JPH07152775A - Method for determining three-dimensional structure of protein - Google Patents

Abstract

PURPOSE:To quickly predict and determine structure close to natural structure by automating a part for setting up and updating a mutually interacting range and optimizing energy. CONSTITUTION:A simulation condition is inputted, the two-face angles of parts having the primary structure and secondary structure of a protein are inputted, the initial two-face angle is inputted to a random coil of the protein so as to obtain extended structure at first, and an updatable mutually interacting range is set up. Under the set condition, energy optimization is executed by using simulated annealing, and after checking that the mutually affected range is equal to the total number of residues of the protein, simulation is executed only by prescribed trial frequency and a two-face angle for predicted structure is outputted. Consequently structure close to natural structure can be quickly predicted and determined.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for automatically determining the three-dimensional structure (that is, tertiary structure) of a protein having a known primary structure and secondary structure.

[0002]

2. Description of the Related Art X-ray crystallography and spectroscopic methods are known as methods for determining the three-dimensional structure of proteins (Protein Chemistry 3, "Higher-order structure", pp. 215-289, p.
324-347, 1973 (Kyoritsu Shuppan); ED Schmid et al. "Spectroscopy of Biological Molecules-New Advance"
S., John Wiley & Sons, New York, 1988).

[0003] Proteins are in dynamic equilibrium between a folded and ordered state and an unfolded and disordered state, and this equilibrium is, in part, the amino acids that try to stabilize the protein's structure. It reflects the interactions between the side chains of the residues and, on the other hand, the thermodynamic forces that try to promote the disorder of the molecule.

A protein is usually composed of 20 kinds of amino acids, and has different three-dimensional structures and properties depending on the difference in the amino acid sequence. Amino acids that cause such a difference are nonpolar amino acids (for example, Ala, Val, Leu, Ile, P, etc.) due to the structure and charge of the side chain.
ro, Phe, Trp, Met), uncharged polar amino acids (eg, Gly, Ser, Thr, Cys, Ty)
r, Asn, Gln) and charged polar amino acids (eg Asp, Glu, Lys, Arg, His).

The primary structure of a protein is determined by the DNA sequencing method (Maxam et al., Methods in Enzymology, 65 , 499 (1
980)) is being decided one after another. further,
Based on the primary structure, it is possible to predict secondary structures such as helix structure and β structure (GESchulz &
Co-authored by RHSchirmer, "Protein-Structure / Function / Evolution", translated by Tatsuo Ooi, Kagaku Dojin, 1980). The secondary structure is a regular arrangement of the protein main chain, which is independent of the side chain type and conformation, and is stabilized by the hydrogen bond between the amide group and the carboxyl group of the peptide. Since the helix structure is a linear structure, a regular helix can be expressed by a pair of dihedral angles (φ, Ψ) (for dihedral angles, see GESchulz &
RHSchirmer (see above)). Is a very stable and well appears to have α- helix in the protein, rarely 3 ₁₀ Herissu can also be seen. On the other hand, the β structure is a structure in which a plurality of extended polypeptide chains are connected by hydrogen bonds to form a sheet-like structure, such as a parallel β sheet and an antiparallel β sheet.

[0006] In particular, the three-dimensional structure of a protein is predicted by the above-mentioned physical methods (X-ray crystallographic analysis method, spectroscopic method),
Although decisions have been made, there are only a few hundred proteins of known conformation due to the large availability of the protein of interest and the difficulty in preparing its crystals. Most of these known three-dimensional structures are based on X-ray crystallography. Under such circumstances, some methods for theoretically predicting the three-dimensional structure of proteins have been proposed (for example, N. Saito et al., Repo
rts on Progress in Polymer Physics in Japan, Vol.
XXXV, 1992; Nobuhiko Saito, Biophysics, 25 (6), 11-20 (198)
5) ； C. Wilson and S. Doniach, PROTEINS: Structurp Fu
nction, and Genetics 6: 193-209 (1989); TL Blun
dell et al., Nature 326 , 347 (1987); with J. Skolnick
A. Kolinski, J. Mol. Biol., 221 , 499 (1991)) are known. Further, regarding the prediction of the three-dimensional structure of a protein, the present inventor has already proposed the method shown in FIG. 2 in the attached drawings (N. Saito et al., Reports on Progress in Polym.
er Physics in Japan, Vol. XXXV, 1992; Nobuhiko Saito, Biophysics, 25 (6), 11-20 (1985)). However, this method has the drawback that the interaction range cannot be updated.

[0007]

The conventional methods for predicting the three-dimensional structure of a protein require large-scale calculation even if a computer is used, and depend on the experience of the operator himself who executes programming. There was a problem that there was a lot to do. As a result of intensive studies to solve such problems, the present invention is based on a three-dimensional structure prediction method using an “island model”.
Hydrophobic groups at short distances (sometimes medium distances) along the protein chain are subjected to simulated annealing (simula) described later.
We have discovered a method of automatically packing secondary structures into tertiary structures by using ted annealing).

[0008]

The present invention is a method for determining the three-dimensional structure of a protein, which comprises (a) inputting simulation conditions, and (b) a part of the protein having a primary structure and a secondary structure. Input the dihedral angle, (c) enter the initial dihedral angle in the random coil of the protein so that the structure initially stretches, and (d) set the renewable interaction range,
(E) Energy optimization was performed using simulated annealing under set conditions, (f) it was confirmed that the interaction range was the total number of residues of the protein, and (g) The method is characterized by comprising the step of: (h) outputting the dihedral angle of the prediction structure by performing simulation for the number of trials.

The island model developed by Saito et al. (Supra) predicts the three-dimensional structure of a protein by one important rule, namely, "a close pair of hydrophobic amino acid pairs are close to each other along the protein chain. Interacting with each other) ”. In other words, it is presumed that proteins are folded with the restriction that short-distance hydrophobic amino acid pairs come close to each other along the protein chain. Generally, a globular protein has a hydrophilic group on the surface and a hydrophobic group inside to form its tertiary structure, and the hydrophobic group plays an important role during folding. "Sex interaction" has attracted attention. Here, the hydrophobic interaction is explained as follows. When the side chain of the hydrophobic amino acid comes into contact with water, water regularly sticks around the side chain, resulting in a low entropy state. When these hydrophobic amino acids approach each other and the water clinging to their side chains contact each other, the regular structure collapses and the entropy increases. As a result, the contact increases the energy a little, but the overall free energy decreases. This is the generally known mechanism of hydrophobic interaction. The hydrophobic amino acids involved in such hydrophobic interactions are mainly tryptophan, phenylalanine, leucine, isoleucine, methionine and valine. In addition, the weaker hydrophobic amino acids, alanine, cysteine and tyrosine, may also be involved in the interaction.

The present invention will be described with reference to the flow chart shown in FIG.

The protein three-dimensional structure of the present invention is automatically predicted.
The determination method is as follows: (a) simulation condition input step, (b) primary and secondary structure data input step, (c) initial dihedral angle input step to random coil, (d) interaction range setting Step, (e) energy optimization step, (f) step of confirming that the interaction range is the total number of residues, (g) step of confirming that the simulation has been performed a designated number of times, and ( h) Comprised of dihedral angle output steps of the prediction structure. The method of the present invention may further include (i) graphing the three-dimensional coordinates of the prediction structure.

The steps (a) to (i) will be described below.

In step (a), simulation conditions are input. Generally, the following simulation conditions can be used.

Number of trials: Shows how many times the simulation is repeated.

Number of residues: Indicates the number of residues in the protein.

Start value of random number: A random number sequence is generated from this value.

Number of random coils: Shows how many random coil parts there are.

Location of random coil: Specify the number of random coils in the form of N residues to M residues.

Initial value of interaction range: Indicates the initial value of the number of residues along the chain to be interacted with.

Update value of interaction range: Indicates how much the interaction range is extended after the energy optimization is completed once.

Number of steps for one energy optimization:
After setting the interaction range, we show how many Monte Carlo steps to perform energy optimization.

Annealing start temperature: Indicates the starting temperature of simulated annealing.

Annealing end temperature: Indicates the end temperature of simulated annealing.

Number of times annealing temperature is updated: Indicates how many times the temperature is updated from the start to the end of annealing.

Further, the following conditions may be added if necessary.

[Lennard-Jones]
For the potential (Van der Waals force): Coefficient for 1-4 interaction: For 1-4 interaction (interaction from an atom with the fourth atom counting along the bond), multiply this coefficient Only the minutes are counted as energy.

Coefficients between atoms outside the interaction range:
Between atoms outside the interaction range, the repulsive force is multiplied by this coefficient to count as energy.

Temperature when entropy is converted to free energy: The temperature for multiplying the entropy value by the temperature and adding the energy to obtain free energy is shown.

Scaling interval: Indicates how many Monte Carlo steps are performed for scaling to change the twist of the dihedral angle using a random number.

Energy fluctuation: It shows how much the maximum value of the twist of the dihedral angle corresponds to the energy fluctuation.

Maximum cutoff value for twisting of dihedral angles: Set to this value when twisting of dihedral angles becomes larger than this value.

Minimum twist-off cutoff value of dihedral angle: Set to this value when the twist of dihedral angle becomes smaller than this value.

List output interval: The energy value is output to a file for each value.

Dihedral angle step value: Indicates the step value used when changing the random number to the dihedral angle twist value.

Presence / absence of SS bond.

Presence / absence of hydrogen bond.

In step (b), data is input for the primary structure of the protein and the portion having the secondary structure (helix structure and β structure). At this time, an amino acid sequence is input as the primary structure, and a dihedral angle that can be actually taken is input as the secondary structure.

In step (c), the value of the dihedral angle initially given to the random coil is input. First, assume that the structure is elongated, and enter the dihedral angle (φ, Ψ) at that time. The setting of this dihedral angle is arbitrary, but normally (180 °, 180
) Or (-120 °, 120 °), and in the case of proline, φ should be fixed at about -75 °.

In step (d), for the range of interaction along the first main chain already determined in step (a) (this is referred to as "initial value of interaction range"), the range remains constant. Perform the operation to increase each radix (this is called the "updated value"). Regarding the range of interaction, the protein'folding 'determines that the above-mentioned'hydrophobic interaction' acts as a broad force, and that short-range (sometimes medium-range) hydrophobic amino acid pairs are specifically determined. It is based on the premise that it will be implemented based on that. "Interaction range" is the following formula: Interaction range = (initial value of interaction range) + (updated value)
Calculated from × (the number of times the interaction range is updated). Here, the number of updates of the interaction range is
Indicates the number of energy optimizations.

In step (e), the energy is optimized by using simulated annealing under the conditions set in step (d).

Simulated annealing is a kind of non-linear optimization method, and since it simulates gradually cooling a metal or the like to crystallize it neatly,
Also called the annealing method. In this method, the temperature is first set, and then the value of the objective function (O
The value (O2) of the objective function in the state variable in 1) and the next step is calculated, and the difference (ΔO = O2-O1) is obtained. If ΔO is negative, it means that the value of O2 is smaller than that of O1, so the state variable in the next step is unconditionally accepted. If ΔO is positive, the state variable in the next step is accepted with a probability corresponding to the magnitude and temperature, and the others are excluded. The acceptance probability at this time is set to be higher as ΔO is smaller and as temperature is higher. After repeating this procedure several times, the temperature is gradually lowered to obtain the final state. When the optimization is performed in this way, even if the state is caught by a local minimum value (minimum value of an objective function having many irregularities), it is possible to exit with a certain probability. For more information on simulated annealing, see Kirkpatrick, S. et al.
Science, 220 , 4598 (1983).

The energy function (value of the objective function) in the energy optimization is the following three energies (unit:
kcal / mol).

(I) Van der Waals Force The following 6-12 type Leonard-Jones potential is adopted.

[0044]

[Equation 1]

Here, A and B are parameters ECCEP
(Momany, FA et al., J. Phys. Chem., 78 , 1595 (1974))
Or OPLS (Jorgensen, WL et al., J. Am. Chem.
Soc., 110 , 1657 (1988)) and other molecular dynamics van der Waals forces are used. Further, r represents the center-to-center distance when the hydrophobic side chain is regarded as a sphere.

(Ii) Hydrophobic Interaction This interaction is introduced between the hydrophobic amino acid side chains.

[0047]

[Equation 2]

Where r _o is the van der of the hydrophobic side chain
The sum of the Walls radii (angstrom), and r represents the center-to-center distance when the hydrophobic side chain is regarded as a sphere.

(Iii) Entropy force Assuming a Gaussian chain, it is introduced between the hydrophobic amino acids Cα.

[0050]

[Equation 3]

Here, R is a gas constant and T is a temperature (°
K) and N are the numbers of monomers between the hydrophobic groups, and r is as defined above.

The above-mentioned three sums are used as an objective function, and the objective function is optimized by using simulated annealing. The annealing temperature is set to about 160 ° K at the beginning and about 60 ° K at the end, and is not set to several thousand degrees as is conventionally done.
This is to proceed with the optimization without breaking the structure created in the short interaction range.

In step (f), it is determined whether the interaction range is the total number of residues. If the total number of residues is not reached, the process returns to step (d) and the subsequent steps (d) to (f) are performed. On the other hand, if the interaction range = the total number of residues, go to step (g).

In step (g), it is determined whether the simulation has been performed the set number of times. If not, the dihedral angle is initialized to change the random number sequence to perform the same simulation, and then the process returns to step (d), and the subsequent steps (d) to (g) are executed. On the other hand, if the specified number of trials is simulated, the process proceeds to step (h).

In step (h), the dihedral angle of the predicted structure is output.

In step (i), the three-dimensional coordinates of the predicted structure are graphed from the dihedral angles output in step (h).

Generally, the three-dimensional coordinates can be obtained as follows.

For example, assume that the dihedral angle to be twisted is between the i-th and i + 1-th atoms of the protein main chain. At this time, even if the dihedral angle is twisted, that is, the bond main chain is twisted, i +
The coordinates of the first atom do not change, but the coordinates of the i + 2nd and subsequent atoms all change. The twisting angle is θ, and the three-dimensional structure to be referred to [a structure other than the dihedral angle (for example, bond length, bond angle) equal to the natural structure] is already input.

The following work is carried out under the above settings.

(I) Determine the twist angle θ. Where θ
Is the difference between the dihedral angle between the i-th and i + 1-th atoms of the reference structure and the predicted dihedral angle between the i-th and i + 1-th atoms of the structure.

(Ii) The direction vector from the i-th atom to the i + 1-th atom is obtained. This is (dx, dy, d
z).

(Iii) A rotation matrix R is created.

[0063]

[Equation 4]

(Iv) Find the vector from the (i + 1) th atom to the (i + 2) th atom. This is (x ₁₂ , y ₁₂ , z
₁₂ ).

(V) By the calculation of the following formula, rotation is performed around the (i + 1) th atom, and a new coordinate of the (i + 2) th atom is determined. Here, the coordinates of the new i + 2nd atom are (X _{i + 2} , Y _{i + 2} , Z _{i + 2} ). The coordinates of the i + 1-th atom are (X _{i + 1} , Y _{i + 1} , Z _{i + 1} ).

[0066]

[Equation 5]

(Vi) The work of (iv) and (v) is sequentially performed by i +
Advance from the third atom to the last atom.

(Vii) The position of the dihedral angle to be rotated is i
Dihedral angle between + 1st and i + 2nd atoms, i + 2 and i
Change the dihedral angle between the + 3rd atoms and so on to determine new coordinates.

(Viii) When all the dihedral angles have been twisted, the process ends.

The calculated coordinates are the atomic coordinate display program [for example, Insight (manufactured by BIOSYM), Mo
1-Graph (manufactured by Daikin Industries, Ltd.), AVS (manufactured by AVS), etc.], for example, in a protein data bank (PDB) format for output. Then, the program as described above is input and displayed graphically.

Alternatively, the C of each amino acid can be determined from the three-dimensional coordinates.
It is also possible to calculate the distance between α and use this to draw a distance map. The distance map can be constructed as follows.

(I) In the case of a protein consisting of N residues, numbers are given from 1 to N-1 on the horizontal axis and from 2 to N from the top on the vertical axis.

(Ii) Calculate the distance between Cα of the 1st amino acid and the 2nd amino acid, and mark the corresponding place (1, 2) depending on whether it is larger or smaller than the cutoff value.

(Iii) Calculate the distance between Cα of the 2nd amino acid and the 3rd amino acid, and mark the corresponding place (1, 3) depending on whether it is larger or smaller than the cutoff value.

(Iv) This operation is repeated up to (N-1, N), and the lower left half of the figure (triangular map) is filled with symbols.

[0076]

The invention is illustrated in more detail by the following non-limiting examples.

Example 1 Folding of crambin
Folding Simulation (1) In this embodiment, the folding simulation of the clambin was performed according to the flowchart of the present invention shown in FIG. The simulation conditions are as follows.

Number of trials: 20 Number of residues: 46 Start value of random number: 1026400755 Number of random coils: 6 Place of random coils: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Initial value of interaction range: 6 Updated value of interaction range: 5 About Leonard Jones potential (van der Waals force) Coefficient for interaction 1-4: 0.5 Interaction range Coefficients for atoms other than 0.5: Number of steps for one energy optimization: 300 Scaling interval: 30 Energy fluctuation: 10 Kcal / mol Maximum cutoff value for torsion of dihedral angle: 9 degrees Dihedral angle Minimum twist-off value: 0.018 degrees Annealing start temperature: 160 Kelvin Annealing end temperature: 60 Kelvin Temperature when converting entropy to free energy: 1
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol.

Further, the primary structure is the amino acid sequence of crambin (see Protein Data Bank 1CRN), and the dihedral angle of the secondary structure part is the dihedral angle (α of the natural structure of crambin.
In case of helix φ = about −57 °, Ψ = about −47 °; 3
_In case of ₁₀ helices φ = about −60 °, Ψ = about −30
°; in case of β structure φ = about −120 °, Ψ = about 120 °)
As the initial dihedral angle of the random coil (-
120 °, 120 °).

Table 1 shows the energies and root mean square (RMS) values of 20 structures obtained as a result. From Table 1, the structure with the lowest energy (-91.1
Kcal / mol) is the minimum root mean square value (3.88 angstrom), and the distance map of the predicted structure (Fig. 3) and the distance map of the natural structure (Fig. 4) obtained by this method. From the comparison, it can be seen that a rough structure prediction can be made by this method. That is, in both of them, (1) α helices are packed almost in parallel, (2) The second β strand interacts with 3 ₁₀ helices, and (3) β strands interact with each other. There are similarities such as what you are doing. Here, R. M. S. The value is an evaluation function indicating how well the prediction was performed, and the distance between the i-th residue and the j-th residue in the predicted structure is Rij,
The distance between the i-th residue and the j-th residue in the natural structure is Tij
Is defined as (<Rij-Tij) ² >) ^1/2 . However, <**> indicates the arithmetic mean of residue pairs.

[0081]

[Table 1]

Example 2 Folding stains on clambin
Simulation (2) The simulation was performed under the same conditions as in Example 1 except that the underlined conditions were changed.

Number of trials: 20 Number of residues: 46 Start value of random number: 1026400755 Number of random coils: 6 Place of random coils: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Initial value of interaction range: 6 Updated value of interaction range: 5 About Leonard Jones potential (van der Waals force) Coefficient for interaction 1-4: 0.5 Interaction range Coefficients between atoms other than 0.5: 0.5 Number of steps for one energy optimization: 300 Scaling interval: 30 Energy fluctuation: 7 Kcal / mol Maximum cutoff value of torsion of dihedral angle: 9 degrees Dihedral angle Minimum twist-off value: 0.018 degrees Annealing start temperature: 160 Kelvin Annealing end temperature: 60 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol.

The inputs of the primary structure, the secondary structure, and the dihedral angles of the random coil are the same as in the first embodiment.

Table 2 shows the energies and root mean squares of 20 structures obtained as a result. From Table 2, although the structure with the lowest energy does not have the root mean square value, a good structure (see FIG. 5) corresponding to the simulation performed in Example 1 is obtained. That is, looking at Table 2, the number of trials was 5, and the R. M. S. Value 3.
It is 80 angstroms, which is a value close to the result of Example 1.

[0086]

[Table 2]

Comparative Example Folding of Granbin by the conventional method
Folding Simulation In this comparative example, a folding simulation of Granvin was performed according to the flowchart shown in FIG. Since the interaction range cannot be updated in this program, when the program ends once, the interaction range and the part of the random coil to be moved are reset by the person, and the dihedral angle previously output is set. Type in, run the program again,
This needs to be repeated.

First Simulation A simulation was performed by inputting the following simulation conditions. The input of the primary structure, the secondary structure, and the dihedral angle of the random coil are the same as in Example 1 (the same applies to the second to fifth times described below).

Number of trials: 10 times Residue number: 46 Random starting value: 1026400755 Number of random coils: 6 Random coil locations: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Number of random coils to move: 1 Location of random coil to move: 31-32 Range of interaction range: 10 For Leonard-Jones potential (van der Waals force) Coefficients for 1-4 interactions : 0.5 Coefficient between atoms outside the interaction range: 0.5 Number of steps for energy optimization: 800 Scaling interval: 50 Energy fluctuation: 1 Kcal / mol Maximum cutoff value of torsion of dihedral angle: 9 degrees Minimum cutoff value for twist of dihedral angle: 0.018 degrees Annealing start temperature: 150 Kelvin Annealing end temperature: 50 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol The energy of 10 structures obtained as a result and the state of interaction of the hydrophobic group were compared, and the input structure for the next simulation was selected.

2nd simulation trial number: 10 times Residue number: 46 Random start value: 1026400755 Number of random coils: 6 Random coil location: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Number of moving random coil: 1 Location of moving random coil: 35-41 Range of interaction range: 10 For Leonard Jones potential (van der Waals force) Coefficient for 1-4 interaction : 0.5 Coefficient between atoms outside the interaction range: 0.5 Number of steps for one energy optimization: 800 Scaling interval: 50 Energy fluctuation: 1 Kcal / mol Maximum cutoff value of torsion of dihedral angle : 9 degrees Minimum cutoff value for dihedral twist: 0.018 degrees Annealing start temperature: 150 Kelvin Annealing end temperature: 50 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol The point different from the previous input is the position of the random coil to move. A simulation was performed using this as an input. The energy of 10 structures obtained as a result and the state of interaction of the hydrophobic group were compared, and the input structure for the next simulation was selected.

Third simulation trial number: 10 times Residue number: 46 Random number starting value: 1026400755 Number of random coils: 6 Random coil location: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Number of random coils to move: 1 Place of random coil to move: 4-6 Range of interaction range: 10 For Lennard Jones potential (van der Waals force) Coefficients for 1-4 interactions : 0.5 Coefficient between atoms outside the interaction range: 0.5 Number of steps for one energy optimization: 800 Scaling interval: 50 Energy fluctuation: 1 Kcal / mol Maximum cutoff value of torsion of dihedral angle : 9 degrees Minimum cutoff value for dihedral twist: 0.018 degrees Annealing start temperature: 150 Kelvin Annealing end temperature: 50 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol The point different from the previous input is the position of the random coil to move. A simulation was performed using this as an input. The energy of 10 structures obtained as a result and the state of interaction of the hydrophobic group were compared, and the input structure for the next simulation was selected.

4th number of simulation trials: 20 times Number of residues: 46 Start value of random number: 1026400755 Number of random coils: 6 Location of random coils: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Number of random coils to move: 1 Location of random coil to move: 18-22 Range of interaction range: 14 For Lennard-Jones potential (van der Waals force) Coefficients for 1-4 interactions : 0.5 Coefficient between atoms outside the interaction range: 0.5 Number of steps for one energy optimization: 1200 Scaling interval: 50 Energy fluctuation: 1 Kcal / mol Maximum cutoff value of torsion of dihedral angle : 9 degrees Minimum cutoff value for dihedral twist: 0.018 degrees Annealing start temperature: 150 Kelvin Annealing end temperature: 50 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol The points that changed from the previous input are the number of trials, the position of the random coil to move, the range of interaction, and the number of steps for energy optimization. A simulation was performed using this as an input. The energy of 20 structures obtained as a result and the state of interaction of the hydrophobic groups were compared, and the input structure for the next simulation was selected.

Fifth simulation trial number: 20 times Residue number: 46 Random number starting value: 1026400755 Random coil number: 6 Random coil location: 1-1, 4-6, 18-22, 31-
32,35-41,45-46 Number of moving random coil: 6 Place of moving random coil: 1-1,4-6,18-2
2, 31-32, 35-41, 45-46 Range of interaction range: 46 About Leonard-Jones potential (van der Waals force) 1-4 Coefficient for interaction: 0.5 For atoms outside the interaction range Coefficient: 0.5 Number of steps for one energy optimization: 3600 Scaling interval: 50 Energy fluctuation: 1 Kcal / mol Maximum cutoff value of torsion of dihedral angle: 9 degrees Minimum cutoff of torsion of dihedral angle Value: 0.018 degrees Annealing start temperature: 150 Kelvin Annealing end temperature: 50 Kelvin Temperature when converting entropy to free energy: 3
00 Kelvin Annealing temperature Number of updates: 50 times Dihedral angle value: 1.8 degrees Leonard Jones potential cutoff value: 10
Kcal / mol The point that changed from the previous input is the position of the random coil to move,
The range of interaction and the number of steps for energy optimization. A simulation was performed using this as an input. Table 3 shows the energies and root mean square values of 20 structures obtained as a result. Although not the lowest in terms of energy, the distance map shown in FIG. 6 was obtained. As a result, the same results as those obtained by the method of the present invention are obtained.

[0094]

[Table 3]

[0095]

INDUSTRIAL APPLICABILITY In protein three-dimensional structure prediction, by performing energy optimization by automating the part for setting and updating the interaction range, the interaction range can be updated in a shorter time than in the conventional method, It has an advantage that a structure closer to a natural structure can be predicted and determined, and moreover, the protein can be folded without depending on the technique of the program executor.

[Brief description of drawings]

FIG. 1 shows a flowchart of the protein tertiary structure prediction system of the present invention.

FIG. 2 shows a flowchart of a conventional protein tertiary structure prediction system.

FIG. 3 is a simulation of Example 1 (1).
3 shows a distance map of the prediction structure of Crambin predicted by. In the figure, white squares are amino acid pairs other than hydrophobic group pairs with a Cα distance of less than 13 Å (cutoff distance), black squares are hydrophobic group pairs with a Cα distance of less than 13 Å, and × is a Cα distance of 13 Å. The above hydrophobic group pairs, and the blanks represent amino acid pairs other than the hydrophobic group pairs in which the distance between Cα is 13 Å or more (FIG. 4).
The same applies to cases of ~ 6).

FIG. 4 This figure shows a distance map of the natural structure of clambin.

FIG. 5 shows a simulation (2) of the second embodiment.
3 shows a distance map of the prediction structure of Crambin predicted by.

FIG. 6 shows a distance map of the prediction structure of clambin predicted by the simulation of the conventional method.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical display location G06F 15/42 A (72) Inventor Nobuhiko Saito 3-4 Okubo, Shinjuku-ku, Tokyo School corporation Waseda University Science and Engineering Research Center

Claims (5)

[Claims] 1. A method for determining a three-dimensional structure of a protein, comprising: (a) inputting simulation conditions;
Input the primary structure of the protein and the dihedral angle of the part that takes the secondary structure, (c) enter the initial dihedral angle in the random coil of the protein so that the structure initially extends, and (d) updatable The interaction range is set, (e) energy optimization is performed using simulated annealing under the set conditions, and (f) the interaction range is confirmed to be the total number of residues of the protein. Then, (g) the simulation is performed a predetermined number of times, and (h) each step of outputting the dihedral angle of the prediction structure is provided. 2. The simulation condition is at least the number of simulation trials, the number of protein residues, the random number start value, the random coil number, the random coil location, the initial value of the interaction range, and the updated value of the interaction range. 1
The method according to claim 1, comprising the number of steps of energy optimization per cycle, an annealing start temperature, an annealing end temperature, and an annealing temperature update frequency. 3. In the step (f), if the interaction range is not the total number of residues, the process returns to the step (d),
The method according to claim 1, wherein the subsequent steps (d) to (f) are performed. 4. In step (g), if the simulation is not performed for the set number of trials, the process returns to step (d) and the subsequent steps (d) to (g).
The method of claim 1 for performing. 5. The method according to claim 1, further comprising the step of graphing the three-dimensional coordinates of the prediction structure from the output dihedral angle.