Search program, search method, and search device

Abstract

To efficiently search for an optimum arrangement combination in a coarse-grained model of a middle molecule.SOLUTION: A search program causes a computer, which instructs to search for an arrangement combination of a plurality of amino acids included in a middle molecule in multiple stages as a combinatorial optimization problem, to execute processing of instructing to search for a first arrangement of the plurality of amino acids by calculating a value of a first cost arithmetic expression that does not distinguish between an L-form and a D-form of the plurality of amino acids, instructing to search for a second arrangement of the plurality of amino acids by calculating a value of a second cost arithmetic expression that distinguishes between the L-form and the D-form of the plurality of amino acids using the searched first arrangement as an initial arrangement, and outputting the second arrangement.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a search program, a search method, and a search device. [Background technology]

[0002] In recent years, in the field of drug discovery, there has been a growing expectation for the development of drugs using medium molecules (molecular weight 500 to 3000) with fewer side effects, and the development of search methods for finding stable structures of medium molecules is underway.

[0003] One example is a search method that applies an interaction potential between multiple amino acids to a coarse-grained model of a medium-sized molecule containing multiple amino acids, and solves the optimal amino acid arrangement that minimizes the value of a cost calculation formula in a lattice space divided into lattices as a combinatorial optimization problem. This search method makes it possible to efficiently search for optimal arrangement combinations in the coarse-grained model. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-159683 [Patent Document 2] Japanese Patent Application Laid-Open No. 2015-007537 [Patent Document 3] Patent Publication No. 2021-082165 Summary of the Invention [Problem to be solved by the invention]

[0005] On the other hand, even when the above search method is used, when searching for the optimal combination of amino acid arrangements in a coarse-grained model of, for example, a peptide with a large number of residues, the number of bits used for the combinatorial optimization problem becomes enormous, resulting in an increase in search time.

[0006] One aspect aims to efficiently search for optimal configuration combinations in a coarse-grained model of a medium-sized molecule. [Means for solving the problem]

[0007] According to one aspect, the search program comprises: A computer is instructed to search for combinations of arrangements of coarse-grained models of molecular structures of medium-sized molecules with molecular weights of 500 to 3000 containing multiple amino acids as a combinatorial optimization problem. The L-body and D-body of the coarse-grained model The term expressing the difference between instructing a search for a first configuration of the coarse-grained model by calculating a value of a first cost calculation formula; The first configuration found is used as an initial configuration, and the L body and the D body of the coarse-grained model are Contains terms that represent the difference between instructing a search for a second configuration of the coarse-grained model by calculating a value of a second cost calculation formula; The second arrangement is output and a process is executed. [Effects of the Invention]

[0008] It is possible to efficiently search for optimal configuration combinations in coarse-grained models of medium-sized molecules. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 illustrates an example of a system configuration of a search system. [Figure 2] FIG. 2 illustrates an example of a hardware configuration of a terminal device. [Figure 3] FIG. 10 is a diagram illustrating an example of cost calculation formula information. [Figure 4] FIG. 10 is a diagram illustrating an example of constraint term calculation formula information. [Figure 5] FIG. 2 is a diagram illustrating an example of a functional configuration of a terminal device. [Figure 6] FIG. 2 is a diagram illustrating an example of a functional configuration of an Ising device. [Figure 7] 10 is a flowchart showing the flow of a search process. [Figure 8] FIG. 10 is a diagram illustrating a specific example of a search process. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, each embodiment will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0011] [First embodiment] <Search system configuration> First, the system configuration of a search system according to the first embodiment will be described. Fig. 1 is a diagram showing an example of the system configuration of the search system. The search system 100 is a system that searches for an optimal arrangement combination that minimizes cost under predetermined constraints in a lattice space, using a coarse-grained model in which the molecular structure of a medium-sized molecule is coarse-grained as a search target.

[0012] As shown in FIG. 1, the search system 100 includes a terminal device 110 (an example of a search device) and an Ising device 120 such as a quantum annealing device.

[0013] A search program is installed in the terminal device 110, and when the program is executed, the terminal device 110 functions as a search target information acquisition unit 111, a first stage execution unit 112, a second stage execution unit 113, an output unit 114, and a third stage execution unit 115.

[0014] The search target information acquisition unit 111 acquires a coarse-grained model of a search target medium molecule for searching for an optimal arrangement combination. In the first embodiment, the coarse-grained model of a search target medium molecule acquired by the search target information acquisition unit 111 is, for example, The backbone of the peptide is centered on the α-carbon of the amino acid that forms the main chain of the peptide, and Side chain particles, are replaced by one coarse-grained particle each, - Arranged in a simple cubic lattice (FCC: face-centered cubic) space (lattice space), Each point in the lattice space is assigned a bit that represents its position, and once the arrangement of the coarse-grained particles is determined, the distance between the coarse-grained particles is uniquely determined.

[0015] The search target information acquisition unit 111 notifies the first stage execution unit 112 of the acquired coarse-grained model of the medium molecule being the search target.

[0016] The first stage execution unit 112 obtains the following information from the optimization information storage unit 116 as information necessary for the Ising device 120 to solve a combinatorial optimization problem to find an optimal combination of amino acid arrangements that minimizes the value of the cost calculation formula under predetermined constraints in the lattice space: Cost calculation formula information (details below), and Constraint expression information (details below), Read out.

[0017] Furthermore, the first stage executing unit 112 instructs the Ising device 120 to search for an optimal combination of arrangements from a predetermined initial arrangement using the cost calculation formula information and the constraint term calculation formula information for the coarse-grained model of the medium molecule to be searched for. When instructing the Ising device 120, the first stage executing unit 112 transmits first search information to the Ising device 120. The first search information here includes the predetermined coarse-grained model of the initial arrangement, the cost calculation formula information, and the constraint term calculation formula information.

[0018] In the first embodiment, the first stage executing unit 112 transforms the cost calculation formula information so as to reduce the amount of calculation (number of bits) of the Ising device 120 when searching for an optimal arrangement combination, and then transmits the transformed cost calculation formula information to the Ising device 120. In other words, when the first stage executing unit 112 transmits the first search information, the cost calculation formula information to be included in the first search information is the transformed cost calculation formula information that has been transformed so as to reduce the amount of calculation (number of bits).

[0019] The second stage execution unit 113 acquires the first search result transmitted from the Ising device 120 in response to an instruction from the first stage execution unit 112 to search for an optimal combination of placements.

[0020] Furthermore, the second stage execution unit 113 reads out cost calculation formula information and constraint term calculation formula information from the optimization information storage unit 116. Furthermore, the second stage execution unit 113 instructs the Ising device 120 to search for a further optimal combination of arrangements for the coarse-grained model of the medium molecule to be searched for, using the unmodified cost calculation formula information and constraint term calculation formula information. When instructing the Ising device 120, the second stage execution unit 113 transmits second search information to the Ising device 120. The second search information here includes the coarse-grained model whose initial arrangement is the first search result, and the unmodified cost calculation formula information and constraint term calculation formula information.

[0021] In response to an instruction from the second stage execution unit 113 to search for the optimal placement combination, the output unit 114 acquires the second search result sent from the Ising device 120 and notifies the third stage execution unit 115.

[0022] Furthermore, the output unit 114 evaluates the acquired second search result and determines whether there is room for improvement in the optimal layout combination. If it is determined that there is no room for improvement, the output unit 114 outputs the second search result as an optimal solution.

[0023] In addition, in response to notifying the third stage execution unit 115 of the second search result, the output unit 114 acquires the third search result transmitted from the Ising device 120 and notifies the third stage execution unit 115 again.

[0024] Furthermore, the output unit 114 evaluates the acquired third search result and determines whether there is room for improvement in the optimal layout combination. If it is determined that there is no room for improvement, the output unit 114 outputs the third search result as the optimal solution.

[0025] The third stage execution unit 115 acquires the second search result or the third search result notified by the output unit 114.

[0026] Furthermore, the third stage execution unit 115 reads out cost calculation formula information and constraint term calculation formula information from the optimization information storage unit 116. Furthermore, the third stage execution unit 115 instructs the Ising device 120 to search for a further optimal arrangement combination for the coarse-grained model of the medium molecule to be searched for, using the cost calculation formula information and the constraint term calculation formula information. When instructing the Ising device 120, the third stage execution unit 115 transmits third search information to the Ising device 120. The third search information here includes a coarse-grained model whose initial arrangement is the second search result (or the third search result), cost calculation formula information, and constraint term calculation formula information.

[0027] In the first embodiment, the third stage executing unit 115 changes the coefficients included in the constraint term calculation formula information so that the Ising device 120 does not fall into a local solution when searching for an optimal placement combination, and then transmits the information to the Ising device 120. In other words, the constraint term calculation formula information included when the third stage executing unit 115 transmits the third search information is the changed constraint term calculation formula information in which the coefficients have been changed so that the Ising device 120 does not fall into a local solution.

[0028] On the other hand, an optimization program is installed in the Ising device 120, and the Ising device 120 functions as a combination optimization unit 121 by executing the program.

[0029] When the combination optimization unit 121 receives the first search information from the terminal device 110, it calculates a first search result by solving an optimal placement combination in the coarse-grained model as a combinatorial optimization problem, and transmits the calculated first search result to the terminal device 110. As described above, the cost calculation formula information included in the first search information is cost calculation formula information after transformation that has been transformed so as to reduce the amount of calculation (number of bits), and therefore the combination optimization unit 121 can quickly calculate the first search result.

[0030] Furthermore, upon receiving the second search information from the terminal device 110, the combination optimization unit 121 calculates a second search result by solving an optimal combination of placements in the coarse-grained model as a combinatorial optimization problem, and transmits the calculated second search result to the terminal device 110. As described above, the coarse-grained model included in the second search information uses the first search result as its initial placement, so the combination optimization unit 121 can calculate the second search result quickly. As described above, the cost calculation formula information included in the second search information is untransformed cost calculation formula information, so the combination optimization unit 121 can calculate a highly accurate second search result.

[0031] Furthermore, upon receiving the third search information from the terminal device 110, the combinatorial optimization unit 121 calculates a third search result by solving an optimal combination of placements in the coarse-grained model as a combinatorial optimization problem, and transmits the calculated third search result to the terminal device 110. As described above, the constraint term operation formula information included in the third search information is the constraint term operation formula information after the coefficients have been changed, and therefore the combinatorial optimization unit 121 can calculate a highly accurate third search result without falling into a local solution.

[0032] <Hardware configuration of terminal device> Next, a description will be given of the hardware configuration of the terminal device 110. Fig. 2 is a diagram showing an example of the hardware configuration of the terminal device.

[0033] 2, the terminal device 110 includes a processor 201, a memory 202, an auxiliary storage device 203, an I / F (Interface) device 204, a communication device 205, and a drive device 206. The hardware components of the terminal device 110 are connected to each other via a bus 207.

[0034] The processor 201 has various arithmetic devices such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc. The processor 201 reads various programs (for example, a search program, etc.) into the memory 202 and executes them.

[0035] The memory 202 has a main storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The processor 201 and the memory 202 form a so-called computer, and the processor 201 executes various programs read onto the memory 202, causing the computer to realize various functions.

[0036] The auxiliary storage device 203 stores various programs and various information used when the various programs are executed by the processor 201. For example, the optimization information storage unit 116 is realized in the auxiliary storage device 203.

[0037] The I / F device 204 is a connection device that connects the terminal device 110 with an operation device 210 and an output device 220, which are examples of external devices.

[0038] The communication device 205 is a communication device for communicating with the Ising device 120, which is an example of another device.

[0039] The drive device 206 is a device for loading a recording medium 230. The recording medium 230 here includes media that record information optically, electrically, or magnetically, such as a CD-ROM, a flexible disk, a magneto-optical disk, etc. The recording medium 230 may also include semiconductor memory that records information electrically, such as a ROM, a flash memory, etc.

[0040] The various programs to be installed in the auxiliary storage device 203 are installed, for example, by setting the distributed recording medium 230 in the drive device 206 and reading the various programs recorded on the recording medium 230 by the drive device 206. Alternatively, the various programs to be installed in the auxiliary storage device 203 may be installed by being downloaded from a network via the communication device 205.

[0041] Note that, here, only the hardware configuration of the terminal device 110 has been described, and the hardware configuration of the Ising device 120 has not been described, but the hardware configuration of the Ising device 120 may be the same as that of the terminal device 110, for example. In this case, the Ising device 120 functions as a combination optimization unit 121 by having the processor execute an optimization program read into the memory.

[0042] Alternatively, the hardware configuration of the Ising device 120 may be the same as that of a so-called quantum computer.

[0043] <Example of cost calculation formula information> Next, a specific example of cost calculation formula information stored in the optimization information storage unit 116 will be described. Fig. 3 is a diagram showing an example of cost calculation formula information. As shown in cost calculation formula information 300 in Fig. 3, in a formulation for solving an optimal combination of placements in a coarse-grained model as a combinatorial optimization problem by an annealing method, the cost is expressed by the Hamiltonian of the following equation (1):

[0044]

number

[0045] Thus, the cost includes a term expressing the interaction energy between the coarse-grained particles and a term expressing the difference between the L-form and D-form of the coarse-grained particles.

[0046] As shown in the interaction energy information 310 between coarse-grained particles in FIG. 3, the Hamiltonian of the term expressing the interaction energy between coarse-grained particles is defined by the following equation (2).

[0047]

number

[0048] In the above formula (2), N: total number of coarse-grained particles (amino acid residues), Q i : a set of bit numbers representing the main chain particle of the i-th coarse-grained particle (amino acid residue), Q SC i : a set of bit numbers representing the side chain particles of the i-th coarse-grained particle (amino acid residue), η(a): A set of bit numbers representing the grid points of side chain particles within the distance where potential data exists from the grid point represented by bit number a. ω(a): Coarse-grained particle (amino acid residue) represented by bit number a, ω(b): Coarse-grained particle (amino acid residue) represented by bit number b, P ω(a)ω(b) : The interaction energy between the coarse-grained particle (amino acid residue) ω(a) and the coarse-grained particle (amino acid residue) ω(b), ·q a : Variable with bit number a, ·q b : Variable with bit number b, is.

[0049] As shown in the library information 311 in FIG. 3, the interaction energy P ω(a)ω(b) is calculated in advance for each distance for each type of coarse-grained particle. The example of library information 311 in FIG. 3 shows the interaction energy P ω(a)ω(b) is calculated in advance for distances 1 to n and stored in a library.

[0050] As shown in information 320 relating to the distinction between L- and D-isomers in FIG. 3, the Hamiltonian of the term expressing the difference between L- and D-isomers of the coarse-grained particles (amino acid residues) is defined by the following formula (3):

[0051]

number

[0052] In the above equation (3), η(a): a bit representing a lattice point of a main chain particle adjacent to the lattice point represented by bit number a, η(b): a bit representing a lattice point of a main chain particle adjacent to the lattice point represented by bit number b, β(b): A bit representing a lattice point of a side chain particle adjacent to the lattice point represented by bit number b. ·S(a,b,c,d): potential value, ·q c : variable with bit number c, ·q d : variable with bit number d, is.

[0053] <Example of constraint expression information> Next, we will explain a specific example of constraint term calculation formula information stored in the optimization information storage unit 116. Fig. 4 is a diagram showing an example of constraint term calculation formula information. As shown in Fig. 4, four types of constraint term calculation formula information are stored in the optimization information storage unit 116.

[0054] Of these, the constraint term calculation formula information 410 is a constraint condition that ensures that each coarse-grained particle (amino acid residue) exists in only one location in the lattice space, and is defined by the following formula (4).

[0055]

number

[0056] In the above equation (4), N: total number of coarse-grained particles (amino acid residues), Q i : a set of bit numbers representing the main chain particle of the i-th coarse-grained particle (amino acid residue), Q SC i : a set of bit numbers representing the side chain particles of the i-th coarse-grained particle (amino acid residue), ·q a : Variable with bit number a, ·q b : Variable with bit number b, λ one :coefficient, is.

[0057] The constraint term calculation formula information 420 is a constraint condition that ensures that multiple coarse-grained particles (amino acid residues) do not exist at the same lattice point, and is defined by the following formula (5).

[0058]

number

[0059] In the above equation (5), ·V: lattice space, θ(ν): the set of all bit numbers representing the lattice point ν, λ olap :coefficient, is.

[0060] The constraint term calculation formula information 430 is a constraint condition that ensures that each coarse-grained particle (amino acid residue) is linked in accordance with the molecular sequence, and is defined by the following formula (6).

[0061]

number

[0062] In the above equation (6), d(a,b): the distance between the lattice points corresponding to the a-th and b-th bit variables, d(a,c): the distance between the lattice points corresponding to the a-th and c-th bit variables, d0: distance between the nearest lattice points, ·q c : variable with bit number c, ·q i : variable with bit number i, ·q j : variable with bit number j, λ conn :coefficient, is.

[0063] The constraint term calculation formula information 440 is a constraint condition that ensures that no unconnected coarse-grained particles (amino acid residues) exist in the adjacent lattice, and is defined by the following formula (7).

[0064]

number

[0065] In the above equation (7), η(a): A set of bit numbers representing adjacent lattice points of the lattice point represented by the a-th bit, c: 1 if cyclic and i=N / 2, otherwise 2 λ nadj :coefficient, is.

[0066] <Details of the functional configuration of the terminal device> Next, a detailed description will be given of the functional configuration of the terminal device 110. Fig. 5 is a diagram showing an example of the functional configuration of the terminal device. As described using Fig. 1, the terminal device 110 functions as a search target information acquisition unit 111, a first stage execution unit 112, a second stage execution unit 113, an output unit 114, and a third stage execution unit 115.

[0067] As shown in FIG. 5, the first stage execution unit 112 further includes an execution instruction unit 511, a cost calculation formula information modification unit 512, and a calculation formula information acquisition unit 513.

[0068] The second stage execution unit 113 also includes a search result acquisition unit 521 , an execution instruction unit 522 , and an arithmetic expression information acquisition unit 523 .

[0069] Moreover, the output unit 114 has a search result acquisition unit 531 and an evaluation unit 532. Furthermore, the third stage execution unit 115 has an execution instruction unit 541, a coefficient change unit 542, and an arithmetic expression information acquisition unit 543. Below, the details of each unit that the first stage execution unit 112, the second stage execution unit 113, the output unit 114, and the third stage execution unit 115 have will be described.

[0070] The calculation formula information acquisition unit 513 reads out the cost calculation formula information and the constraint term calculation formula information from the optimization information storage unit 116 and notifies the cost calculation formula information modification unit 512 of them.

[0071] The cost calculation formula information modification unit 512 acquires cost calculation formula information from the calculation formula information acquisition unit 513, and modifies the acquired cost calculation formula information so as to reduce the amount of calculation (number of bits) of the Ising device 120 when searching for an optimal arrangement combination. Specifically, the cost calculation formula information modification unit 512 generates modified cost calculation formula information shown in the following formula (8) by omitting a term that expresses the difference between the L body and D body of the coarse-grained particles in the cost calculation formula information 300.

[0072]

number

[0073] In the case of the above formula (8), the amount of calculation (number of bits) can be reduced compared to the cost calculation formula information shown in the above formula (1), so the search time when quickly calculating the first search result can be significantly reduced.

[0074] Furthermore, the cost calculation formula information transformation unit 512 notifies the execution instruction unit 511 of the transformed cost calculation formula information and the constraint term calculation formula information.

[0075] The execution instruction unit 511 acquires a coarse-grained model of a predetermined initial placement from the search target information acquisition unit 111, and also acquires cost calculation formula information and constraint term calculation formula information after transformation from the cost calculation formula information transformation unit 512. Furthermore, the execution instruction unit 511 transmits first search information including the coarse-grained model of a predetermined initial placement and the cost calculation formula information and constraint term calculation formula information after transformation to the Ising device 120, and instructs it to search for an optimal placement combination.

[0076] The search result acquisition unit 521 acquires the first search result from the Ising device 120 and notifies the execution instruction unit 522 of the first search result.

[0077] The calculation formula information acquisition unit 523 reads out the cost calculation formula information and the constraint term calculation formula information from the optimization information storage unit 116 and notifies the execution instruction unit 522 of them.

[0078] The execution instruction unit 522 acquires the first search result from the search result acquisition unit 521, and acquires cost calculation formula information and constraint term calculation formula information from the calculation formula information acquisition unit 523. The execution instruction unit 522 also transmits second search information including a coarse-grained model with the acquired first search result as an initial arrangement and untransformed cost calculation formula information and constraint term calculation formula information to the Ising device 120, and instructs the Ising device 120 to search for an optimal arrangement combination. In the case of the second search information, an optimal arrangement combination is searched for with the first search result as the initial arrangement. Therefore, even if a term expressing the difference between the L-body and D-body of the coarse-grained particles is included (that is, even for an arrangement taking the L-body / D-body into consideration), a highly accurate second search result can be calculated quickly.

[0079] The search result acquisition unit 531 acquires the second search result or the third search result from the Ising device 120 and notifies the evaluation unit 532 of the result.

[0080] The evaluation unit 532 evaluates the second search result or the third search result notified by the search result acquisition unit 531, and determines whether there is room for improvement in the optimal layout combination. If it is determined that there is no room for improvement, the evaluation unit 532 outputs the immediately preceding search result as the optimal solution.

[0081] The immediately preceding search result is, for example, when the third search result is notified after the second search result is notified, and if the optimal placement combination has not been updated between the second and third search results, the second search result is output as the optimal solution. Also, for example, when the kth (k is any integer) third search result is notified after the k+1th third search result is notified, and if the optimal placement combination has not been updated between the kth and k+1th search results, the kth third search result is output as the optimal solution.

[0082] Furthermore, if the evaluation unit 532 determines that there is room for improvement in the optimal layout combination, it notifies the execution instruction unit 541 of the second search result or the third search result notified by the search result acquisition unit 531.

[0083] The calculation formula information acquisition unit 543 reads out the cost calculation formula information and the constraint term calculation formula information from the optimization information storage unit 116 and notifies the coefficient change unit 542 of them.

[0084] The coefficient change unit 542 acquires the cost calculation formula information and the constraint term calculation formula information from the calculation formula information acquisition unit 543, and changes the coefficients included in the constraint term calculation formula information. In addition, the coefficient change unit 542 notifies the execution instruction unit 541 of the constraint term calculation formula information and the cost calculation formula information after the change.

[0085] The execution instruction unit 541 acquires the second search result or the third search result from the evaluation unit 532, and acquires the changed constraint term operation formula information and cost operation formula information from the coefficient change unit 542. Furthermore, the execution instruction unit 541 transmits to the Ising device 120 third search information including a coarse-grained model whose initial arrangement is the acquired second search result or the third search result, cost operation formula information, and changed constraint term operation formula information, and instructs the Ising device 120 to search for an optimal arrangement combination. In the case of the third search information, the optimal arrangement combination is searched for using the changed constraint term operation formula information and cost operation formula information, so that it is possible to calculate a highly accurate search result without falling into a local solution.

[0086] This is because when coarse-grained particles are placed in an artificial lattice space, the position and shape are different from those of actual molecules, and depending on the contribution of constraints, this can increase the local energy barrier, making it difficult to move to the vicinity of the optimal solution. However, according to the third search information, The coefficient change unit 542 changes the coefficients of the constraint term calculation formula information, -Temporarily weakening constraints, This makes it easier to move to the vicinity of the optimal solution, increasing the possibility of arriving at a more accurate solution.

[0087] <Details of the functional configuration of the Ising device> Next, the functional configuration of the Ising device 120 will be described in detail. FIG. 6 is a diagram showing an example of the functional configuration of the Ising device. As described using FIG. 1, the Ising device 120 functions as a combination optimization unit 121. As shown in FIG. 6, the combination optimization unit 121 further includes a QUBO generation unit 611, a conversion unit 612, and an optimal solution calculation unit 613.

[0088] The QUBO generation unit 611 acquires first search information, second search information, and third search information from the terminal device 110. The QUBO generation unit 611 also converts the transformed cost calculation formula information and constraint term calculation formula information included in the first search information into a QUBO (Quadratic Unconstrained Binary Optimization). The QUBO generation unit 611 also converts the untransformed cost calculation formula information and constraint term calculation formula information included in the second search information into a QUBO. The QUBO generation unit 611 also converts the cost calculation formula information and changed constraint term calculation formula information included in the third search information into a QUBO.

[0089] The conversion unit 612 converts the converted QUBO into a format suitable for Digital Annealer.

[0090] The optimal solution calculation unit 613 searches for an optimal placement combination by using Digital Annealer to solve the QUBO converted into a format suitable for Digital Annealer. The optimal solution calculation unit 613 also transmits the optimal placement combination searched for based on the first search information to the terminal device 110 as a first search result. The optimal solution calculation unit 613 also transmits the optimal placement combination searched for based on the second search information to the terminal device 110 as a second search result. The optimal solution calculation unit 613 also transmits the optimal placement combination searched for based on the third search information to the terminal device 110 as a third search result.

[0091] <Search process flow> Next, we will explain the flow of search processing by search system 100. Fig. 7 is a flowchart showing the flow of search processing.

[0092] In step S701, the search target information acquisition unit 111 of the terminal device 110 acquires a coarse-grained model of the medium molecule to be searched for.

[0093] In step S702, the first stage execution unit 112 of the terminal device 110 generates transformed cost calculation formula information in which the term expressing the difference between the L body and the D body of the coarse-grained particles is omitted from the read cost calculation formula information 300.

[0094] In step S703, the first-stage execution unit 112 of the terminal device 110 instructs the Ising device 120 to search for an optimal arrangement combination for the coarse-grained model of the medium molecule to be searched for. At this time, the first-stage execution unit 112 of the terminal device 110 instructs the Ising device 120 to search for an optimal arrangement combination using the transformed cost operation formula information and constraint term operation formula information 410 to 440.

[0095] In step S704, the combination optimization unit 121 of the Ising device 120 calculates the first search result, and the second stage execution unit 113 of the terminal device 110 acquires the first search result from the combination optimization unit 121 of the Ising device 120.

[0096] In step S705, the second-stage execution unit 113 of the terminal device 110 instructs the Ising device 120 to search for an optimal combination of arrangements for the coarse-grained model of the medium molecule being searched for, using the first search result as the initial arrangement. At this time, the second-stage execution unit 113 of the terminal device 110 instructs the Ising device 120 to search for an optimal combination of arrangements using the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440.

[0097] In step S706, the combination optimization unit 121 of the Ising device 120 calculates the second search result, and the output unit 114 of the terminal device 110 acquires the second search result from the combination optimization unit 121 of the Ising device 120.

[0098] In step S707, the third stage execution unit 115 of the terminal device 110 changes one or more coefficients of the read constraint term computation equation information 410 to 440, and generates the changed constraint term computation equation information.

[0099] In step S708, the third-stage execution unit 115 of the terminal device 110 instructs the Ising device 120 to search for an optimal combination of arrangements for the coarse-grained model of the medium molecule to be searched for, using as the initial arrangement the latest search result acquired by the output unit 114. At this time, the third-stage execution unit 115 of the terminal device 110 instructs the Ising device 120 to search for an optimal combination of arrangements using the cost calculation formula information 300 and the changed constraint term calculation formula information.

[0100] In step S709, the combination optimization unit 121 of the Ising device 120 calculates the third search result, and the output unit 114 of the terminal device 110 acquires the third search result from the combination optimization unit 121 of the Ising device 120.

[0101] In step S710, the output unit 114 of the terminal device 110 determines whether there is room for improvement for the third search result acquired in step S709. If it is determined that there is room for improvement in step S710 (YES in step S710), the process returns to step S707.

[0102] In this case, the third stage execution unit 115 of the terminal device 110 generates constraint term operation formula information after the second change (step S707), and issues an instruction to search for an optimal combination of placements using the first third search result as the initial placement (step S708). As a result, the output unit 114 of the terminal device 110 obtains the second third search result (step S709).

[0103] Thereafter, in the search system 100, the processing of steps S707 to S710 is repeated until the output unit 114 of the terminal device 110 determines in step S710 that there is no room for improvement.

[0104] On the other hand, if it is determined in step S710 that there is no room for improvement (NO in step S710), the process proceeds to step S711.

[0105] In step S711, the output unit 114 of the terminal device 110 outputs the search result immediately before it is determined that there is no room for improvement as the optimal solution.

[0106] <Example of search processing> Next, we will explain a specific example of the search process by the search system 100. Fig. 8 is a diagram showing a specific example of the search process, and shows how the optimal arrangement combination is searched for using coarse-grained models of three types of peptides (peptide A (number of residues = 12), peptide B (number of residues = 10), and peptide C (number of residues = 15)) as search targets.

[0107] Of these, FIG. 8(a) is a comparative example. -Without using the cost calculation formula information after transformation, and Without using the constraint expression information after the change, It shows a case where an optimum combination of placements is searched for (that is, it shows a case where an optimum combination of placements is searched for using the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440).

[0108] Specifically, when the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide A is "60,671", and by searching for the optimal arrangement combination, -Cost reduced by "1282" -The search time took 255 seconds. This shows that...

[0109] Similarly, when the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide B is "42,918", and by searching for the optimal arrangement combination, -Cost reduced by "966" - The search time was 188 seconds. This shows that...

[0110] Similarly, when the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide B is "42,918", and by searching for the optimal arrangement combination, -Cost reduced by "926" -The search time was "855" seconds. This shows that...

[0111] On the other hand, FIG. 8(b) shows the case where the search process shown in FIG. 7 is executed (i.e., using the transformed cost calculation formula information and the changed constraint term calculation formula information) to search for the optimal placement combination.

[0112] Specifically, in the first stage, when the transformed cost calculation formula information and constraint term calculation formula information 410 to 440 are used, the number of bits of peptide A becomes "8,192", and by searching for the optimal arrangement combination, -Cost reduced by "1387" -The search time took "41" seconds. This shows that...

[0113] Similarly, in the first stage, when the transformed cost calculation formula information and constraint term calculation formula information 410 to 440 are used, the number of bits of peptide B becomes "8,192", and by searching for the optimal arrangement combination, -Cost reduced by "960" -The search time took 42 seconds. This shows that...

[0114] Similarly, in the first stage, when the transformed cost calculation formula information and constraint term calculation formula information 410 to 440 are used, the number of bits of peptide C becomes "14,298", and by searching for the optimal arrangement combination, -Cost reduced by "1172" -The search time was 712 seconds. This shows that...

[0115] Furthermore, in the second stage, when the unmodified cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide A is "60,671", but by searching for the optimal combination of arrangements using the first search result as the initial arrangement, -Cost reduced by "1357" - Search time became "0" seconds, This shows that...

[0116] Similarly, in the second stage, when the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide B is "42,918", but by searching for the optimal arrangement combination using the first search result as the initial arrangement, -Cost reduced by "930" - Search time became "0" seconds, This shows that...

[0117] Similarly, in the second stage, when the cost calculation formula information 300 and the constraint term calculation formula information 410 to 440 are used, the number of bits of peptide C is "167,256", but by searching for the optimal combination of arrangements using the first search result as the initial arrangement, -Cost reduced by "1112" - Search time is now 1 second. This shows that...

[0118] In addition, in the third stage, when the cost calculation formula information 300 and the constraint term calculation formula information after the change in which the coefficient is multiplied by "0.6" are used, the second search result is used as the initial placement to search for the optimal placement combination, and the following is obtained: -Cost reduced by "1357" - Search time became "0" seconds, This shows that...

[0119] Similarly, in the third stage, when the cost calculation formula information 300 and the constraint term calculation formula information after the change in which the coefficient is multiplied by "0.6" are used, the second search result is used as the initial placement to search for the optimal placement combination, and the following is obtained: - Costs reduced by 1001 - Search time became "0" seconds, This shows that...

[0120] Similarly, in the third stage, when the cost calculation formula information 300 and the constraint term calculation formula information after the change in which the coefficient is multiplied by "0.8" are used, the second search result is used as the initial placement to search for the optimal placement combination, and the following is obtained: -Cost reduced by "1132" - Search time is now 1 second. This shows that...

[0121] Comparing Figure 8(a) and Figure 8(b), it is clear that peptides A to C all have The cost reduction is larger in Figure 8(b) (see "Cost" in "Third Stage" in Figure 8(b)). The search time is shorter in Figure 8(b) (see "Total time" in Figure 8(b)). In other words, the search system 100 can efficiently (highly accurately and quickly) search for optimal combinations of arrangements in a coarse-grained model of a medium-sized molecule.

[0122] As is clear from the above description, the terminal device 110 according to the first embodiment instructs a multi-stage search for arrangement combinations in a coarse-grained model of a medium molecule as a combinatorial optimization problem. In this case, the terminal device 110 according to the first embodiment instructs a search for an arrangement combination that minimizes cost in the first stage by calculating the cost of the coarse-grained model while omitting distinction between L- and D-forms of amino acids. Furthermore, the terminal device 110 according to the first embodiment instructs a search for an arrangement that minimizes cost in the second stage by calculating the cost of the coarse-grained model while distinguishing between L- and D-forms of amino acids, using the arrangement searched in the first stage as the initial arrangement. Furthermore, the terminal device 110 according to the first embodiment outputs the arrangement searched in the second stage as the optimal arrangement combination in the coarse-grained model.

[0123] As a result, according to the first embodiment, it is possible to quickly search for a highly accurate combination of arrangements. In other words, according to the first embodiment, it is possible to efficiently search for an optimal combination of arrangements in a coarse-grained model of a medium-sized molecule.

[0124] [Second embodiment] In the first embodiment, the QUBO generation unit 611 has been described as sequentially converting the first search information and the second search information. However, the timing of conversion by the QUBO generation unit 611 is not limited to this. For example, the transformed cost calculation formula information included in the first search information and the untransformed cost calculation formula information included in the second search information may be converted before the optimal solution calculation unit 613 searches for an optimal arrangement combination.

[0125] Furthermore, in the first embodiment described above, the search system 100 has been described as including the terminal device 110 and the Ising device 120. However, the search system 100 may be formed by devices other than the terminal device 110 and the Ising device 120 (i.e., three or more devices). Alternatively, the search system 100 may be formed by an integrated device (i.e., one device). In this case, the search process (FIG. 7) is realized, for example, by the one device executing a search program (search program + optimization program).

[0126] In the above first embodiment, the terminal device 110 has been described as having the search target information acquisition unit 111 to the third stage execution unit 115, and the Ising device 120 has the combination optimization unit 121. However, some of the functions of the terminal device 110 may be realized in the Ising device 120. Similarly, some of the functions of the Ising device 120 may be realized in the terminal device 110.

[0127] The disclosed technology may take the following forms as described below. (Appendix 1) The computer is instructed to search for combinations of amino acid arrangements contained in the medium molecule in a multi-stage combinatorial optimization problem. instructing to search for a first configuration of the plurality of amino acids by calculating a value of a first cost calculation formula that does not distinguish between L- and D-isomers of the plurality of amino acids; instructing to search for a second configuration of the plurality of amino acids by calculating a value of a second cost calculation formula that distinguishes between the L-form and the D-form of the plurality of amino acids, using the first configuration found as an initial configuration; outputting the second configuration; A search program for executing processing. (Appendix 2) instructing to search for a third arrangement of the plurality of amino acids while changing a coefficient of any one of a plurality of constraint terms that constrain the arrangement of the plurality of amino acids; outputting the second arrangement if the third arrangement has not been updated from the second arrangement; 2. A search program according to claim 1, further causing the computer to execute a process. (Appendix 3) 3. The search program according to claim 2, wherein the search program instructs the program to search for the third arrangement while lowering the coefficient of any one of the plurality of constraint terms. (Appendix 4) 3. The search program of claim 2, wherein the plurality of constraint terms include a constraint term for ensuring that each of the plurality of amino acids contained in the medium molecule is present at only one position in lattice space. (Appendix 5) 3. The search program of claim 2, wherein the plurality of constraint terms include a constraint term for ensuring that each of the plurality of amino acids contained in the medium molecule does not exist at the same lattice point. (Appendix 6) 3. The search program of claim 2, wherein the plurality of constraints include a constraint for ensuring that each of the plurality of amino acids contained in the medium molecule is linked to each other in a predetermined sequence. (Appendix 7) 3. The search program of claim 2, wherein the plurality of constraint terms include a constraint term for ensuring that, for the plurality of amino acids contained in the medium molecule, no unlinked amino acids exist in adjacent lattices. (Appendix 8) 4. A search program according to claim 3, wherein, in the search for the third arrangement, if the arrangement searched when the coefficient of the constraint term is lowered the kth time (k is any integer) and the arrangement searched when the coefficient of the constraint term is lowered the k+1th time has not been updated, the arrangement searched when the coefficient of the constraint term is lowered the kth time is output. (Appendix 9) A computer is instructed to search for combinations of amino acid arrangements contained in a medium molecule in a multi-stage manner as a combinatorial optimization problem. instructing to search for a first configuration of the plurality of amino acids by calculating a value of a first cost calculation formula that does not distinguish between L- and D-isomers of the plurality of amino acids; instructing to search for a second configuration of the plurality of amino acids by calculating a value of a second cost calculation formula that distinguishes between the L-form and the D-form of the plurality of amino acids, using the first configuration found as an initial configuration; outputting the second configuration; The discovery method to perform the operation. (Appendix 10) A search device that instructs a multi-stage search for combinations of arrangements of a plurality of amino acids contained in a medium molecule as a combinatorial optimization problem, the search device comprising: a first stage execution unit that instructs to search for a first arrangement of the plurality of amino acids by calculating a value of a first cost calculation formula that does not distinguish between L- and D-isomers of the plurality of amino acids; a second stage execution unit that instructs to search for a second arrangement of the plurality of amino acids by calculating a value of a second cost calculation formula that distinguishes between the L-form and the D-form of the plurality of amino acids, using the searched first arrangement as an initial arrangement; an output unit that outputs the second arrangement; A search device having the above.

[0128] The present invention is not limited to the configurations described in the above embodiments, but may be combined with other elements, etc. These aspects can be changed without departing from the spirit of the present invention, and can be appropriately determined depending on the application form. [Explanation of symbols]

[0129] 100: Exploration System 110: Terminal device 111: Search target information acquisition unit 112: First stage execution unit 113: Second stage execution unit 114: Output section 115: Third Stage Execution Unit 121: Combinatorial Optimization Department 300: Cost calculation formula information 310: Interaction energy information between coarse-grained particles 311: Library Information 320: Information on the distinction between L-isomer and D-isomer 410~440: Constraint term operation expression information 511: Execution instruction section 512: Cost calculation formula information transformation part 513: Arithmetic expression information acquisition unit 521:Search result acquisition unit 522: Execution instruction section 523: Arithmetic expression information acquisition unit 531:Search result acquisition unit 532: Evaluation section 541: Execution instruction section 542: Coefficient change unit 543: Arithmetic expression information acquisition unit 611 :QUBO generation part 612: Conversion section 613: Optimal solution calculation unit

Claims

1. a computer that is instructed to search for combinations of arrangements of coarse-grained models of molecular structures of medium-sized molecules containing a plurality of amino acids and having a molecular weight of 500 to 3000 as a combinatorial optimization problem; instructing to search for a first configuration of the coarse-grained model by calculating a value of a first cost calculation formula in which a term expressing a difference between an L-body and a D-body of the coarse-grained model is omitted; instructing to search for a second arrangement of the coarse-grained model by calculating a value of a second cost calculation formula including a term expressing a difference between the L-body and the D-body of the coarse-grained model using the searched first arrangement as an initial arrangement; outputting the second configuration; A search program for executing processing.

2. instructing a search for a third placement of the coarse-grained model while changing a coefficient of any one of a plurality of constraint terms that constrain the placement of the coarse-grained model; outputting the second arrangement if the third arrangement has not been updated from the second arrangement; The search program according to claim 1 , further causing the computer to execute a process.

3. 3. The search program according to claim 2, wherein the search program instructs that the third arrangement be searched for while decreasing the coefficient of any one of the plurality of constraint terms.

4. 4. The search program according to claim 3, wherein, in the search for the third arrangement, if the arrangement searched when the coefficient of the constraint term is lowered the kth time (k is an arbitrary integer) and the arrangement searched when the coefficient of the constraint term is lowered the k+1th time has not been updated, the arrangement searched when the coefficient of the constraint term is lowered the kth time is output.

5. a computer that instructs the computer to search for combinations of arrangements of a coarse-grained model of a molecular structure of a medium molecule having a molecular weight of 500 to 3000 and containing a plurality of amino acids as a combinatorial optimization problem; instructing to search for a first configuration of the coarse-grained model by calculating a value of a first cost calculation formula in which a term expressing a difference between an L-body and a D-body of the coarse-grained model is omitted; instructing to search for a second arrangement of the coarse-grained model by calculating a value of a second cost calculation formula including a term expressing a difference between the L-body and the D-body of the coarse-grained model using the searched first arrangement as an initial arrangement; outputting the second configuration; The discovery method to perform the operation.

6. A search device that instructs searching for combinations of arrangements of a coarse-grained model of a molecular structure of a medium molecule having a molecular weight of 500 to 3000 and containing a plurality of amino acids, as a combinatorial optimization problem, comprising: a first stage execution unit that instructs searching for a first arrangement of the coarse-grained model by calculating a value of a first cost calculation formula in which a term expressing a difference between an L-body and a D-body of the coarse-grained model is omitted; a second stage execution unit that instructs to search for a second arrangement of the coarse-grained model by calculating a value of a second cost calculation formula including a term that expresses a difference between the L-body and the D-body of the coarse-grained model, using the searched first arrangement as an initial arrangement; an output unit that outputs the second arrangement; A search device having the above.

Images