Variable-structure path prediction device, variable-structure path prediction result acquisition method, variable-structure path prediction method, and program

Abstract

A variable-structure route prediction device according to an embodiment includes a network map generation unit and a route calculation unit. The network map generation unit generates a network map on the basis of stereostructure information of a protein, the network map having vertices corresponding to at least the amino acid residues constituting the protein and any binding substance that binds to the protein, and the edges of the network map having weights assigned to the vertices and the edges of the vertices, the vertices corresponding to at least the amino acid residues constituting the protein and any binding substance that binds to the protein. The weight is based on the interaction between the amino acid residues and at least the amino acid residues in the arbitrary binding substance. The path calculation unit calculates, on the basis of an evaluation function based on the weights, a path connecting vertices on the network diagram generated by the network diagram generation unit.

Description

Technical Field

[0001] The present invention relates to an allosteric pathway prediction device, an allosteric pathway prediction result acquisition method, an allosteric pathway prediction method and a program.

[0002] This application claims priority based on Japanese Patent Application No. 2022-102155 filed in Japan on June 24, 2022, the contents of which are cited herein. Background Art

[0003] In recent years, in the drug development market, research and development investment has tended to expand, but the exhaustion of drug development targets has become significant. Therefore, it is expected that new drug development targets will be derived using IT technology. For example, through allosteric drug development, proteins that are difficult to develop in the prior art may become drug development targets. Allosteric drug development refers to a drug development approach that targets the mechanism of allosteric control of proteins. Allosteric control refers to the control of the structure, activity or reaction of the active site by the signal of the allosteric control site of the protein. In allosteric drug development, the allosteric control site of the protein, or the mechanism (allosteric pathway) by which the signal to the allosteric control site is transmitted to the active site is targeted.

[0004] The prediction of allosteric control sites, or how signals to allosteric control sites are transmitted to the active site (called allosteric pathway), is very important for the development of allosteric drugs.

[0005] Prior art literature

[0006] Patent Literature

[0007] Patent Document 1: Chinese Patent No. 108830043

[0008] Non-patent literature

[0009] Non-patent document 1: "Biophysical Journal", February 9, 2007, Vol. 92, pp. 3052-3062 ("Biophysical Journal", February 9, 2007, Vol. 92, pp. 3052-3062)

[0010] Non-patent document 2: “NATURE COMUNICATIONS”, July 31, 2020, Vol. 11, p. 3862 (Nature Communications, July 31, 2020, Vol. 11, p. 3862)

[0011] Non-patent document 3: "Scientific Reports", February 22, 2016, Vol. 6, p. 21686 (Scientific Reports, February 22, 2016, Vol. 6, p. 21686) Summary of the invention

[0012] Technical problem to be solved by the invention

[0013] The technical problem to be solved by the present invention is to provide an allosteric pathway prediction device, an allosteric pathway prediction result acquisition method, an allosteric pathway prediction method and a program that can predict amino acid residues that contribute to allosteric control based on the three-dimensional structural information of a protein.

[0014] Means for solving technical problems

[0015] The allosteric pathway prediction device of the embodiment comprises a network graph generation unit and a pathway calculation unit. The network graph generation unit generates a network graph based on the three-dimensional structural information of the protein, wherein the network graph has vertices corresponding to at least one of the amino acid residues constituting the protein and any binding substance binding to the protein, and its edges are assigned weights based on the interaction between the amino acid residue and at least one of the amino acid residues in the any binding substance. The pathway calculation unit calculates a pathway formed by connecting the vertices on the network graph generated by the network graph generation unit according to an evaluation function based on the weights. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a diagram showing an example of the three-dimensional structure of the protein according to the first embodiment of the present invention.

[0017] Figure 2 This is a diagram showing an example of a network diagram expressing the three-dimensional structure of the protein according to the first embodiment of the present invention.

[0018] Figure 3 This is a diagram showing an example of a path according to the first embodiment of the present invention.

[0019] Figure 4 It is a diagram for explaining the signal transmission probability according to the first embodiment of the present invention.

[0020] Figure 5 This is a diagram for explaining the probability of signal transmission from the start point to the end point of a path according to the first embodiment of the present invention.

[0021] Figure 6 It is a diagram showing an example of the functional structure of the allosteric pathway prediction device according to the first embodiment of the present invention.

[0022] Figure 7 It is a diagram showing an example of the allosteric pathway prediction process according to the first embodiment of the present invention.

[0023] Figure 8 This is a diagram showing an example of the network graph generation process according to the first embodiment of the present invention.

[0024] Fig. 9 This is a diagram showing an example of the Ising calculation process according to the first embodiment of the present invention.

[0025] Fig.10 It is a diagram for explaining the oriented edges according to the first embodiment of the present invention.

[0026] Fig.11 This is a diagram showing an example of the Hamiltonian for a single path according to the first embodiment of the present invention.

[0027] Fig.12 This is a diagram showing an example of the calculation result of the path according to the first embodiment of the present invention.

[0028] Fig.13 This is a diagram showing an example of the outline of the method for searching for amino acid residues contributing to allosteric control according to the second embodiment of the present invention.

[0029] Fig.14 This is a diagram showing an example of the value of the residue score according to the second embodiment of the present invention.

[0030] Fig.15 It is a diagram showing an example of the functional structure of the allosteric pathway prediction device according to the second embodiment of the present invention.

[0031] Fig.16 It is a diagram showing an example of the allosteric pathway prediction process according to the second embodiment of the present invention.

[0032] Fig.17 This is a diagram showing an example of the important amino acid residue search process according to the second embodiment of the present invention.

[0033] Fig.18 This is a diagram showing a network graph of the search results of important amino acid residues in the first example of the second embodiment of the present invention.

[0034] Fig.19 This is a diagram showing the search results of important amino acid residues in the first example of the second embodiment of the present invention.

[0035] Fig. 20 This is a diagram showing a network graph of search results for important amino acid residues in the second example of the second embodiment of the present invention.

[0036] Fig.21 This is a diagram showing the search results of important amino acid residues in the second example of the second embodiment of the present invention.

[0037] Fig. 22 This is a diagram showing a network diagram showing the search results of important amino acid residues in the second example of the second embodiment of the present invention.

[0038] Fig.23 This is a diagram showing the search results of important amino acid residues in the second example of the second embodiment of the present invention.

[0039] Fig.24 This is a diagram showing an example of the outline of the allosteric region search method according to the third embodiment of the present invention.

[0040] Fig.25 This is a diagram showing an example of the results of evaluating the influence of each of a plurality of starting points on the end point according to the third embodiment of the present invention.

[0041] Fig.26 It is a diagram showing an example of the functional structure of the allosteric pathway prediction device according to the third embodiment of the present invention.

[0042] Fig. 27 It is a diagram showing an example of the allosteric pathway prediction process according to the third embodiment of the present invention.

[0043] Fig.28 This is a diagram showing an example of Ising calculation processing according to the third embodiment of the present invention.

[0044] Fig.29 It is a diagram showing an example of the area search process according to the third embodiment of the present invention.

[0045] Fig.30 This is a diagram showing the search results of the allosteric control region of the amino acid region of KRAS with amino acid numbers 1 to 67 according to the examples of the third embodiment of the present invention.

[0046] Fig.31 This is a diagram showing the search results of the allosteric control region of the amino acid region of KRAS from amino acid number 114 to 169 according to the examples of the third embodiment of the present invention.

[0047] Fig.32 This is a diagram showing the search results of the allosteric control region of the entire KRAS in Examples of the third embodiment of the present invention together with the positions of α-helical chains.

[0048] Fig.33 This is a diagram showing an example of the outline of a method for searching for an allosteric region according to a modification of the third embodiment of the present invention.

[0049] Fig.34 This is a diagram showing an example of the outline of a method for searching for an allosteric region according to a modification of the third embodiment of the present invention.

[0050] Fig.35 It is a diagram showing an example of the functional structure of an allosteric path prediction device according to a modified example of the third embodiment of the present invention.

[0051] Fig.36 It is a diagram showing an example of the allosteric pathway prediction process according to the modification example of the third embodiment of the present invention.

[0052] Fig.37 It is a diagram showing an example of a residue set search process according to a modification of the third embodiment of the present invention.

[0053] Fig.38 It is a diagram showing an example of a residue set search Ising calculation process according to a modification of the third embodiment of the present invention.

[0054] Fig.39 This is a diagram showing an example of an outline of a process for calculating a residue set score according to a modification of the third embodiment of the present invention.

[0055] Fig.40 This is a diagram showing an example of distances between residues according to a modification of the third embodiment of the present invention.

[0056] Fig.41 This is a diagram showing an example of an overview of search results of a residue set according to a modification of the third embodiment of the present invention.

[0057] Fig.42 The search results of the allosteric control region of the modification example of the third embodiment of the present invention are shown.

[0058] Fig.43 It is a diagram showing a ribbon model of a KRAS / RAF complex according to a modification of the third embodiment of the present invention.

[0059] Fig.44 It is a diagram showing a belt model of a KRAS / RAF complex according to a modification of the third embodiment of the present invention.

[0060] Fig.45 It is a diagram showing a belt model of a KRAS / RAF complex according to a modification of the third embodiment of the present invention.

[0061] Fig.46 This is a diagram showing an example of multipath according to the fourth embodiment of the present invention.

[0062] Fig.47 1 is a diagram showing an example of the functional configuration of an allosteric pathway prediction device according to the fourth embodiment of the present invention.

[0063] Fig.48 This is a diagram showing an example of the Hamiltonian for multiple paths according to the fourth embodiment of the present invention.

[0064] Fig.49 This is a diagram showing the search results of important amino acid residues in Examples of the fourth embodiment of the present invention.

[0065] Fig.50 This is a diagram showing an example of allosteric pathways calculated using the Hamiltonian for multiple pathways in the example of the fourth embodiment of the present invention.

[0066] Fig.51 This is a diagram showing the search results of the allosteric control region of the second example of the fourth embodiment of the present invention.

[0067] Fig.52 This is a diagram showing the search results of the allosteric control region of the second example of the fourth embodiment of the present invention.

[0068] Fig.53 : is a diagram showing an example of the configuration of an allosteric pathway prediction system according to the fifth embodiment of the present invention.

[0069] Fig.54 It is a diagram showing an example of the allosteric pathway prediction process according to the fifth embodiment of the present invention.

[0070] Fig.55 This is a diagram showing an example of a network graph generation process according to the fifth embodiment of the present invention.

[0071] Fig.56 This is a diagram showing an example of the setting process of QUBO and constraint list according to the fifth embodiment of the present invention.

[0072] Fig.57 It is a diagram showing an example of the system configuration of a first modified example of the fifth embodiment of the present invention.

[0073] Fig.58 It is a diagram showing an example of the configuration of a system according to a second modified example of the fifth embodiment of the present invention. DETAILED DESCRIPTION

[0074] In the present specification and claims, "protein" refers to a molecule having a polypeptide chain formed by a plurality of amino acids linked by peptide bonds. The number of amino acids constituting a protein is not particularly limited, and may be less than 100, less than 1000, less than 2000, less than 3000, or less than 5000. The upper limit is not particularly limited, and as an example, less than 5000 is listed as a standard. The amino acids constituting a protein are not limited to naturally occurring amino acids, but may also be non-natural artificially synthesized amino acids. Non-natural amino acids include amino acids after the amino acid residues are arbitrarily chemically modified. The amino acid sequence possessed by a protein is not limited to naturally occurring amino acid sequences, but may also be non-natural artificially designed amino acid sequences. Proteins are generally roughly divided into water-soluble proteins and water-insoluble proteins. The protein as the object of the present invention may be any protein, or may be a water-insoluble membrane protein. The protein as the object of the present invention may be any protein that can obtain a three-dimensional structure. Even if it is a protein that is difficult to obtain a three-dimensional structure experimentally, as long as it is a protein that can predict the three-dimensional structure by a three-dimensional structure prediction system such as AlphaFold2. Acquiring a three-dimensional structure experimentally means acquiring a three-dimensional structure by X-ray structural analysis, analysis based on imaging results of a cryo-electron microscope, NMR analysis, or the like.

[0075] (First Embodiment)

[0076] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0077] First, refer to Figures 1 to 3 , describing the modeling overview of allosteric control in the allosteric pathway prediction method of this embodiment. In this embodiment, allosteric control refers to the signal control of the allosteric control site of the protein, either one or more of the three-dimensional structure or reactivity (also referred to as activity) of the active site.

[0078] As an example of positive control of the reaction that has attracted attention, the following examples are listed. That is, in a state where allosteric control is not performed, the active site is not a three-dimensional structure that can bind to the substrate that should specifically bind to the active site, or is not in a state suitable for reaction. On the other hand, in a state where allosteric control is performed by causing an exogenous allosteric regulatory molecule such as a drug to act on an allosteric control site that is located far from the active site of the target protein, any one or more of the three-dimensional structure or reactivity of the active site changes. The active site after the change binds to the specific substrate, etc. and reacts.

[0079] As an example of negative control that inhibits the reaction of interest, an external allosteric regulatory molecule such as a drug is allowed to act on the allosteric control site of the target protein to change one or more of the three-dimensional structure and reactivity of the active site, thereby inhibiting the binding and reaction that should have occurred.

[0080] Figure 1 The stereostructure T1 shown is an example of a stereostructure of a protein. The stereostructure T1 includes an allosteric control site T11 and an active site T12. In the allosteric control in the stereostructure T1, a signal to the allosteric control site T11 changes one or more of the stereostructure or reactivity of the active site T12 located away from the allosteric control site T11.

[0081] Figure 2 The network diagram G2 shown is a network diagram representing the three-dimensional structure of a protein. The network diagram G2 includes a plurality of vertices and edges connecting the plurality of vertices. The plurality of vertices correspond to amino acid residues, respectively. Vertex N22 corresponds to the allosteric control site. Vertex N210 corresponds to the active site T12. Weights based on the interactions between amino acid residues are assigned to the edges connecting the plurality of vertices. The edges connecting the plurality of vertices include edges representing the main chain and edges representing the interactions between side chains. Figure 2 In FIG, the edges representing the main chain are represented by solid lines, and the edges representing the interactions between the side chains are represented by dashed lines.

[0082] In the allosteric pathway prediction method of the present embodiment, allosteric control is modeled as signal transmission. Signal transmission is represented as a path between two vertices in a network diagram. This path is referred to as a signal transmission path or an allosteric path. The allosteric control in the three-dimensional structure of a protein is modeled as, for example, signal transmission from vertex N22 to vertex N210 using a network diagram G2. Figure 3 The path P21 shown is a signal transmission path corresponding to signal transmission from the vertex N22 to the vertex N210.

[0083] In the allosteric pathway prediction method of the present embodiment, the signal transmission pathway is predicted based on the Ising model. In the allosteric pathway prediction method of the present embodiment, based on the predicted signal transmission pathway, the region important for allosteric control is identified in the three-dimensional structure of the protein.

[0084] Next, refer to Figure 4 , describing the weights assigned to the edges connecting multiple vertices in the network graph based on the interactions between amino acid residues. In the allosteric pathway prediction method of the present embodiment, the signal transmission probability between amino acid residues is used as the weight. Figure 4 is a diagram for explaining the signal transmission probability of this embodiment. Figure 4 In the above example, the ith amino acid residue in the protein is taken as the amino acid residue R i, taking the jth amino acid residue as the amino acid residue R j The amino acid residue R i The number of atoms C i is 9, constituting the amino acid residue R j The number of atoms C j is 8.

[0085] The amino acid residue R i The position vector representing the position of the a-th atom is set as vector r a i . The amino acid residue R j The position vector representing the position of the b-th atom is set to vector r b j . Will constitute the amino acid residue R i The one atom of the amino acid residue R j The number of pairs of atoms in which the distance between the two atoms constituting the group is less than the cutoff value (also called threshold) r0 is set as the pairing number C. ij . Pairing number C ij The Heaviside step function H is used and represented by equation (1).

[0086] [Formula 1]

[0087]

[0088] exist Figure 4 In the example shown, the number of pairs C ij is 3.

[0089] In addition, in this embodiment, the pairing number C ij In the calculation of the distance between two atoms, the distance between the atoms of the main chain and the side chain is described as an example, but the distance between two atoms can be calculated including the atoms of the main chain together with the atoms of the side chain.

[0090] Here, the amino acid residue R i The other amino acid residue R j The number of pairs C ij Divide by the number of amino acid residues R i The number of atoms C i The obtained number is the contact number N ij In other words, the number of contacts N ij The amino acid residue R i The number of pairs of size C ij . Contact number N ij It is expressed by formula (2).

[0091] [Formula 2]

[0092]

[0093] In addition, the contact number N ij It can be regarded as the element of row i and column j of the matrix, which will also touch the number N ij It is called the interaction matrix.

[0094] From the amino acid residue R i To amino acid residue R j The signal transmission probability P ij Use contact number N ij It is expressed by formula (3). In formula (3), parameter α is a constant.

[0095] [Formula 3]

[0096] P ij =1-exp(-α×N ij )…(3)

[0097] Therefore, from amino acid residue Ri to amino acid residue R j The signal transmission probability P ij Become the contact number N ij The larger the value, the closer it is to 1, the contact number N ij The smaller the value, the closer it is to 0, especially when the contact number N ij When , it becomes 0. In addition, the signal transmission probability P ij It can be regarded as the element of the i-row and j-column of the matrix, and the signal transmission probability P ij Also called the transfer probability matrix.

[0098] Next, refer to Figure 5 , describes the probability of signal transmission from the starting point to the end point of the path between vertices on the connecting network graph. Figure 5 This is a diagram for explaining the probability of signal transmission from the starting point to the end point of the pathway according to the present embodiment. i With amino acid residue R j In the adjacent case, as described above, from the amino acid residue R i To amino acid residue R j The signal transmission probability is expressed as the signal transmission probability P ij , from the amino acid residue R j The probability of signal transmission to amino acid residue Ri is expressed as the signal transmission probability P ji . Take vertex N22 as the starting point s of the path and vertex N210 as the end point e of the path.

[0099] The signal transmission probability from the starting point s to the end point e is represented by the product of the signal transmission probabilities between the vertices constituting the path. The product of the signal transmission probabilities between the vertices constituting the path is called the allosteric path score APS of the path. s，e n .

[0100] Path P21 is the first path from starting point s to end point e, passing through vertex N22 (starting point s), vertex N21, vertex N24, vertex N26, vertex N211 and vertex N210 (end point e) in sequence. Allosteric path score APS of path P21 s，e 1 It is expressed by formula (4).

[0101] [Formula 4]

[0102]

[0103] Path P22 is the second path from the starting point s to the end point e, and is a path that passes through vertex N22 (starting point s), vertex N23, vertex N25, vertex N211, and vertex N210 (end point e) in sequence. Allosteric path score APS of path P22 s，e 2 It is expressed by formula (5).

[0104] [Formula 5]

[0105]

[0106] The allosteric path from the starting point s to the end point e is scored according to the allosteric path APS s，e n Because the allosteric pathway score APS s，e n The higher it is, the higher the probability of signal transmission from the starting point s to the end point e is, and thus the allosteric pathway is evaluated as being important for allosteric control.

[0107] Here, the allosteric path is a combination of directed edges connecting vertices on the network graph. Therefore, the search allosteric path score APS s，e n The high allosteric path problem is a combinatorial optimization problem of searching for allosteric paths that are combinations of directed edges connecting vertices on a network graph.

[0108] In addition, in this embodiment, the allosteric pathway is a single pathway. A single pathway is a pathway that has no branches or confluences and has one starting point and one end point. In other words, a single pathway is a pathway that does not branch or interrupt.

[0109] In the allosteric pathway prediction method of the present embodiment, a combinatorial optimization problem is solved based on the Ising model. In the allosteric pathway prediction method of the present embodiment, a simulated bifurcation machine (SBM) is used to calculate the Ising model. In the allosteric pathway prediction method of the present embodiment, a simulated bifurcation machine (SBM) is used to predict the allosteric pathway score APS. s，e n The highest optimal pathway and allosteric pathway scores APS s，e n The quasi-optimal path with a high ranking.

[0110] Here, the problem of using the Ising model to find an approximate solution with the smallest possible energy (i.e., an approximate solution with the value of the objective function as close to the optimal value as possible) is called the Ising problem. On the other hand, the combinatorial optimization problem of finding a solution that minimizes a quadratic objective function with discrete variables (bits) that take a value of 0 or 1 as variables is called a quadratic unconstrained binary optimization problem (QUBO: Quadratic Unconstrained Binary Optimization). Here, the discrete variables in the QUBO problem can be transformed into spins in the Ising model by variable transformation, and the QUBO problem can be transformed into the Ising problem.

[0111] The information processing device used to calculate the energy of the ground state of the Ising model is called an Ising machine. As hardware implementations of the Ising machine, quantum annealing machines, coherent Ising machines, or quantum branching machines have been proposed. However, although these hardware implementations may significantly shorten the calculation time, they have technical problems such as difficulty in large-scale and stable operation.

[0112] Therefore, it is considered to use a widely used digital computer to solve the Ising problem. Compared with the above-mentioned hardware installation, digital computers are easy to scale up and operate stably. As an example of an algorithm for solving the Ising problem using a digital computer, simulated annealing is known. Simulated annealing is an iterative update algorithm that iteratively updates multiple variables, so it is difficult to speed up the calculation process by parallelization.

[0113] A simulated bifurcation algorithm has been proposed that can solve large-scale combinatorial optimization problems at high speed through parallel computing in a digital computer. A simulated bifurcation machine is a combinatorial optimization algorithm that calculates solutions by numerically solving the equations of motion of a classical model of a quantum bifurcation machine (for example, refer to non-patent document 3).

[0114] As described above, in the allosteric pathway prediction method of the present embodiment, an analog branching machine is used in order to solve the Ising problem using a digital computer.

[0115] [Functional structure of the allosteric pathway prediction device 1]

[0116] Next, refer to Figure 6 , a functional structure of an allosteric pathway prediction device 1 that performs a process of predicting an allosteric pathway, namely, an allosteric pathway prediction process, will be described. Figure 6 1 is a diagram showing an example of the functional configuration of the metamorphic path prediction device 1 according to the present embodiment. As an example, the metamorphic path prediction device 1 is a computer such as a personal computer (PC), a workstation, or a server.

[0117] The allomorphic path prediction device 1 includes a control unit 10 and a storage unit 20. As an example, the control unit 10 includes a CPU (Central Processing Unit), and each functional unit of the control unit 10 is realized by the CPU reading a program from a ROM (Read Only Memory) and executing the process. The control unit 10 may include a graphics processing unit (GPU) or a field programmable gate array (FPGA). The control unit 10 includes a three-dimensional structure information acquisition unit 11, a network diagram generation unit 12, a path calculation unit 13, an evaluation unit 14, and an output unit 15.

[0118] The three-dimensional structure information acquisition unit 11 acquires the three-dimensional structure information of the protein (also referred to as protein three-dimensional structure information). Protein three-dimensional structure information is information representing the three-dimensional structure of the protein. Protein three-dimensional structure information can be obtained from a well-known database such as the Protein Data Bank (PDB). The PDB registers the coordinates of atoms constituting proteins analyzed by various methods such as X-ray crystal structure analysis, NMR, and electron microscopy. In addition, protein three-dimensional structure information can also be predicted by three-dimensional structure prediction systems such as AlphaFold2. The protein three-dimensional structure information includes not only the atoms that constitute the protein, but also the coordinates of any atoms or atomic groups (hereinafter collectively referred to as arbitrary atomic groups) analyzed at the same time as the protein. As arbitrary atomic groups, for example, proteins (including antibodies), peptides (including artificial peptides), nucleic acids (including artificial nucleic acids), sugar chains, organic compounds such as drugs, natural ligands, and organic solvents, metals contained in solvents, various ions, water molecules, and antibodies, peptides, and nucleic acids (including artificial nucleic acids) and other middle molecules and polymers are listed.

[0119] The protein stereostructure information used may include all the atoms constituting the protein, or may include only the arbitrary atoms of interest. Among the atomic coordinates of proteins registered in the PDB, there are many proteins that do not record the hydrogen atoms that originally existed. The reason is that it is difficult to determine the coordinates of hydrogen atoms due to the characteristics and resolution of the stereostructure analysis method. Therefore, it is also possible to use only the atomic coordinates of carbon atoms, oxygen atoms, nitrogen atoms, sulfur atoms, etc. other than hydrogen atoms.

[0120] In addition, the number of proteins included in the protein three-dimensional structure information can be one or more than two. When the number of proteins is more than two, the two or more proteins can be connected (shared) by a linker or the like to form a polypeptide. For example, when proteins form a complex with each other, three-dimensional structure information containing more than two proteins can be used. When containing more than two proteins, the types of each protein can be the same or different.

[0121] The network graph generation unit 12 generates a network graph based on the three-dimensional structural information of the protein. The network graph generated by the network graph generation unit 12 is a graph in which amino acid residues are used as vertices and weights based on the interactions between amino acid residues are assigned to the edges.

[0122] The path calculation unit 13 calculates a path connecting the vertices on the network graph generated by the network graph generation unit 12 based on an evaluation function based on the weights assigned to the edges of the network graph. The evaluation function based on the weights refers to the Hamiltonian of the Ising model. The paths calculated by the path calculation unit 13 include a quasi-optimal path as a quasi-optimal solution, as well as an optimal path as an optimal solution. In addition, the optimal solution refers to a local optimal solution.

[0123] As described above, in the present embodiment, the path is a single path. Therefore, the path calculated by the path calculation unit 13 has no branches or confluences, and the path is an uninterrupted path connecting a starting point and an end point. In addition, in the present embodiment, the starting point and the end point of the path are respectively specified in advance. Therefore, in the present embodiment, the starting point and the end point of the path are respectively specified vertices on the network diagram.

[0124] The path calculation unit 13 includes a Hamiltonian generation unit 130 and an Ising calculation unit 131 .

[0125] The Hamiltonian generation unit 130 generates a Hamiltonian based on the network graph. The Hamiltonian generated by the Hamiltonian generation unit 130 is expressed in the QUBO format. QUBO is a problem that can be solved by an Ising machine. In the following description, the Hamiltonian generated by the Hamiltonian generation unit 130 is also referred to as QUBO.

[0126] The Ising calculation unit 131 performs Ising calculation on the Hamiltonian generated by the Hamiltonian generation unit 130 based on the simulated bifurcation machine. The simulated bifurcation machine is a parallel update algorithm, and the calculation of the Ising calculation unit 131 preferably uses parallel calculation.

[0127] The evaluation unit 14 evaluates the paths calculated by the path calculation unit 13. The evaluation unit 14 extracts a set of paths having the highest K evaluation values ​​(K is a natural number greater than or equal to 1) from among the paths calculated by the path calculation unit 13.

[0128] The output unit 15 outputs the prediction result of the allosteric pathway. The prediction result includes information indicating a set of pathways with the highest K evaluation values ​​(K is a natural number greater than or equal to 1) extracted by the evaluation unit 14 and the evaluation result of the evaluation unit 14.

[0129] The storage unit 20 stores various information. The information stored in the storage unit 20 includes Ising model information 21. The Ising model information 21 indicates information about the objective function and constraint conditions used by the Hamiltonian generation unit 130 to generate the Hamiltonian. The storage unit 20 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device.

[0130] [Allosteric pathway prediction processing]

[0131] Next, refer to Figures 7 to 9 , the allosteric pathway prediction process is explained. Figure 7 2 is a diagram showing an example of the allosteric pathway prediction process according to the present embodiment. The allosteric pathway prediction process is executed by the control unit 10 .

[0132] Step S10: The three-dimensional structure information acquisition unit 11 acquires the three-dimensional structure information of the protein. The three-dimensional structure information acquisition unit 11 acquires, for example, the three-dimensional structure information input by the user of the allosteric pathway prediction device 1. Then, the control unit 10 executes the process of step S20.

[0133] Step S20: The network diagram generation unit 12 generates a network diagram based on the three-dimensional structure information acquired by the three-dimensional structure information acquisition unit 11. Thereafter, the control unit 10 executes the process of step S30.

[0134] Here, refer to Figure 8 , details of the process of generating a network diagram by the network diagram generating unit 12, namely, the network diagram generating process, will be described. Figure 8 This is a diagram showing an example of the network diagram generation process according to the present embodiment. Figure 8 The network diagram generation process shown in Figure 7 The process is performed in step S20 shown.

[0135] Step S110: The network diagram generation unit 12 generates vertices of the network diagram. The network diagram generation unit 12 generates vertices for amino acid residues based on the three-dimensional structure information. The network diagram generation unit 12 generates vertices by making a one-to-one correspondence between amino acid residues and vertices. Thereafter, the network diagram generation unit 12 performs the processing of step S120.

[0136] Step S120: The network diagram generation unit 12 generates an interaction network diagram. Generating an interaction network diagram means calculating an interaction matrix. After that, the network diagram generation unit 12 performs the processing of step S130.

[0137] Step S130: The network diagram generation unit 12 generates a signal transmission network diagram. Generating a signal transmission network diagram refers to calculating a signal transmission matrix.

[0138] As described above, the network diagram generation unit 12 ends the network diagram generation process.

[0139] Back to Figure 7 The allosteric pathway prediction process will be described below.

[0140] Step S30: The pathway calculation unit 13 performs a process of predicting an allosteric pathway based on the Ising model. Thereafter, the control unit 10 performs a process of step S40.

[0141] Here, reference Fig. 9 The Ising calculation process, which is a process of predicting an allosteric pathway based on the Ising model by the pathway calculation unit 13 , will be described. Fig. 9 is a diagram showing an example of Ising calculation processing according to this embodiment. Figure 7 In step S30 shown in FIG. Fig. 9 Ising calculation process shown.

[0142] Step S210: The path calculation unit 13 obtains site specifying information. The site specifying information is information indicating an allosteric control site and an allosteric control target site. The site specifying information is input by, for example, a user of the allosteric path prediction device 1. Then, the path calculation unit 13 performs the process of step S220.

[0143] Step S220: The path calculation unit 13 determines the starting point and the end point of the path based on the site designation information. The path calculation unit 13 determines the vertex corresponding to the allosteric control site indicated by the site designation information as the starting point of the path. The path calculation unit 13 determines the vertex corresponding to the allosteric control target site in the table indicated by the site designation information as the end point of the path. Thereafter, the path calculation unit 13 performs the processing of step S230.

[0144] Step S230: The Hamiltonian generation unit 130 generates a Hamiltonian based on the designed mathematical model. The mathematical model is designed by the network diagram generated by the network diagram generation unit 12, the specified objective function, constraints and site designation information. The objective function and constraints are pre-stored in the storage unit 20 as Ising model information 21. In addition, the objective function and constraints can also be input by the user of the allosteric pathway prediction device 1. In addition, the mathematical model can be designed by the Ising model. In addition, the mathematical model is a model that expands the Ising model, and can also be designed by a model including a high-order expression, a rational expression or a general real-valued function.

[0145] Here, refer to Fig.10 and Fig.11 The Hamiltonian generated by the Hamiltonian generation unit 130 will be described. An example of the Hamiltonian generated by the Hamiltonian generation unit 130 is represented by equation (6). In equation (6), A, B, C, D, and E are constants.

[0146] [Formula 6]

[0147]

[0148] In formula (6), V represents the number of vertices. ij Represents the vertex V i With vertex V j The interaction between them.

[0149] q ij Represents and connects the vertex V i and vertex V j The spin value is 1 if the edge corresponding to the spin is used as part of the path, and 0 if the edge corresponding to the spin is not used as part of the path. i and vertex V j The edge is like Fig.10 Thus, the edges connecting the vertices V i and vertex V j Among the edges of i As the starting point, with the vertex V j The edge with the endpoint and the vertex V j As the starting point, with the vertex V i Here, we use the vertex V i As the starting point, with the vertex V j When the edge with the endpoint is a path, it is represented by the vertex V i From the vertex V j The outflow, or vertex V j From the vertex V iinflow.

[0150] V i start Represents the vertex V i The set of vertices whose endpoints are the edges with starting points. i end Represents the vertex V i The set of vertices whose edges are the end points and whose starting points are the edges. In formula (6), the starting point s represents the specified starting point, and the end point e represents the specified end point.

[0151] The Hamiltonian generated by the Hamiltonian generation unit 130 includes an objective function and constraints. Fig.11 As shown, item 61 is the objective function. Item 62, item 63, item 64, item 65, item 66, item 67 and item 68 are constraints respectively. The constraints are constraints required by graph theory.

[0152] The objective function is equivalent to the product of the scores of the edges taken in the path. As mentioned above, p ij Represents the vertex V i and vertex V j The interaction between the above signal transmission probability P ij In formula (6), as an example, for the connected vertex V i and vertex V j The weights of the edges are assigned using p ij The logarithm of .

[0153] Item 62 indicates the constraint that there is only an outflow from the starting point s to one vertex (the number of outflows is 1). Item 63 indicates the constraint that there is only an inflow from one vertex at the end point e (the number of inflows is 1). Item 64 indicates the constraint that there is no inflow from any vertex at the starting point s (the number of inflows is 0). Item 65 indicates the constraint that there is no outflow from any vertex at the end point e (the number of outflows is 0).

[0154] Item 66 indicates a constraint condition that, with respect to a vertex in the middle of a path, there is only an outflow from the vertex to one or less vertices (the number of outflows is 1 or less). Item 67 indicates a constraint condition that, with respect to a vertex in the middle of a path, there is only an inflow from one or less vertices to the vertex (the number of inflows is 1 or less). Item 68 indicates a constraint condition that, with respect to a vertex in the middle of a path, the number of inflows to the vertex is equal to the number of outflows from the vertex. In the present embodiment, items 66, 67, and 68 correspond to the path being a single path.

[0155] Thereafter, the path calculation unit 13 executes the process of step S240 .

[0156] Step S240 : The Ising calculation unit 131 performs Ising calculation on the Hamiltonian generated by the Hamiltonian generation unit 130 based on the analog branch.

[0157] As described above, the path calculation unit 13 completes the Ising calculation process.

[0158] Back to Figure 7 The allosteric pathway prediction process will be described below.

[0159] Step S40: The evaluation unit 14 calculates an evaluation value for the path calculated by the path calculation unit 13. The evaluation value is equivalent to the energy represented by the Hamiltonian, and the lower the value, the higher the evaluation as the optimal path. The evaluation unit 14 obtains the energy of the Hamiltonian for each path calculated by the Ising calculation unit 131 as the evaluation value for the path. In addition, the evaluation unit 14 determines the ranking of the path based on the evaluation value. The evaluation unit 14 extracts the path with the first ranking as the optimal path. The evaluation unit 14 extracts the paths from the second rank or below to the specified rank as quasi-optimal paths.

[0160] In addition, the evaluation unit 14 may determine whether the path calculated by the path calculation unit 13 is a single path before calculating the evaluation value. In this case, the evaluation unit 14 calculates the evaluation value for the path determined to be a single path among the paths calculated by the path calculation unit 13. The evaluation unit 14 does not calculate the evaluation value for the path determined not to be a single path among the paths calculated by the path calculation unit 13.

[0161] The number of paths extracted by the evaluation unit 14 may be 1 or more, and may be 2 or more, 3 or more, 5 or more, 10 or more, 20 or more, 30 or more, 40 or more, 50 or more, 100 or more, 500 or more, or 1000 or more.

[0162] Thereafter, the control unit 10 executes the process of step S50 .

[0163] Step S50: The output unit 15 outputs the prediction result of the allosteric pathway. The output unit 15 outputs the optimal pathway extracted by the evaluation unit 14, information indicating one or more quasi-optimal pathways, and the evaluation result of the evaluation unit 14 as the prediction result ... to an external server such as a database server, a display device (on a computer), or a computer program product. Figure 6 The prediction result is output to the storage unit 20, etc.

[0164] With the above, the control unit 10 ends the allosteric pathway prediction process.

[0165] exist Fig.12 In , as the calculation result of the path calculated by the path calculation unit 13, the path, its ranking and its evaluation value are shown. Fig.12In FIG. 1 , as an example, the top 10 pathway is shown for the pathway from the 14th amino acid residue to the 36th amino acid residue of KRAS.

[0166] In addition, in the present embodiment, an example of generating a network diagram by making the vertices of the network diagram correspond to the amino acid residues constituting the protein is described, but it is not limited to this. The vertices of the network diagram can also include vertices corresponding to substances bound to the protein (referred to as binding substances) to generate a network diagram. In addition, as the weights of the edges of the network diagram, weights based on the interactions between the binding substances and the amino acid residues, or the interactions between the binding substances and the binding substances can also be allocated to generate the network diagram.

[0167] The binding substance includes any atom, any ion, any molecule, any atomic group, or any compound. More specifically, the binding substance includes proteins (including antibodies), peptides (including artificial peptides), nucleic acids (including artificial nucleic acids), sugar chains, drugs, organic compounds, natural ligands, metals, water, solvents, ions, etc. In addition, when the atomic coordinates of atoms are included in the protein three-dimensional structure information, the vertices corresponding to the atoms can also be included in the vertices of the network graph.

[0168] When a vertex or atom corresponding to a binding substance is included in a vertex of a network graph, in the allosteric pathway prediction process, the vertex corresponding to the binding substance or atom may be designated as one or more of a start point or an end point of the pathway.

[0169] When the number of proteins included in the protein three-dimensional structure information is two or more, the vertex corresponding to the amino acid residue constituting the first protein molecule (or the binding substance bound to the first protein molecule) can be designated as the starting point of the path, and the vertex corresponding to the amino acid residue constituting the second protein molecule bound to the first protein molecule (or the binding substance bound to the second protein molecule) can be designated as the end point of the path.

[0170] (Second Embodiment)

[0171] Hereinafter, a second embodiment of the present invention will be described in detail with reference to the drawings.

[0172] In the above-mentioned first embodiment, the quasi-optimal path is calculated together with the optimal path, but they are calculated as candidates for one allosteric path. That is, in the above-mentioned first embodiment, allosteric control is modeled as a signal transmission path from a specified starting point to a specified end point. In allosteric control, it is believed that signal transmission is generated by a combination of multiple paths. The amino acid residues contained in multiple paths with high probability of signal transmission play an important role in allosteric control. In this embodiment, the situation that the amino acid residues (i.e., vertices) important in allosteric control are extracted based on the number of times the vertex is included in the predicted multiple allosteric paths is described.

[0173] In addition, the same reference numerals are given to the same structures as those in the first embodiment described above, and the description of the same structures and operations may be omitted.

[0174] Fig.13 : is a diagram showing an example of a summary of a method for searching amino acid residues that contribute to allosteric control of the present embodiment. In the present embodiment, the contribution of an amino acid residue to allosteric control is defined by the number of times the amino acid residue is included in a path within the evaluation value of the highest n positions. A score for evaluating the contribution of the amino acid residue to allosteric control is determined based on the number of times the vertex corresponding to the amino acid residue is included in the allosteric path from the starting point s to the end point e. If the score is set as the residue score RS s，e i , then the residue score RS s，e i It is expressed as formula (7).

[0175] [Formula 7]

[0176]

[0177] In formula (7), w (k) is the weight function corresponding to the allosteric path with position k in the allosteric path from the starting point s to the end point e. In formula (7), q i k As shown in formula (8), it is a number whose value is 0 when the i-th amino acid residue is not included in the k-th path, and whose value is 1 when the i-th amino acid residue is included in the k-th path.

[0178] [Formula 8]

[0179]

[0180] Fig.13 The allosteric pathway shown in (A) is P131, Fig.13 The allosteric pathway shown in (B) is P132, Fig.13 The allosteric pathway P133 shown in (C), and Fig.13The allosteric pathways P134 shown in (D) are examples of allosteric pathways from the starting point s to the 4th position at the top of the end point e. The residue score RS for each amino acid residue included in these four allosteric pathways is s，e i The value is shown in Fig.14 .like Fig.14 As shown, the residue score RS of the 11th amino acid residue among the 10 amino acid residues except the 2nd amino acid residue corresponding to the starting point s and the 10th amino acid residue corresponding to the end point e is s，e i Therefore, it is predicted that the 11th amino acid residue among the 10 amino acid residues is most conducive to allosteric control.

[0181] [Functional structure of the allosteric pathway prediction device 1a]

[0182] Fig.15 1 is a diagram showing an example of the functional structure of the allosteric path prediction device 1a of the present embodiment. The allosteric path prediction device 1a includes a control unit 10a and a storage unit 20. Fig.15 ) and the allosteric pathway prediction device 1 of the first embodiment ( Figure 6 ) are compared, the control unit 10a is different. The description of the same functions as those in the first embodiment is omitted, and the second embodiment is mainly described with the difference from the first embodiment.

[0183] The control unit 10a includes a three-dimensional structure information acquisition unit 11, a network graph generation unit 12, a path calculation unit 13, a first evaluation unit 14a, and an output unit 15a. Fig.15 ) and the control unit 10( Figure 6 ), the first evaluation unit 14a and the output unit 15a are different. Here, the functions of other components (three-dimensional structure information acquisition unit 11, network graph generation unit 12 and path calculation unit 13) are the same as those of the first embodiment.

[0184] In this embodiment, the path calculation unit 13 calculates a plurality of paths. The first evaluation unit 14a evaluates the vertex included in the network graph based on the number of times the optimal path calculated by the path calculation unit 13 and one or more quasi-optimal paths are included. The first evaluation unit 14a calculates the above-mentioned residue score RS s，e i to make an evaluation.

[0185] The output unit 15a outputs the search result of the important amino acid residues. The search result includes the evaluation result of the vertices by the first evaluation unit 14a.

[0186] [Search process for important amino acid residues]

[0187] Next, refer to Fig.16 and Fig.17 Next, the process of searching for amino acid residues that contribute to allosteric control, that is, the important amino acid residue search process, will be described. The important amino acid residue search process is included in the allosteric pathway prediction process and is executed. Fig.16 is a diagram showing an example of the allosteric pathway prediction process of this embodiment. Figure 7 The processing of steps S10 to S40 in the same manner is omitted. Fig.16 In the process of step S340 shown in the figure, the ranking of each path may be evaluated, and the extraction of the optimal path and one or more suboptimal paths may be omitted.

[0188] Step S350: The first evaluation unit 14a performs an important amino acid residue search process. Then, the control unit 10a executes the process of step S360.

[0189] Here, refer to Fig.17 Detailed mechanical description of the search process for important amino acid residues. Fig.17 This is a diagram showing an example of the important amino acid residue search process according to the present embodiment. Fig.17 The search process for important amino acid residues shown in Fig.16 The process is performed in step S350 as shown.

[0190] Step S410: The first evaluation unit 14a calculates the residue score RS for each vertex included in the network graph. s，e i Thereafter, the first evaluation unit 14a executes the process of step S420.

[0191] Step S420: The first evaluation unit 14a calculates the residue score RS s，e i The first evaluation unit 14a, for example, evaluates the vertices in the network graph other than the starting point and the end point of the path according to the residue score RS. s，e i The vertices up to the predetermined positions are determined as important amino acid residues in descending order.

[0192] With the above, the first evaluation unit 14a ends the important amino acid residue search process.

[0193] Back to Fig.16 The allosteric pathway prediction process will be described below.

[0194] Step S360: The output unit 15a outputs the evaluation result of the first evaluation unit 14a on the vertex as the search result of important amino acid residues. The output unit 15a may also output the prediction result of the allosteric pathway together with the search result of important amino acid residues.

[0195] As described above, the control unit 10a ends the allosteric pathway prediction process.

[0196] (First Example of the Second Embodiment)

[0197] In the present embodiment, the search for the important amino acid residues in the allosteric control of HRAS is performed. The protein of the known RAS family is activated by the GDP combined with RAS being replaced by GTP, and the displacement of GDP / GTP in normal RAS is produced by GEF and RAS reaction. That is, the displacement of GDP / GTP is allosterically controlled. In this allosteric control, the GEF binding region is the allosteric control region, and the GDP / GTP binding region is the allosteric control target region. Therefore, in the present embodiment, the important amino acid residue search process of the second embodiment is used, and the GDP / GTP displacement reaction of HRAS is confirmed to be able to extract this situation of amino acid residues that contribute to allosteric control.

[0198] For the HRAS structural information (PDBID: 3k8y), the starting point was set to A719 and the end point was set to A528, and the allosteric pathway prediction process was implemented. In addition, A719 is the compound placed in the GEF binding region in the PDB file, namely ACT, and A528 is a GTP analog, namely GNP. Here, the expression Ai (i is a natural number) in principle refers to the i-th "amino acid" from the N-terminus. The same is true in the following description. However, for atoms or compounds other than amino acids that constitute proteins, it is also convenient to use the expression Aj (j is a natural number other than the natural number assigned to the existing amino acids).

[0199] The evaluation value of the pathway obtained by the allosteric pathway prediction process (allosteric pathway score APS) is used. s，e n ) high top 10 paths, calculate the residue score RS of each amino acid residue s，e i In addition, as the weighting function w corresponding to the allosteric pathway (k) , as shown in formula (9), the allosteric pathway score APS is used s，e k .

[0200] [Formula 9]

[0201]

[0202] Fig.18 and Fig.19 The search results for the important amino acid residues of this example are shown in Fig.18 In the figure, the top 10 pathways with high evaluation values ​​are shown together with the amino acid residues. Fig.19 Regarding the residue score RS s，e i , the amino acid residues up to the top 10 and the residue score RS s，e i Express together.

[0203] The search results of important amino acid residues in this embodiment are compared with the results of Ohm, which is an existing method (refer to non-patent document 2). A97, A96, A10, A16 and A17 among the amino acid residues up to the top 10 in the search results of this embodiment are also extracted as important amino acid residues in the results of Ohm. That is, the results of Ohm, which is an existing method, can be reproduced by using the important amino acid residue search process of the second embodiment. Furthermore, according to the search results of this embodiment, amino acid residues such as 101, A100, A98, A94 and A14 that are not extracted in Ohm are extracted as important amino acid residues.

[0204] It is believed that the extraction of A101, A100, A98, A94, and A14 as important amino acid residues suggests that the combination of branched routes contributes to allosteric control compared to the results of Ohm. This result is considered to be the result that, in the search for each path, Ohm sequentially performs path search using the presence or absence of signal transmission between each vertex as an index, whereas the important amino acid residue search process of the second embodiment relies on the fact that the signal transmission probability of the entire path is used as an index.

[0205] (Second Example of the Second Embodiment)

[0206] In this embodiment, a search for important amino acid residues in the allosteric control of KRAS is performed. Normal KRAS does not react with effector proteins in the state bound to GDP, but can react with effector proteins by replacing GDP with GTP. In this way, the reaction of KRAS with effector proteins is allosterically controlled by GDP / GTP. In this allosteric control, the GDP binding region is the allosteric control region, and the effector binding region is the allosteric control target region. On the other hand, KRAS with a G12C mutation does not function based on the allosteric control of GDP / GTP and can always react with effector proteins (activating mutation).

[0207] In this activating mutation, the allosteric pathway from the GDP binding region to the effector binding region may be greatly changed compared to the wild type. Therefore, regarding the change in activity of the protein caused by this activating mutation, it was confirmed that the change in the allosteric pathway and the change in the important amino acid residues can be evaluated as indicators using the allosteric pathway prediction process and the important amino acid residue search process of the second embodiment.

[0208] For two KRAS structural information (PDB ID: 4obe: wild type and PDB ID: 4ldj: G12C mutant), the allosteric pathway prediction process was performed with the starting point set to the amino acid residue A14 and the end point set to the amino acid residue A36. For the two KRAS structural information, the evaluation values ​​(allosteric pathway score APS) in the pathways obtained by the allosteric pathway prediction process were used. s，e n ) The residue score RS of each amino acid residue is calculated by the top 10 paths s，e i In addition, as the weighting function w corresponding to the allosteric pathway (k) , the allosteric pathway score APS is used in the same manner as in the first embodiment described above s，e k .

[0209] Figure 20 to Figure 23 The search results of the important amino acid residues of this embodiment are shown in FIG. Fig. 20 In the example, the top 10 pathways with high evaluation values ​​are shown together with the amino acid residues for the wild type. Fig.21 For the wild type, for the residue score RS s，e i , amino acid residues up to the top 10 and residue score RS s，e i are shown together. Fig. 22 In the figure, regarding the G12C mutant, the top 10 pathways with high evaluation values ​​are shown together with the amino acid residues. Fig.23 For the G12C mutant, the residue score RS s，e i , amino acid residues up to the top 10 and residue score RS s，e i are shown together.

[0210] Combine search results with Fig.21 and Fig.23 By comparison, it can be seen that in the G12C mutant, an allosteric pathway via A12 (the amino acid residue that produces the G12C mutation) was detected. The allosteric pathway via A12 is a new pathway that was not detected in the wild type. Fig. 22As shown, the apex has the amino acid residues included in region R1. The fact that a new pathway not detected in the wild type was detected in the G12C mutant is consistent with the fact that the allosteric control of KRAS is changed by the G12C mutation. The results suggest that the amino acid residues important in the allosteric control of proteins can be extracted from the allosteric pathway.

[0211] (Third Embodiment)

[0212] Hereinafter, a third embodiment of the present invention will be described in detail with reference to the drawings.

[0213] In the first and second embodiments described above, the case where the start point and the end point of the path are set as the prescribed vertices to predict the allosteric path is described. In this embodiment, only one of the vertices of the start point and the end point is set as the prescribed vertex, and the other vertex is changed. In this embodiment, the case where the allosteric control region or the allosteric control target region is searched by extracting a vertex that contributes greatly to the reactivity of the prescribed vertex from the other changed vertex based on the evaluation value of the allosteric path is described. In the following description, the allosteric control region or the allosteric control target region that is the search object is sometimes collectively referred to as the allosteric region.

[0214] In addition, the same reference numerals are given to the same structures as those in the first embodiment or the second embodiment described above, and the description of the same structures and operations may be omitted.

[0215] Fig.24 : is a diagram showing an example of the summary of the allosteric region search method of the present embodiment. In the search method of the present embodiment, the starting point is set to a predetermined vertex and the end point is changed, or the end point is set to a predetermined vertex and the starting point is changed. For a specific starting point, the allosteric path score APS is used. s，e n The amino acid residue corresponding to the end point of the high-pathway connection is considered to have a strong influence on the reactivity of the amino acid residue corresponding to the starting point. Similarly, for a specific end point, the allosteric pathway score APS s，e n The amino acid residue corresponding to the starting point of a high pathway connection is considered to have a strong influence on the reactivity of the amino acid residue corresponding to the end point.

[0216] In the present embodiment, the index representing the degree of these effects is referred to as an allosteric score. The allosteric score is an index representing the degree of influence of each amino acid residue on the reactivity of a specific region. The allosteric score indicates the size of the possibility of being the allosteric control region of each amino acid residue when the allosteric control target region is specified as a specific region, and indicates the size of the possibility of being the allosteric control target region of each amino acid residue when the allosteric control region is specified as a specific region. An example of an allosteric score is represented as formula (10). In addition, the above-mentioned allosteric path score can also be used directly as an allosteric score.

[0217] [Formula 10]

[0218]

[0219] The allosteric score AS shown in formula (10) s，e Indicates the influence of the starting point s on the end point e. In formula (10), correctedAPS s，e k Scoring APS for allosteric pathways s，e n The corrected APS is expressed by equations (11), (12) and (13) respectively. s，e k Example.

[0220] [Formula 11]

[0221]

[0222] [Formula 12]

[0223]

[0224] [Formula 13]

[0225]

[0226] Allosteric Pathway Scoring APS s，e k is the allosteric pathway score of the kth pathway between the starting point s and the end point e, and is the product of the signal transmission probabilities between the vertices constituting the pathway, similarly to the first embodiment described above. s，e k is the length of the kth path from starting point s to end point e. Here, the length of a path refers to the number of edges contained in the path. As corrected APS s，e k , just correct the APS according to the length of the path s，e kAlternatively, it may be expressed by a formula other than formula (11), formula (12) or formula (13).

[0227] In addition, in formula (10), w (k) is a weight function corresponding to the rank (k) of the path. Therefore, the allosteric score AS s，e is corrected APS s，e k In formula (10), n is the allosteric score AS s，e The number of paths used in the calculation.

[0228] exist Fig.24 , an example of a case where a starting point is changed to evaluate the influence of the starting point on a specific end point is shown. Fig.24 The allosteric path P241 shown in (A) is a path having the vertex N21 corresponding to the first amino acid residue as the starting point (starting point "1") and the vertex N210 corresponding to the tenth amino acid residue as the end point (end point "10"). Fig.24 The allosteric path P242 shown in (B) is a path having the vertex N22 corresponding to the second amino acid residue as the starting point (starting point "2") and the vertex N210 corresponding to the tenth amino acid residue as the end point (end point "10"). Fig.24 The allosteric path P243 shown in (C) is a path having the vertex N21 corresponding to the third amino acid residue as the starting point (starting point "3") and the vertex N210 corresponding to the tenth amino acid residue as the end point (end point "10"). Fig.24 The allosteric path P244 shown in (D) is a path having the vertex N21 corresponding to the seventh amino acid residue as the starting point (starting point "7") and the vertex N210 corresponding to the tenth amino acid residue as the end point (end point "10").

[0229] Fig.25 The results of evaluating the influence of the four starting points, starting point "1", starting point "2", starting point "3" and starting point "7" on the end point "10" are shown in FIG. Fig.25 In the example shown, the allosteric score AS is calculated s，e When , one path is used (i.e., n=1). Fig.25 In the example shown, the starting point and the path correspond to each other in a one-to-one manner. In addition, the weight function w (k) The value of is 1 for all paths. In addition, as the corrected APS s，e k , using the expression of formula (11). In addition, Fig.25 In , the length of the path is denoted as "l". Fig.25 In the ASs，e The allosteric score AS is shown together s，e The standardized value. Fig.25 The evaluation results shown show that the second amino acid residue corresponding to the starting point "2" is judged to have the greatest influence on the amino acid residue corresponding to the end point "10".

[0230] [Functional structure of the allosteric pathway prediction device 1b]

[0231] Fig.26 1 is a diagram showing an example of the functional structure of the allosteric path prediction device 1b of the present embodiment. The allosteric path prediction device 1b includes a control unit 10b and a storage unit 20. Fig.26 ) and the allosteric pathway prediction device 1 of the first embodiment ( Figure 6 ) are compared, the control unit 10b is different. The description of the same functions as those in the first embodiment is omitted, and in the third embodiment, the description will focus on the parts that are different from the first embodiment.

[0232] The control unit 10b includes a three-dimensional structure information acquisition unit 11, a network graph generation unit 12, a path calculation unit 13b, a second evaluation unit 14b, and an output unit 15b. Fig.26 ) and the control unit 10( Figure 6 ), the path calculation unit 13b, the second evaluation unit 14b and the output unit 15b are different. Here, the functions of other components (three-dimensional structure information acquisition unit 11 and network graph generation unit 12) are the same as those of the first embodiment.

[0233] In the present embodiment, one of the starting point and the end point of the path is a predetermined vertex on the network diagram, and the other is an evaluation target vertex. The evaluation target vertex is a vertex that is the object of evaluation by the second evaluation unit 14b. The evaluation target vertex is a vertex for evaluating the degree of influence on the reactivity of the amino acid residue corresponding to the predetermined vertex for the amino acid residue corresponding to the evaluation target vertex. In other words, the evaluation target vertex is a vertex corresponding to the allosteric region.

[0234] The path calculation unit 13b calculates the path connecting the specified vertex in the vertex on the network diagram with the evaluation object vertex for each combination of the specified vertex in the vertex on the network diagram and the evaluation object vertex. Here, in the present embodiment, there is one specified vertex. When one specified vertex is specified, the combination of the specified vertex and the evaluation object vertex is specified by the evaluation object vertex. Therefore, in the present embodiment, the path calculation unit 13b calculates the path connecting the specified vertex in the vertex on the network diagram with the evaluation object vertex for each evaluation object vertex. In addition, when there are multiple specified vertices, it is explained in a modified example of the third embodiment.

[0235] The second evaluation unit 14b calculates evaluation values ​​for each of the plurality of evaluation target vertices based on the evaluation values ​​of the paths calculated by the path calculation unit 13b. The evaluation values ​​calculated by the second evaluation unit 14b are the above-mentioned allosteric scores.

[0236] The output unit 15b outputs the search result of the allosteric region. The search result includes the evaluation value of the second evaluation unit 14b.

[0237] [Area search process]

[0238] Next, refer to Figure 27 to Figure 29 , the process of searching for an allosteric region, that is, the region search process will be described. The region search process is included in the allosteric pathway prediction process and is executed. Fig. 27 is a diagram showing an example of the allosteric pathway prediction process of this embodiment. In addition, each process of step S510, step S520 and step S540 is similar to Figure 7 The processing of step S10, step S20 and step S40 in the embodiment is the same, so the description is omitted. Fig. 27 In the process of step S540 shown in the figure, the ranking of each path may be evaluated, and the extraction of the optimal path and one or more suboptimal paths may be omitted.

[0239] In addition, Figure 27 to Figure 29 In the area search process shown, as an example, the case where the end point of the path is a predetermined vertex and the starting point of the path is an evaluation target vertex is described. Even in the case where the starting point of the path is a predetermined vertex and the end point of the path is an evaluation target vertex, the area search process is the same as the case where the end point of the path is a predetermined vertex and the starting point of the path is an evaluation target vertex.

[0240] When the end point of the path is a predetermined vertex and the start point of the path is an evaluation target vertex, the degree of influence from the allosteric control site on the designated allosteric control target site is evaluated with the allosteric control site as the evaluation target. On the other hand, when the start point of the path is a predetermined vertex and the end point of the path is an evaluation target vertex, the degree of influence from the designated allosteric control site on the allosteric control target site corresponds to evaluating the allosteric control target site as the evaluation target.

[0241] Step S530: The pathway calculation unit 13b performs a process of predicting an allosteric pathway based on the Ising model. Thereafter, the control unit 10b performs a process of step S540.

[0242] Here, reference Fig.28 , the Ising calculation process, which is a process in which the pathway calculation unit 13b predicts an allosteric pathway based on the Ising model, will be described. Fig.28 is a diagram showing an example of Ising calculation processing according to this embodiment. Fig. 27 In step S530 shown in FIG. Fig.28 The Ising calculation process shown in FIG. 1 is performed in step S640 and step S650. Fig. 9 The processing of step S230 and step S240 is the same, so the description is omitted.

[0243] Step S610: The path calculation unit 13b obtains search region designation information. In this embodiment, the search region designation information is information that specifies the allosteric control target site and the allosteric control site to be evaluated. The search region designation information is input by a user of the allosteric path prediction device 1b, for example. Then, the path calculation unit 13b performs the process of step S620.

[0244] Step S620: The path calculation unit 13b starts the process of predicting the metamorphic path for each evaluation target vertex. That is, the path calculation unit 13b specifies one or more evaluation target vertices as the starting point or the end point of the path, and predicts the allocation path for each evaluation target vertex. Then, the path calculation unit 13b performs the process of step S630.

[0245] Step S630: The path calculation unit 13b determines the starting point and the end point of the path based on the search area designation information. The path calculation unit 13b designates the evaluation target vertex indicated by the search area designation information as the starting point of the path. The path calculation unit 13b determines the specified vertex indicated by the search area designation information as the end point of the path. Thereafter, the path calculation unit 13b performs the processing of step S640.

[0246] Step S660: The path calculation unit 13b ends the process of predicting the allosteric path for each evaluation target vertex.

[0247] As described above, the path calculation unit 13b ends the Ising calculation process.

[0248] Back to Fig. 27 The allosteric pathway prediction process will be described below.

[0249] Step S550: The second evaluation unit 14b performs area search processing. Then, the control unit 10b executes the processing of step S560.

[0250] Here, refer to Fig.29 , the area search process is explained. Fig.29 It is a diagram showing an example of the area search process according to the present embodiment. Fig.29 The area search process shown in Fig. 27 The process is performed in step S550 as shown.

[0251] Step S710: The second evaluation unit 14b calculates an evaluation value for each evaluation target vertex. The second evaluation unit 14b calculates an allosteric score as the evaluation value. Then, the second evaluation unit 14b performs the process of step S720.

[0252] Step S720: The second evaluation unit 14b evaluates the evaluation target vertex based on the calculated evaluation value. For example, the second evaluation unit 14b determines the evaluation target vertex with an evaluation value higher than a predetermined value. The second evaluation unit 14b may also determine the evaluation target vertex with the highest evaluation value.

[0253] With the above, the second evaluation unit 14b ends the area search process.

[0254] Back to Fig. 27 Continue with the description of the allosteric pathway prediction process.

[0255] Step S560: The output unit 15b outputs the evaluation result of the second evaluation unit 14b as the search result of the allosteric control site. As described above, the search result of the allosteric control site is the result of searching for the allosteric control site evaluated to have a high degree of influence on the designated allosteric control target site.

[0256] As described above, the control unit 10b ends the allosteric pathway prediction process.

[0257] (Example of the third embodiment)

[0258] In this example, an exhaustive search for allosteric control regions that control the reaction with the effector of KRAS was performed. Normal KRAS does not react with effector proteins when bound to GDP, but can react with effector proteins by replacing GDP with GTP. In this way, the reaction of KRAS with effector proteins is allosterically controlled by GDP / GTP. In this allosteric control, the GDP binding region is the allosteric control region, and the effector binding region is the allosteric control target region. In this example, the allosteric path score between I36, which is located at the center of the RAF binding region, one of the effector proteins, in the crystal structure of PDB ID: 4obe of normal KRAS, and each amino acid residue is calculated, and the allosteric score is calculated. I36 corresponds to the specified vertex, and each amino acid residue corresponds to the evaluation object vertex.

[0259] Figure 30 to Figure 32 The search results for the allosteric control region of this example are shown. Fig.30 This is a diagram showing the search results for the allosteric control region related to the amino acid region numbered 1 to 67 of KRAS. Fig.31 This is a diagram showing the search results for the allosteric control region related to the amino acid region from amino acid number 114 to 169 of KRAS. Fig.30 and Fig.31 In FIG, the allosteric score is shown by a bar graph for each amino acid number of KRAS. Fig.32 This is a diagram showing the search results of the allosteric control region for the entire KRAS together with the positions of the α-helical chains.

[0260] according to Fig.30 In the region from amino acids 1 to 67, high allosteric scoring signals were observed in the GTP binding region, the vicinity of the GTP binding region, the RAF / RBD binding region, and the RAF / CRD binding region. Fig.31 In the region with amino acid numbers 114 to 169, signals with high allosteric scores were observed in the GTP binding region and the RAF / CRD binding region. The regions where signals with high allosteric scores were observed throughout KRAS were concentrated in these interaction regions and their vicinities. It was confirmed that the allosteric score is an effective indicator for detecting the allosteric control region for the interaction between KRAS and RAF.

[0261] like Fig.32 As shown, there are five α helices in KRAS. Fig.32 , a signal with a high allosteric score was observed only in the interaction region of the α helix that interacts with RAF. The observation of a signal with a high allosteric score only in the interaction region of the α helix is ​​considered to indicate the high specificity of the allosteric region detected by the allosteric region search method of this embodiment.

[0262] (Variation of the Third Embodiment)

[0263] Here, a modification of the third embodiment is described in detail. In this modification, a case where an allosteric control region or an allosteric control target region is searched as a set of a plurality of start points or end points by using the method described in the third embodiment in which only one vertex of the start point and the end point is set as a predetermined vertex and the other vertex is changed is described.

[0264] Fig.33 and Fig.34 : is a diagram showing an example of the outline of the method for searching the allosteric region of this modification. The multiple residues contained in the allosteric control region should be close to each other in the three-dimensional structure. The method for searching the allosteric region of this modification has obtained a viewpoint based on this investigation. First, Fig.33 As shown in FIG. 1 , a specific end point is determined, and multiple starting points connected to the end point through an arbitrary path are extracted. The starting point can be exposed on the protein surface or buried inside the protein, but it is preferably exposed on the protein surface (including the matrix binding pocket). The allosteric score is calculated for each residue that serves as the starting point. Fig.34 As shown in Figure 1, based on the allosteric score calculated for each residue, a score of a residue set consisting of a plurality of residues close to each other is defined. The score of the residue set based on these allosteric scores is recorded as a residue set score. On the premise that the residue set with the highest residue set score is an allosteric control region, an allosteric control region is searched.

[0265] exist Fig.34 In the example shown, the following is shown: as a result of defining a residue set score that takes into account the allosteric score for each residue and the proximity of each residue in the three-dimensional structure, three vertices, vertex N21, vertex N22, and vertex N23, are extracted as the residue set with the highest residue set score. The residue set score is defined to be higher when the allosteric score of each residue included in the set is high and the residues are densely packed in the three-dimensional structure, and therefore, residues such as vertex N27 with a negative allosteric score are excluded from the allosteric control region.

[0266] [Functional structure of the allosteric pathway prediction device 1d]

[0267] Fig.35 1d is a diagram showing an example of the functional structure of the allosteric path prediction device 1d of this modification. The allosteric path prediction device 1d includes a control unit 10d and a storage unit 20. Fig.35 ) and the allosteric pathway prediction device 1b of the third embodiment ( Fig.26), the control unit 10d is different from the residue set search Ising information 22d stored in the storage unit 20. The description of the same functions as those in the third embodiment is omitted, and in this modification, the description is centered on the parts different from the third embodiment.

[0268] The control unit 10d includes a three-dimensional structure information acquisition unit 11, a network graph generation unit 12, a path calculation unit 13, a second evaluation unit 14b, a residue set search unit 140d, a third evaluation unit 16d, and an output unit 15d. Fig.35 ) and the control unit 10b( Fig.26 ), the difference is that the residue set search unit 140d, the third evaluation unit 16d, the output unit 15d and the Ising model information 21 are stored together with the Ising model information 21 in the storage unit 20. Here, the functions of other structural elements (stereostructure information acquisition unit 11 and network graph generation unit 12) are the same as those in the first embodiment.

[0269] The residue set search unit 140d includes a second Hamiltonian generation unit 141d and a second Ising calculation unit 142d.

[0270] The second Hamiltonian generator 141d generates a Hamiltonian for searching a residue set. The objective function and constraint conditions used by the second Hamiltonian generator 141d to generate the Hamiltonian are pre-stored in the storage unit 20 as the residue set search Ising information 22d. A specific example of this Hamiltonian will be described later.

[0271] The second Ising calculation unit 142 d performs Ising calculation on the Hamiltonian generated by the second Hamiltonian generation unit 141 d based on a pseudo bifurcation machine.

[0272] The third evaluation unit 16d calculates an evaluation value for a partial set of multiple evaluation target vertices based on the evaluation value calculated for each evaluation target vertex by the second evaluation unit 14b. The residue set described below is an example of a partial set of multiple evaluation target vertices. In addition, the residue set score described below is an example of the evaluation value calculated by the third evaluation unit 16d.

[0273] The output unit 15d outputs the search result of the residue set. The search result includes the evaluation value of the third evaluation unit 16d.

[0274] [Residue set search process]

[0275] Here, refer to Figure 36 to Figure 38 , the residue set search process is described. The residue set search process is included in the allosteric pathway prediction process. Fig.36is a diagram showing an example of the allosteric pathway prediction process of this modification. Fig. 27 The processing of step S510, step S520 and step S540 is the same, so the description is omitted.

[0276] In addition, Figure 36 to Figure 38 In the residue set search process shown in FIG. 1 , as an example, a case where the set at the end of the path is an active center and the set at the starting point of the path is an allosteric control region is described. Even in the case where the set at the starting point of the path is an active center and the set at the end of the path is an allosteric control region, the residue set search process is the same.

[0277] Step S1530: The path calculation unit 13 performs a process of predicting an allosteric path based on the Ising model (Ising calculation process). Then, the control unit 10d performs the process of step S1540. The Ising calculation process of this modification example and the Ising calculation process of the third embodiment ( Fig.28 ), the region specified by the search region specifying information is different. In the Ising calculation processing of this variant, the search region specifying information is information that specifies the active center (allosteric control target site) and the residue set (allosteric control site) as the evaluation object. Except that the region specified by the search region specifying information is different, the Ising calculation processing of this variant is different from the Ising calculation processing of the third embodiment ( Fig.28 ) are the same, so the description is omitted.

[0278] Step S1550: The residue set search unit 140d performs a residue set search process. Then, the control unit 10d executes the process of step S1560.

[0279] Here, refer to Fig.37 , the residue set search process is explained. Fig.37 This is a diagram showing an example of the residue set search process according to this modification. Fig.37 The residue set search process shown in Fig.36 The process is performed in step S1550 as shown.

[0280] Step S1710: The residue set search unit 140d causes the second evaluation unit 14b to calculate an evaluation value for each evaluation target vertex. Here, the evaluation target vertex refers to a residue included in the residue set as the evaluation target, as described above. The second evaluation unit 14b calculates an allosteric score as an evaluation value. In addition, an example of an allosteric score is represented by formula (14).

[0281] [Formula 14]

[0282]

[0283] The allosteric score shown in formula (14) specifies the residues contained in the active center. j Except for the endpoint, the allosteric score (formula (10)) described in the third embodiment is the same. s，e k , any one of the above formulas (11), (12) and (13) can be used.

[0284] Thereafter, the third evaluation unit 16d executes the process of step S1720.

[0285] Step S1720: The third evaluation unit 16d calculates an evaluation value for the active center for each evaluation target vertex (residue). This evaluation value is represented by, for example, equation (15).

[0286] [Formula 15]

[0287]

[0288] As shown in formula (15), the evaluation value for the active center calculated by the third evaluation unit 16d for each evaluation target vertex (residue) is the residue e included in the active center with respect to the amount normalized for the allosteric score shown in formula (14): j Here, when calculating the evaluation value, when the allosteric score is not standardized, the allosteric score of the path with an extremely low score sometimes dominates. Therefore, when calculating the evaluation value, it is preferred to standardize the allosteric score as shown in formula (15). In addition, when calculating the evaluation value, the allosteric score may not be standardized.

[0289] Fig.39 The outline of the process of calculating the evaluation value of each evaluation target vertex (residue) with respect to the active center by the third evaluation unit 16d is shown in FIG. Fig.39 , the evaluation value for the active center R2 is calculated for the vertex N31. The active center R2 includes three vertices, namely, vertex E1, vertex E2, and vertex E3. The evaluation value for the active center R2 of vertex N31 is calculated as the sum of the allosteric score of the path with vertex N31 as the starting point and vertex E1 as the end point, the allosteric score of the route with vertex N31 as the starting point and vertex E2 as the end point, and the allosteric score of the route with vertex N31 as the starting point and vertex E3 as the end point. The evaluation value for the active center R2 is calculated by making various changes to the vertex N31 as the starting point.

[0290] Step S1730: The residue set search unit 140d calculates the distance between the residues based on the three-dimensional structure information. Then, the control unit 10d executes the process of step S1740.

[0291] In addition, the stereostructure information is acquired by the stereostructure information acquisition unit 11 as described above. The distance between residues is, for example, the distance between a representative atom among atoms constituting one residue and a representative atom among atoms constituting the other residue. Here, the representative atom is a predetermined atom among atoms constituting the residue. The representative atom is, for example, α carbon. The representative atom may also be an atom other than α carbon. Information specifying the representative atom is, for example, included in the residue set search Ising information 22d.

[0292] Alternatively, the distance between residues can be based on the signal transmission probability P ij For example, the distance between residues can be the signal transmission probability P ij The distance between residues can also be calculated by subtracting the signal transmission probability P from 1. ij The value obtained by the value of .

[0293] In addition, the distance between residues may be the smallest distance between the two atoms of the group of atoms constituting one residue and the other residue. The distance between residues may be the average value of the distance between the two atoms of the group of atoms constituting one residue and the other residue. In addition, the distance between residues may be the distance between the centers of gravity of the residues. Fig.40 As an example of the distance between residues, the distance d between vertex N31 and vertex N32 is shown. 2，1 .

[0294] Step S1740: The residue set search unit 140d searches for a residue set based on the Ising machine.

[0295] Here, refer to Fig.38 The process of searching for a residue set based on the Ising machine by the residue set search unit 140d, that is, the residue set search Ising calculation process, will be described. Fig.38 : is a diagram showing an example of Ising calculation processing for residue set search in this modification. Fig.37 The step S1740 shown is executed Fig.38 The residue set shown is searched for by Ising calculations.

[0296] Step S1810: The second Hamiltonian generation unit 141d generates a Hamiltonian based on the designed mathematical model. In the residue set search Ising calculation process, the objective function and constraints included in the Hamiltonian generated by the second Hamiltonian generation unit 141d are pre-stored in the storage unit 20 as the residue set search Ising information 22d. In addition, the objective function and the constraints can also be input by the user of the allosteric pathway prediction device 1. The mathematical model can be designed by the Ising model. In addition, the mathematical model is a model that expands the Ising model, and can also be designed by a model including a high-order expression, a rational expression or a general real-valued function.

[0297] An example of the Hamiltonian generated by the second Hamiltonian generation unit 141d is represented by equation (16).

[0298] [Formula 16]

[0299]

[0300] In formula (16), q i represents the spin corresponding to the residue. When the residue is selected as an element of the residue set, q i The value of q is 1, and if the residue is not selected as an element of the residue set, i The value of is 0. In formula (16), the first term on the right side represents the objective function, and the second term on the right side represents the constraint condition.

[0301] The objective function shown in the first item represents the following condition: for a certain residue set, the larger the sum of the allosteric scores of the residues contained in the residue set is, the easier it is for the residue set to be selected.

[0302] The constraint condition represented by the second term indicates that residues close to each other are easily selected. ij It is expressed by formula (17). In addition, in the second term, A is a constant.

[0303] [Formula 17]

[0304]

[0305] Therefore, the constraint condition expressed by the second term on the right side of equation (16) indicates that the residue q on one side i With the other side residue q j The distance between ij When the distance between the residues is greater than a predetermined distance ("threshold"), it is difficult for the residues to be selected simultaneously. That is, in the Hamiltonian represented by equation (16), a limit is placed on the maximum distance between the residues to be selected.

[0306] In addition, the Hamiltonian represented by equation (18) may be used instead of the Hamiltonian represented by equation (16).

[0307] [Formula 18]

[0308]

[0309] In the Hamiltonian represented by formula (18), there is no restriction on the longest distance between selected residues. The constraint condition shown in the second term on the right side of formula (18) indicates that for a certain residue set, the one with a smaller sum of distances between residues is more likely to be selected. Therefore, according to the Hamiltonian represented by formula (18), for a certain residue set, the larger the sum of allosteric scores for each residue included in the residue set and the smaller the sum of distances between residues, the easier it is for the residue set to be selected.

[0310] Thereafter, the residue set search unit 140d executes the process of step S1820.

[0311] Step S1820: The second Ising calculation unit 142d performs Ising calculation on the Hamiltonian generated by the second Hamiltonian generation unit 141d based on the analog branch.

[0312] With the above, the residue set search unit 140d ends the residue set search Ising calculation process.

[0313] Back to Fig.37 Continue with the description of the residue set search process.

[0314] Step S1750: The third evaluation unit 16d evaluates the residue set based on the residue set score calculated by the residue set search unit 140d. The residue set score is equivalent to the energy represented by the Hamiltonian shown in formula (16), and the lower the value, the higher the evaluation as a path set. The third evaluation unit 16d, for example, determines the residue set with the highest evaluation value based on the residue set score. The third evaluation unit 16d may also determine a specified number of residue sets in descending order of the evaluation values ​​based on the residue set score. The third evaluation unit 16d may also determine a residue set with an evaluation value based on the residue set score higher than a specified value.

[0315] Here, a residue set is associated with a set of pathways that connect the residues included in the residue set and the residues included in the active center, that is, a pathway set. Therefore, the evaluation function is a function that provides an evaluation value of the pathway set.

[0316] exist Fig.41 , as an overview of the search results of the residue sets, an example of the residue set R3 determined to be the residue set with the highest evaluation value is shown. The residue set R3 corresponds to the allosteric control region.

[0317] With the above, the control unit 10d ends the residue set search process.

[0318] Back to Fig.36 Continue with the description of the allosteric pathway prediction process.

[0319] Step S1560: The output unit 15d outputs the prediction result of the residue set search process. The output unit 15d outputs information indicating one or more residue sets determined by the third evaluation unit 16d and the evaluation result of the third evaluation unit 16d as the prediction result. The information indicating the residue set includes, for example, information indicating the residues included in the residue set that is the starting point of the allosteric path and information indicating the residues included in the active center that is the end point of the allosteric path.

[0320] With the above, the control unit 10d ends the allosteric pathway prediction process.

[0321] (Example of Modification of Third Embodiment)

[0322] In the example of the third embodiment described above, the allosteric score was calculated for I36 located at the center of the binding region with RAF, one of the effector proteins, using the crystal structure of normal KRAS (PDBID: 4obe). In this example, the allosteric score calculated in the example of the third embodiment was not used to search for an allosteric control region that controls the reaction with the effector of KRAS.

[0323] Figure 42 to Figure 45 The search results for the allosteric control region of this example are shown. Fig.42 In the figure, the allosteric score of each amino acid residue of KRAS is shown by a bar graph. Fig.42 In the figure, the amino acid residues constituting the allosteric control region obtained in this example are represented by "result 1", "result 2" and "result 3", respectively. Fig.43 , Fig.44 , Fig.45 The diagrams show the band models of the KRAS / RAF complex corresponding to "result 1", "result 2", and "result 3", respectively. Fig.43 , Fig.44 , Fig.45 In FIG. 1 , the amino acids constituting the searched region 41 , the searched region 42 , and the searched region 43 are filled and shown respectively.

[0324] KRAS and RAF bind to two domains (RBD and CRD) of RAF, and then the reaction of KRAS and RAF is controlled by the displacement of GDP / GTP. This embodiment uses the I36 residue present in the center of the KRAS and RAF / RBD binding region as the end point, and implements a search for a residue set including a plurality of starting points connected by a path with a large impact on I36, i.e., an allosteric control region. The region searched by "result 1" represents the region (prescribed vertex periphery) associated with the binding of KRAS and RAF / RBD. The region searched by "result 2" represents the region associated with the binding of KRAS and RAF / CRD. The region searched by "result 3" represents the binding region of GDP and KRAS. The three regions searched by "result 1", "result 2" and "result 3" are regions that are already known to be important regions for the reaction of KRAS and RAF. As described above, it is confirmed that this allosteric control region search method is an effective method for determining the allosteric control region for the interaction of KRAS and RAF.

[0325] (Fourth Embodiment)

[0326] Hereinafter, a fourth embodiment of the present invention will be described in detail with reference to the drawings.

[0327] In the first to third embodiments described above, the case where the allosteric path is a single path is described. In this embodiment, the case where the allosteric path is a multiple path is described. Multiple paths are a concept included in a partial graph of a network graph together with a single path. In multiple paths, there are cases where the paths branch or merge. In addition, in multiple paths, there are cases where the paths have multiple starting points or multiple end points.

[0328] Fig.46 An example of a multiple path is shown. Path P3 is a multiple path from starting point s to end point e. Path P3 is composed of path P31 and path P32. Path P31 passes through vertex N22 (starting point s), vertex N23, vertex N25, vertex N212, vertex N211 and vertex N210 (end point e) in sequence. Path P32 passes through vertex N22 (starting point s), vertex N24, vertex N25, vertex N211 and vertex N210 (end point e) in sequence. Therefore, path P3 is a multiple path that branches to path P31 and path P32 at vertex N22 (starting point s), merges at vertex N25, branches again and merges again at vertex N211. As described above, a multiple path includes branches or merges.

[0329] [Functional structure of the allosteric pathway prediction device 1c]

[0330] Fig.471 is a diagram showing an example of the functional structure of the allosteric path prediction device 1c of the present embodiment. The allosteric path prediction device 1c includes a control unit 10c and a storage unit 20. Fig.47 ) and the allosteric pathway prediction device 1a of the second embodiment ( Fig.15 ) are compared, the control unit 10c and the Ising model information 21c stored in the storage unit 20 are different. The description of the same functions as those in the second embodiment is omitted, and in the fourth embodiment, the description will focus on the parts different from the second embodiment.

[0331] The control unit 10c includes a three-dimensional structure information acquisition unit 11, a network graph generation unit 12, a path calculation unit 13c, a first evaluation unit 14a, and an output unit 15a. Fig.47 ) and the control unit 10a( Fig.15 Here, the functions of the other components (three-dimensional structure information acquisition unit 11, network graph generation unit 12, first evaluation unit 14a and output unit 15a) are the same as those of the second embodiment.

[0332] The path calculation unit 13 c includes a Hamiltonian generation unit 130 c and an Ising calculation unit 131 .

[0333] The function of the Hamiltonian generation unit 130c is the same as that of the Hamiltonian generation unit 130, except that the Hamiltonian to be generated is the Hamiltonian for multiple paths. The objective function and constraint conditions used by the Hamiltonian generation unit 130c to generate the Hamiltonian are pre-stored in the storage unit 20 as Ising model information 21c. The constraint conditions represented by the Ising model information 21c are the constraint conditions for multiple paths.

[0334] Here, refer to Fig.48 , the Hamiltonian for multiple paths generated by the Hamiltonian generating unit 130c is described. An example of the Hamiltonian generated by the Hamiltonian generating unit 130c is represented by equation (14). In equation (14), A, B, C, D, E, and F are constants.

[0335] [Formula 19]

[0336]

[0337] In the Hamiltonian for a single path shown in equation (6), the spin is simply related to the vertex V i and vertex V j The spin corresponding to the connected edge is q ij In the Hamiltonian for multiple paths shown in equation (19), the spin has not only ijIt also represents the spin of the vertex state. In the spin representing the vertex state, there is q i start ,q i end ,q i pass , and q i non -pass These 4. i start It has a value of 1 when the vertex is the starting point, and has a value of 0 when the vertex is not the starting point. i end It has a value of 1 if the vertex is an endpoint, and a value of 0 if the vertex is not an endpoint. i pass It has a value of 1 if the vertex is a through point, and a value of 0 if the vertex is not a through point. i non-pass It has a value of 1 when the vertex is a non-through point, and has a value of 0 when the vertex is not a non-through point.

[0338] like Fig.48 As shown, item 71 is the objective function. Items 72, 73, 74, 75, 76, 77, 78 and 79 are constraints. Items 72 and 73 in the constraints are constraints required by graph theory. Items 74, 75, 76, 77, 78 and 79 in the constraints are constraints required by biological investigations.

[0339] In formula (19), the objective function is the sum of the scores of the edges used in the path. i and vertex V j The weights of the edges are assigned using p ij .

[0340] Item 72 represents the constraint that the state of each vertex is either a start point, an end point, a through point, or a non-through point. i status is the set of spins for each state of vertex Vi (i.e., q i start ,q i end ,q i pass , and q i non-pass ).

[0341] Item 73 represents a constraint for specifying the multiplicity of a path. The multiplicity of a path is represented by the number of starting points, the number of end points, and the number of through-points. The multiplicity of a path is also referred to as the ratio of the number of starting points, the number of end points, and the number of through-points. Parameter α is a parameter that affects the number of specified through-points and is the ratio of through-points to the total number of vertices.

[0342] Item 74 represents a constraint condition for making the distances between vertices of the same type as close as possible (also called aggregation). Item 74 includes a condition for making the distances between starting points as close as possible, a condition for making the distances between end points as close as possible, and a condition for making the distances between passing points as close as possible. The sign of the constraint condition shown in item 74 is negative when the distance between vertices is closer than the average distance between vertices, and positive when the distance between vertices is farther than the average distance between vertices. The constraint condition shown in item 74 is that the greater the number of vertices in the same state, the greater the absolute value.

[0343] Item 75 represents the constraint that the path passes through vertices with the most inflow or outflow as much as possible.

[0344] Item 76 indicates a constraint that the starting point does not become the end point. Item 77 indicates a constraint that the end point does not become the starting point. Item 78 indicates a constraint that a non-through point does not become the end point. Item 79 indicates a constraint that a non-through point does not become the starting point.

[0345] In addition, any one or more of the terms 74, 75, 76, 77, 78, and 79 as constraints required by biological studies may be omitted from the Hamiltonian shown in equation (19). In addition, as constraints required by biological studies, constraints other than terms 74, 75, 76, 77, 78, and 79 may be added. That is, the Hamiltonian shown in equation (19) is an example, and the Hamiltonian may also be changed (customized). The customization of the Hamiltonian will be described in the fifth embodiment below.

[0346] (First Example of the Fourth Embodiment)

[0347] In the second example of the second embodiment, as described above, the reaction between KRAS and the effector protein is allosterically controlled by GDP / GTP. Regarding this allosteric control, the results of searching for important amino acid residues using the Hamiltonian for a single pathway are described in the second example of the second embodiment. In this example, it was confirmed whether the same calculation as the combination of single pathways can be performed using the Hamiltonian for multiple pathways.

[0348] In this embodiment, regarding KRAS structural information (PDB ID: 4obe), the starting point is set to A14, the end point is set to A36, and the allosteric path prediction process is implemented using the Hamiltonian for multiple paths. In addition, in this embodiment, in the Hamiltonian for multiple paths shown in formula (19), a first-order term in which the specified starting point and end point become favorable is added, and the starting point and end point are specified. The calculated path becomes a multiple path that allows branching and confluence in the middle of the path. The amino acid residues contained in the path of the multiple paths are compared with the results obtained by combining a single path to evaluate the appropriateness of the allosteric path prediction process using the Hamiltonian for multiple paths.

[0349] Fig.49 The search results for the important amino acid residues of this example are shown in Fig.49 In the embodiment, the search results of important amino acid residues by the combination of single paths are shown as a comparative example together with the search results using the Hamiltonian for multiple paths in this embodiment. In the optimal path (allowing branching and confluence) calculated using the Hamiltonian for multiple paths, the amino acid residues searched by using the Hamiltonian for a single path are also shown. The amino acid residues searched using the Hamiltonian for a single path are A8, A15 to A18, and A57 to A59, etc. Fig.50 The allosteric pathways calculated using the Hamiltonian for multiple paths are shown. Fig.50 In the figure, the allosteric pathway is represented by superimposing the three-dimensional structure of the protein.

[0350] The search results of this example suggest that by using the Hamiltonian for multiple paths, the important amino acid residues in allosteric control can be determined using only the optimal solution of the Hamiltonian.

[0351] (Second Example of the Fourth Embodiment)

[0352] In the example of the third embodiment, as described above, the reaction between KRAS and the effector protein is allosterically controlled by GDP / GTP. Regarding this allosteric control, the results of searching for an allosteric control region using a Hamiltonian for a single pathway are described in the example of the third embodiment. In this example, it was confirmed that the allosteric control region search can be performed using a Hamiltonian for multiple pathways.

[0353] In this embodiment, regarding KRAS structural information (PDB ID: 4obe), the end point was set to I36, and the allosteric control region prediction process was implemented using the Hamiltonian for multiple paths. In addition, in this embodiment, the starting point was not specified. The calculated path becomes a multiple path that allows multiple starting points and branching and confluence in the middle of the path. By comparing the selected multiple starting points with the known allosteric control regions of KRAS, the appropriateness of the allosteric control region prediction process using the Hamiltonian for multiple paths is evaluated.

[0354] Figure 51 to Figure 52 The search results for the allosteric control region of this example are shown. Fig.51 In FIG. 1 , a region in KRAS that is important for the reaction with RAF, one of the effector proteins, and a starting point obtained in this example are indicated as “result”. Fig.52 is a diagram showing a band model of the KRAS / RAF complex corresponding to "result". Fig.52 In the figure, the amino acids constituting the starting point of the searched path are filled and shown in regions 51, 52, and 53, respectively. The amino acids constituting the starting point of the searched path represent the binding region of GDP and KRAS. This result is consistent with the fact that the reaction between KRAS and the effector protein is allosterically controlled by GDP / GTP.

[0355] The search results of this embodiment suggest that by using the Hamiltonian for multiple paths, the allosteric control region can be determined using only the optimal solution of the Hamiltonian.

[0356] In addition, in this embodiment, an example of a case where the Hamiltonian for multiple paths is used instead of the Hamiltonian for a single path to perform the same processing as the important amino acid residue search processing described in the second embodiment and the region search processing described in the modified example of the third embodiment is described, but the present invention is not limited thereto. In the allosteric path prediction processing described in the first embodiment or the region search processing described in the third embodiment, the Hamiltonian for multiple paths may be used instead of the Hamiltonian for a single path to perform the processing.

[0357] In addition, in each of the above-mentioned embodiments, in the allosteric path prediction devices 1, 1a, 1b, and 1c, as described above, the Ising calculation is performed by the analog branch machine, so that the Ising calculation can be performed at a high speed compared to the case where the analog branch machine is not used. Since the calculation speed is fast, the allosteric path prediction process can be repeatedly performed within a practical time. In the allosteric path prediction devices 1, 1a, 1b, and 1c, by repeatedly performing the allosteric path prediction process, an exhaustive search can be performed.

[0358] Exhaustive search refers to, for example, calculating a set of candidates for the end point of a certain starting point (end point candidate set), or a set of candidates for the starting point of a certain end point (start point candidate set). For example, the starting point may be modified in various ways, and the end point candidate set may be calculated for all the starting points included in the network diagram. In another example, the end point may be modified in various ways, and the start point candidate set may be calculated for all the end points included in the network diagram.

[0359] In another example, the process of calculating the candidate set of important amino acid residues for the specified starting point and the specified end point (important amino acid residue candidate set) in the second embodiment is described, but the process can also be repeatedly performed on all the groups of starting points and end points included in the network diagram, and the groups of all the starting points and end points included in the network diagram are corresponding to the important amino acid residue candidate set. The important amino acid residue candidate set for all the groups of starting points and end points included in the network diagram can be databased as all the information related to the allosteric control in the target protein. The databaseization will be described in the fifth embodiment below.

[0360] In addition, in this embodiment, an example of using the Hamiltonian for multiple paths is described, but in the exhaustive search, the Hamiltonian for multiple paths may be repeatedly used, or the Hamiltonian for a single path may be repeatedly used. It is also possible to combine the processing using the Hamiltonian for multiple paths and the processing using the Hamiltonian for a single path.

[0361] In addition, in each of the above-mentioned embodiments, an example of a case where each functional unit or storage unit 20 of the control unit 10, 10a, 10b, 10c in the metamorphic path prediction device 1, 1a, 1b, 1c is implemented as a function of one computer is described, but it is not limited to this. Each functional unit or storage unit 20 of the control unit 10, 10a, 10b, 10c may be distributed in a plurality of servers, for example. In addition, each functional unit or storage unit 20 of the control unit 10, 10a, 10b, 10c may be implemented as a cloud server. In addition, in the case where each functional unit or storage unit 20 of the control unit 10, 10a, 10b, 10c is distributed in a plurality of servers, or in the case of being implemented as a cloud server, these plurality of servers or cloud servers may also be distributed in different countries.

[0362] In the fifth embodiment described below, an example of a case where the metamorphic path prediction device is implemented as a plurality of servers will be described. In the fifth embodiment, details of the user interface will be described.

[0363] (Fifth Embodiment)

[0364] [Structure of the Allosteric Pathway Prediction System 3]

[0365] Fig.53 1 is a diagram showing an example of the structure of the allosteric path prediction system 3 of the present embodiment. The allosteric path prediction system 3 includes an interface unit 31, a network graph generation unit 32, a Hamiltonian generation unit 33, and an allosteric control information calculation unit 34. As an example, the interface unit 31, the network graph generation unit 32, the Hamiltonian generation unit 33, and the allosteric control information calculation unit 34 are servers, respectively.

[0366] The interface unit 31 performs input and output, pre-processing, and post-processing of various information.

[0367] The interface unit 31 obtains input information from the user terminal 5. The input information is information input by the user to the user terminal 5. The input information includes, for example, three-dimensional structure information 41. The three-dimensional structure information 41 is information indicating the three-dimensional structure of a protein. The user terminal 5 is a terminal device such as a PC used by the user.

[0368] In addition, the interface unit 31 pre-processes the input information. The pre-processing includes converting the information input from the user into a format that can be processed by the allosteric pathway prediction system 3 .

[0369] In addition, the interface unit 31 performs post-processing on the prediction result of the allosteric pathway prediction system 3. The post-processing includes converting the prediction result into a form desired by the user. The prediction result includes allosteric control information 42. The allosteric control information 42 includes one or more of allosteric pathway information, allosteric pathway score, residue score or allosteric score.

[0370] In addition, the interface unit 31 outputs the prediction result to the user terminal 5 .

[0371] The network diagram generation unit 32 generates a network diagram based on the three-dimensional structure information 41. The network diagram generation unit 32 has the same function as the network diagram generation unit 12.

[0372] The Hamiltonian generation unit 33 generates a Hamiltonian based on the network graph generated by the network graph generation unit 32. The Hamiltonian generation unit 33 has the same function as that of the Hamiltonian generation unit 130.

[0373] The allosteric control information calculation unit 34 performs Ising calculation on the Hamiltonian generated by the Hamiltonian generation unit 130 based on the analog branch. The allosteric control information calculation unit 34 has the same function as the Ising calculation unit 131. The allosteric control information calculation unit 34 generates allosteric control information 42 based on the execution result of the Ising calculation.

[0374] The user terminal 5 obtains the prediction result of the structural path prediction system 3 output from the interface unit 31. That is, the user terminal 5 obtains the calculation result of the path connecting the vertices on the network graph. In addition, the user terminal 5 and the structural path prediction system 3 may be installed in different countries.

[0375] [Allosteric Pathway Prediction Processing in Allosteric Pathway Prediction System 3]

[0376] Next, refer to Figure 54 to Figure 56 , describing the allosteric pathway prediction process in the allosteric pathway prediction system 3. Fig.54 : is a diagram showing an example of the allosteric pathway prediction process according to the present embodiment.

[0377] Step S810: The interface unit 31 obtains the three-dimensional structure information 41 as input information from the user terminal 5. The three-dimensional structure information 41 is three-dimensional structure information about the protein to be predicted for the allosteric pathway. Then, the allosteric pathway prediction system 3 executes the process of step S820.

[0378] Step S820: The network diagram generation unit 32 executes the network diagram generation process. After that, the network diagram generation unit 32 executes the process of step S830.

[0379] Here, refer to Fig.55 , the network graph generation process is explained. Fig.55 This is a diagram showing an example of the network diagram generation process according to the present embodiment. Fig.55 The network diagram generation process shown in Fig.54 The process is performed in step S820 as shown.

[0380] Step S910: The network diagram generation unit 32 determines whether the generated network diagram is a standard diagram. The network diagram generation unit 32 determines whether it is a standard diagram based on the diagram setting information. The diagram setting information includes information indicating the type of the network diagram. In addition, the diagram setting information can be obtained from the user terminal 5 by the interface unit 31 in step S920, or can be obtained from the user terminal 5 by the interface unit 31 in step S810 by including it in the input information.

[0381] When the network diagram generation unit 32 determines that the diagram setting information indicates a standard diagram (step S910; Yes), it executes the process of step S930. On the other hand, when the network diagram generation unit 32 determines that the diagram setting information does not indicate a standard diagram (step S910; No), it executes the process of step S920.

[0382] Step S920: The network diagram generation unit 32 specifies the expression generation method of the generated network diagram. The information indicating the expression generation method is included in the above-mentioned graph setting information. The network diagram generation unit 32 specifies the expression generation method based on the graph setting information. The graph setting information, for example, includes information for setting weights other than the weights represented by the transfer probability matrix as weights assigned to the edges of the network diagram. This information is a formula for calculating weights other than the weights represented by the transfer probability matrix. Thereafter, the network diagram generation unit 32 performs the processing of step S930.

[0383] Step S930: The network graph generation unit 32 generates a network graph based on the three-dimensional structure information 41. The network graph generation unit 32 generates a network graph including the generation of vertices, interaction matrices, and signal transfer matrices for amino acid residues. In addition, when the representation generation method is specified, the network graph generation unit 32 calculates weights other than the weights represented by the transfer probability matrix.

[0384] As described above, the network diagram generation unit 32 ends the network diagram generation process.

[0385] Back to Fig.54 Continue with the description of the allosteric pathway prediction process.

[0386] Step S830: The network diagram generation unit 32 performs the addition of search conditions and the setting of search exclusion areas. The search condition information includes the above-mentioned site designation information (information indicating the allosteric control site and the allosteric control target site, and therefore information indicating the starting point and the end point. In addition, the starting point and the end point do not have to be a single point, and the starting point and the end point can be a set of points composed of multiple points, respectively), information indicating the additional search conditions, and information indicating the search exclusion area. The additional search conditions are search conditions specified by the user. The search exclusion area is an area in the three-dimensional structure of the protein shown in the three-dimensional structure information 41 that is excluded from the search. The search exclusion area is specified by the user. In addition, the search condition information can be obtained from the user terminal 5 by the interface unit 31 in step S830, or it can be obtained from the user terminal 5 by the interface unit 31 in step S810 and included in the input information.

[0387] When the search condition information includes information indicating an additional search condition, the network diagram generation unit 32 adds the additional search condition to the search condition. In addition, when the search condition information includes information indicating a search exclusion area, the network diagram generation unit 32 excludes the search exclusion area indicated by the information from the search. Thereafter, the Hamiltonian generation unit 33 performs the processing of step S840.

[0388] In addition, the interface unit 31 may also change the interface to the user's desired mode according to the user's setting. For example, the interface unit 31 may change the interface to specify information indicating a domain instead of the site specifying information included in the search condition information. A domain is a unit indicating a structural part in the three-dimensional structure of a protein.

[0389] Step S840: The Hamiltonian generation unit 33 sets the Hamiltonian. Then, the allosteric control information calculation unit 34 executes the process of step S850.

[0390] Here, refer to Fig.56 , the process of setting the Hamiltonian is explained. Fig.56 : is a diagram showing an example of the Hamiltonian setting process according to the present embodiment. Fig.56 The setting of the Hamiltonian shown in Fig.54 The process is performed in step S840 as shown.

[0391] Step S1010: The Hamiltonian generation unit 33 determines whether to specify the use of the standard QUBO as the Hamiltonian to be generated. The Hamiltonian generation unit 33 determines whether to specify the use of the standard QUBO as the Hamiltonian based on the Hamiltonian designation information. The Hamiltonian designation information includes information for designating the Hamiltonian. The Hamiltonian designation information is specified by the user. In addition, the Hamiltonian designation information can be obtained by the interface unit 31 from the user terminal 5 in step S1010, or can be obtained by the interface unit 31 from the user terminal 5 in step S810 by being included in the input information.

[0392] When the Hamiltonian generation unit 33 determines that the standard QUBO is specified as the generated Hamiltonian (step S1010; Yes), the processing of step S1030 is performed. On the other hand, when the Hamiltonian generation unit 33 determines that the standard QUBO is not used as the generated Hamiltonian (step S1010; No), the processing of step S1020 is performed.

[0393] Step S1020: The Hamiltonian generation unit 33 adds information specifying the calculation formula represented by the Hamiltonian designation information to the Hamiltonian information. In addition, the calculation formula may be a quadratic formula based on the Ising model, a high-order formula after the Ising model is expanded, a rational formula, or a general real-valued function. As described above, the Hamiltonian information is information including the objective function and the constraint conditions, and is pre-stored in the storage unit.

[0394] For example, when the Hamiltonian designation information includes a calculation formula specified by the user, the Hamiltonian generation unit 33 adds the calculation formula to the Hamiltonian information. For example, when the Hamiltonian designation information includes information specifying a calculation rule for weights assigned to edges of the network graph, the Hamiltonian generation unit 33 adds the information to the Hamiltonian information. The calculation rule for weights is to multiply a plurality of edges by p as weights. ij (Including p in the objective function ij logarithm of ), or addition (the objective function does not include p ij The logarithm of p ij Thereafter, the Hamiltonian generation unit 33 performs the process of step S1030.

[0395] Step S1030: The Hamiltonian generation unit 33 determines whether it is specified to use the standard constraint formula as the term of the constraint formula of the Hamiltonian to be generated. The Hamiltonian generation unit 33 determines whether the term of the constraint formula is a standard constraint formula based on the constraint formula designation information. The constraint formula designation information is specified by the user. The constraint formula designation information contains information for specifying the constraint formula. The constraint formula may also represent a constraint condition related to the shape of a local graph of the network diagram. As a restriction related to the shape of a local graph of the network diagram, it may also be a constraint related to forming a path without branches and inflows between 1 point of the starting point set and 1 point of the end point set, or a constraint related to the distance between vertices related to the local graph. In addition, the constraint formula designation information may be obtained by the interface unit 31 from the user terminal 5 in step S1030, or may be obtained by the interface unit 31 from the user terminal 5 in step S810 and included in the input information.

[0396] When the Hamiltonian generation unit 33 determines that the standard constraint formula is specified as the term of the constraint formula of the generated Hamiltonian (step S1030; Yes), the processing of step S1050 is performed. On the other hand, when the Hamiltonian generation unit 33 determines that the standard constraint formula is not used as the term of the constraint formula of the generated Hamiltonian (step S1030; No), the processing of step S1040 is performed.

[0397] Step S1040: The Hamiltonian generation unit 33 adds information for specifying the constraint formula indicated by the constraint formula specifying information to the Ising model information. Thereafter, the Hamiltonian generation unit 33 performs the process of step S1050.

[0398] Step S1050: The Hamiltonian generation unit 33 generates a Hamiltonian based on the site designation information, the information indicating the search exclusion area, and the Hamiltonian information. When the Hamiltonian information includes information specifying a calculation formula specified by the user, the Hamiltonian generation unit 33 generates a Hamiltonian based on the information. When the Hamiltonian information includes information specifying a constraint specified by the user, the Hamiltonian generation unit 33 generates a constraint based on the information. When the Hamiltonian information does not include information specified by the user, the Hamiltonian generation unit 33 generates a Hamiltonian based on the standard QUBO and standard constraint represented by the pre-stored Hamiltonian information. When the Hamiltonian searches for important areas based on the calculation of the allosteric control information, it can also be composed of a vector composed of one or more indicators that preferably take smaller values, and the inner product of vectors that assign weights to each indicator. The Hamiltonian can be designed to preferably have a large value, and a local graph with a large value of the Hamiltonian can be searched through the optimization in step S860. As described above, the Hamiltonian generation unit 33 completes the process of setting the Hamiltonian.

[0399] Back to Fig.54 Continue with the description of the allosteric pathway prediction process.

[0400] Step S850: The allosteric control information calculation unit 34 sets and automatically tunes the calculation parameter list. The calculation parameter list includes hyperparameters used when performing optimization calculations through the simulation branch machine and constants of each item included in the Hamiltonian set in the Hamiltonian generation unit 33. In the automatic tuning, for these multiple parameters, a combination of values ​​that appropriately perform optimization calculations based on the simulation branch machine is automatically calculated and set. Automatic tuning is performed, for example, through conditional branches based on the characteristics of the target protein and the network graph. Alternatively, it can also be performed by an optimization method such as Bayesian optimization. Thereafter, the allosteric control information calculation unit 34 performs the processing of step S860.

[0401] Step S860: The allosteric control information calculation unit 34 performs optimization calculation on the Hamiltonian generated by the Hamiltonian generation unit 33. The optimization calculation is a calculation as follows: a local graph of the network graph is made to correspond to a Hamiltonian representing the allosteric control correlation degree related to the specified allosteric control site and the allosteric control target site, and a local graph having a small value of the Hamiltonian is searched, thereby calculating a set of local graphs that may be associated with allosteric control.

[0402] The allosteric control information calculation unit 34 can use an optimization calculation unit called an Ising machine as a unit for searching for a local graph with a small value of the Hamiltonian, or can use software that extends the Ising machine and processes high-order expressions, rational expressions or general real-valued functions. In addition, calculations can also be performed based on a quasi-optimal solution given by a result given in a manner with a probability change. As a means of these calculations, a simulated bifurcation machine can also be used. The allosteric control information calculation unit 34 can calculate multiple solution candidates by repeatedly executing the Ising machine. In addition, the allosteric control information calculation unit 34 can calculate multiple solution candidates based on the Hamiltonian that exhibits the best optimization for a path set. A path set is a set of multiple paths that include quasi-optimal paths in addition to the optimal path.

[0403] In addition, the allosteric control information calculation unit 34 may use the natural number k as a threshold value to calculate the set of k local graphs with the smallest Hamiltonian value, or may multiply the Hamiltonian by a weight determined for each local graph and sort them, and calculate the set of local graphs in this order. In addition, another threshold m may be determined, and a local graph with a Hamiltonian value less than the threshold m may be selected. The allosteric control information calculation unit 34 provides the calculation result to the interface unit 31. Thereafter, the interface unit 31 performs the processing of step S870.

[0404] Step S870: The interface unit 31 outputs the allosteric control information. For example, the interface unit 31 outputs the allosteric control information to the user terminal 5. In addition, the interface unit 31 may store the allosteric control information in a database ( Fig.53 After that, the interface unit 31 executes the process of step S880.

[0405] Here, the allosteric control information is generated as post-processing by the interface unit 31. For example, the allosteric control information calculation unit 34 arranges the calculated partial graphs in descending order of the evaluation values ​​(allosteric path scores) calculated from the Hamiltonian based on the calculation results of the allosteric control information calculation unit 34, and generates the allosteric control information as a set of the partial graph and the evaluation value of the path.

[0406] In addition, as the type of local graph, only a local graph of a single path may be selected. In another example, the allosteric control information calculation unit 34 arranges the evaluation values ​​(residue scores) of the vertices included in a plurality of local graphs with high evaluation values ​​in descending order, and generates allosteric control information by combining the vertices and the evaluation values.

[0407] In addition, in another example, the allosteric control information calculation unit 34 arranges the endpoint sets with high evaluation values ​​(allosteric scores) indicating the degree of influence on the starting point set specified by the user, or the starting point sets with high evaluation values ​​(allosteric scores) indicating the degree of influence on the endpoint set specified by the user, in order of the evaluation values ​​(allosteric scores) from high to low, and generates allosteric control information by grouping the starting point set or the endpoint set with the evaluation value of the starting point set or the endpoint set.

[0408] Step S880: The user terminal 5 graphically displays the calculation result of the allosteric control information calculation unit 34. The interface unit 31 outputs graphic information for graphic display to the user terminal 5. The graphic information is, for example, information representing the amino acid residues corresponding to the vertices included in the predicted path in the three-dimensional structure of the protein. The graphic information can also be customized as post-processing by the interface unit 31 before being output to the user terminal 5. For example, the interface unit 31 changes the graphic information according to the user's settings so that the calculation result of the allosteric control information calculation unit 34 is graphically displayed in the form desired by the user.

[0409] As described above, the allosteric pathway prediction system 3 ends the allosteric pathway prediction process.

[0410] Next, as a modification of the present embodiment, an example in which the allomorphic pathway prediction system 3 is used will be described.

[0411] (First Modification of Fifth Embodiment)

[0412] Fig.57 1 is a diagram showing an example of the structure of a system B1 of a first variant of the present embodiment. The system B1 is operated as a computing system. The system B1 includes a local allosteric pathway prediction system 3, an inquiry interface 6, an effect function database 7, a compound database 8, a computing processing server 9, a computing experience database 16, and a rights processing system 17. The allosteric pathway prediction system 3, the inquiry interface 6, the computing processing server 9, and the rights processing system 17 are servers. As an example, the effect function database 7, the compound database 8, and the computing experience database 16 are database servers.

[0413] The allosteric pathway prediction system 3 obtains input information D1 from the user terminal 5 via the query interface 6. The input information D1 includes, for example, one or more of the target protein, the allosteric control target site (end point), the allosteric control site (start point), the domain, the effect or function to be searched, the candidate compound that produces the effect, and the designated mutation information.

[0414] The allosteric pathway prediction system 3 performs an allosteric pathway prediction process based on the input information D1. Fig.57The computing processing server 9 can use the experience information stored in the computing experience database 16 in the execution of the optimization calculation. The experience information includes, for example, information indicating the effective Hamiltonian or effective parameter settings based on past analysis results. By using the experience information, the user can try to use the analysis of various conditions stored as experience information even if the conditions (Hamiltonian or parameter settings, etc.) are not reconsidered. That is, in the allosteric pathway prediction system 3, by storing the past analysis results in the computing experience database 16, the past analysis results can be reused.

[0415] The allosteric pathway prediction system 3 refers to the effect / function database 7 to establish a correspondence between the calculated allosteric control information and the effect or function. In addition, the allosteric pathway prediction system 3 refers to the compound database 8 to establish a correspondence between the calculated allosteric control information and the compound. The allosteric pathway prediction system 3 outputs the allosteric control information corresponding to one or more of the effect, function or compound to the query interface 6. In addition, the allosteric pathway prediction system 3 may also output only the allosteric control information to the query interface 6.

[0416] Here, sometimes the data related to the effect or function stored in the effect / function database 7 and the data related to the compound stored in the compound database 8 include data that cannot be provided to the user. The data that cannot be provided to the user is, for example, data that cannot be provided to the user due to safety, experience, or rights relationship when the user and other users (pharmaceutical companies, etc.) jointly develop drugs. The authority processing system 17 determines whether one or more of the effects, functions, or compounds corresponding to the allosteric control information output from the allosteric pathway prediction system 3 to the inquiry interface 6 is information that can be provided to the user.

[0417] The query interface 6 outputs the information of the allosteric control that is determined by the rights processing system 17 to be provided to the user as output information D2 to the user terminal 5, which is output from the allosteric pathway prediction system 3 and corresponds to one or more of the effects, functions, or compounds. The output information D2 includes, for example, a candidate list of allosteric control target sites (end points), a candidate list of allosteric control sites (starting points), a list of important amino acid residues, a domain, an effect / function candidate list, and one or more of the candidate compound candidate lists. The effect / function candidate list is a list of candidates for effects or functions predicted (expected) based on the search results. The candidate compound list is a list of candidates for compounds predicted based on the search results when allosteric control is exerted.

[0418] In addition, when the input information D1 includes information specifying an effect or function to be searched, the query interface 6 may also include a candidate compound that acts on the effect or function in the output information D2. In this case, the query interface 6 establishes a correspondence between the effect or function corresponding to the calculated allosteric control information and the compound corresponding to the allosteric control information via the allosteric control information. Based on the result of the correspondence, the query interface 6 includes a candidate compound corresponding to the information specifying the effect or function to be searched contained in the input information D1 in the output information D2.

[0419] Similarly, when the input information D1 includes information specifying a compound to be searched, the query interface 6 may include candidates for the effects or functions caused by the compound in the output information D2.

[0420] (Second Modification of Fifth Embodiment)

[0421] Fig.58 1 is a diagram showing an example of the structure of the system B2 of the second variant of the present embodiment. The system B2 is used as a database. The system B2 includes an allosteric pathway prediction system 3, an inquiry interface 6, an effect / function database 7, a compound database 8, a calculation processing server 9, a calculation experience database 16, a rights processing system 17, a target protein database 18, and an allosteric control information database 19. If the system B2 of the second variant ( Fig.58 ) and the system B1 according to the first modification ( Fig.57 ) are compared, the target protein database 18 and the allosteric control information database 19 are different. The description of the same functions as those in the first modification example is omitted, and the second modification example is mainly described with the difference from the first modification example.

[0422] As an example, the target protein database 18 and the allosteric control information database 19 are database servers, respectively.

[0423] The allosteric pathway prediction system 3 generates allosteric control information in advance. Here, the allosteric control information calculation unit 34 (in Fig.57 The calculation processing server 9 performs optimization calculations (not shown). The calculation processing server 9 performs allosteric pathway prediction processing based on the three-dimensional structural information of the protein stored in the target protein database 18. As a result of the allosteric pathway prediction processing, the calculation processing server 9 generates allosteric control information.

[0424] The calculation processing server 9 stores the generated allosteric control information in advance in the allosteric control information database 19. That is, in the system B2, the results of the allosteric pathway prediction process performed in advance are stored as a database.

[0425] When the allosteric pathway prediction process is newly performed, the calculation processing server 9 newly adds the allosteric control information, which is the result of the allosteric pathway prediction process, to the allosteric control information database 19. For example, when the stereostructure information is newly added to the target protein database 18, the calculation processing server 9 performs the allosteric pathway prediction process based on the added stereostructure information.

[0426] The computation processing server 9 may also cause the rights processing system 17 to determine whether the allosteric control information is publicly available before adding the allosteric control information to the allosteric control information database 19. The computation processing server 9 stores the allosteric control information determined by the rights processing system 17 as publicly available in the allosteric control information database 19. The computation processing server 9 may discard the allosteric control information determined by the rights processing system 17 as not publicly available, or may store the allosteric control information in the allosteric control information database 19 with a non-discloseable flag.

[0427] In addition, the calculation processing server 9 can use the allosteric control information stored in the allosteric control information database 19 as an existing result in the allosteric pathway prediction process.

[0428] When receiving a request from the user terminal 5, the query interface 6 outputs the provided data D3 to the user terminal 5. The query interface 6 refers to the allosteric control information database 19, and outputs the allosteric control information stored in the allosteric control information database 19 to the user terminal 5 as the provided data D3. When the query interface 6 refers to the allosteric control information database 19, the rights processing system 17 determines whether the allosteric control information of the reference destination can be disclosed. The provided data D3 only includes the allosteric control information that can be disclosed.

[0429] (Summary of each embodiment)

[0430] The functions of the allosteric pathway prediction device or the allosteric pathway prediction system of each embodiment are summarized. The allosteric pathway prediction device or the allosteric pathway prediction system of each embodiment has a basic function of performing allosteric pathway prediction processing, and has a calculation function based on aggregation and a customization function.

[0431] In the function of performing allosteric path prediction processing, the optimal path or important amino acid residues in the path are searched based on a predetermined criterion. The predetermined criterion is to minimize the energy represented by the Hamiltonian.

[0432] In the set-based operation function, not only the optimal path is searched, but also the path set is searched. The path set can be calculated by repeatedly performing the allosteric path prediction process on the Hamiltonian of a single path, or by performing the allosteric path prediction process on the Hamiltonian of multiple paths.

[0433] In addition, in the calculation function based on the aggregation, it is possible to calculate the path set and multiple paths under various conditions according to the setting of the Hamiltonian. The various conditions are obtained by specifying various conditions as the constraints required for biological studies in the Hamiltonian for multiple paths.

[0434] In addition, in the set-based calculation function, various calculations are performed based on the path set and the multi-path set. The various calculations include an important amino acid residue search process for searching for vertices with higher evaluation values ​​(residue scores) among the vertices included in the multiple paths and the multi-path set, and a region search process for searching for vertices with higher evaluation values ​​(allosteric scores) indicating the degree of influence on a certain vertex.

[0435] In the allosteric path prediction device or allosteric path prediction system of each embodiment, Ising calculation, optimization calculation of extended Ising calculation and simulation branch machine can be used to calculate path sets and multiple path sets at high speed. Therefore, it is possible to exhaustively search for a candidate set of the end point and end point set (allosteric control target site) for the starting point and the starting point set, a candidate set of the starting point set (allosteric control site) for the end point and the end point set, etc. In addition, the search results can be databased. Experience related to the search of tuned parameters used in the search can also be databased. In addition, the search results can be set to information combined with the effect / candidate list and the candidate compound list.

[0436] In the customized function, the evaluation function used to evaluate the path can be changed. The evaluation function is a Hamiltonian. The customized function is realized by changing the objective function included in the Hamiltonian or changing or adding the constraint condition.

[0437] In the customization function, various criteria can be set according to the user's specification to search for a path. For example, the Hamiltonian for multiple paths is changed according to the user's specification. In the customization function, it is also possible to specify a user's own objective function, specify a set of avoidance vertices, or specify a set of mandatory pass vertices. The avoidance vertex set is a set of vertices that are not included in the path. The mandatory pass vertex set is a set of vertices that must be included in the path.

[0438] In the custom function, newly acquired biological knowledge can be reflected in the allosteric pathway prediction process by changing the Hamiltonian.

[0439] As described above, in the allosteric path prediction device or allosteric path prediction system of each embodiment, one or more of the evaluation function (Hamiltonian) or the conditions for calculating the path set and the multiple path set (by specifying the avoidance vertex set, or forcing the passing vertex set, etc.) can be changed based on the external specification. The external refers to the user's operation, the file in which the instruction for customization is shown, or the allosteric path prediction device, or the information processing device separate from the allosteric path prediction system, etc.

[0440] In addition, in each of the above-mentioned embodiments, an example of using a simulated branching machine for calculation of the Ising model in the allosteric pathway prediction process is described, but it is not limited to this. Algorithms other than simulated branching machines can be used for calculation of the Ising model. For example, an algorithm such as simulated annealing can be used to calculate the Ising model. However, as described above, in order to achieve high speed and perform exhaustive search, a simulated branching machine is preferred.

[0441] In the above embodiment, an example of calculating the solution of the combinatorial optimization problem of QUBO based on the Ising model in the allosteric pathway prediction process is described, but the present invention is not limited thereto. The solution of the combinatorial optimization problem may be calculated based on a method other than the Ising model.

[0442] As described above, the allosteric path prediction device of each embodiment (in each embodiment, the allosteric path prediction device 1 , 1a , 1b , 1c or the allosteric path prediction system 3 ) includes a network graph generation unit 12 , 32 and a path calculation unit 13 , 13b , 13c .

[0443] The network diagram generating unit 12 generates a network diagram based on the three-dimensional structural information of the protein, wherein the network diagram has vertices corresponding to at least one of the amino acid residues constituting the protein and any binding substance binding to the protein, and weights of interactions between the amino acid residues and at least one of the amino acid residues in the arbitrary binding substance are assigned to the edges.

[0444] The path calculation units 13 , 13 b , and 13 c calculate paths and multipaths connecting vertices on the network graph generated by the network graph generation unit 12 , based on a weighted evaluation function (in each embodiment, a Hamiltonian in the QUBO format).

[0445] According to this structure, the allosteric path prediction device of this embodiment can calculate the path and multiple paths connecting the vertices on the network graph based on the weighted evaluation function, and thus can predict the amino acid residues that contribute to allosteric control based on the three-dimensional structural information of the protein.

[0446] Conventionally, several in silico prediction methods have been proposed as prediction methods for allosteric pathways. Conventional prediction methods for allosteric pathways are roughly divided into two types: molecular dynamics methods (dynamic methods) and prediction methods based on network models (static methods).

[0447] In the molecular dynamics method, the structural changes of the protein are captured by molecular dynamics simulation to predict the allosteric control site. On the other hand, in the prediction method in the network model, the allosteric control site is predicted on the network reflecting the specific structure of the protein. As the prediction method in the network model, for example, a prediction method based on machine learning of network information (Patent Document 1) and a prediction method in an information transfer model on the network (Non-Patent Document 1, Non-Patent Document 2) are known.

[0448] The movement of molecules related to protein function may cause delays ranging from milliseconds to seconds, and even the most advanced computers cannot calculate all of them from first principles when using molecular dynamics simulations.

[0449] In the prediction in the network model, in addition to the low prediction accuracy, there are also technical problems such as the selection of teacher data and the adaptability (scalability) to larger proteins (protein complexes, etc.). Specifically, for example, in the prediction method described in patent document 1, there are multiple parameters such as evolutionary characteristics and physical characteristics, and the protein is represented by a network diagram reflecting the structure of the protein, and the important amino acid residues in the allosteric control are calculated using the random forest method according to the parameters in the network diagram. However, in the prediction method described in patent document 1, the parameters used are many, and it is unclear whether it can be applied to all proteins. In addition, in the prediction method described in patent document 1, the prediction accuracy is affected by the quality of the teacher data, so it is believed that the prediction result of the protein that does not have a structure similar to the protein used in the preparation of the teacher data or the physical property becomes low.

[0450] In addition, for example, in the prediction method recorded in non-patent literature 1 or non-patent literature 2, a network diagram reflecting the structural information of the protein is produced, and the allosteric path is predicted, thereby calculating the important amino acid residues in the allosteric control. However, in the prediction method recorded in non-patent literature 1 or non-patent literature 2, the prediction accuracy is not high. In addition, in the prediction method recorded in non-patent literature 1 or non-patent literature 2, if it is a problem of large scale, the solution is difficult to converge, and the expansibility to large structures such as protein complexes is low.

[0451] On the other hand, in the allosteric path prediction device of the present embodiment, since the paths and multiple paths connecting the vertices on the network graph can be calculated based on the weighted evaluation function (Hamiltonian of the Ising model), the prediction accuracy can be improved, and the scalability to large structures is high. The network graph with amino acid residues as vertices and weights assigned to the edges based on the interactions between amino acid residues is a model simplified to a degree that can be realistically calculated. In addition, in the allosteric path prediction device of the present embodiment, since multiple paths including optimal paths and quasi-optimal paths can be calculated based on the weighted evaluation function, the prediction accuracy can be improved.

[0452] In addition, the allosteric pathway prediction device 1, 1a, 1b, 1c, 1d or a part of the allosteric pathway prediction system 3 in the above-mentioned embodiment, such as the three-dimensional structure information acquisition unit 11, the network diagram generation unit 12, 32, the path calculation unit 13, 13b, 13c, 13d, the Hamiltonian generation unit 33, the allosteric control information calculation unit 34, the evaluation unit 14, the first evaluation unit 14a, the second evaluation unit 14b, the third evaluation unit 16d and the output unit 15, 15a, 15b, 15d can also be implemented by a computer. In this case, it can also be implemented by recording the program for realizing the control function in a computer-readable recording medium, and causing the computer system to read and execute the program recorded in the recording medium. In addition, the "computer system" mentioned here is a computer system built into the allosteric pathway prediction device 1, 1a, 1b, 1c, 1d or the allosteric pathway prediction system 3, including hardware such as OS and peripheral devices. In addition, "computer-readable recording medium" refers to removable media such as floppy disks, optical magnetic disks, ROMs, CD-ROMs, and storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording medium" may also include recording media that dynamically retain programs for a short period of time, such as communication lines in the case of sending programs via networks such as the Internet or communication lines such as telephone lines, and recording media that retain programs for a certain period of time, such as volatile memories inside computer systems that serve as servers or clients in this case. In addition, the above-mentioned program can be used to implement part of the above-mentioned functions, and the above-mentioned functions can also be implemented by combining with programs already recorded in the computer system.

[0453] In addition, part or all of the allosteric path prediction device 1, 1a, 1b, 1c, 1d or the allosteric path prediction system 3 in the above-mentioned embodiment can also be implemented as an integrated circuit such as LSI (Large Scale Integration). Each functional block of the allosteric path prediction device 1, 1a, 1b, 1c, 1d or the allosteric path prediction system 3 can be independently processed, or part or all of them can be integrated and processed. In addition, the method of integrated circuitization is not limited to LSI, and it can also be implemented by a dedicated circuit or a general-purpose processor. In addition, in the case where an integrated circuit technology that replaces LSI emerges due to the advancement of semiconductor technology, an integrated circuit based on this technology can also be used.

[0454] As mentioned above, one embodiment of the present invention has been described in detail with reference to the drawings, but the specific structure is not limited to the above structure, and various design changes and the like can be made within the scope not departing from the gist of the present invention.

[0455] [Explanation of Reference Numerals]

[0456] 1, 1a, 1b, 1c, 1d...allosteric pathway prediction device, 3...allosteric pathway prediction system, 12...network graph generation unit, 13, 13b, 13c, 13d...path calculation unit.

Claims

1. An allosteric pathway prediction device, comprising: A network graph generating unit, based on the three-dimensional structural information of the protein, generates a network graph having vertices and edges to which weights are assigned, wherein the vertices correspond to amino acid residues constituting the protein and at least the amino acid residues in any binding substance bound to the protein, and the weights are based on interactions between the amino acid residues and at least the amino acid residues in the any binding substance; and The path calculation unit calculates a path connecting vertices on the network graph generated by the network graph generation unit, based on an evaluation function based on the weight.

2. The allosteric pathway prediction device according to claim 1, The starting point and the end point of the path are respectively specified vertices on the network graph.

3. The allosteric pathway prediction device according to claim 1 or 2, The path calculation unit calculates the plurality of paths.

4. The allosteric pathway prediction device according to claim 3, The plurality of paths are ranked based on evaluation values.

5. The allosteric pathway prediction device according to claim 3, The method further includes a first evaluation unit that evaluates the vertex based on the number of times the vertex is included in the plurality of paths calculated by the path calculation unit.

6. The allosteric pathway prediction device according to claim 3, There is also a second evaluation department. One of the starting point and the end point of the path is a predetermined vertex on the network graph, and the other is a vertex to be evaluated by the second evaluation unit, that is, an evaluation target vertex. The path calculation unit calculates the path connecting the predetermined vertex and the evaluation target vertex for each combination of the predetermined vertex and the evaluation target vertex. The second evaluation unit calculates an evaluation value for each of the plurality of evaluation target vertices based on the evaluation value of the path calculated by the path calculation unit.

7. The allosteric pathway prediction device according to claim 6, The method further includes a third evaluation unit that calculates an evaluation value for the local set of the plurality of evaluation target vertices based on the evaluation values ​​calculated for the evaluation target vertices by the second evaluation unit.

8. The allosteric pathway prediction device according to claim 1, The paths may include branches or junctions.

9. The allosteric pathway prediction device according to claim 1, There are multiple starting points or end points of the path.

10. The allosteric pathway prediction device according to claim 1, The evaluation function is a function that gives an evaluation value of a local graph.

11. The allosteric pathway prediction device according to claim 1, The evaluation function is a function that assigns an evaluation value to a path set.

12. The allosteric pathway prediction device according to claim 1, The evaluation function is a Hamiltonian.

13. The allosteric pathway prediction device according to claim 1, The evaluation function is the Ising model.

14. The allosteric pathway prediction device according to claim 1, The path calculation unit calculates the path by solving the evaluation function using an optimization algorithm based on Hamiltonian.

15. The allosteric pathway prediction device according to claim 1, The path calculation section calculates the path by solving the evaluation function using an Ising machine.

16. The allosteric pathway prediction device according to claim 1, The path calculation unit calculates the path by solving the evaluation function based on a simulated bifurcation algorithm.

17. The allosteric pathway prediction device according to claim 1, The system further includes a three-dimensional structure information acquisition unit that acquires the early three-dimensional structure information from a database that stores the early three-dimensional structure information in advance.

18. The allosteric pathway prediction device according to claim 1, The preliminary three-dimensional structure information is calculated based on the three-dimensional structure simulation.

19. The allosteric pathway prediction device according to claim 1, The predicted results of the paths are stored in a database.

20. The allosteric pathway prediction device according to claim 1, Output an effect / function candidate list and a group of one or more of the candidate compound list and the prediction results of the pathway, wherein the effect / function candidate list represents candidates for effects or functions predicted based on the prediction results of the pathway, and the candidate compound list represents candidates for compounds predicted based on the prediction results of the pathway when allosteric control is exerted.

21. The allosteric pathway prediction device according to claim 1, One or more of the evaluation function and the conditions for calculating the path can be changed based on external designation.

22. A method for obtaining a prediction result of a structural path, obtaining a calculation result of a path connecting vertices on a network graph, The network diagram is generated based on the three-dimensional structural information of the protein, and has vertices corresponding to the amino acid residues constituting the protein and at least the amino acid residues in any binding substance bound to the protein, and edges are assigned weights based on the interactions between the amino acid residues and at least the amino acid residues in the any binding substance. The path is a path connecting vertices on the network graph calculated according to an evaluation function based on the weight.

23. A method for predicting an allosteric pathway, comprising: A network graph generating step, based on the three-dimensional structural information of the protein, generating a network graph having vertices and edges to which weights are assigned, wherein the vertices correspond to amino acid residues constituting the protein and at least the amino acid residues in any binding substance binding to the protein, and the weights are based on the interactions between the amino acid residues and at least the amino acid residues in the any binding substance; and The path calculation step calculates a path connecting the vertices on the network graph generated by the network graph generation step, based on an evaluation function based on the weight.

24. A program for causing a computer to execute the following steps: A network graph generating step, based on the three-dimensional structural information of the protein, generating a network graph having vertices and edges to which weights are assigned, wherein the vertices correspond to amino acid residues constituting the protein and at least the amino acid residues in any binding substance binding to the protein, and the weights are based on the interactions between the amino acid residues and at least the amino acid residues in the any binding substance; and The path calculation step calculates a path connecting the vertices on the network graph generated by the network graph generation step, based on an evaluation function based on the weight.

Images