) Doubly resonant optical parametric oscillator network, used to simulate the spin interaction in the Ising model.

[0095] (2) Phase sensitive amplifier, used to amplify signals with specific phases.

[0096] (3) Field-programmable gate array (FPGA), used to control the doubly resonant optical parametric oscillator network.

[0097] The FPGA can be programmed to configure its internal circuit, which can be used to adjust the parameters of the coherent Ising machine system in real time according to the measurement results during the evolution of the coherent Ising machine system.

[0098] (4) Phase / intensity measurer, used to measure the phase and intensity of the light field in the cavity.

[0099] (5) optical modulator, for changing the phase difference between the light beams, thereby simulating the interaction strength between different spin states.

[0100] (6) beam splitter, for generating interference effects, simulating the connection between spins in the Ising model.

[0101] (7) optical fiber, used as a transmission medium for optical signals.

[0102] As can be seen from the above, compared with the prior art, the drug molecule screening method provided by the application has the following beneficial effects:

[0103] 1. Guarantee of global optimal solution: The method can systematically search all possible matching solutions, ensuring that the solution found is globally optimal. This is particularly important in the NP-hard MCS problem, because many traditional heuristic algorithms can quickly find approximate solutions, but cannot guarantee the optimality of the solution.

[0104] 2. Fast and accurate solution: The method is implemented based on a coherent Ising machine system. The coherent Ising machine can explore the solution space more quickly due to the use of coherence effects, improving the solving speed, especially for large graph clique problems. The acceleration advantage of the Ising machine exceeds the field of complexity theory, providing a polynomial improvement in scaling or constant pre-factor advantage, which has a great impact on the running time of large problems. As the size of the system increases, the advantage of the method becomes more and more obvious, which is crucial for drug screening and molecular structure analysis.

[0105] 3. Efficient and innovative path: The application solves the MCS question through the coherent Ising machine, not only improving the accuracy of the screening process, but also greatly enhancing our ability to quickly identify molecules with expected biological activity from massive data. By focusing on the identification of MCS, the challenge of drug screening is transformed into a more specific and easier to calculate problem, thereby opening up an efficient and innovative path for the discovery of new drugs.

[0106] The application also provides a drug molecule screening device 300, a schematic diagram of which is shown in Figure 3 , comprising:

[0107] The acquisition module 301 is configured to acquire a chemical structure graph corresponding to each of the candidate molecules and the target molecules; each molecule comprises a plurality of atoms and chemical bonds connecting the atoms; the chemical structure graph comprises an atom vertex set and a chemical bond edge set.

[0108] The graph generation module 302 is configured to generate a correlation graph according to the two chemical structure graphs; the correlation graph comprises a correlation vertex set and a correlation edge set; the correlation vertex set is a Cartesian product of two atomic vertex sets corresponding to the two chemical structure graphs; any pair of correlation vertices connected by a correlation edge satisfy that, among four atomic vertices corresponding to the pair of correlation vertices, there is a chemical bond edge between each pair of atomic vertices from the same chemical structure graph, or there is no chemical bond edge between each pair of atomic vertices from the same chemical structure graph.

[0109] The solving module 303 is configured to search a solution space represented by the correlation graph by using a coherent Ising machine system, and solve a ground state of a Hamiltonian H, where the Hamiltonian H is:

[0110]

[0111] wherein subscripts and respectively denote vertices and in the chemical structure graph of the candidate molecule, subscripts and respectively denote vertices and in the chemical structure graph of the target molecule, represents whether a correlation vertex corresponding to subscript is selected, is a weight of , represents whether a correlation vertex corresponding to subscript is selected, represents whether there is a correlation edge between a correlation vertex corresponding to subscript and a correlation vertex corresponding to subscript , and K1 is a coefficient of .

[0112] The labeling module 304 is configured to label a most similar part of the candidate molecule and the target molecule according to the ground state of the Hamiltonian H.

[0113] The screening module 305 is configured to screen a drug molecule.

[0114] It should be noted that the above device can perform the aforementioned drug molecule screening method, and the functions of each module can be referred to the aforementioned description of the method, and will not be repeated here.

[0115] In the description of the embodiments of the present application, the words "exemplary", "for example", or "e.g." are used to mean "an example of" or "an example, only". Any embodiment or design solution described as "exemplary", "for example" or "e.g." in the embodiments of the present application should not be construed as preferred or superior over other embodiments or design solutions. In fact, the use of the words "exemplary", "for example" or "e.g." is intended to present related concepts in a specific manner.

[0116] In the description of the embodiments of the present application, the term "and / or" is only to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent three cases of A alone, B alone, and A and B together. In addition, unless otherwise specified, the term "a plurality of" means two or more.

[0117] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more features. The terms "include", "contain", "have" and their variants mean "include but are not limited to", unless otherwise specifically emphasized.

[0118] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made on the basis of the technical solutions of the present application should be included in the protection scope of the present application.

Claims

1. A method of drug molecule screening, characterized by, The method comprises the following steps: obtaining respective chemical structure diagrams of the candidate molecule and the target molecule; each molecule comprises a plurality of atoms and chemical bonds connecting the atoms; the chemical structure diagram comprises an atom vertex set and a chemical bond edge set; generating an association graph according to the two obtained chemical structure diagrams; the association graph comprises an association vertex set and an association edge set; the association vertex set is a Cartesian product between two atom vertex sets corresponding to the two chemical structure diagrams; any pair of association vertices connected by an association edge satisfies: among four atom vertices corresponding to the pair of association vertices, there is a chemical bond edge between each pair of atom vertices from the same chemical structure diagram, or there is no chemical bond edge between each pair of atom vertices from the same chemical structure diagram; using a coherent Ising machine system to search the solution space represented by the association graph and solve the ground state of a Hamiltonian H, wherein the Hamiltonian H is: wherein subscript and denote vertex and vertex in the chemical structure graph of the candidate molecule, respectively, subscript and denote vertex and vertex in the chemical structure graph of the target molecule, respectively, represents whether the associated vertex corresponding to subscript , is selected, is the weight of , represents whether the associated vertex corresponding to subscript , is selected, represents whether there is an associated edge between the associated vertex corresponding to subscript , and the associated vertex corresponding to subscript , , K1 is the coefficient of . According to the ground state of the Hamiltonian H, the most similar part of the candidate molecule and the target molecule is labeled, and a drug molecule is screened according to the most similar part.

2. The method of claim 1, wherein, The method of using a coherent Ising machine system to search the solution space represented by the association graph and solve the ground state of the Hamiltonian H specifically comprises: According to the Hamiltonian H, initializing the coherent Ising machine and starting the coherent Ising machine system evolution process; During the coherent Ising machine system evolution process, the phase and intensity of the optical field in the coherent Ising machine system are measured multiple times, and the coherent Ising machine system parameters are adjusted in real time according to the measurement results, and finally a collective oscillation mode much higher than the threshold value is obtained, and the quantum spin state in this mode corresponds to the ground state of the Hamiltonian H.

3. The method of claim 1, wherein, The coherent Ising machine system comprises: an optical parametric oscillator network for simulating spin interactions in the Ising model; a phase-sensitive amplifier for amplifying signals of a specific phase; a field programmable gate array for controlling the optical parametric oscillator network; a phase / intensity measurer for measuring the phase and intensity of the optical field in the cavity; an optical modulator for changing the phase difference between light beams to simulate the interaction strength between different spin states; a beam splitter for generating an interference effect to simulate the connection relationship between spins in the Ising model; optical fibers serving as transmission media for optical signals.

4. The method of claim 1, wherein, The candidate molecule is from a molecular library; the target molecule is a biological molecule acted on by a drug.

5. The method of claim 4, wherein the screening of the drug molecule specifically comprises: All candidate molecules in the molecular library are traversed, and the most similar part of each candidate molecule and the target molecule is labeled, and a drug molecule with the highest similarity to the target molecule and having a drug active functional group in the most similar part is screened.

6. A drug molecule screening device characterized by, The method comprises: an obtaining module for obtaining respective chemical structure diagrams of the candidate molecule and the target molecule; each molecule comprises a plurality of atoms and chemical bonds connecting the atoms; the chemical structure diagram comprises an atom vertex set and a chemical bond edge set; a graph generating module for generating an association graph according to the two obtained chemical structure diagrams; the association graph comprises an association vertex set and an association edge set; the association vertex set is a Cartesian product between two atom vertex sets corresponding to the two chemical structure diagrams; Any pair of associated vertices connected by any associated edge meets the following condition: among the four atomic vertices corresponding to the pair of associated vertices, there is a chemical bond edge between each pair of atomic vertices from the same chemical structure graph, or there is no chemical bond edge between each pair of atomic vertices from the same chemical structure graph; A solving module configured to search the solution space of the associated graph representation by using a coherent Ising machine system, and solve the ground state of a Hamiltonian H, the Hamiltonian H being: wherein subscript and denote vertices and vertices in the chemical structure graph of the candidate molecule, respectively, subscript and denote vertices and vertices in the chemical structure graph of the target molecule, respectively, represents whether the associated vertex corresponding to subscript , is selected, is the weight of , represents whether the associated vertex corresponding to subscript , is selected, represents whether there is an associated edge between the associated vertex corresponding to subscript , and the associated vertex corresponding to subscript , , K1 is the coefficient of . A labeling module configured to label the maximum similar part of the candidate molecule and the target molecule according to the ground state of the Hamiltonian H; A screening module configured to screen drug molecules.

7. The apparatus of claim 6, wherein, The solving module is specifically configured to: Initialize the coherent Ising machine according to the Hamiltonian H, and start the coherent Ising machine system evolution process; During the coherent Ising machine system evolution process, the phase and intensity of the light field in the coherent Ising machine system are measured multiple times, and the coherent Ising machine system parameters are adjusted in real time according to the measurement results, so that a collective oscillation mode far higher than the threshold value is finally obtained, and the quantum spin state under the mode corresponds to the ground state of the Hamiltonian H.

8. The apparatus of claim 6, wherein, The coherent Ising machine system includes: An optical parametric oscillator network configured to simulate the spin interaction in the Ising model; A phase-sensitive amplifier configured to amplify signals of a specific phase; A field programmable gate array configured to control the optical parametric oscillator network; A phase / intensity measurer configured to measure the phase and intensity of the light field in the cavity; An optical modulator configured to change the phase difference between light beams, thereby simulating the interaction strength between different spin states; A beam splitter configured to generate an interference effect to simulate the connection relationship between spins in the Ising model; An optical fiber configured to serve as a transmission medium for optical signals.

9. The apparatus of claim 6, wherein, The candidate molecule is from a molecular library; and the target molecule is a biological molecule acted on by a drug.

10. The apparatus of claim 9, wherein, The screening module is specifically configured to traverse all candidate molecules in the molecular library, label the maximum similar part of all candidate molecules and the target molecule, and screen drug molecules with the highest similarity to the target molecule and having a drug active functional group in the similar part.

Images