JP2003206246A - Program for searching steric structure of compound, steric structure searching apparatus and steric structure searching method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a program for searching the steric structure of a compound, an apparatus for searching the steric structure and a searching method therefor. <P>SOLUTION: The program contains a step to calculate the molecular potential energy of an assumed steric structure of a specified compound X, a step to memorize the calculation result in connection with the steric structure, a step to select two bonded adjacent segments A1 and A2 from plural segments A1-A4 of the specified compound, a step to divided the specified compound into fixed rigid units A2-4 and a mobile rigid unit A1 and parallelly translating the mobile rigid unit A1 at a randomly selected specific moving distance, a step to calculate the molecular potential energy of the deformed structure obtained by the translation, a step to compare the molecular potential energy of the deformed structure with the memorized energy and judge and calculation value to be adopted, and a step to memorize the adopted molecular potential energy in connection with the steric structure. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a compound three-dimensional structure search program, a three-dimensional structure search device, and a three-dimensional structure search method. According to the present invention, for example, the three-dimensional structure of a compound can be effectively predicted, structural simulation, or molecular design can be effectively performed. By using a molecular deformation method (parallel transfer method) different from the angle method, Monte Carlo simulation can be performed more efficiently.

[0002]

BACKGROUND ART Currently, the decoding of the genomes of various organisms including human beings is being vigorously pursued, and as a result, it has become possible to easily obtain the primary sequence information of a vast amount of proteins. . In such a situation, if we could clarify the principle of constructing a three-dimensional structure from the primary sequence of a protein and predict the tertiary structure from the primary structure, we would analyze genomic information as a life phenomenon that occurs in a three-dimensional space. It becomes possible to use it while associating it directly. For these reasons, elucidation of protein construction principles and establishment of structure prediction methods have become important themes in modern life sciences.

The structure prediction of a protein can be rephrased as finding the structure having the minimum free energy of a peptide chain having a designated sequence, focusing on its energy aspect. However, the number of structures that a peptide chain can take is enormous, and it is impossible with current computer capabilities to find the free energy for all structures and select the smallest one. Further, since the free energy curved surface has many local minimum values, the optimization cannot be performed smoothly. Due to these circumstances, solving this minimum value search problem in general is an unsolved and difficult problem. Various structural prediction methods have been attempted toward this solution. Broadly speaking, they are empirical methods that aim at highly accurate predictions using known structural information, and inexperienced methods that deductively reproduce the folding process based on physicochemical principles such as atomic interactions. Method.

Examples of empirical methods include homology modeling and 3D-1D method. These methods are
It is effective when a structure similar to the structure to be predicted has already been solved experimentally, but it cannot be used without such structural information, and it is not a direct aim to elucidate the principle of the folding mechanism. There is. As an ab initio method, a lattice model method [Hinds, DA and Lev
itt, M. A lattice model for protein structure predi
ction at low resolution. Proc. Natl. Acad. Sci. US
A 89, 2536-2540 (1992)] and the molecular dynamics method (MD method) originally used to investigate the dynamic properties of proteins [Momany, FA, McGuire, RF, Burges]
s, AW and Scheraga, HA Energy parameters in
VII.geometric parameters, partial c
harges, non-bonded interactions, hydrogen bond int
eractions and intrinsic torsional potentials for n
aturally ocurring amino acids. J. Phys. Chem. 79,
2361-2381 (1975); and McCammon, JA, Gelin, B.
R. and Karplus, M. Dynamics of folded proteins. Na
ture 267, 585-590 (1977)] and Monte Carlo method (MC method) for investigating statistical mechanical properties [Northrup, SH an
d McCammon, JA Simulation methods for protein s
tructure fluctuations. Biopolymers 19, 1001-1016
(1980); and Noguti, T. and Go, N. Efficient Monte C
arlo method for simulation of fluctuating conforma
tions of native proteins.Biopolymers 24, 527-546
(1985)] is applied to a folding simulation.

The lattice model method replaces the protein structure with a simpler lattice model, and various energy calculations and M
This is a method of performing a simulation in combination with the C method, and its theoretical treatment is easier than other methods. However, although it is considered that there are many factors involved in the structure formation such as tight packing of side chains in actual protein structure formation, there is a weakness that the lattice model cannot directly deal with them.

As a simulation method that uses a model that is not a simplified model and that is true to the actual molecular structure, the MD method and the MC method that use an all-atom model are used. The structure energy curved surface of an actual protein is considered to have a shape called a funnel model which is convenient for folding in which the energy of the natural structure is significantly lower than that of other structures. If such an energy phase can be reproduced on a computer, it is expected that the MD or MC method can be used to reach the natural structure of the protein without conducting a lice-destructive structure search. In the structure prediction by the MD method, a structure having a low energy that can be taken by a protein-containing system is obtained by a simulation based on Newtonian mechanics. This method well reflects the actual physicochemical properties, but on the other hand, it has a problem that it requires a long calculation time. Despite the fact that actual protein folding is a phenomenon of the order of milliseconds, seconds, and minutes, the MD simulation of sub-microseconds can finally be performed with modern computational capabilities.

Structure prediction by Monte Carlo method (MC method) [eg Hansmann, UHE and Okamoto, Y. New Mo
nte Carlo algorithms for protein folding. Current
Opinion in Structural Biology 9, 177-183 (1999);
And Shimada, J., Kussell, EL and Shakhnovich, E.
I. The folding thermodynamics and Kinetics of cr
ambin using an all-atom Monte Carlo simulation.J.
Mol.Biol. 308, 79-95 (2001)], the molecule is artificially deformed by calculation and the structure is changed according to the energy change. This molecular deformation is a drawback in the sense that it does not necessarily correspond to reality, but it is an advantage in the sense that this deformation makes it possible to quickly and extensively search for structures with low free energy. In that sense, the MC method is considered to be an excellent method capable of simulating the peptide chain folding process using a model corresponding to reality at the atomic level within a feasible calculation time.

[0008]

Generally, in the molecular deformation operation of the MC method, the dihedral angle method for randomly changing the dihedral angle of a protein is often adopted. This dihedral method
When the peptide chain is short, it is an effective method that is easy to handle and can efficiently search for a stable structure, but it is not always the best molecular modification method for reproducing the natural structure of a long peptide chain. The present inventor has conducted extensive studies to develop an effective method capable of efficiently searching for a stable structure when treating a peptide chain having a long peptide chain and a complicated three-dimensional structure. When searching for a stable structure of a peptide chain or a long peptide chain containing many amino acid residues, rather than using the conventional dihedral angle method,
It has been found that it is often better to use a new molecular transformation method, the "translation method". Therefore,
The present invention, in a particular aspect, is a Monte Carlo method (MC method).
The present invention provides a new translation method in place of the conventional dihedral angle method in the molecular deformation operation of proteins in. In addition, the present inventor has found that the “parallel transfer method”, which is a new molecular transformation method, is an effective means that can be used for searching for three-dimensional structures of not only proteins but also widely general compounds. Also found.

[0009]

Therefore, the present invention is a program for causing a computer to search for a three-dimensional structure of a compound. A step of calculating a molecular potential energy, (2) a step of storing the result of the calculation step (1) in a calculated value storage area in association with the virtual three-dimensional structure used at that time, (3) the specified compound in advance A process of selecting two adjacent segments that are connected to each other from a plurality of segments divided by a specified method,
(4) The specified compound includes a fixed rigid body unit that is a unit that includes one of the bonded adjacent segments selected in the selection step (3), and the other segment of the bonded adjacent segments. A step of dividing the moving rigid body unit into two units, and moving the moving rigid body unit in parallel with the fixed rigid body unit at a specific moving distance randomly selected within a range of a maximum moving distance designated in advance,
(5) calculating the molecular potential energy of the virtual three-dimensional structure deformed by the moving step (4),
(6) A step of comparing the molecular potential energy of the deformed three-dimensional structure calculated in the calculation step (5) with the molecular potential energy stored in the calculated value storage area to determine the calculated value to be adopted. (7) In the comparison and determination step (6), a step of newly storing one of the molecular potential energies determined to be adopted in the calculated value storage area in association with the virtual three-dimensional structure at that time is included. The present invention relates to a program for causing a computer to search for a three-dimensional structure of a compound.

The present invention is also an apparatus for searching a three-dimensional structure of a compound, comprising (1) an input section for inputting search conditions,
(2) A three-dimensional structure storage area for storing the virtual three-dimensional structure of the designated compound, (3) a calculator for calculating the molecular potential energy of the virtual three-dimensional structure of the designated compound, (4)
A calculation value storage area for storing the calculation result of the calculation unit in association with the virtual three-dimensional structure used in the calculation, (5) from among a plurality of segments obtained by dividing the specified compound in a predetermined method, A segment selection unit for selecting two adjacent segments that are connected to each other, (6) A movement distance selection unit for randomly selecting a specific movement distance within a range of a maximum movement distance designated in advance, (7) the specified compound A fixed rigid body unit that is a unit including one of the adjacent segments connected by the segment selection unit,
The moving rigid body unit is divided into two units, a moving rigid body unit which is a unit including the other segment of the joining adjacent segments, and the moving rigid body unit with respect to the fixed rigid body unit at a specific moving distance selected by the moving distance selecting unit. (8) The molecular potential energy of the three-dimensional structure after deformation by the parallel movement operating unit and the molecular potential energy already stored in the calculated value storage area are compared and adopted. A comparison determination unit for determining a calculated value, and (9) one of the molecular potential energies determined to be adopted in the comparison determination unit is newly associated with the calculated three-dimensional storage area in the calculated value storage area. The present invention relates to the compound three-dimensional structure searching apparatus, including a control unit for storing the compound.

Furthermore, the present invention is a method for searching a three-dimensional structure of a compound, which is (1) calculating a molecular potential energy of a certain virtual three-dimensional structure with respect to a predesignated compound, and (2) the calculation. Storing the result of step (1) in the calculated value storage area in association with the virtual three-dimensional structure used at that time, (3) from the plurality of segments obtained by dividing the specified compound in a prespecified manner, Selecting two adjacent segments coupled to each other, (4)
The specified compound is a fixed rigid body unit that is a unit that includes one of the bonded adjacent segments selected in the selection step (3), and a unit that includes the other of the bonded adjacent segments. Moving the rigid body unit in parallel with the fixed rigid body unit at a specific moving distance randomly selected within a range of a maximum moving distance designated in advance, (5) the moving Calculating the molecular potential energy of the virtual three-dimensional structure deformed by the step (4), (6) the molecular potential energy of the three-dimensional structure after the deformation calculated in the calculation step (5), and the calculated value storage area. The step (7) of comparing the stored molecular potential energy with the stored molecular potential energy and judging the calculated value to be adopted is adopted in the comparison and judgment step (6). One of the molecular potential energies determined to be related to the virtual three-dimensional structure at that time, and a step of newly storing in the calculated value storage area is also included in the above-mentioned compound three-dimensional structure search method. Concerned.

In a preferred program according to the present invention, the compound is a protein, and the bond between the fixed rigid body unit and the movable rigid body unit is represented by the formula: -C (= O) -N (H)-. The peptide bond is represented, and the parallel movement of the moving rigid body unit with respect to the fixed rigid body unit is represented by the general formula (1):

[Chemical 2] (In the formula, R is a side chain contained in an amino acid residue, and R_n
Is an α carbon atom C contained in the fixed rigid body unit^α _nSide chain of
And R_{n + 1}Is the α carbon source contained in the moving rigid body unit
Child C^α _{n + 1}Is the side chain of
Α carbon atom forming the peptide-bonded main chain of
C ^α _n, Carbon atom C, nitrogen atom N, and second α carbon source
Child C^α _{n + 1}And an oxygen atom O bonded to the carbon atom C
And a hydrogen atom H bonded to the nitrogen atom N.
And the nitrogen atom C on the plane of the peptide
Child N and the second α carbon atom C^α _{n + 1}Exercise and
(B) the carbon atom C is
First α carbon atom C^α _nWithout changing the interatomic distance between
First α carbon atom C^α _nIs the center of rotation and the bond angle is
By changing, the maximum transfer on the peptide plane
A moving distance that changes randomly within the range of the moving distance,
The circular motion of an arc is performed, and (c) the nitrogen atom N is
Without changing the interatomic distance from the carbon atom C, the carbon atom
By changing the bond angle with C as the center of rotation,
Setting on the peptide plane independently of the step (b)
Change randomly within the maximum movement distance
At the moving distance, independent of the circular motion of the carbon atom C,
At the same time as the circular motion of carbon atom C, circular arc motion is performed.
And (d) the second α carbon atom C^α _{n + 1}Is
Without changing the interatomic distance from the nitrogen atom N,
By changing the bond angle with the atom N as the center of rotation,
From the nitrogen atom N on the peptide plane.
Second α carbon atom C^α _{n + 1}Direction of direction vector toward
Is the above-mentioned first α carbon atom C^α _nFrom the above carbon atom C
Circle so that it always matches the direction of the direction vector to
It is carried out by performing a circular motion in an arc shape.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION In the present invention, when a three-dimensional structure of a compound is searched, the compound is divided into a plurality of segments and the relative position of each segment is maintained without changing the internal structure of each segment. The basic operation is to shift in parallel. After performing this basic operation once, the segment dividing position is changed and the same parallel movement operation is repeated.

For example, as shown schematically in FIG. 1, first, a designated compound X [FIG. 1 (a)] is divided into four segments A.
It is divided into 1 to A4 [Fig. 1 (b)], and the molecular potential energy of this virtual three-dimensional structure is calculated. Next, a combination A2 of two adjacent segments that are connected to each other
A2 is selected and the unit A1 containing the segment A1 and the unit A2-4 containing the other segment A2 (ie,
It is divided into two segments (A2, A3, A4).

These two units A1 and A2-4 are treated as a rigid body whose internal structure does not change, and are relatively parallel by changing the bond angle and / or the bond distance between the segment A1 and the segment A2. To move. For example, as shown in FIG. 1C, the moving rigid body unit A2-4 including the segment A2 is translated in parallel with the fixed rigid body unit A1 including the segment A1. Alternatively, the movable rigid body unit A1 is moved in parallel with respect to the fixed rigid body unit A2-4.

Here, the moving direction is not limited to one plane and can be moved in a three-dimensional space. Also,
Although the maximum moving distance is not particularly limited, it is preferable to predict the maximum moving distance with excellent calculation efficiency by a short-time pilot test, use the predicted maximum moving distance as a search condition, and move within the range. The molecular potential energy of the virtual three-dimensional structure deformed by this parallel movement is calculated, the molecular potential energies before and after the deformation are compared, and the calculated value and the three-dimensional structure to be adopted are stored.

Next, as shown in FIG. 1 (d), a combination A of two other segments that are mutually connected and are adjacent to each other.
2, A3 is selected, and a fixed rigid body unit A1-2 including the segment A2 (that is, the segments A1 and A2) and a moving rigid body unit A3-4 including the other segment A3 (that is, the segments A3 and A4) are selected. The movable rigid body unit A3-4 is moved in parallel with respect to the fixed rigid body unit A1-2. Also in this case, the moving rigid body unit A1-2 can be moved in parallel with the fixed rigid body unit A3-4. The molecular potential energy of the virtual three-dimensional structure deformed by this parallel movement is calculated, the stored molecular potential energy is compared with the molecular potential energy after the deformation, and the calculated value and the three-dimensional structure to be adopted are stored.

Subsequently, another two combinations A3 and A4 of two adjacent segments that are connected to each other are selected, and similarly to the above, the fixed rigid body unit (or the moving rigid body unit) including the segment A3 and the other one are selected. And a moving rigid body unit (including a fixed rigid body unit) including the segment A4, the moving rigid body unit is translated in relation to the fixed rigid body unit, and the molecular potential energy of the virtual three-dimensional structure deformed by this parallel movement is divided. Is calculated, the stored molecular potential energy is compared with the molecular potential energy after deformation, and the calculated value and the three-dimensional structure to be adopted are stored. Thus, as shown in FIG. 1E, the basic operation is an operation of shifting the relative positions of all the segments A1 to A4 included in the compound X in parallel. After carrying out this basic operation once or a plurality of times, for example, as shown schematically in FIG.
Divide into 4 and repeat the same parallel movement operation.

In the present invention, the compound for which the three-dimensional structure can be searched is not particularly limited, and includes both inorganic compounds and organic compounds. Particularly preferred compounds are those that are expected to have complex stereostructures,
For example, organic polymer compounds, biopolymer compounds, especially proteins. The bond between adjacent segments is not particularly limited, and includes, for example, covalent bond, ionic bond, hydrogen bond and the like. The covalent bond is not limited to a single bond, and includes double bond and triple bond. Further, a bond between the constituent atoms of the aromatic ring is also included.

The parallel transfer method according to the present invention will be described below in connection with the case where it is used in a Monte Carlo (MC) simulation for predicting the three-dimensional structure of a protein, but this does not limit the scope of the present invention. The molecular deformation operation in the protein parallel displacement method according to the present invention will be described below with reference to the accompanying drawings while comparing it with the conventional dihedral angle method. FIG. 3 is an explanatory view schematically showing the structural deformation operation by the parallel displacement method and the structural deformation operation by the dihedral method according to the present invention. 3 (a) is a structure before a structural deformation operation, FIG. 3 (b) is a structure after a structural deformation operation by the parallel displacement method according to the present invention, and FIG. 3 (c) is a dihedral angle method. It is the structure after the structural deformation operation in.

In the parallel displacement method according to the present invention, FIG.
And, as shown in FIG. 3 (b), the peptide chain is divided into two peptide moieties 11 and 12 before and after a suitable covalent bond (covalent bond between atom A and atom B), and one of the peptide moieties 12 is arbitrarily separated. The molecule is deformed by translating in the direction. The moving direction [arrows s1 and s2 in FIG. 3 (b)] is determined from all directions using random numbers without bias. The moving distance is also determined using random numbers. The maximum distance that can be moved per deforming operation is set for each simulation trial, and this is called the "maximum movement distance". The actual movement distance is randomly determined within a range greater than 0 and less than or equal to the maximum movement distance. It

On the other hand, in the conventional dihedral angle method, as shown in FIGS. 3 (a) and 3 (c), the dihedral angle is rotated about an appropriate covalent bond portion as a rotation axis [FIG. 3 (c)]. [Arrow r1 and arrow r2]], the deformation operation is performed. The maximum rotation angle that can be moved by one deformation operation is hereinafter referred to as “maximum rotation angle”, and the actual rotation angle is determined in the range of more than 0 and less than or equal to the maximum rotation angle.

As is apparent from FIG. 3, in the dihedral angle method, as the distance from the rotation axis d increases, the moving distance of the atoms due to the deformation operation increases, and the collision of atoms easily occurs. It is thought that atoms move at the same distance at each position regardless of whether the distance from is short or long and collision between atoms is unlikely to occur. In the dihedral angle method, only the dihedral angle among the structural parameters changes, but in the parallel displacement method, the bond angle also changes. Since there is potential energy due to this bond angle, the total energy when the parallel transfer method is adopted tends to be slightly higher than when the dihedral angle method is adopted.

A concrete transformation operation will be described with reference to FIG. That is, FIG. 4 is an explanatory diagram showing the transformation operation on the peptide plane P that is the target of the transformation operation. Figure 4 (a)
Is a structural deformation operation in the translation method according to the present invention,
FIG. 4B shows a structural deformation operation by the conventional dihedral angle method. As shown in FIGS. 4A and 4B, the peptide plane P
In the main chain portion of the, "CA1-C" bond in order from the N-terminal side,
There are three covalent bonds, a "CN" bond and an "N-CA2" bond. In the translation method according to the present invention, among these three covalent bonds, the “CA1-C” bond [FIG.
One bond shown in black in (a)] and the two bonds of "C-N" bond [bond shown in gray in FIG. 4 (a)] are independently and simultaneously deformed. Figure 4 (a)
Arrow x and arrow y]. Specifically, this deformation operation is performed by randomly changing the directions of the direction vector from the carbon atom CA1 toward the carbon atom C and the direction vector from the carbon atom C toward the nitrogen atom N within a predetermined range. The nitrogen atom N to the carbon atom CA2
Although the direction vector of the direction vector toward C. is also changed, the change is not performed independently, but is changed so as to always coincide with the direction vector from the carbon atom CA1 to the carbon atom C. Therefore, "CA1-C" bond and "N-CA"
The 2 "bond, i.e. the two bonds shown in black in Figure 4 (a), move in parallel.

Note that, due to the change in the direction vector of the direction from the carbon atom CA1 to the carbon atom C, the carbon atom C
Moves three-dimensionally with the carbon atom CA1 as the center of rotation. That is, the movement of the carbon atom C is caused by the peptide plane P.
Not limited to two-dimensional movement only in the
It also moves three-dimensionally up and down. At this time, without changing the interatomic distance between the carbon atom CA1 and the carbon atom C,
Only change the bond angle. Moreover, the range of the moving distance of the carbon atom C is randomly changed within the range of the maximum moving distance or less. Further, due to the change in the direction vector of the direction from the carbon atom C to the nitrogen atom N, the nitrogen atom N also three-dimensionally with the carbon atom C as the center of rotation, that is, not only in the peptide plane P but also in the peptide plane P. Move up and down the plane P. At this time, only the bond angle is changed without changing the interatomic distance between the carbon atom C and the nitrogen atom N. This movement of the nitrogen atom N is caused by the above-mentioned carbon atom C.
And the movement of the nitrogen atom N are randomly changed within the range of the maximum movement distance or less. Furthermore, due to the direction change of the direction vector from the nitrogen atom N to the carbon atom CA2, the carbon atom CA2 is three-dimensionally with the nitrogen atom N as the center of rotation, that is, not only in the peptide plane P but also in the peptide plane. Move up and down the plane P. At this time, only the bond angle is changed without changing the interatomic distance between the nitrogen atom N and the carbon atom CA2. This carbon atom CA
The movement of 2 moves from the nitrogen atom N to the carbon atom CA2.
The direction of the direction vector toward the direction is the carbon atom CA1 described above.
The motion is performed so as to always match the direction of the direction vector from the above to the carbon atom C.

As described above, in the method for translating a protein according to the present invention, the relative position between the carbon atom CA1 and the carbon atom CA2 is changed only by changing the direction of the "CA1-C" bond vector. be able to. That is, the direction of the vector of the “N-CA2” combination is “C
It is maintained in the same direction as the vector of the "A1-C" combination. To change the overall conformation of a protein,
The carbon atom C with respect to the peptide plane at various places
It is sufficient to change the relative position of A1 and the carbon atom CA2 in various ways. Therefore, in the present invention, the object can be sufficiently achieved even by the parallel movement method by changing the relative position of the carbon atom CA1 and the carbon atom CA2.

However, although it is possible to change the overall structure of the protein only by changing the relative position of the carbon atom CA1 and the carbon atom CA2, when the protein is partially observed, each peptide plane The inclination of can not be changed at all. Therefore, "C-
It is preferred to rotate the peptide plane P by moving the direction of the vector of the "N" bonds independently of the direction of the vector of the "CA1-C" bonds. That is, how each peptide plane P of the protein is inclined is a structurally important feature, and in addition to changing the relative position of the carbon atom CA1 and the carbon atom CA2,
By changing the vector of the "C-N" bond, it is possible to carry out a simulation in line with the actual molecular structure.

On the other hand, as shown in FIG. 4 (b), in the conventional dihedral angle method, a "CA1-C" bond [bonding shown in black in FIG. 4 (b)] is used as a rotation axis, and " The rotation operation with the "N-CA2" bond [bond shown in gray in FIG. 4 (b)] as the rotation axis is simultaneously executed. The rotation angles of the two rotation operations are independently determined by random numbers. That is, "CA1-C"
Deformation angle ψ of bond and “N-CA2” deformation angle φ of bond
Are randomly determined within the range of the maximum rotation angle or less.

The outline of the execution procedure of the MC simulation using the above-mentioned parallel movement method found by the present inventor will be described below. For example, when the method is used for the structural deformation method of the main chain portion, the following procedure 1 to procedure 3 is repeatedly executed. (1) Procedure 1: Deformation step in arbitrary peptide bond of peptide molecule Since there are "planar number of amino acid residues-1" by the peptide bond on the main chain of each peptide chain, one plane is selected from them. Is selected to perform the deformation operation by the parallel movement method. (2) Procedure 2: Calculation step of structural energy difference before and after deformation operation The potential energy of each of the structures of the two peptide molecules before and after the deformation operation by the parallel transfer method is calculated and obtained. As a method of calculating the potential energy, a normal calculation method similar to the calculation method of the conventional dihedral angle method can be used. (3) Step 3: Step of deciding adoption / non-adoption of deformation based on energy difference Based on the difference in potential energy obtained in step 2, the structure after the deformation operation is adopted or rejected and rejected. Determine whether to adopt. For this criterion, an ordinary judgment method similar to the conventional dihedral angle method can be used. For example, like the conventional dihedral method, the Metropolis method [Metropolis, N., Rosenbluth, AW, Ros
enbluth, MN, Teller, AH and Teller, E. Equat
ion of state calculations by fast computer machine
s. J. Chem. Phys. 21, 1087-1092 (1953)].

In the metropolis method, when the energy value becomes small due to deformation, the structure after deformation is unconditionally adopted. On the other hand, when the energy value becomes high due to the deformation, basically the structure after the deformation is mostly discarded, but the structure after the deformation may be adopted a little with a probability according to the energy value. It corresponds to the change in the potential energy that rarely occurs even if the energy change of each molecule is noted even in nature. That is, since molecules capable of changing a certain shape have various structures even in actual nature, and there is not necessarily only a structure having the minimum potential energy, the existence of such an actual molecular structure The purpose is to realize the situation on a computer and aim for a realistic simulation.

A round of work from "Procedure 1" to "Procedure 3" is considered as one unit of the simulation progress process, and will be referred to as "1 step" hereinafter. In the simulation based on the present invention, the selection of the peptide plane portion on the main chain in “Procedure 1” is sequentially selected from the N-terminal as the steps proceed, and when the C-terminal is reached, the selection is returned to the N-terminal and repeated. Can be done in that order. Alternatively, it is possible to sequentially select from the C-terminal and return to the C-terminal and repeat when the N-terminal is reached, or it is possible to start from an intermediate portion between the N-terminal and the C-terminal. . Here, while the selection and modification operations of all the peptide planes included in the target peptide chain make one round (for example, one round from the N terminus to the C terminus)
The series of operations performed in step 1 will be collectively referred to as one sweep. For example, in a simulation of a 28-mer peptide chain, one sweep includes 27 steps of operation.

The reason why a suitable number of steps are collectively counted as one sweep is to make it easy to grasp the calculation process by combining a plurality of step operations. For example, rather than describing the calculation progress of a certain peptide chain by the number of steps, one step is defined as a step operation with the same number of times as the number of variable parts in the peptide chain,
If the degree of progress is represented by the number of sweeps, it can be seen that each variable portion has undergone an average of m deformation operations per location at the time of m sweeps. This utility is, for example,
Perform MC calculation of molecules with different sizes,
When comparing the calculation progresses of both molecules, it is immediately apparent that if the number of sweeps of calculation of both molecules is the same, the same number of deformation operations are performed per variable portion.

The above-mentioned parallel displacement method found by the present inventor can be used not only for the structural modification of the main chain part of the peptide chain but also for the structural modification of the side chain part. For example, with respect to the main chain part of the target peptide chain, the steps by the above-mentioned steps 1 to 3 are modified in two consecutive sweeps, and then the side chain part is modified from N-terminal to C-terminal one residue at a time. The simulation can be performed by the combination of doing. The number of sweeps for the main chain portion and the number of sweeps for the side chain portion can be arbitrarily selected. These sweeps are performed a predetermined number of times (for example, 100,
000 times or 200,000 times) is repeated to make one trial of Monte Carlo simulation. Each time one such trial is completed, the setting of the maximum moving distance can be changed and another trial can be performed. The above-mentioned one trial can be carried out only by the structural modification of the main chain portion, or can be carried out by combining the structural modification of the side chain portion.

When a certain residue is modified in the modification operation of the side chain, the whole residue is set as a moving rigid body unit of 1 unit, and the other parts are set as a fixed rigid body unit of 1 unit, and the parallel moving method is used. (For example, the maximum moving distance = 0.5 Å), it is possible to randomly deform, and it is possible to determine whether or not to adopt the deformation by the method of Metropolis. However, for parts that can be considered to actually cause rotational motion, such as the methyl groups and aromatic rings present in the side chains, the dihedral angle method (for example, maximum rotational angle = 5 °) is used instead of the parallel displacement method. It is preferable to perform a deforming operation. As described above, the above-mentioned parallel displacement method found by the present inventor can be used not only for the structural modification of the main chain part of the peptide chain but also for the structural modification of the side chain part. Further, the above-mentioned parallel displacement method found by the present inventor can be used in combination with the dihedral angle method for structural deformation of the main chain portion of the peptide chain.

Next, the three-dimensional structure searching apparatus for compounds according to the present invention will be described with reference to FIG. FIG. 5 is a configuration diagram of the three-dimensional structure searching apparatus 1 according to the present invention. The three-dimensional structure searching apparatus 1 includes an input unit 2, a central processing unit 3, and an output unit 4, and the central processing unit 3 includes a control unit 5, a storage unit 6, a processing unit 7, and a counting unit 8.

From the input section 2, the information of the specified compound to be searched for the three-dimensional structure, the method of dividing the specified compound into segments, the virtual initial three-dimensional structure, the maximum moving distance, the number of steps, the number of sweeps, etc. are inputted. Can be specified. These search conditions are stored in the storage unit 6 via the input processing unit 2a in the central processing unit 3 and the control unit 5. For example, information on a specified compound (eg, virtual initial three-dimensional structure)
Is stored in the designated compound storage area 61, and similarly, the method of dividing the designated compound into segments and each segment divided by the method is stored in the segment storage area 6
2. The maximum moving distance is stored in the maximum moving distance storage area 65, and the step number and the sweep number are stored in the step number / sweep number storage area 66, respectively.

The processing unit 7 in the central processing unit 3 includes a segment division processing unit 71, a segment selection unit 72, and a movement distance selection unit 73. The segment division processing unit 7
In 1, the specified compound is divided according to the method specified in advance,
The divided segment can be stored in the segment storage area 62. The segment selection unit 72 selects two adjacent segments that are connected to each other from the plurality of segments. Further, in the movement distance selection unit 73,
Within the range of the maximum moving distance stored in the maximum moving distance storage area 65, the moving distance is randomly selected by a random number.

The processing unit 7 in the central processing unit 3 further includes a parallel movement operating unit 74 and a molecular potential energy calculating unit 7.
5 and a comparison and determination section 76. Translation operation unit 74
Performs an operation of translating the moving rigid body unit with respect to the fixed rigid body unit at a specific moving distance randomly selected by the moving distance selecting unit 73. The molecular potential energy of the three-dimensional structure or the molecular potential energy of the virtual three-dimensional structure deformed by the parallel movement operation is calculated. The molecular potential energy of the virtual initial three-dimensional structure of the designated compound is stored in the molecular potential energy storage area 63 of the storage unit 6 as it is. Further, the comparison / determination unit 76 compares the molecular potential energy of the deformed three-dimensional structure with the molecular potential energy already stored in the molecular potential energy storage area 63 of the storage unit 6, and determines the molecular potential energy to be adopted. The calculated calculated molecular potential energy to be adopted is newly stored in the molecular potential energy storage area 63. The virtual three-dimensional structure having the calculated calculated molecular potential energy is stored in the three-dimensional structure storage area 64 in association with the calculated calculated molecular potential energy.

The processing unit 7 in the central processing unit 3 further includes a first
A branch determination unit 77 and a second branch determination unit 78 may be included. The first branch determination unit 77 compares the number of steps counted by the counting unit 8 described later with the designated number of steps stored in the number-of-steps / sweep-number storage area 66 and repeats the step processing, or It is determined and controlled whether to end the step process and proceed to the next process. In addition, the second branch determination unit 78 is configured to store the number of sweeps counted by the counting unit 8, which will be described later, and the number of steps / sweep number storage area 6
The number of designated sweeps stored in 6 is compared, and it is determined whether the sweep processing is repeated or whether the sweep processing is terminated and the next step is performed.

The central processing unit 3 may include the counting unit 8 as described above, and the counting unit 8 may include the step number counting unit 81 and the sweep number counting unit 82.
The step number counting unit 81 is
(B) Translation process of moving rigid unit, (c) Molecular potential energy calculation process of deformed virtual three-dimensional structure, (d)
A step of comparing the molecular potential energy of the three-dimensional structure after deformation with the molecular potential energy stored in the molecular potential energy storage area 63, and (e)
The molecular potential energy determined to be adopted is
In association with the virtual three-dimensional structure at that time, the steps (a) to (e) including the step of newly storing in the molecular potential energy storage area 63 are defined as one step, and the number of steps is counted. This count value is the first value of the processing unit 3.
In the branch determination unit 77, the number of steps in the storage unit 6
It is determined whether or not it is an integral multiple of the number of designated steps stored in the sweep number storage area 66. If it is not an integral multiple, the process returns to the segment selection step (a). Proceed to the process.

The designated step number stored in the step number / sweep number storage area 66 is, for example, the number obtained by subtracting 1 from the total number of segments included in the designated compound.
For example, when the protein is the specified compound and the segment is divided by all peptide bonds contained in the peptide chain, the specified number of steps is the total number of amino acid residues minus one. That is, in the simulation of a 28-mer peptide chain, the specified number of steps is 2
7

The sweep number counting unit 82 counts the number of times the first branch determination unit 77 has advanced to the next step as the number of sweeps. In the second branch determination unit 78 of the processing unit 3, it is determined whether or not this count value is the designated sweep number stored in the step number / sweep number storage area 66 of the storage unit 6, and the designation is performed. If the number of sweeps is less than the specified number of sweeps, the process returns to the selection step (a). If the number of sweeps is greater than or equal to the specified number of sweeps, the next step, that is, the output step is performed. The number of designated sweeps stored in the step number / sweep number storage area 66 is not particularly limited, but is, for example, 10,000 to 1,000.
It is 10,000 times.

When the process proceeds to the output step, the output means 4 outputs the data to the three-dimensional structure storage area 64 in association with the calculated value stored in the molecular potential energy calculation unit 75 after the final step by the output processing unit 4a. The stored virtual three-dimensional structure is output. The output unit 4 can output the entire storage stored in the storage unit 6. For example, the designated compound storage area 61, the segment storage area 62, the molecular potential energy storage area 6 of the storage unit 6
3, the three-dimensional structure storage area 64, and the maximum moving distance storage area 65, all the storages that have been stored in the storage area 67 are accumulated and stored, and the content is output from the output unit 4. can do.

Next, using the translation method according to the present invention,
The operation of predicting the three-dimensional structure of a protein by Monte Carlo simulation will be described according to a flowchart. FIG. 6 is a flowchart showing the operation of one trial. First, the simulation condition is designated from the input means (step S1). The simulation conditions include the primary structure of a protein whose structure is to be predicted, an imaginary initial three-dimensional structure (for example, a linear shape), the range of the maximum migration distance, the number of main chain sweep processes, and the side chain sweep process. Or the number of all sweep processes. These are stored in, for example, the storage unit 6 of the three-dimensional structure searching apparatus shown in FIG.

Next, for the previously designated protein, the molecular potential energy of the previously designated virtual initial three-dimensional structure is calculated (step S2). As the virtual initial three-dimensional structure, any structure can be virtualized, and examples thereof include a linear structure, an α helix structure, and a β sheet structure. The above calculation result is
It is stored in the calculated value storage area (for example, the molecular potential energy storage area 63) in association with the virtual three-dimensional structure used at that time (that is, the virtual initial three-dimensional structure) (step S3). The virtual initial three-dimensional structure is stored in the three-dimensional structure storage area 64, for example. The calculation result and the virtual initial three-dimensional structure can be stored in the molecular potential energy storage area 63 and the three-dimensional structure storage area 64, and can also be stored in the separately provided cumulative storage area 67.

Subsequently, a main chain sweep process is carried out (step S4). FIG. 7 shows the specific content of the main chain sweep processing. In the main chain sweep process, first, peptide bonds are selected (step S41). The peptide bond is
An arbitrary peptide bond is sequentially selected from all peptide bonds contained in the designated protein. This selection splits the designated protein in two. That is, it is divided into a fixed rigid body unit on the side containing one α carbon atom and a moving rigid body unit on the side containing the other α carbon atom.

Then, using a random number, a specific movement distance is randomly selected within the designated maximum movement distance (step S).
42). This specific moving distance can also be stored in the cumulative storage area 67. Next, the moving rigid body unit is translated in parallel with the fixed rigid body unit existing on both sides of the peptide bond selected in the step S41 by the specific moving distance selected in the step S42 to deform the three-dimensional structure (step. S43). In this modification step S43, it is sufficient to change the relative position of one α carbon atom existing on both sides of the selected peptide bond to the other α carbon atom. As I explained along
"CA1-C" bond, "C-N" bond, and "N-CA"
It is preferred to move each of the three covalent bonds of the 2 "bond. Specifically, the "CA1-C" bond randomly changes the direction vector from the carbon atom CA1 to the carbon atom C within a specified range, and the "CN" bond also changes from the nitrogen atom N to the carbon atom CA2. Randomly change the direction vector toward the specified range. Changes in these direction vectors are made independently. Further, the "N-CA2" bond changes the direction vector from the nitrogen atom N toward the carbon atom CA2 so as to always match the direction vector from the carbon atom CA1 toward the carbon atom C.

Next, the molecular potential energy of the virtual three-dimensional structure of the protein transformed by the transformation step S43 is calculated (step S44). The calculated value obtained is
The cumulative storage area 6 is associated with the virtual three-dimensional structure at that time.
It can be stored in 7. Subsequently, the molecular potential energy of the three-dimensional structure of the deformed protein calculated in the calculation step S44 and the virtual initial three-dimensional structure stored in the calculated value storage area (for example, the molecular potential energy storage area 63) in the storage step S3. The molecular potential energy is compared (step S45). Based on this difference in molecular potential energy, the molecular potential energy to be adopted and the virtual three-dimensional structure at that time are determined according to, for example, the metropolis method.

The molecular potential energy adopted in the determination in step S45 is newly stored in the calculated value storage area (for example, the molecular potential energy storage area 63), and the virtual three-dimensional structure at that time is also stored. It is stored in the three-dimensional structure storage area 64 in association with the molecular potential energy value. These data can also be stored in the cumulative storage area 67. Next, the number of times of storage in the calculated value storage area (for example, the molecular potential energy storage area 63) is counted as the number of steps and integrated, and it is determined whether the integrated value is an integral multiple of the specified number of steps. (Step S47). The designated step number is the total number of peptide bonds and is stored in the step number / sweep number storage area 66. If the integrated value is not an integral multiple of the specified number of steps, the peptide bond selection step S4
When the value is returned to 1 and is an integral multiple, the main chain sweep processing is ended, and the integrated value is erased.

Subsequently, returning to FIG. 6, the cumulative number of main chain sweep processes is compared with the designated main chain sweep process frequency (m) (step S5). The total number of main chain sweep processing is
It is the number of times the integrated value is erased, assuming that the integrated value is an integral multiple of the specified number of steps in the determination step S47. On the other hand, as the designated main chain sweep processing number (m), for example, a number of 1 or more (for example, 2 to 4) is stored in advance in the step number / sweep number storage area 66. When the integrated value is less than the designated main chain sweep processing number (m), the main chain sweep processing S4 is repeated. When the integrated value is equal to or greater than the designated main chain sweep processing number (m), the integrated value of the main chain sweep processing is deleted, and the process proceeds to the next side chain sweep processing step S6.

The side chain sweep processing step S6 comprises a side chain step processing step S61 and a processing number determination step S62, as shown in FIG. The side chain step processing step S61 includes
For each side chain, the designated protein is divided into two parts, each side chain part (moving rigid body unit) and the other part (fixed rigid body unit), and randomly selected by random numbers within the specified maximum moving distance. Each side chain portion (moving rigid body unit) is translated in parallel with the other portion (fixed rigid body unit) at a specific moving distance. The designated maximum movement distance can be stored in the maximum movement distance storage area 65, for example.

After performing the side chain step processing step S61 for one side chain portion, the side chain step processing step S6 is performed.
The number of times of 1 is integrated, and the integrated number is compared with the designated step number (b). As the specified number of steps (b),
For example, the total number of side chains in the designated protein (corresponding to the number of amino acids that make up the designated protein)
It can be stored in the sweep number storage area 66.
When the integrated number of the side chain step processing step S61 is less than the designated step number (b), the side chain step processing step S6
Return to step 1, select another side chain, and select the side chain step processing step S
Carry out 61. When the integrated number of the side chain step processing step S61 is equal to or larger than the designated step number (b), the integrated value of the side chain step processing is deleted and the process proceeds to the next step S7.

In the next step S7, the cumulative number of side chain sweep processes is compared with the designated number of side chain sweep processes (n). The integration count of the side chain sweep process is the number of times the integration value is erased in the determination step, assuming that the integration value is the designated step number (b) or more. On the other hand, the number of designated side chain sweep processes (n) is, for example, 1 or more (for example, 2 to
4) is the number of steps / sweep number storage area 66 in advance.
Remembered in. When the integrated value is less than the designated side chain sweep processing number (n), the side chain sweep processing S6 is repeated. When the integrated value is equal to or more than the designated side chain sweep processing number (n), the integrated value of the side chain sweep processing is erased and the process proceeds to the next sweep processing number determination step S9.

In the next sweep processing number determination step S9,
When the number of main chain sweep processes, the number of side chain sweep processes, or the total number of these processes is compared with the specified number of sweep processes (p), and the integrated number is less than the specified number of sweep processes (p), Then, returning to the main chain sweep processing step S4, another peptide is selected to carry out the main chain sweep processing step. When the integrated number is equal to or more than the designated sweep processing number (p), the main chain sweep processing number, the side chain sweep processing number, or the integrated number of these total numbers is deleted, and the process proceeds to the next output step S10. proceed.

In the output step S10, the calculated value storage area (for example, the molecular potential energy storage area 6) is finally obtained.
The virtual three-dimensional structure having the molecular potential energy stored in 3) is output. At that time, the above-mentioned molecular potential energy can also be output together. Alternatively, all or part of the data stored in the cumulative storage area 67 can be selected and output.

The present invention can be effectively used for, for example, predicting the three-dimensional structure of a compound, structural simulation, or molecular design by conducting a three-dimensional structure search of the compound. For example, as described above, in particular, it can be used in the Monte Carlo method for simulating the protein folding mechanism.

[0057]

The present invention will be described in detail below with reference to examples, but these do not limit the scope of the present invention. (A) Outline of execution procedure of MC simulation (1) Origin of polypeptide chain used for MC simulation As a target of MC simulation in the parallel displacement method and the dihedral method such as Reference Example 1 and Example 1 below, respectively. Two types of polypeptides were used. One is a C peptide corresponding to 13 amino acid residues from the N-terminal of ribonuclease A, and it has been experimentally reported to have the ability to form an α-helix [Brown, JE and Klee, WA Helix.
-coil transition of the isolated amino terminals o
f ribonuclease. Biochemistry 10,470-476 (1971)].
The amino acid sequence of C peptide is as follows. Lys Glu Thr Ala Ala Ala L
ys Phe GluArg Gln His Met
(SEQ ID NO: 1) Another is a polypeptide of 28 amino acid residues (hereinafter 28
It has been reported by NMR structural analysis that it has a ββα structure with two strands and one helix [Dahiyat, BI, Sarisky, C.
A. and Mayo, SL De Novo protein design: Towards
fully automated sequence selection.JM Biol. 27
3, 789-796 (1997)]. The amino acid sequence of this 28mer is as follows. Lys Pro Tyr Thr Ala Arg I
le Lys GlyArg Thr Phe Ser
Asn Glu Lys Glu LeuArg A
sp Phe Leu Glu Thr Phe Th
r GlyArg (SEQ ID NO: 2)

(2) Software and hardware used for MC simulation The MC method program ORT used for calculation in the parallel movement method and the dihedral angle method of the following Reference Example 1 and Example 1 is
It was developed by the inventor using C language. This MC method program ORT can deform molecules by freely combining both the dihedral angle method and the parallel displacement method. The calculation is L
D with Linux (RedHat 6.1) installed
OS / V PC (cpu is PentiumIII800
MHz equivalent).

(3) Molecular model used for MC simulation In the parallel displacement method and dihedral angle method of Reference Example 1 and Example 1 below, the molecular model used for calculation is the all-atom model, and the N-terminal is NH ₃ ⁺ and C-terminal were COO ⁻ . When the molecular model was deformed, the length of the covalent bond was fixed and the bond angle and dihedral angle were changed. All simulations started with a linear initial structure.

(4) Calculation method of potential energy In the parallel transfer method and dihedral angle method of Reference Example 1 and Example 1 below, the force field parameter for calculating the potential energy of the molecule is AMBER parm96. A dat file was used [Kollman, P., Dixon, R., Cornel
l, W., Fox, T., Chipot, C. and Pohorille, A. The d
evelopment / application of a "minimalist" organic / b
iochemical molecular mechanic force field using a
combination of ab initio calculations and experime
ntal data. In: "Computer simulation of biomolecula
r systems: Theoretical and experimental applicatio
ns "(Van Gunsteren, WF and Weiner, PK Eds.)
p83-p96 (1997) Escom, The Netherlands]. However,
In order to avoid the bias toward α-helix structure and β-strand structure, the energy terms related to the dihedral angle of the main chain were excluded from the calculation. The cut-off radius when calculating each energy item was 8.0Å. Regarding hydrogen bond, NH
… At an angle of O of 90 ° or less, the attractive force becomes weak (0.
(0 times), the attraction becomes stronger near 180 ° (2.5
Doubled). Although the water molecule is not explicitly treated in the calculation, the dielectric constant is set to the square of the distance in order to take the effect of water into consideration. In order to incorporate the hydrophobic interaction into the energy calculation formula, an attractive term corresponding to the hydrophobic interaction was added to the energy calculation formula between the hydrophobic side chains. In many cases, the parameter T corresponding to the temperature was 2000K and the temperature was constant. This value is larger than that at room temperature, which is 300 K. However, this simulation uses a system that does not explicitly include water (a system that does not explicitly include water). This is the result of balancing against the fact that the energy gain due to the formation of the action is largely estimated. The system that does not explicitly contain water is a term for a system that explicitly contains water (a system in which water molecules around protein molecules are actually calculated as individual molecules).

(5) Items to be set throughout the whole simulation process In the parallel movement method and the dihedral angle method of the following Reference Example 1 and Example 1, 1000 per trial simulation
A 00 sweep (when only the deformation operation of the main chain portion is counted and the deformation operation of the side chain portion is not counted) was calculated. Every time the main chain portion was calculated by two sweeps, the side chain portion was deformed by one round from the N terminal to the C terminal, and the pair list was updated. In the parallel movement method, the movement distance and direction are randomly updated by random numbers, and in the dihedral angle method, φ value and ψ
The values and were randomly updated with random numbers. In the parallel displacement method and the dihedral angle method such as Reference Example 1 and Example 1 below, simulations were performed under various conditions. To avoid this, simulations in which only the random number sequence is different under each condition were repeated 10 trials, and the results were comprehensively interpreted.

[0062]

Reference Example 1 (1) MC Simulation of C Peptide MC simulation of C peptide was performed using the dihedral angle method and the parallel displacement method, and the results were compared and examined. The number of residues forming a helix structure was used as an index for judging the presence or absence of structure formation. The DSSP method [Kabsch, W. and San] is used to determine whether each residue has a helix structure.
der, C. Dictionary of protein secondary structure:
Pattern recognition of hydrogen-bonded and geomet
rical features. Biopolymers 22, 2577-2637 (1983)]
I went there. In the MC method, the size of the molecular deformation has a great influence on the efficiency of the structure search. Therefore, the simulation was performed by changing the degree of the deformation regardless of which molecular deformation method is used.

The results are shown in FIG. FIG. 9A shows the results when the parallel displacement method is used, and FIG. 9B shows the results when the dihedral angle method is used.
In FIGS. 9 (a) and 9 (b), the horizontal axis represents the magnitude of deformation, and the vertical axis represents the number of helix-forming residues at the time when the MC simulation was performed 100,000 sweeps. Temperature parameter (1000K,
The results of the simulation under 2000K, 3000K, and 4000K) are overwritten. Appropriate conditions in which the helix is considered to be formed efficiently when the parallel displacement method is used are that the temperature parameter is about 1000 K to 2000 K, and the maximum movement distance that represents the magnitude of deformation is
0.3 Å to 1.0 when the temperature parameter is 1000K
At Å and 2000K, it was about 0.3Å to 0.7Å. Almost no helix was formed when the temperature parameter was 3000 K or higher. On the other hand, when the dihedral method is used, suitable conditions for forming a helix are that the temperature parameter is 2000K and the maximum rotation angle is 20 ° to 120 °.
It was °. Although helices were formed even at 1000K and 3000K, the number of helix-forming residues was obviously smaller than that at 2000K. Almost no helix was formed at 4000K.

A comparison of the conditions which seem to have the highest helix-forming efficiency in both methods revealed that the parallel transfer method (temperature 2
At 000K-maximum migration distance of 0.25Å), only about 6 residues form a helix, whereas the dihedral method (temperature parameter 2000K-maximum rotation angle of 35 °).
Then, 10 or more residues among all 13 residues formed a helix. Also, the number of helix-forming residues in the dihedral angle method (temperature parameter 2000K) is such that the maximum rotation angle is 15 ° to 120 °.
The number of residues was 8 or more in most of the ranges up to °, which was higher than the value under the optimum conditions of the parallel displacement method. From this, it is understood that the dihedral angle method is more advantageous than the parallel displacement method in forming the α helix. This is because the collision due to atom movement does not pose a problem when targeting a simple structure such as a single-stranded helix, so the advantage that the atom movement by one deformation operation of the dihedral angle method is large remains the same. This is considered to be reflected in the good search efficiency of the entire simulation.

In the parallel transfer method, the number of helix-forming residues at the 100,000th sweep is small, but this does not mean that the helix structure does not develop when the parallel transfer method is used. In this reference example 1, 1
We performed MC simulations for 10 trials by changing only the random number sequence per condition, and in each trial we investigated how much the maximum number of helix-forming residues increased during the entire process from the beginning to the 100,000 sweep point. It can be seen that even when the parallel transfer method is used, there is a period in which the number of helix-forming residues is 8 residues or more in all the trials out of the 10 trials. In some trials, there were times when the number of helix-forming residues was 10 or more. That is, even with the parallel displacement method, a long helix is momentarily formed, but the helix is immediately destroyed, so when taking an average of 10 trials at an appropriate sweep time, a low helix forming residue is obtained. You can only get the number. This problem that the helix easily changes to another shape is closely related to the force field function used for energy calculation, and can be improved by using more correct force field parameters.

(2) Helix formation rate of C-peptide when each modification method is used. Next, in order to know the helix formation rate when each modification method is used, (a) the parallel displacement method, the maximum at 2000K With the condition of moving distance 0.3 Å and (b) dihedral angle method,
Regarding the MC simulation under the condition that the maximum rotation angle was 30 ° at 2000K, the change with time in the number of helix-forming residues from the start to the end was graphed (Fig. 1
0). In FIG. 10, ● indicates the result of the dihedral angle method, and ■
Is the result of the translation method. When an MC simulation of 500 sweeps by the dihedral angle method and about 5000 sweeps by the parallel displacement method was performed, the graph had a substantially constant value, and no significant increase or decrease was observed thereafter. From the above,
Regarding the helix formation rate, it was found that the dihedral angle method is more efficient than the parallel displacement method and can form a helix structure 10 times or more faster. It is estimated that the actual calculation time to reach the equilibrium state is about 12 seconds per 200 sweeps in both methods, and it is estimated that the dihedral angle method is 30 seconds, and the parallel movement method is about 300 seconds.
It can be said that the simulation reaches equilibrium in an extremely short calculation time.

[0067]

Example 1 and Comparative Example 1 (1) MC Simulation of 28mer Targeting a 28mer polypeptide having a natural structure more complicated than the C peptide used in Reference Example 1, the parallel transfer method (Example 1) and MC using the dihedral angle method (Comparative Example 1)
Simulations were performed and the results were compared and examined. It is known from NMR experiments that this polypeptide has a ββα structure as described above. Generally lower RMSD (root mean s) relative to natural structure
Since the molecular deformation method capable of frequently generating a structure having a square displacement (root mean square displacement) value is considered to be efficient in aiming the structure prediction, the present inventor found that the RM of the structure obtained by the simulation.
The SD value was obtained and used as an index showing the effectiveness of the simulation.

First, it was examined to what extent the RMSD was lowered at the time of 100,000 sweeps when the parallel displacement method and the dihedral angle method were used. The results are shown in Fig. 11. FIG. 11A shows the result of performing the MC simulation by the parallel movement method under various conditions of the constant temperature of 2000 K and variously changing the maximum movement distance. Similarly, FIG. 11B shows the result of performing MC simulation by variously changing the maximum rotation angle under the constant temperature condition of 2000 K by the dihedral angle method.

When the parallel displacement method was used, the RMSD value was 5.6Å and the minimum value when the maximum displacement was 0.4Å. Also, the maximum movement distance is 0.2Å to 0.9Å
The RMSD value in the range was around 6.5Å.
On the other hand, when the dihedral angle method was used, the RMSD value was 6.7Å and the minimum value when the maximum rotation angle was 40 °.
Also, the maximum rotation angle is RM within the range of 10 ° to 110 °.
The SD value was around 7Å. From this, 2
In the 8-mer MC simulation, it is considered that the parallel displacement method can efficiently generate a structure having a lower RMSD value than the dihedral angle method.

The RMSD value here is a value obtained by averaging 10 trials for each condition, and the accuracy of the force field function used in the Monte Carlo method in Example 1 and Comparative Example 1 is 5 is not always enough
Although it is a large value of Å or more, RMSD values of 3 Å units also appear in individual simulations. Table 1 shows the RMSD values for 10 trials at the time of 100,000 sweeps, arranged in ascending order without averaging. The upper row shows the maximum movement distance of 0.4 Å by the parallel displacement method, and the lower row shows the maximum rotation angle of 40 ° by the dihedral angle method. The result is. In Table 1, the lowest RMSD value was 3.4 Å in the parallel displacement method and 4.6 Å in the dihedral angle method.

Table 1 Calculated structure and NMR of 28mer polypeptide at 100,000 sweeps
Distributed Minimum Maximum Mean value translation method of RMSD values between the structure by (Å): 3.4 4.3 4.3 4.5 5.0 6.5 6.8 7.0 7.0 7.5 5.6 dihedral angle method: 4.6 4.8 5.9 6.4 6.8 7.1 7.5 7.8 7.9 8.4 6.7

(2) 28 me when each modification method is used
Helix formation rate of r Next, in order to investigate the rate at which the structure is formed, (a) by the parallel movement method, the condition of the maximum movement distance 0.4 Å,
(B) The change with time of the average value and the minimum value of RMSD under the condition of the maximum rotation angle of 40 ° was examined by the dihedral angle method. 12
Is the change with time of the average value of RMSD. From this graph, it can be seen that the parallel movement method (black line a) initially has a higher RMSD value than the dihedral angle method (grey line b), but finally has a lower RMSD value than the dihedral angle method. . In the parallel displacement method, all the atoms that move by the deformation operation move by the same distance little by little, so when the simulation progresses and the structure becomes more complicated, the number of atom collisions is smaller than in the dihedral angle method, and it is efficient. It is thought that the molecule can be deformed.

FIG. 13 shows changes with time of the minimum value of RMSD for 10 trials under each condition. The minimum RMSD value in the translation method (line a) is higher than the dihedral angle method (line b) at the beginning of the simulation, but is lower than the dihedral angle method in most cases after 10,000 sweeps.
From 20000 sweep onwards, it often takes a value of 2Å. FIG. 14 shows a predicted three-dimensional structure when such a value on the order of 2Å is taken. That is, FIG. 14 shows the 28-mer arrangement (black line a) at the low RMSD (2.85Å) obtained by the MC simulation using the translation method and the corresponding NMR structure (grey line b) in an overlapping manner. . Note that FIG. 14 shows only carbon atoms in the skeleton. Moreover, this FIG. 14 shows RasMo.
l [Sayle, RA and Milner-White, EJ RasMol: Bi
omolecular graphics for all. Trends Biochem. Sci.
20, 373-375 (1995)]. 28m
In the er parallel translation method, the calculation time per 100 sweeps is about 30 seconds, so the calculation time required for the first appearance of a structure with an RMSD value of 2Å is about 100 minutes.

From the above, it is considered that the parallel transfer method is the most efficient for searching the structure near the natural structure in the 28-mer MC simulation. This result contrasts with the result that the dihedral angle method is more efficient in forming the structure of the C peptide. This seems to be due to the advantage of the parallel transfer method, in which collisions of atoms due to deformation operations are less likely to occur as the target structure becomes more complex. Judging from the results of the various MC simulations described above, the dihedral angle method is suitable for the one that forms a simple helix structure such as C peptide, but the one having a slightly complicated structure such as 28mer. It is clear that the translation method is suitable for.

[0075]

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to efficiently search for a stable structure of a peptide chain having a complicated three-dimensional structure or a long peptide chain containing many amino acid residues.

[0076]

[Sequence list] <110> YAMATO, Ichiro <120> Compound three-dimensional structure search program, three-dimensional structure search device, and three-dimensional structure search Method <130> SUT01001P <160> 2 <170> PatentIn Ver. 2.1 <210> 1 <211> 13 <212> PRT <213> unknown <220> <223> N terminal fragment of unknown ribonuclease A <300> <301> Brown, J. E. and Klee, W. A. <302> W. A. Helix-coil transition of the isolated amino terminals of rib onuclease <303> Biochemistry <305> 10 <306> 470-476 <307> 1971 <400> 1 Lys Glu Thr Ala Ala Ala Lys Phe Glu Arg Gln His Met   1 5 10 <210> 2 <211> 28 <212> PRT <213> Artificial Sequence <220> <223> 28 mer polypeptide having a beta-beta-alpha structure <300> <301> Dahiyat, B. I., Sarisky, C. A. and Mayo, S. L. <302> De Novo protein design; Towards fully automated sequence selection <303> J. M. Biol. <305> 273 <306> 789-796 <307> 1997 <400> 2 Lys Pro Tyr Thr Ala Arg Ile Lys Gly Arg Thr Phe Ser Asn Glu Lys   1 5 10 15 Glu Leu Arg Asp Phe Leu Glu Thr Phe Thr Gly Arg              20 25

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a basic operation of a translation method according to the present invention.

FIG. 2 is an explanatory diagram of a segment division method in the parallel movement method according to the present invention.

FIG. 3 is an explanatory view schematically showing a structural deformation operation by the parallel displacement method and a structural deformation operation by the dihedral method according to the present invention. FIG. 3A shows the structure before the structural deformation operation.
FIG. 3B is the structure after the structural deformation operation by the parallel displacement method according to the present invention, and FIG. 3C is the structure after the structural deformation operation by the dihedral angle method.

FIG. 4 is an explanatory diagram schematically showing the movement of each atom on the peptide plane in the parallel displacement method according to the present invention, and the movement of each atom on the peptide plane in the dihedral angle method. FIG. 4 (a) shows the parallel displacement method, and FIG. 4 (b) shows the dihedral angle method.

FIG. 5 is a configuration diagram of a three-dimensional structure searching apparatus according to the present invention.

FIG. 6 is a flowchart showing an operation of one trial.

FIG. 7 is a flowchart of main chain sweep processing.

FIG. 8 is a flowchart of a side chain sweep process.

FIG. 9: MC simulation of C peptide (1000
(00 sweep) is a graph showing the degree to which the number of residues forming an α-helix structure depends on the size of deformation. FIG. 9A shows the results when the parallel displacement method is used, and FIG. 9B shows the results when the dihedral angle method is used.

FIG. 10 is a graph showing development of the number of residues forming an α-helix structure in MC simulation of C peptide. ● indicates the result of the dihedral angle method (2000K; maximum rotation angle = 30 °), and ■ indicates the parallel displacement method (2000
K; maximum movement distance = 0.3Å).

FIG. 11: MC simulation of 28mer (100
000 sweep) is a graph showing the degree to which the RMSD of the calculated 28-mer structure and the NMR 28-mer structure depends on the magnitude of deformation. Figure 11
(A) shows the RMSD calculated using the translation method, and FIG. 11 (b) shows the RMSD calculated using the dihedral angle method.
D is shown. R described in each simulation condition
MSD is the average of 10 MC trials.

FIG. 12 is a graph showing developments of RMSD values of a calculated 28-mer structure and a 28-mer structure by NMR. The black line a indicates the RMSD value obtained using the parallel displacement method (2000K, maximum displacement = 0.4Å),
And the gray line b shows the RMSD value obtained by using the dihedral angle method (2000K, maximum rotation angle = 40 °). RM
The SD value is an average value of 10 MC trials.

FIG. 13 is a graph showing the development of RMSD values for the calculated 28-mer structure and the NMR 28-mer structure. This FIG. 13 shows the minimum RMSD value in each sweep during 10 MC trials. This eliminates the effects of failed MC trials and predicts the rate at which better structures will be formed. The black line a is the translation method (200
0K, maximum movement distance = 0.4Å), and the gray line b indicates the dihedral angle method (2000K, maximum rotation angle = 40 °).

FIG. 14: 28 me at low RMSD (2.85Å) obtained by MC simulation using translation method
It is explanatory drawing which overlapped and showed the arrangement | positioning of r, and the structure of 28 mer by the corresponding NMR.

[Explanation of symbols]

1 ... Three-dimensional structure search device; 2 ... Input unit; 2a ...
・ Input processing unit; 3 ... Central processing unit; 4 ... Output unit;
4a ... Output processing unit; 5 ... Control unit; 6 ... Storage unit; 7 ... Processing unit; 8 ... Counting unit; 11, 12 ...
・ Peptide portion; 61 ... designated compound storage area; 62
... Segment storage area; 63 ... Molecular potential energy storage area; 64 ... Steric structure storage area;
65 ... Maximum moving distance storage area; 66 ... Step number / sweep number storage area; 67 ... Cumulative storage area; 7
1 ... Segment division processing unit; 72 ... Segment selection unit; 73 ... Movement distance selection unit; 74 ... Parallel movement operation unit; 75 ... Molecular potential energy calculation unit; 76 ... Comparison Judgment unit; 77 ... First branch judgment unit; 78 ... Second branch judgment unit; 81 ... Step number counting unit; 82 ... Sweep number counting unit.

Claims (6)

[Claims] 1. A program for causing a computer to execute a search for a three-dimensional structure of a compound, the method comprising: (1) calculating a molecular potential energy of a certain virtual three-dimensional structure of a compound designated in advance; ) A step of storing the result of the calculation step (1) in a calculated value storage area in association with the virtual three-dimensional structure used at that time, (3) of a plurality of segments obtained by dividing the specified compound in a predetermined method. A step of selecting two adjacent segments bound to each other from among them, (4) a unit containing one of the bound adjacent segments selected in the selecting step (3) of the specified compound It is divided into a fixed rigid body unit and a moving rigid body unit, which is a unit including the other of the connected adjacent segments, and is divided within a range of a maximum movement distance specified in advance. With the specific travel distance selected for the random
Translating the moving rigid body unit with respect to the fixed rigid body unit, (5) calculating the molecular potential energy of the virtual three-dimensional structure deformed by the moving step (4), (6) the calculating step (5) ) Comparing the molecular potential energy of the three-dimensional structure after deformation calculated in) with the molecular potential energy stored in the calculated value storage area, and determining the calculated value to be adopted (7) The comparison and determination step (7) In 6), the method further comprises a step of newly storing one of the molecular potential energies determined to be adopted in the calculated value storage area in association with the virtual three-dimensional structure at that time, and storing the compound in the computer. A program for executing the search for the three-dimensional structure of. 2. The compound is a protein and the solid
The connection between the constant rigid body unit and the moving rigid body unit is expressed by the formula: -C (= O) -N (H)- Is a peptide bond represented by
The parallel movement of the moving rigid body unit is represented by the general formula (1): [Chemical 1] (In the formula, R is a side chain contained in an amino acid residue, and R_n
Is an α carbon atom C contained in the fixed rigid body unit^α _nSide chain of
And R_{n + 1}Is the α carbon source contained in the moving rigid body unit
Child C^α _{n + 1}Is the side chain of
Α carbon atom forming the peptide-bonded main chain of
C ^α _n, Carbon atom C, nitrogen atom N, and second α carbon source
Child C^α _{n + 1}And an oxygen atom O bonded to the carbon atom C
And a hydrogen atom H bonded to the nitrogen atom N.
And the nitrogen atom C on the plane of the peptide
Child N and the second α carbon atom C^α _{n + 1}Exercise and
(B) the carbon atom C is
First α carbon atom C^α _nWithout changing the interatomic distance between
First α carbon atom C^α _nIs the center of rotation and the bond angle is
By changing, the maximum transfer on the peptide plane
A moving distance that changes randomly within the range of the moving distance,
The circular motion of an arc is performed, and (c) the nitrogen atom N is
Without changing the interatomic distance from the carbon atom C, the carbon atom
By changing the bond angle with C as the center of rotation,
Setting on the peptide plane independently of the step (b)
Change randomly within the maximum movement distance
At the moving distance, independent of the circular motion of the carbon atom C,
At the same time as the circular motion of carbon atom C, circular arc motion is performed.
And (d) the second α carbon atom C^α _{n + 1}Is
Without changing the interatomic distance from the nitrogen atom N,
By changing the bond angle with the atom N as the center of rotation,
From the nitrogen atom N on the peptide plane.
Second α carbon atom C^α _{n + 1}Direction of direction vector toward
Is the above-mentioned first α carbon atom C^α _nFrom the above carbon atom C
Circle so that it always matches the direction of the direction vector to
The method according to claim 1, wherein the circular movement is performed in an arc shape.
The listed program. 3. A program for causing a computer to execute a search for a three-dimensional structure of a compound, the method comprising: (1) calculating a molecular potential energy of a certain virtual initial three-dimensional structure of a compound designated in advance; 2) a step of storing the result of the calculation step (1) in a calculated value storage area in association with the virtual initial three-dimensional structure used at that time,
(3) a step of selecting two adjacently bonded segments from a plurality of segments obtained by dividing the specified compound by a method specified in advance, (4) the specified compound,
A fixed rigid body unit, which is a unit including one of the joining adjacent segments selected in the selecting step (3), and a moving rigid body unit, which is a unit including the other segment of the joining adjacent segments. A step of parallelly moving the movable rigid body unit with respect to the fixed rigid body unit at a specific movement distance randomly selected within a range of a maximum movement distance designated in advance, (5) by the moving step (4) Calculating the molecular potential energy of the deformed virtual three-dimensional structure, (6) the molecular potential energy of the three-dimensional structure after the deformation calculated in the calculation step (5), and the molecule stored in the calculated value storage area Step (7) of comparing with potential energy One molecular potential energy determined to be adopted in the comparison step (6) , In conjunction with a virtual three-dimensional structure when the step of storing newly on the calculated value memory area (8)
The selection step (3) to the storage step (7) are defined as one step, and it is determined whether or not the number of steps is an integral multiple of the designated step number designated in advance.
Returning to the selection step (3), if it is an integer multiple, it has the molecular potential energy stored in the branching step of advancing to the next step (9), and (9) the storing step (7) of the final step. A program for causing a computer to search for a three-dimensional structure of a compound, the method including the step of outputting a virtual three-dimensional structure. 4. A program for causing a computer to execute a search for a three-dimensional structure of a compound, the method comprising: (1) calculating a molecular potential energy of a virtual initial three-dimensional structure of a compound designated in advance; 2) a step of storing the result of the calculation step (1) in a calculated value storage area in association with the virtual initial three-dimensional structure used at that time,
(3) a step of selecting two adjacently bonded segments from a plurality of segments obtained by dividing the specified compound by a method specified in advance, (4) the specified compound,
A fixed rigid body unit, which is a unit including one of the joining adjacent segments selected in the selecting step (3), and a moving rigid body unit, which is a unit including the other segment of the joining adjacent segments. A step of parallelly moving the movable rigid body unit with respect to the fixed rigid body unit at a specific movement distance randomly selected within a range of a maximum movement distance designated in advance, (5) by the moving step (4) Calculating the molecular potential energy of the deformed virtual three-dimensional structure, (6) the molecular potential energy of the three-dimensional structure after the deformation calculated in the calculation step (5), and the molecule stored in the calculated value storage area Step (7) of comparing with potential energy One molecular potential energy determined to be adopted in the comparison step (6) , Step (8) of newly storing in the calculated value storage area in association with the virtual three-dimensional structure at that time, the selection step (3) to the storage step (7) are defined as one step, and the number of steps is designated in advance. It is determined whether or not it is an integer multiple of the designated number of steps specified. If it is not an integer multiple, the process returns to the selection step (3), and if it is an integer multiple, the process proceeds to the next step (9). Step (9) Counting the number of times of proceeding from the branching step (8) to the main step (9) as the number of sweeps, determining whether or not the number of sweeps is a preset designated sweep number, If it is less than the specified number of sweeps, the process returns to the selection step (3), and if it is more than the specified number of sweeps, the process proceeds to the next step (10), the second branching step, (10) the storage step of the final step. Molecular potential stored in (7) Characterized in that it comprises a step of outputting a virtual three-dimensional structure having a Energy, the computer program for executing the search of the three-dimensional structure of the compound. 5. An apparatus for searching a three-dimensional structure of a compound, comprising: (1) an input unit for inputting search conditions, (2) a three-dimensional structure storage area for storing a virtual three-dimensional structure of a designated compound,
(3) A calculation unit that calculates the molecular potential energy of the virtual three-dimensional structure of the specified compound, and (4) a calculation value storage that stores the calculation result of the calculation unit in association with the virtual three-dimensional structure used in the calculation. A region, (5) a segment selection unit that selects two adjacent segments that are connected to each other from a plurality of segments obtained by dividing the specified compound by a predetermined method, (6) a maximum movement distance specified in advance (7) A fixed rigid body unit that is a unit that includes one of the bonded adjacent segments selected by the segment selection unit, the movement distance selection unit that randomly selects a specific movement distance within the range And a moving rigid body unit that is a unit including the other of the connected adjacent segments, and the specific moving distance selected by the moving distance selecting unit. In translating operation unit to translate the movement rigid unit with respect to the fixed rigid unit,
(8) A comparison determination unit that compares the molecular potential energy of the three-dimensional structure after deformation by the parallel movement operation unit with the molecular potential energy already stored in the calculated value storage area to determine the calculated value to be adopted, And (9)
In the comparison / determination unit, one of the molecular potential energies determined to be adopted is associated with a virtual three-dimensional structure at that time, and a control unit for newly storing in the calculated value storage area is included, Compound three-dimensional structure search device. 6. A method for searching a three-dimensional structure of a compound, comprising: (1) calculating a molecular potential energy of a certain virtual three-dimensional structure of a compound designated in advance,
(2) a step of storing the result of the calculation step (1) in a calculated value storage area in association with the virtual three-dimensional structure used at that time, (3) a plurality of divided the specified compounds by a method specified in advance A step of selecting two adjacent segments that are bonded to each other from among the segments; (4) a unit containing one of the bound adjacent segments selected in the selecting step (3), the specified compound; It is divided into a fixed rigid body unit and a moving rigid body unit which is a unit including the other segment of the connected adjacent segment, and at a specific movement distance randomly selected within the range of the maximum movement distance specified in advance. A step of translating the moving rigid body unit with respect to the fixed rigid body unit, (5) calculating the molecular potential energy of the virtual three-dimensional structure deformed by the moving step (4) Step (6) Comparing the molecular potential energy of the three-dimensional structure after deformation calculated in the calculation step (5) with the molecular potential energy stored in the calculated value storage area to determine the calculated value to be adopted. Step (7) including the step of newly storing in the calculated value storage area one of the molecular potential energies determined to be adopted in the comparison and determination step (6) in association with the virtual three-dimensional structure at that time. The method for searching for a three-dimensional structure of a compound as described above, wherein