Molecular model construction device, method, and program

Abstract

According to the present invention, a binding candidate input means receives input of a binding location of each binding candidate that binds to a target molecule, which is a target of binding of a molecule, and a binding score indicating the strength of binding between the target molecule and each binding candidate, the binding location and the binding score being estimated on the basis of information about the target molecule. An optimization means, under a constraint condition in which the position indicated by the binding location is associated with a point that is passed through in a path optimization problem and a condition for combining the binding candidates is associated with a condition for connecting the points corresponding to the binding candidates in the path optimization problem, optimizes a molecular model to be constructed by solving, through quantum calculation, a path optimization problem that minimizes an objective function representing the sum of the binding scores corresponding to a plurality of points that are passed through. An output means outputs the optimized molecular model.

Description

Molecular model construction apparatus, method, and program

[0001] This disclosure relates to a molecular model construction apparatus, a molecular model construction method, and a molecular model construction program for constructing molecular models by combining molecules.

[0002] Many universities and companies are developing molecular designs with the aim of drug discovery and molecular sensing. Screening methods are generally used in the development of biomolecular elements such as nucleic acids and peptide aptamers to detect optimal compounds. High-throughput screening, an advanced version of screening methods, is also known (see Non-Patent Literature 1).

[0003] Yoshifumi Arai, "Chemistry and Biology," Chemistry and Biology, Taisho Pharmaceutical Co., Ltd. Drug Discovery Research Institute, Vol. 38, No. 4, pp. 264-269, 2000.

[0004] Screening methods are, in a sense, brute-force approaches, requiring significant time and cost for development. While designing target molecules from 3D structures is also a possibility, this method involves what is known as the folding problem, which also requires considerable time and cost to resolve.

[0005] Therefore, the purpose of this disclosure is to provide a molecular model construction device, a molecular model construction method, and a molecular model construction program that can rapidly construct molecular models by combining molecules.

[0006] The molecular model construction apparatus according to this disclosure is a molecular model construction apparatus for constructing a molecular model by combining molecules, comprising: a binding candidate input means that accepts input of binding sites for each binding candidate that bind to the target molecule, estimated based on information of the target molecule which is the target molecule to which the molecules are bound, and a binding score indicating the strength of the binding between the target molecule and each binding candidate; an optimization means that optimizes the molecular model constructed by each binding candidate using quantum computation; and an output means that outputs the optimized molecular model, wherein the optimization means optimizes the constructed molecular model by solving a path optimization problem using quantum computation that minimizes an objective function representing the sum of binding scores corresponding to the points traversed when passing through multiple points, under constraints that correspond the positions indicated by the binding sites to points traversed in a path optimization problem, and the conditions for combining the binding candidates to conditions for connecting the points corresponding to the binding candidates in a path optimization problem.

[0007] The molecular model construction method according to this disclosure is a molecular model construction method for constructing a molecular model by combining molecules, characterized in that it accepts inputs of the binding sites of each binding candidate that bind to the target molecule, estimated based on information of the target molecule to which the molecules bind, and binding scores indicating the strength of the binding between the target molecule and each binding candidate, the positions indicated by the binding sites are associated with points traversed in a path optimization problem, and the conditions for combining the binding candidates are associated with conditions for connecting the points corresponding to the binding candidates in a path optimization problem, and the path optimization problem is solved by quantum computation to minimize an objective function that represents the sum of the binding scores corresponding to the points traversed when passing through multiple points, thereby optimizing the molecular model constructed by each binding candidate and outputting the optimized molecular model.

[0008] The molecular model construction program described herein is a molecular model construction program applied to a computer that constructs molecular models by combining molecules, and is characterized in that it causes the computer to perform a binding candidate input process that accepts input of binding sites for each binding candidate that binds to the target, estimated based on information of the target molecule which is the target to which the molecules bind, and a binding score indicating the strength of the binding between the target molecule and each binding candidate; an optimization process that optimizes the molecular model constructed by each binding candidate using quantum computation; and an output process that outputs the optimized molecular model, and in the optimization process, the molecular model to be constructed is optimized by solving a path optimization problem using quantum computation, under constraints that correspond the positions indicated by the binding sites to points traversed in a path optimization problem, and the conditions for combining binding candidates to conditions for connecting the points corresponding to the binding candidates in a path optimization problem, thereby minimizing an objective function that represents the sum of the binding scores corresponding to the points traversed when passing through multiple points.

[0009] According to this disclosure, molecular models can be constructed at high speed by combining molecules.

[0010] This is a block diagram showing an example configuration of one embodiment of the molecular model construction system of this disclosure. This is an explanatory diagram showing an example of defining information used in quantum computing. This is an explanatory diagram showing an example of the process of calculating the sum of binding scores. This is an explanatory diagram showing an example of a state represented by the first constraint condition. This is an explanatory diagram showing an example of a state represented by the second constraint condition. This is an explanatory diagram showing an example of a state represented by the third constraint condition. This is an explanatory diagram showing an example of a state represented by the fourth constraint condition. This is an explanatory diagram showing an example of the process of setting according to the distance between binding candidates. This is a flowchart showing an example of operation of the molecular model construction system. This is a block diagram showing a modified example of the molecular model construction system of this disclosure. This is an explanatory diagram showing another example of the process of calculating the sum of binding scores. This is a block diagram showing an overview of the molecular model construction apparatus according to this disclosure.

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] FIG. 1 is a block diagram showing a configuration example of an embodiment of a molecular model construction system of the present disclosure. The molecular model construction system 1 of the present disclosure includes a molecular model construction device 100 and a quantum annealing machine 200.

[0013] Further, the molecular model construction device 100 includes a binding site estimation unit 10, a storage unit 20, a binding candidate input unit 30, an optimization unit 40, and an output unit 50.

[0014] The storage unit 20 stores various information used by the molecular model construction system 1 of the present embodiment for processing. The storage unit 20 is realized by, for example, a magnetic disk or the like.

[0015] The binding site estimation unit 10 estimates the binding site of a candidate for a molecule (hereinafter referred to as a target molecule) to which a molecule is to be bound (hereinafter referred to as a binding candidate). Further, the binding site estimation unit 10 also estimates the strength (hereinafter referred to as a binding score) with which the binding candidate binds to the target molecule.

[0016] The binding site may be represented, for example, by a relative position with respect to the target molecule, or may be represented by an absolute position including the target molecule. Also, as long as the binding score is a value represented in the same unit system through processing, its content is arbitrary. In the present embodiment, it is assumed that the smaller the binding score, the higher the binding force.

[0017] The method by which the binding site estimation unit 10 estimates the binding site and the binding score is arbitrary. The binding site estimation unit 10 may use, for example, a docking simulation that computationally estimates the stable structure of a complex in the interaction between a low molecular weight compound and a biopolymer on a computer.

[0018] In this case, the binding site estimation unit 10 may estimate the binding site, for example, using a biopolymer such as thrombin as a target molecule and a low molecular weight compound such as a mononucleotide (A: adenine, G: guanine, C: cytosine, T: thymine) as a binding candidate. In the following description, a nucleic acid is exemplified as a binding candidate. However, the binding candidate is not limited to a nucleic acid, and may be a molecular compound such as an amino acid.

[0019] The connection location estimation unit 10 may store the estimated connection locations and connection scores in the storage unit 20, or it may input them to the connection candidate input unit 30, which will be described later.

[0020] The binding candidate input unit 30 accepts input of binding sites and binding scores for binding candidates to the target molecule. The binding candidate input unit 30 may accept input of binding sites and binding scores stored in the storage unit 20, or it may accept input of binding sites and binding scores estimated by the binding site estimation unit 10.

[0021] The optimization unit 40 optimizes the molecular model formed by each binding candidate. Here, the molecular model is, for example, a base sequence.

[0022] In this regard, the inventors conceived the idea of ​​treating the problem of selecting a preferred combination from among the candidate bonds to construct a molecular model as a path optimization problem. The path optimization problem assumed in this embodiment is the problem of selecting the route that passes through points where bond scores are set, under given conditions, and which results in the smallest sum of bond scores.

[0023] Specifically, the inventors conceived the idea of ​​relating a method for constructing a molecular model by selecting more preferable bond candidates while satisfying the conditions required when combining each bond candidate to a method for optimizing a path in a path optimization problem. Below, we will detail how to solve a path optimization problem by relating a method for optimizing a molecular model to that problem.

[0024] First, the locations indicated by the connection points of the connection candidates are mapped to points traversed in the path optimization problem. These points can be represented, for example, by coordinates. Once this mapping is established, the connection score of the connection candidate is also mapped to each of these points. Furthermore, the conditions for combining the connection candidates are treated as constraints that correspond to the conditions for connecting the points corresponding to those connection candidates in the path optimization problem. Specific examples of constraints will be described later.

[0025] Furthermore, the objective function to be used in the path optimization problem is represented by the sum of the binding scores corresponding to the plurality of binding candidates passed through. Normalization may or may not be performed when obtaining the sum of these binding scores. In the molecular model construction apparatus according to the present disclosure, normalization is performed by dividing the binding score of the binding candidate by the sum of the absolute values of the maximum value and the minimum value of the binding scores of the binding candidates, plus 1. This is done in consideration of the fact that positive and negative values are mixed in the binding scores of the binding candidates and to avoid division by zero, in order to make it easier to handle the constraint conditions. Under such correspondence, the optimization unit 40 optimizes the constructed molecular model by solving the path optimization problem that minimizes the above objective function.

[0026] Furthermore, since the path optimization problem is a type of combinatorial optimization problem, it can be solved by quantum computing using a quantum computer or the like. That is, by describing the above objective function and constraint conditions with a Hamiltonian that can be used by a quantum computer, it can be solved by quantum computing. Therefore, the optimization unit 40 optimizes the constructed molecular model by solving the path optimization problem that minimizes the above objective function by quantum computing under the given constraint conditions.

[0027] FIG. 2 is an explanatory diagram showing an example of defining information used in quantum computing. The m ｍｓ，ｍｔ , m ２ , m ３ shown in FIG. 2 indicate binding candidates. Hereinafter, the spin that becomes 1 when the bond b ｓ , m ｔ between two binding candidates m is selected is represented as x ｍｓ，ｍｔ ｂｍｓ，ｍｔ . For example, the spin that becomes 1 when the bond b ２ １ , m ｍ１，ｍ２ between two binding candidates m is selected is represented as x ｂｍ１，ｍ２ ｍ . Also, the spin that becomes 1 when the binding candidate m is selected and 0 when it is not selected is represented as y ｎ ｍ , and the auxiliary spin is denoted as z ｍ １ .

[0028] Furthermore, the join score obtained when a join candidate m is selected is S. ｍ G is the set of joins b that include the candidate join m. ｍ , combination b １ , b ２ The angle formed by A ｂ１，ｂ２ This is expressed as follows. Furthermore, a function called CheckAngle(A) is defined as a function that returns 0 if angle A is within a certain value, and 1 if it is not within a certain value, and the combination candidate is m １ ,m ２ A function called CheckOverlap(m) returns 1 if they overlap and 0 if they do not. １ ,m ２ ) is defined as follows.

[0029] At this time, the objective function (Hamiltonian H) described above Ｏ ) is the weighting W of the objective function Ｏ Taking this into account, it can be expressed by Equation 1, which is exemplified below.

[0030]

[0031] Figure 3 is an explanatory diagram illustrating an example of the process for calculating the sum of binding scores. Figure 3 shows an example of designing the loop portion of a stem-loop structure, which is a typical structure of an aptamer. Aptamers are artificial nuclear sequences that specifically bind to a variety of targets, like antibodies, and have the property of returning to their original structure even after thermal transformation. Examples of substances to which aptamers bind include hazardous substances such as narcotics, explosives, pesticides, and allergens, as well as influenza viruses and food poisoning bacteria. The target molecule 101 illustrated in Figure 3 is, for example, thrombin.

[0032] Furthermore, Figure 3 shows an example where multiple binding candidates 102a, 102b, 102c, etc. exist for the target molecule 101. For example, binding candidate 102a illustrated in Figure 3 is a mononucleotide (A: adenine) and has a binding score of -3. Similarly, binding candidate 102c is a mononucleotide (T: thymine) and has a binding score of -5.

[0033] As illustrated in Figure 3, a join score is associated with each join candidate, and when that join candidate m is selected, y ｍWhen this is set to 1, the combined score S ｍ The result is obtained. The optimization unit 40 selects a combination candidate using, for example, the equation 1 described above, thereby obtaining the combination score S. ｍ The goal is to minimize the sum of the binding scores. In this case, the binding scores may be treated as normalized binding scores. This process corresponds to the problem of selecting the path with the minimum cost in a path optimization problem. And determining a molecular model in accordance with a path optimization problem can be said to be equivalent to determining a loop sequence in which the binding energy is low in a closed loop.

[0034] While Figure 3 shows an example of designing the loop portion of a stem-loop structure, the structures that can be designed are not limited to stem-loop structures. For example, bulge structures, mismatches, and internal loops can also be designed in a similar manner.

[0035] The optimization unit 40 uses the Hamiltonian H, which is the objective function exemplified in Equation 1. Ｏ By minimizing this, the sum of the bond scores of the selected bond candidates is minimized. At this time, the optimization unit 40 can have the quantum annealing machine 200 optimize the molecular model.

[0036] The quantum annealing machine 200 is a dedicated device for finding the ground state of the Hamiltonian of the Ising model, and it performs annealing based on the Ising model. More specifically, the quantum annealing machine is a device that probabilistically finds the values ​​of a binary variable that minimize or maximize the objective function (i.e., the Hamiltonian) of the Ising model, which takes a binary variable as an argument. The binary variable may be realized using classical bits or quantum bits.

[0037] Next, we will explain specific examples of the constraints mentioned above (i.e., the conditions for combining connection candidates). The first constraint is the constraint that the connection candidates must be connected by a cycle. This constraint is, for example, the constraint weight W. Ｃ１ Taking this into account, it can be expressed by Equation 2, which is exemplified below.

[0038]

[0039] The reason why the first constraint is expressed by equation 2 above will be explained below. Figure 4 is an explanatory diagram showing an example of the state represented by the first constraint. The candidate connection m illustrated in Figure 4 １ Candidates for joining are those that join b １ , b ２ , b ３ Let's assume there are three types. Combination candidate m １ is selected (i.e., y ｍ１ If = 1, then constraint condition H Ｃ１ This can be expressed by equation 3, which is exemplified below.

[0040]

[0041] This constraint has the effect of selecting two connection lines from one candidate connection. Therefore, by selecting two connection lines for each candidate connection, it has the effect of creating cycles like pattern P11. In other words, this constraint makes it less likely to select candidates like patterns P12 and P13 that include candidate connections or connection lines that incur a penalty.

[0042] The second constraint is that the sequence length of the selected join candidates must be within a specified range. The maximum number of specified join candidates is n. ｍａｘ This is expressed as, and the smallest number is n ｍｉｎ Expressed as such, this constraint condition is, for example, the weighting W of the constraint condition. Ｃ２ Taking this into account, it can be expressed by equation 4, which is exemplified below.

[0043]

[0044] Figure 5 is an explanatory diagram illustrating an example of the state represented by the second constraint. Figure 5 shows an example where the sequence length of the candidate link is 5.

[0045] A third constraint is that the angle formed by the three candidate connections must be within a specified range. This is because the angle formed by the three candidate connections must remain constant. This constraint can be applied, for example, to the constraint weight W. Ｃ３ Taking this into account, it can be expressed by equation 5, which is exemplified below.

[0046]

[0047] This constraint results in a penalty if a combination of connection lines is selected in which the angles formed by the three candidate connections do not fall within a specific range of angles. Figure 6 is an explanatory diagram illustrating an example of the state represented by the third constraint.

[0048] In the example shown in Figure 6, in pattern P21, coupling b １ and combined b ２ Since the angle formed is within the acceptable range, the constraint equation H in equation 5 shown above Ｃ３ The value becomes 0, and no penalty occurs. On the other hand, in pattern P22, combination b １ and combined b ３ Since the angle formed is outside the acceptable range, the constraint equation H in equation 5 shown above Ｃ３ The value of x becomes greater than 0, resulting in a penalty. ｂ１ and x ｂ３ This has the effect of making it less likely that both will be selected simultaneously. In other words, this constraint makes it less likely that candidates such as pattern P23, which includes connection candidates that are penalized or connection lines, will be selected.

[0049] A fourth constraint is that the candidate joins must not overlap. This is because a certain distance must be maintained between the two candidate joins. This constraint can be expressed, for example, by the constraint weight W. Ｃ４ Taking this into account, it can be expressed by equation 6, which is exemplified below.

[0050]

[0051] This constraint imposes a penalty if a join candidate is selected whose distance from other join candidates is less than or equal to a specific value. Figure 7 is an explanatory diagram illustrating an example of the state represented by the fourth constraint.

[0052] In the example shown in Figure 7, the candidate for joining m １ and candidate combination m ２ Since they do not overlap, the value of the function CheckOverlap in equation 6 shown above becomes 0, and constraint equation H Ｃ４ The value of becomes 0. On the other hand, the candidate for joining m ２ and candidate combination m３ Since they overlap, in equation 6 shown above, constraint equation H Ｃ４ The value of becomes greater than 0, resulting in a penalty. Due to this constraint, a penalty is given to y. ｍ２ and y ｍ３ It can be said that it becomes less likely for both to be selected simultaneously.

[0053] The above describes specific constraints, but the number and content of constraints can be combined arbitrarily, and are not limited to the constraints exemplified above. Other constraints could include, for example, a constraint that prevents connection lines from intersecting.

[0054] In addition to the constraints exemplified above, other settings may be pre-defined to exclude certain conditions from the spin definition. For example, one such exclusion setting is to not define a spin if the distance between two candidate bonds is outside a specified range.

[0055] Figure 8 is an explanatory diagram illustrating an example of a process that sets parameters according to the distance between candidate joins. Pattern P31 shown in Figure 8 is a combination of candidate joins m １ and candidate combination m ２ b with ｍ１，ｍ２ Since the distance is within the acceptable range, x ｂｍ１，ｍ２ We define this as spin. On the other hand, pattern P32 is a coupling candidate m １ and candidate combination m ２ b with ｍ１，ｍ２ Because the distance is outside the acceptable range, x ｂｍ１，ｍ２ We do not define spin as such.

[0056] By performing filtering at the spin definition stage in this way, it becomes possible to reduce the number of coupling lines. For example, in pattern P33 illustrated in Figure 8, the shaded area is within the acceptable distance range for coupling candidate 103. In this case, coupling candidates 103a and 103b are within the acceptable distance range for coupling candidate 103. On the other hand, coupling candidates 103c and 103d are outside the acceptable distance range for coupling candidate 103.

[0057] Therefore, by not defining the combination between combination candidate 103 and combination candidate 103c, and the combination between combination candidate 103 and combination candidate 103d, it becomes possible to exclude combinations between these combination candidates.

[0058] In addition, to prevent the number of potential merge candidates from becoming too large, merge candidates that meet predetermined conditions may be excluded beforehand.

[0059] The output unit 50 outputs an optimized molecular model. The method by which the output unit 50 outputs the molecular model is arbitrary. For example, the output unit 50 may output selected binding candidates and information showing the binding relationships between each binding candidate.

[0060] The binding site estimation unit 10, the binding candidate input unit 30, the optimization unit 40, and the output unit 50 are all implemented by a computer processor (e.g., a CPU (Central Processing Unit) or GPU (Graphics Processing Unit)) that operates according to a program (molecular model construction program).

[0061] For example, the program may be stored in the memory unit 20 of the molecular model construction device 100, and the processor may read the program and operate as a binding site estimation unit 10, a binding candidate input unit 30, an optimization unit 40, and an output unit 50 according to the program. Alternatively, the functions of the molecular model construction device 100 may be provided in SaaS (Software as a Service) format.

[0062] Furthermore, the connection point estimation unit 10, the connection candidate input unit 30, the optimization unit 40, and the output unit 50 may each be implemented with dedicated hardware. Also, some or all of the components of each device may be implemented by general-purpose or dedicated circuits, processors, etc., or combinations thereof. These may be composed of a single chip or multiple chips connected via a bus. Some or all of the components of each device may be implemented by a combination of the above-mentioned circuits, etc., and programs.

[0063] Furthermore, if some or all of the components of the molecular model construction apparatus 100 are realized by multiple information processing devices or circuits, these multiple information processing devices or circuits may be centrally located or distributed. For example, the information processing devices or circuits may be realized in a form in which each is connected via a communication network, such as a client-server system or a cloud computing system.

[0064] Next, the operation of the molecular model construction system 1 of this embodiment will be described. Figure 9 is a flowchart showing an example of the operation of the molecular model construction system 1 of this embodiment.

[0065] The binding site estimation unit 10 estimates the binding sites and binding scores of binding candidates that are expected to bind to the target molecule (step S11). The binding candidate input unit 30 receives input of the estimated binding sites and binding scores of the binding candidates (step S12). The optimization unit 40 optimizes the molecular model formed by each binding candidate using quantum computation (step S13). At this time, the optimization unit 40 may have the quantum annealing machine 200 optimize the molecular model. The output unit 50 then outputs the optimized molecular model (step S14).

[0066] As described above, in this embodiment, the binding candidate input unit 30 receives input of binding locations and binding scores of binding candidates for the target molecule, the optimization unit 40 optimizes the molecular model by quantum computation, and the output unit 50 outputs the optimized molecular model. At this time, constraints are set such that the positions indicated by the binding locations correspond to the points traversed in the path optimization problem, and the conditions for combining binding candidates correspond to the conditions for connecting the points corresponding to the binding candidates in the path optimization problem. In this situation, the optimization unit 40 optimizes the constructed molecular model by solving a path optimization problem by quantum computation that minimizes an objective function representing the sum of binding scores corresponding to the points traversed when multiple points are passed through.

[0067] Such a configuration allows for the rapid construction of molecular models.

[0068] Next, a modified example of the molecular model construction system of this embodiment will be described. Figure 10 is a block diagram showing a modified example of the molecular model construction system of the present disclosure. The molecular model construction system 2 illustrated in Figure 10 comprises a molecular model construction apparatus 110 and a quantum annealing machine 200. The quantum annealing machine 200 is the same as in the above embodiment.

[0069] The molecular model construction apparatus 110 includes a binding site estimation unit 10, a storage unit 20, a binding candidate input unit 31, an optimization unit 40, and an output unit 50. In other words, compared to the above embodiment, the contents of the binding candidate input unit 31 differ from the contents of the binding candidate input unit 30. Otherwise, it is the same as the above embodiment.

[0070] The binding candidate input unit 31 accepts input of binding sites and binding scores for binding candidates to the target molecule, as well as input of binding sites and binding scores for dummy binding candidates. Here, dummy binding candidates are different from binding candidates and do not necessarily have to be assumed to bind to the target molecule. Furthermore, the binding scores for dummy binding candidates are predetermined by the designer or other relevant parties.

[0071] The optimization unit 40 optimizes the molecular model formed by each bond candidate, including dummy bond candidates, using quantum computation. Figure 11 is an explanatory diagram showing another example of the process for calculating the sum of bond scores. The example shown in Figure 11 differs from the process illustrated in Figure 3 in that dummy bond candidates 102d and 102e are present. By adding such dummy bond candidates, it becomes possible to avoid significant constraint violations and the like.

[0072] Next, an overview of the present disclosure will be described. Figure 12 is a block diagram illustrating an overview of the molecular model construction apparatus according to the present disclosure. The molecular model construction apparatus 80 according to the present disclosure is a molecular model construction apparatus (for example, molecular model construction apparatus 100) that constructs a molecular model by combining molecules, and includes a binding candidate input means 81 (for example, binding candidate input unit 30, binding candidate input unit 31) that receives input of the binding sites of each binding candidate (for example, mononucleotide) that bind to the target molecule (for example, thrombin) estimated based on information of the target molecule which is the target to which molecules are bound, and a binding score indicating the strength of the binding between the target molecule and each binding candidate; an optimization means 82 (for example, optimization unit 40) that optimizes the molecular model constructed by each binding candidate by quantum computation; and an output means 83 (for example, output unit 50) that outputs the optimized molecular model.

[0073] The optimization means 82 optimizes the constructed molecular model by solving a path optimization problem using quantum computing, which minimizes an objective function (for example, equation 1 shown above) that represents the sum of the connection scores corresponding to the points traversed when multiple points are traversed, under constraints that correspond the positions indicated by the connection sites to the points traversed in the path optimization problem, and the conditions for combining candidate connections to the conditions for connecting the points corresponding to those candidate connections in the path optimization problem.

[0074] Such a configuration allows for the rapid construction of molecular models by combining molecules.

[0075] Specifically, the binding candidates may be nucleic acids or amino acids.

[0076] Furthermore, the binding candidate input means 81 may accept input of binding sites and binding scores for dummy binding candidates, in addition to input of binding sites and binding scores for binding candidates to the target molecule. The optimization means 82 may then optimize the molecular model constructed from each binding candidate, including the dummy binding candidates, by quantum computation.

[0077] Furthermore, the molecular model construction device 80 may also include a binding site estimation means (for example, a binding site estimation unit 10) that estimates the binding sites of candidate bindings that are expected to bind to the target molecule, and the binding score of those candidate bindings.

[0078] Alternatively, the optimization means 82 may have a quantum annealing machine (for example, a quantum annealing machine 200) perform quantum computations to optimize the molecular model.

[0079] Specifically, the selection status of a bond candidate may be represented by a spin of 1 or 0, and the objective function may be defined by a Hamiltonian that represents the sum of the bond scores obtained when that bond candidate is selected.

[0080] Although the present invention has been described above with reference to the embodiments and examples, the present invention is not limited to the above embodiments and examples. Various modifications to the structure and details of the present invention can be made, as can be understood by those skilled in the art within the scope of the present invention.

[0081] This application claims priority based on Japanese Patent Application No. 2024-174053, filed on 3 October 2024, and incorporates all of its disclosures herein.

[0082] 1,2 Molecular Model Construction System 10 Binding Site Estimation Unit 20 Memory Unit 30,31 Binding Candidate Input Unit 40 Optimization Unit 50 Output Unit 100,110 Molecular Model Construction Apparatus

Claims

1. A molecular model construction device for constructing a molecular model by combining molecules, comprising: a binding candidate input means that receives input of binding sites for each binding candidate that binds to the target molecule and a binding score indicating the strength of the binding between the target molecule and each binding candidate, estimated based on information of the target molecule which is the target to which molecules are bound; an optimization means that optimizes the molecular model constructed by each binding candidate using quantum computation; and an output means that outputs the optimized molecular model, wherein the optimization means optimizes the constructed molecular model by solving a path optimization problem using quantum computation, under constraints that correspond the positions indicated by the binding sites to points traversed in a path optimization problem, and the conditions for combining the binding candidates correspond to the conditions for connecting the points corresponding to the binding candidates in the path optimization problem, thereby minimizing an objective function that represents the sum of the binding scores corresponding to the points traversed when passing through a plurality of such points.

2. The molecular model construction apparatus according to claim 1, wherein the binding candidate is a nucleic acid or an amino acid.

3. The molecular model construction apparatus according to claim 1 or 2, wherein the binding candidate input means accepts input of binding sites and binding scores for binding candidates to a target molecule, as well as input of binding sites and binding scores for dummy binding candidates, and the optimization means optimizes the molecular model constructed by each binding candidate, including the dummy binding candidates, by quantum computation.

4. A molecular model construction apparatus according to claim 1 or 2, comprising a binding site estimation means for estimating the binding sites of candidate bindings expected to bind to a target molecule, and the binding scores of said candidate bindings.

5. The molecular model construction apparatus according to claim 1 or 2, wherein the optimization means causes a quantum annealing machine that performs quantum computation to optimize the molecular model.

6. A molecular model construction apparatus according to claim 1 or 2, wherein the selection of a binding candidate is represented by a spin of 1 or 0, and the objective function is defined by a Hamiltonian representing the sum of binding scores obtained when the binding candidate is selected.

7. A molecular model construction method for constructing a molecular model by combining molecules, characterized in that the method accepts inputs of binding sites for each binding candidate that binds to the target, estimated based on information of the target molecule to which the molecules are bound, and binding scores indicating the strength of the binding between the target molecule and each binding candidate; the method corresponds the positions indicated by the binding sites to points traversed in a path optimization problem; the method solves a path optimization problem using quantum computation to minimize an objective function that represents the sum of the binding scores corresponding to the points traversed when passing through multiple such points, under constraints that correspond the conditions for combining the binding candidates to the conditions for connecting the points corresponding to the binding candidates in the path optimization problem; the method optimizes the molecular model constructed by each binding candidate; and outputs the optimized molecular model.

8. A molecular model construction program applied to a computer for constructing molecular models by combining molecules, the program causing the computer to perform a binding candidate input process that accepts input of binding sites for each binding candidate that binds to the target, estimated based on information of the target molecule which is the target to which the molecules are bound, and binding scores indicating the strength of the binding between the target molecule and each binding candidate; an optimization process that optimizes the molecular model constructed by each binding candidate using quantum computation; and an output process that outputs the optimized molecular model, wherein in the optimization process, the positions indicated by the binding sites are associated with points traversed in a path optimization problem, and the conditions for combining the binding candidates are associated with the conditions for connecting the points corresponding to the binding candidates in the path optimization problem, and the program solves a path optimization problem using quantum computation to minimize an objective function that represents the sum of the binding scores corresponding to the points traversed when passing through multiple such points, thereby optimizing the constructed molecular model.

Images