JP2930851B2 - Method and apparatus for calculating prediction accuracy of three-dimensional structure of protein - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for calculating the prediction accuracy of a three-dimensional structure of a protein in the field of biotechnology such as protein engineering.

[0002]

2. Description of the Related Art A protein is formed by binding 20 kinds of amino acids shown in Table 1 such as glutamic acid and lysine to a peptide, for example, methionine-glutamic acid-glycine-lysine-
It is composed of tens to hundreds of amino acids connected like proline-..., Which is called an amino acid sequence or a primary structure of a protein. Information, especially information on its three-dimensional structure and function. Such a unit that forms a protein by adjoining amino acids with a peptide bond is particularly called an amino acid residue.

[0003]

[Table 1] 名称 Name of amino acid 3 letter code 1 letter code ────────── Glycine GLY G alanine ALA A serine SERS cysteine CYS C methionine MET M lysine LYS K valine VAL V threonine THR T isoleucine ILE I leucine LEUL L aspartic acid ASP D Asparagine ASNN N Glutamic acid GLU E Glutamine GLN Q Arginine ARG R Proline PRO P Histidine HISH H Phenylalanine PHE F Tyrosine TYRY Tryptophan TRP W ────────────────────── ─────

In general, it is relatively easy to determine the amino acid sequence of a protein, and up to the present, the amino acid sequences of about 20,000 types of proteins have been determined and stored in databases. On the other hand, regarding the three-dimensional structure of these proteins, the spatial coordinates (x, y, z) of each atom of amino acid residues constituting the protein have been conventionally determined by, for example, X-ray crystallography or NMR. Is required.
Hereinafter, the protein database of the spatial coordinates is referred to as PDB.
(Protein Data Bank). Therefore, it is possible to calculate statistical data relating to the relative distance between 20 types of amino acid residues from a protein whose known three-dimensional structure is known.

On the other hand, in recent years, with the increase in data on the three-dimensional structure of a protein determined by the X-ray analysis, the amino acid sequence similar to the amino acid sequence of a protein to be predicted (hereinafter referred to as the target protein), that is, homology The possibility of finding a protein to be a reference (homology) having a known three-dimensional structure (hereinafter referred to as a reference protein) is increasing. Under such circumstances, there has been proposed a homology modeling method of predicting the three-dimensional structure of a target protein by referring to the three-dimensional structure of a reference protein. That is, it is based on the assumption that if the amino acid sequences are similar, the three-dimensional structure will also be very similar.

As an example of the above homology modeling method, there is a method using three-dimensional coordinates of an α carbon atom (hereinafter referred to as a _Cα atom) which is a main atom constituting a reference protein and forms a center of an amino residue. Proposed. Absolute coordinates of the C _alpha atoms in order that depends on the orientation and position of the protein molecule, the calculation of the three-dimensional coordinates have a problem of the very complicated.

On the other hand, Crippen (GMCrippen) and Hebel (T.Havel) obtains the metric matrix from the relative distance between the C _alpha atoms by solving the eigenvalue problem, the top three eigenvalues of C _alpha atom in descending order Finding three-dimensional coordinates, the so-called distance geometr
y). However, such a method has a problem that the obtained protein is smaller than the actual protein.

[0008] In order to solve this problem, using the relative distance between two C _alpha atoms of the reference protein, a method for predicting the three-dimensional structure of the target protein, e.g., Wamizuumihaku (Hiroshi
Wako and Harold A. Scherag
This is proposed in the following document according to a) (hereinafter referred to as the Waiko-Sheraga method). “Distance-Conatraint Approach to Protein Folding.
I. Statistical Analysis of Protein Conformations i
n Terms of Distances Between Residues ”, Journal o
f Protein Chemistry, Vol.1, No.1, pp5-45, 1982 (hereinafter referred to as Reference 1)

However, in the method of Wahu and Shelaga disclosed in Reference 1, X-ray crystallography and N
Since the average statistical data based on the three-dimensional structure of the reference protein obtained by the analysis using MR is used, there is a large deviation from the coordinates of the actual protein to be predicted, and it is difficult to predict the actual three-dimensional structure. could not.

In order to solve this problem, the present inventor proposed in Japanese Patent Application No. 5-252895 a method and apparatus for predicting and calculating the three-dimensional structure of a protein. The calculation method and the calculation device select at least one protein having a high degree of homology with the target protein whose three-dimensional structure is to be predicted from among the proteins whose coordinates of the atoms of the amino acid residues are known, as the reference protein, A method for predicting and calculating the three-dimensional structure of a protein, wherein the coordinates of the atoms of the amino acid residues of the target protein are calculated based on the coordinates of the atoms of the amino acid residues of the reference protein. A plurality of homologous residue pairs are cut out by comparing the amino acid sequences between them, and the relative distance between atoms of each residue having a different residue number in the reference protein based on the cut out plurality of residue pairs is determined. Calculate and set to the corresponding first relative distance of the target protein, and based on the coordinates of the initial three-dimensional structure of the predetermined target protein, the mutuals in the initial three-dimensional structure are determined. Calculating a second relative distance between atoms of each residue having a different residue number, the first relative distance between atoms of each residue having a different residue number in the target protein, and The method is characterized in that optimization is performed so that the sum of the squares of the difference from the second relative distance is minimized, and the coordinates of the atoms of the amino acid residues of the target protein are predicted and calculated.

[0011]

By the way, the three-dimensional structure of a protein can be determined by, for example,
When predicted by a method such as homology modeling,
It is known that the deviation from the actual three-dimensional structure due to X-ray crystal analysis or the like is not uniform along the amino acid sequence. Usually, since the result of X-ray crystal analysis or the like is not known for the target protein whose tertiary structure is to be predicted, it is difficult to estimate the local prediction accuracy from the amino acid sequence information. On the other hand, when a reference protein having a sequence homologous (similar) to the amino acid sequence of the target protein is found, a three-dimensional structure is predicted by a technique such as homology modeling, but generally the amino acid sequence is similar. If the three-dimensional structure is expected to be very similar, it suggests the possibility of estimating the local prediction accuracy from the amino acid sequence information.

However, in the prior art, since the degree of homology is not defined quantitatively along the amino acid sequence and reflects the three-dimensional structure, the local prediction accuracy of the three-dimensional structure along the amino acid sequence is low. Could not be estimated quantitatively. That is, there is a problem in that the definition of homology is quantitative along the amino acid sequence, and if the homology of the amino acid sequence of a certain region is high, the three-dimensional structure of the region is not very similar.

In order to solve this problem, regarding the problem of the definition of homology which is not quantitative and does not reflect the three-dimensional structure, the present inventor, Yasushi Kubota,
The following document shows that they can solve the problem by using a correlation function (hereinafter referred to as Kubota et al.'S method). “Correspondence of Homologies in Amino Acid Seque
nce and Tertiary Structure of Protein Molecules "
, Biochimica et Biophysica Acta, Vol.701, pp242-252
(1982) (hereinafter referred to as Reference 2)

However, in the method of Kubota et al. Disclosed in Reference 2, since two amino acid sequences are developed two-dimensionally in the form of a matrix, homology is one-dimensionally determined along the amino acid sequences. There was a disadvantage that it could not be expressed quantitatively.

[0015] An object of the present invention is to solve the above-mentioned problems, and to provide a three-dimensional structure of a protein which can be calculated by estimating the accuracy of the three-dimensional structure of the target protein predicted by a predetermined method quantitatively along the amino acid sequence. It is an object of the present invention to provide a prediction accuracy calculation method and a prediction accuracy calculation device for a structure.

[0016]

According to the present invention, there is provided a method for predicting and calculating a three-dimensional structure of a protein, wherein the target protein whose three-dimensional structure is to be predicted among proteins whose amino acid residue atom coordinates are known. At least one with a high degree of homology with
Three proteins are selected as reference proteins, and the coordinates of the atoms of the amino acid residues of the target protein are predicted by a predetermined method based on the coordinates of the atoms of the amino acid residues of the selected reference protein. A method for calculating the prediction accuracy of a three-dimensional structure of a protein for calculating prediction accuracy, wherein the juxtaposed sequence of the reference protein and the target protein, based on physicochemical parameters of a plurality of types of amino acids contained in the target protein, By changing both the residue number of the reference protein and the residue number of the target protein so that the amino acid residue of the reference protein corresponds to the amino acid residue of the target protein, the correlation with each corresponding amino acid residue is changed. The value of the correlation function indicating the property is determined by using at least one selected physicochemical parameter previously selected from the physicochemical parameters of the plurality of amino acids. Each calculated for over data, characterized by using the average value of the correlation function value is calculated by averaging the correlation function calculated above as the prediction accuracy of the three-dimensional structure of the target protein is the prediction. Further, the method for calculating the prediction accuracy of the three-dimensional structure of the protein according to claim 2 is the method for calculating the prediction accuracy of the three-dimensional structure of the protein according to claim 1, wherein the atom of the amino acid residue is an α-carbon atom. I do. Further, the method for calculating the prediction accuracy of the three-dimensional structure of the protein according to claim 3 is the method for calculating the prediction accuracy of the three-dimensional structure of the protein according to claim 1 or 2, wherein the selected physicochemical parameters are polarities and partial ratios. Volume, degree of turn formation, α
It is characterized by including at least one of a pK value of an amino group, a pK value of an α-carboxyl group, and a degree of mutation.

According to a fourth aspect of the present invention, there is provided an apparatus for predicting and calculating the three-dimensional structure of a protein, wherein the degree of homology with a target protein whose three-dimensional structure is to be predicted among proteins having known coordinates of atoms of amino acid residues. Is selected as a reference protein, and the coordinates of the atoms of the amino acid residues of the target protein are predicted by a predetermined method based on the coordinates of the atoms of the amino acid residues of the selected reference protein. An apparatus for predicting the three-dimensional structure of a protein for calculating the three-dimensional structure prediction accuracy, comprising: juxtaposition of the reference protein and the target protein based on physicochemical parameters of a plurality of types of amino acids contained in the target protein. In the sequence, the residue number of the reference protein and the residue number of the target protein are set so that the amino acid residue of the reference protein corresponds to the amino acid residue of the target protein. By changing both the base number and the value of the correlation function indicating the correlation for each corresponding amino acid residue, at least one selected physicochemical parameter previously selected from the physicochemical parameters of the plurality of amino acids A first calculating means for calculating each of them; and a second calculating means for averaging the values of the correlation functions calculated by the first calculating means to calculate an average value of the correlation functions. The average value of the correlation function calculated by the calculation means is used as the predicted accuracy of the predicted three-dimensional structure of the target protein. Also,
According to a fifth aspect of the present invention, there is provided an apparatus for predicting the three-dimensional structure of a protein, wherein the amino acid residue atom is an α-carbon atom. The apparatus for predicting the three-dimensional structure of a protein according to claim 6 is the apparatus for predicting the three-dimensional structure of a protein according to claim 4 or 5, wherein the selected physicochemical parameters are polarities and partial ratios. It is characterized by including at least one of the following: the volume, the degree of turn formation, the pK value of the α amino group, the pK value of the α carboxyl group, and the degree of mutation.

[0018]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a prediction calculation and prediction accuracy calculation device for a three-dimensional structure of a protein according to one embodiment of the present invention.

The method used in the apparatus for predicting and predicting the three-dimensional structure of a protein according to the present embodiment is characterized in that the method includes the following steps, which are roughly divided. (A) A reference protein having a sequence similar to the amino acid sequence of the target protein and having a known three-dimensional structure is searched based on the PDB, and the degree of homology indicating the ratio of the corresponding homologous residue pairs to the whole is searched. Is selected as a reference protein from among the reference protein candidates having a high degree of homology, for example, 30% or more, with the maximum (or relatively high) degree of homology. (B) Compare the amino acid sequences between the selected reference protein and the target protein, and cut out a plurality of corresponding homologous residue pairs without insertion or deletion of 5 or more residues. (C) calculating a relative distance between atoms of residues having different residue numbers among a plurality of homologous regions corresponding to a reference protein based on a plurality of cut out residue pairs; Set as the corresponding relative distance <d _ij > of the protein. (D) A random number is generated based on a predetermined random number seed IR, a value proportional to the random number value is set as a coordinate value of an initial three-dimensional structure (hereinafter, referred to as an initial structure) of the target protein, and based on the coordinate value. Then, a relative distance d _ij corresponding to a relative distance between atoms of residues having different residue numbers between the plurality of homologous regions in the initial structure of the target protein is calculated. (E) For example, using the quasi-Newton method, the sum of the squares of the difference between the relative distance d _ij and the relative distance <d _ij > between the atoms of the residues having different residue numbers in the target protein is minimized. That is, the function F is optimized so that the function value of the function F of the target protein expressed by the following equation 1 is minimized, and the predicted coordinates of the _Cα atom of the target protein, which is a variable of the function F, Is calculated. (F) The amino acid sequence is determined by, for example, the degree of polarity, the specific volume, the degree of turn formation, the pK value of the α-amino group,
Sequence using the physicochemical parameters p unique to the 20 amino acids in Table 1, such as value, mutation degree, etc.
In an alignment showing the amino acid sequence correspondence between the target protein X and the reference protein Y, the residue number i of the target protein X and the residue number of the reference protein Y corresponding to the residue number i j is calculated by defining the correlation function Cp (i) with the following equation (5). Further, in order to improve the signal-to-noise ratio S / N for noise reduction, as shown in the following equation (6) The prediction accuracy is estimated and calculated by calculating an average value <C (i)> for the plurality of parameters p with respect to the correlation function Cp (i).

First, the basic principle relating to the prediction calculation and the prediction accuracy calculation of the three-dimensional structure of the target protein will be described. In this operation, it is assumed that one target protein is modeled by residues represented by only _Cα atoms. Therefore, 1 composed of M natural residues
The three-dimensional structure of one target protein is represented by M points. Here, the three-dimensional structure including the spatial positions of the M points can be obtained by optimizing the function shown in the following Expression 1.

[0021]

## EQU1 ## F = Σ Σw _ij · (d _ij − <d _ij >) ² / mi <j

The function F is represented by the sum of i and j between atoms of residues having different residue numbers under the condition of i <j as shown in the above equation (1). Here, when one point is represented by coordinates using an XYZ rectangular coordinate system, the function F is a function of 3M coordinate variables as shown in the following Expression 2.

[0023]

F = F (x ₁ , y ₁ , z ₁ , x ₂ , y ₂ , z ₂ ,...,
x _M , y _M , z _M )

In the above formula 1, m is the total number M (M−M−) of the corresponding residue pairs between the target protein and the reference protein.
1) / 2. In addition, w _ij is a weighting coefficient. In the conventional method of Wahu and Shelaga, a weighting coefficient based on average statistical data is used. In the present embodiment, w _ij = 1 is always set. Further, _dij is the i-th residue and j
<D _ij > is the corresponding i-th residue from the average statistical data obtained from many proteins based on PDB in the conventional Wahu-Shelaga method. And the relative distance assigned to the j-th residue pair. In this example, the three-dimensional structure selected so as to have the maximum degree of homology with the target protein as described above is a known reference protein. Are the relative distances assigned to the i-th and j-th residue pairs corresponding to the target protein described above.

Given the amino acid sequence, the function F of Equation 1 is uniquely determined. In this embodiment, a random number seed IR is given to generate a random number, which is proportional to the random number value. By substituting the relative distance between the i-th residue and the j-th residue in the initial structure of the target protein using the values as coordinate values into _dij of the function F, for example, using the quasi-Newton method, The function F is optimized so that the function F is minimized, and the coordinates of the _Cα atom of the target protein are calculated using the optimized function F. That is, the relative distance between the i-th residue and the j-th residue in the initial structure of the target protein, and the i of the reference protein corresponding to the target protein
Coordinate variables x ₁ , y ₁ , z ₁ , x ₂ , y ₂ , z _{2 of the} target protein such that the sum of the squares of the differences between the relative distances assigned to the jth and jth residue pairs is minimized. _{, ..., x M, y M} , z M
To determine the three-dimensional structure of the target protein. In addition,
For residues in regions where there is an insertion or deletion or no homology, the relative distance of the statistical data of the conventional method is used.

On the other hand, in the conventional Wahu-Sheraga method, relative distances assigned to the corresponding i-th and j-th residue pairs based on average statistical data obtained from a large number of proteins based on PDBs When <d _ij > is used, since the data is very coarse, the three-dimensional structure of the target protein obtained from the function F is far from the actual one.

In this embodiment, a predetermined random number seed IR
When the initial structure of the target protein is determined by generating random numbers and giving coordinates that are proportional to the random numbers, a linear regression analysis of 14 globular proteins is performed based on the statistical data of the proteins. The following equation 3 showing the radius of gyration Rg (Å) of the protein obtained in the following way:
3M coordinates (x ₁ , y ₁ , z ₁ , x ₂ , y) of the initial structure
₂ , z ₂ ,..., X _M , y _M , z _M ) (hereinafter, the sign is represented as q _lm , where l = 1, 2, 3;
..., M. ) Is used.

[0028]

## EQU3 ## logRg = 0.824 + 0.367.log
M

## EQU4 ## q _lm = 2 × Rg × rand (IR)

Here, rand () is a known random number generation function that generates a random number value (0 ≦ random number value <1) when a random number seed is set as an initial value. Using Equation 4 above, in order to approximate the coordinates of the initial structure of the target protein with actual coordinates, the random number value rand (IR) is multiplied by the radius of gyration Rg of the protein, and then the coordinates are used to obtain the coordinates of the diameter. Is doubled to obtain the coordinate value of the initial structure.

Further, a method for estimating and calculating the prediction accuracy of the three-dimensional structure of the target protein predicted by the above-described prediction method along the amino acid sequence will be described in detail. Since amino acid sequences are represented by "symbol strings" using the one-letter codes shown in Table 1, conventional and commonly used homologies are similarities in the physical and chemical properties of amino acid residues, such as isoleucine ( This is based on a qualitative method based on similarities such as similarity between I) and leucine (L) and similarity between glutamic acid (E) and aspartic acid (D). In the method of the present invention, the amino acid sequence is converted into a sequence using a certain parameter, and the correlation function between the target protein and the reference protein is determined using Equation 5 described later.
Is defined quantitatively as follows. That is, the method defines homology between the target protein and the reference protein quantitatively along the amino acid sequence and reflects the three-dimensional structure. It is intended to estimate a precise prediction accuracy.

As is known, the amino acid sequence may be, for example, a polarity, a partial specificity, a degree of turn formation, a pK value of an α amino group, a pK value of an α carboxyl group, a degree of mutation, etc. In addition, as shown in Tables 2 to 7, numerical sequences can be formed using physicochemical parameters p unique to the 20 kinds of amino acids shown in Table 1. In Tables 2 to 7, the one-letter codes of the 20 amino acids shown in Table 1 in alphabetical order (A, C, D, E, F, G, H, I,
K, L, M, N, P, Q, R, S, T, V, W, and Y), and two of the parentheses in Tables 2 to 7 indicate the numerical value of each parameter. Is the author's name and the year of publication. (A) polarity: the degree to which amino acid residues are exposed on the protein surface. (B) Partial specific volume: the change in volume of a solution when a unit mass of amino acid is dissolved in an infinite amount of solution. (C) Turn formation (propensipy to form reverse tur)
n): Degree of amino acid residue bending in the protein. (D) pK value of α-amino group (pK-N): α of amino acid
The value obtained by adding the negative sign to the logarithm of the dissociation constant K of the amino group, that is, -logK. (E) pK value of α carboxyl group (pK−C): a value obtained by adding a negative sign to the logarithm of the dissociation constant K of α carboxyl group of amino acid, that is, −logK. (F) Relative mutability: relative substitution frequency of amino acid residues.

[0032]

[Table 2] Polarity (Grantham, 1974) ──────────────────────────── 8.10,5.50,13.00,12.30,5.20,9.00 , 10.40,5.20,11.30,4.90,5.70,11.60,8.00,10.50,10.50,9.20,8.60,5.90,5.40,6.20 ────────────────────── ──────

[0033]

[Table 3] Uneven volume (Cohn and Edsall, 1943) ──────────────────────────── 0.75, 0.61, 0.60, 0.66, 0.77,0.64,0.67,0.90,0.82,0.90,0.75,0.61,0.76,0.67,0.70,0.68,0.70,0.86,0.74,0.71 ─────────────────── ─────────

[0034]

[Table 4] Degree of turn formation (Levitt, 1978) ──────────────────────────── 0.77, 0.81, 1.41, 0.99, 0.59, 1.64,0.68,0.51,0.96,0.58, 0.41,1.28,1.91,0.98,0.88,1.32,1.04,0.47,0.76,1.05 ───────────────────── ───────

[0035]

[Table 5] pK value of α-amino group (Sober, 1970) ──────────────────────────── 9.69, 8.35, 9.60, 9.67 , 9.18,9.78,9.17,9.68,9.18,9.60,9.21,8.80,10.64,9.13,8.99,9.21,9.10,9.62,9.44,9.11 ────────────────── ──────────

[0036]

[Table 6] pK value of α-carboxyl group (Sober, 1970) ──────────────────────────── 2.34, 1.92, 1.88, 2.10 , 2.16,2.35,1.82,2.36,2.16,2.36,2.28,2.02,1.95,2.17,1.82,2.19,2.09,2.32,2.43,2.20 ────────────────── ──────────

[0037]

Table 7 Mutation degree (Dayhoff, 1978) ──────────────────────────── 1.00, 0.20, 1. 06, 1.02, 0.41, 0.49, 0.66
0.96, 0.56, 0.40, 0.94, 1.34, 0.56, 0.93, 0.65, 1.20, 0.97,
0.74, 0.18, 0.41────────────────────────────

Therefore, in the juxtaposed sequence showing the correspondence of the amino acid sequence between the target protein X and the reference protein Y,
A correlation function Cp (i) indicating the correlation between the residue number i of the target protein X and the residue number j of the reference protein Y corresponding to the residue number i is defined by the following equation 5.

[0039]

(Equation 5)

Here, x _p (m) and y _p (n) are, respectively, the residue number m of the target protein X and the physicochemical parameter p for the amino acid residue at the residue number n with the reference protein Y. a value, also <c _p> is the average value of the parameter p in the 20 amino acids of e.g. Table 1,
For example, 5 is given as k. In addition, the value of the correlation function is set to 0 for a portion where there is no corresponding residue pair due to insertion or deletion.

The target protein X and the reference protein Y are converted into a sequence using a certain physicochemical parameter, and when k = 5, 11
The value of the correlation function Cp (i) is obtained using the above equation (5) in a correspondence relationship as shown in FIG. 5 with respect to the segment (fragment) of the residue. Note that the sequence sum Σ of Equation 5 is from l = −k to l = + k. For example, when k = 5, the correlation function Cp
Find the value of (i). (A) When i = 1 (N-terminal), as shown at 31 in FIG. 5, residue numbers i, i + 1,..., I + 5 of the target protein X
Respectively, the residue numbers j and j + of the reference protein Y
The value of the correlation function Cp (i) between six residues of 1,..., J + 5 is calculated. (B) When i = 2, as shown at 32 in FIG. 5, the residue numbers i−1, i, i + 1,..., I + 5 of the target protein X are respectively assigned to the residue numbers j− 1, j, j +
The value of the correlation coefficient Cp (i) between 7 residues of 1,..., J + 5 is calculated. (C) When i is the last residue (C-terminal), conversely, residue numbers i-5,.
−1 and i, respectively, the residue number j of the reference protein Y
Correlation coefficient Cp (i) with 6 residues of −5,..., J−1, j
Is calculated.

Further, in the present invention, noise is removed and S / N
In order to improve the ratio, the above calculated correlation function Cp
Regarding (i), an average value for a plurality of physicochemical parameters p (hereinafter, referred to as an average value of a correlation function) <C
(I)> is expressed by the following equation (6).

[0043]

(Equation 6)

Therefore, in the present invention, the amino acid sequence can be obtained by using the above equation (6) for the above six physicochemical parameters (in this case, n = 6) which reflects the three-dimensional structure and is as independent as possible, ie, uncorrelated. Is calculated, which indicates the degree of one-dimensional homology along. Hereinafter, an example in which the target protein is elastase and the reference protein is chymotrypsinogen will be described.

FIG. 6 is a stereo diagram showing the optimal superposition of the three-dimensional structure (solid line) of the target protein elastase predicted by the above-mentioned prediction method and the three-dimensional structure (dotted line) measured by X-ray crystallography. It is shown by. Also,
Figure 7 is based on Figure 6, the conformation of the target protein elastase, the position of the C _alpha atoms of the predicted the conformation, deviation from the C _alpha atoms of the three-dimensional structure by the corresponding X-ray crystallography, i.e. , And the distance between the positions of each _Cα atom (hereinafter simply referred to as deviation) l _ii is plotted against the residue number using the following _equation (7).

[0046]

## EQU7 ## l _ii = √ ｛(x ₁ −x ₂ ) ² + (y ₁ −y ₂ ) ² +
(Z ₁ −z ₂ ) ² ｝

Here, (x ₁ , y ₁ , z ₁ ) is the three-dimensional coordinates of the target protein, and (x ₂ , y ₂ , z ₂ ) is the three-dimensional coordinates of the reference protein. As is clear from FIG. 7, it is recognized that the value of the deviation l _ii is small in the region where the prediction accuracy is high, and is large in the region where the prediction accuracy is low.

FIG. 8 is a plot of the average <C (i)> of the correlation function calculated using Equation 6 with respect to the residue number. Compared to FIG. 7, the homology between elastase, which is the target protein, and chymotrypsinogen, which is the reference protein, is high, that is, in the region of the residue number having a high correlation value, the prediction accuracy of the three-dimensional structure is high, Although the value of l _ii is small, the homology is low, that is, in the region of the residue number having a low correlation value, the prediction accuracy of the three-dimensional structure is low, that is, the value of the deviation l _ii is large.
As described above, according to the definition of homology based on Equation 6, it is possible to estimate the local prediction accuracy of the three-dimensional structure predicted only from the amino acid sequence information.

In the above embodiment, the above-mentioned six physicochemical parameters are used for estimating the local prediction accuracy of the three-dimensional structure. As a result of searching for a parameter that increases the correlation value in the region of the residue number corresponding to the portion, the parameter was empirically found. As the physicochemical parameters used in the present invention, preferably, at least one of the above six physicochemical parameters may be used.

The present inventor has determined the following relationship besides the relationship between the deviation l _ii between elastase as the target protein and chymotrypsinogen as the reference protein and the average value <C (i)> of the correlation function. It has been confirmed that local prediction accuracy of the three-dimensional structure can be estimated between the target protein and the reference protein. (A) Hemoglobin alpha chain (hemoglob)
in α-chain) and the reference protein hemoglobin β-chain. (B) Target protein alpha-listic protease (α-lytic protease) and reference protein protease A
(Protease A). (C) The target protein endothiapepsiin
n) and the reference protein penicillopepsin
n).

Next, with reference to FIG. 1, the configuration of the prediction calculation and prediction accuracy calculation device of the three-dimensional structure of the target protein of the present embodiment will be described. As shown in FIG. 1, the following devices are connected via a bus to a microprocessing unit (hereinafter, referred to as an MPU) 10, which is an arithmetic and control unit that executes the processes in FIGS. (A) ROM 11: Stores the program data of FIGS. 2 and 3 and data necessary for executing the program. (B) RAM 12: A memory for a storage area serving as a work area for executing the processing of FIGS. (C) Keyboard 13: Input for inputting an instruction command for executing the processing of FIGS. 2 and 3 or the name of the target protein, an amino acid sequence, and a value of a random seed in the predicted coordinate calculation processing. Device. (D) CRT display 14: a display device that displays the contents of the processing of FIGS. 2 and 3 and the calculation results. (E) Printer 15: a printer for printing the calculation results of the processing in FIGS. (F) Calculation result file 20: This is a storage device that is constituted by, for example, a hard disk and stores data of the calculation results of the processing in FIGS. (G) Protein database (PDB) file 21: composed of, for example, a hard disk, and analyzed by known X-ray crystallography and NMR to analyze the spatial coordinates (x ,
y, z), which is a database file containing the results of the determination of the protein name, amino acid sequence (three-letter code display and one-letter code display), the number of residues, This is a storage device that stores spatial coordinates for each residue.

FIG. 2 is a flowchart showing the three-dimensional structure prediction and accuracy calculation processing of a protein, which is executed by the apparatus shown in FIG. Hereinafter, the process of predicting the three-dimensional structure of the protein and calculating the accuracy thereof will be described with reference to an example when the target protein is elastase and the reference protein is chymotrypsinogen. Note that Tables 8 to 24, which are printout results, are described at the end of the embodiment for the sake of editing the specification, and for each table whose data length exceeds one page, It is divided into tables.

In FIG. 2, first, in step S1, the name and the amino acid sequence of the target protein whose coordinates are to be predicted are input using the keyboard 13, and then the process proceeds to step S1.
In 2, a homology search process is executed. In the homology search process, based on the PDB, a reference protein having a sequence similar to the amino acid sequence of the target protein and having a known tertiary structure is searched, and the homology indicating the ratio of the corresponding residue pair to the whole is determined. Among the reference protein candidates having a degree of, for example, 30% or more, in this embodiment, the protein having the highest degree of homology is selected as the reference protein. Then
In step S3, as shown in Tables 8 to 11,
The following items are printed. (A) Table 8: Name and number of residues of target protein (b) Tables 9 to 11: For each of the above-mentioned selected reference protein candidates (b-1) Name of reference protein (b-2) Target protein and reference Corresponding residue number between protein (b-3) Degree of homology (%) (b-4) Correspondence of amino acid sequence between target protein and reference protein: where amino acids are one-letter code , The upper row is the target protein and the lower row is the reference protein,
An asterisk is added to every 10 residue numbers. In addition, “:” is added to a position where the amino acid of the target protein matches the amino acid of the reference protein, and “−” is added to a position where the amino acid is deleted.

In this example, [1] chymotrypsinogen A, [2] trypsinogen, and [3] chymotrypsinogen * A are searched as three reference protein candidates, and in this embodiment, these reference protein candidates are Chymotrypsinogen * A with the highest degree of homology
(Hereinafter simply referred to as chymotrypsinogen) is selected as the reference protein.

Next, in step S4, the amino acid sequence between the selected reference protein (chymotrypsinogen) and the target protein (elastase) is compared. A corresponding homologous residue pair without-) is cut out, and in step S5, as shown in Table 12, the number of the corresponding residue pair,
A correspondence table between the target protein and the residue number of the reference protein is printed. The corresponding residue pairs that are cut out are as follows, as apparent from Table 11. (A) Residue numbers 1 to 23 of the target protein are homologous to residue numbers 16 to 38 of the reference protein, and will be referred to as a residue pair RP1. (B) Residue numbers 28-53 of the target protein are homologous to residue numbers 40-65 of the reference protein, and will be referred to as a residue pair RP2.

Hereinafter, similarly, the residue pairs RP3 to RP8
Exist up to. These corresponding residue pairs RP1 to RP
8 is shown in FIG. In this embodiment, as shown in FIG. 4, the relative between the coordinates residue number i of a reference protein known 'of th C _alpha atom coordinates and residue number j' th C _alpha atom coordinates The distance is represented by the residue number i of the corresponding target protein.
The three-dimensional structure of the target protein is predicted by setting the relative distance <d _ij > between the coordinate of the C _α atom at the residue and the coordinate of the C _α atom at the residue number j. Further, in step S6,
The predicted coordinate calculation process shown in FIG. 3 is executed to obtain C _{α of the} target protein.
The coordinates of the atoms are calculated, the calculation result is displayed on the CRT display 14, and printed using the printer 15.

Next, in step S7, the above 20
The value of the correlation function Cp (i) between the target protein and the reference protein is calculated based on the value of the physicochemical parameter p peculiar to the type of amino acid using Equation 5 for one physicochemical parameter p. Then, the value of the correlation function Cp (i) between the target protein and the reference protein is calculated using Equation 5 for the other five physicochemical parameters p. Chemical parameter p
For the correlation function Cp between the target protein and the reference protein
After calculating the value of (i), the average value <C (i)> of the correlation function is calculated with respect to the residue number i of the target protein using Expression 6.

Further, in step S8, the average <C (i)> of the correlation function calculated in step S7.
Is displayed on the CRT display 14 in the format of FIG. 8, that is, the average value <C (i)> of the correlation function with respect to the residue number i of the target protein is displayed on the CRT display 14,
Printing is performed using the printer 15. From FIG. 8, it is possible to estimate the local prediction accuracy one-dimensionally along the residue number with respect to the coordinates of the predicted three-dimensional structure of the target protein based only on the amino acid sequence information. By using this method, the primary structures can be compared, and a portion having a high possibility that the three-dimensional structure is similar can be extracted. According to this method, it is possible to examine how long the value of the correlation function is, and how long the chain is expected to be in conformity with the three-dimensional structure.

FIG. 3 is a flowchart showing the predicted coordinate calculation processing (step S6) which is a subroutine of FIG. As shown in FIG. 3, first, in step S11, the residue pair RP1 cut out in step S4
RP8, the relative distance d _i ' _j ' between the corresponding homologous regions in the reference protein is calculated, set as the corresponding relative distance <d _ij > of the target protein, and It repeatedly executes for i and j under the condition of j. Next, in step S12, a seed IR of a random number is input. In step S13, a random number value (0 ≦ random number value <1) is generated based on the input seed IR of the random number, and the coordinates of the initial structure of the target protein are approximated to actual coordinates using the above equation (4). After multiplying the random number value by the radius of gyration Rg of the protein, doubling it to obtain the coordinates of the diameter, the coordinates q _lm of the initial structure of the target protein are calculated, and based on this, the coordinates of the initial structure are calculated. The relative distance d _ij is calculated, and this calculation is performed under the condition of i <j.
Is repeatedly executed. Then, in step S14, a function value F of Equation 1 described later in the initial structure is calculated.

Next, in step S15, for example, only the first-order partial derivative is calculated, but the function is minimized without calculating the second-order partial derivative. The function F is optimized so that the function value of the function F is minimized, and the function value F is calculated. Further, in step S16, based on the optimized function F, predicted coordinate variables x ₁ , y ₁ , z ₁ , x ₂ , y ₂ , z ₂ ,..., X _M of the target protein, which are variables of the function F , Y _M and z _M are calculated. The reason for calculating the function value F in the initial structure of the target protein and the function value F of the target protein after optimization in steps S14 and S15 is to know the degree of optimization. When the function value F of the target protein after optimization is sufficiently smaller than the function value F in the initial structure of the target protein, it can be determined that the degree of optimization is large.

Further, in step S17, the table 13
Then, the output result is printed in the standard output format shown in Table 14, and in step S18, the calculation results of the predicted coordinates of the target protein shown in Tables 15 to 24 are printed.

In the standard output format, the following information is printed as shown in Tables 13 and 14. (A) Name and amino acid sequence of target protein (indicated by one letter code) (b) Name and amino acid sequence of reference protein (indicated by three letter code) (c) Homologous residues corresponding between target protein and reference protein No. (d) Seed value of random number (e) Number of iterations of optimization calculation in optimization in step S15 (f) Function value F of initial structure of target protein calculated in step S14 (g) Calculation in step S15 Function value F of the predicted structure of the target protein after optimization

Further, in the calculation results of the predicted coordinates, the following information is printed as shown in Tables 15 to 24. (A) Name of target protein (b) The following data for each residue of target protein (b-1) Amino acid sequence (three-letter code display) (b-2) Residue number (b-3) Predicted x , Y, z coordinate values

As described above, according to this embodiment,
Based on the PDB, a protein having the highest or relatively high degree of homology is selected as a reference protein, the amino acid sequences between the selected reference protein and the target protein are compared, and a plurality of homologous residue pairs are cut out. Calculate the relative distance between atoms of each residue having a different residue number in the reference protein based on the plurality of cut out residue pairs, and use that as the corresponding relative distance <d _ij > of the target protein. After a random number is generated based on a predetermined random number seed IR, a value proportional to the random number value is set as a coordinate value of the initial structure of the target protein, and a relative distance d _ij of the initial structure is calculated. The function F is optimized so that the function value of the function F is minimized, and the predicted coordinates of the target protein, which is a variable of the function F, are calculated. Therefore, the calculation can be simplified as compared with the conventional method, and the coordinates of the target protein can be calculated with higher accuracy, that is, closer to the actual one, as compared with the conventional method such as the method of Wahu and Shelaga. Coordinates can be calculated.

Further, by calculating and displaying the average value <C (i)> of the correlation function for residue number i of the target protein using the above formulas 5 and 6, only the amino acid sequence information is used. In addition, local prediction accuracy can be estimated one-dimensionally along the residue numbers with respect to the coordinates of the predicted three-dimensional structure of the target protein. By using this method, the primary structures can be compared, and a portion having a high possibility that the three-dimensional structure is similar can be extracted. According to this method, it is possible to examine how long the value of the correlation function is, and how long the chain is expected to be in conformity with the three-dimensional structure.

[0066]

[Table 8] Target protein: elastase porcine ec 3.4.21.36 No. of residues: 240 Fragmentation from 1 to 240

[0067]

[Table 9] [1] (1-240) 240 elastase porcine ec 3.4.21.36 (16-226) 226 CHYMOTRYPSINOGEN A code name: 1CHG cor. Coef. = 0.415 (37.9%) + * + * + * + * + * + * + * VVGGTEAQRNSWPSQISLQYRSGSSWAHTCGGTLIRQNWVMTAAHCVDRELTFRVVVGEHNLNQNNGTEQ ::::::::::::::::::::::::::::::: IVNGEEAVPGSWPWQVSLQDKTG--F-HFCGVVVK + * + * + * + * + * + * YVGVQKIVVHPYWNTDDVAAGYDIALLRLAQSVTLNSYVQLGVLPRAGTILANNSPCYITGWGLTRTNGQ:::: :: :::: ::::: ::::: -VF-KNS-KY-NSLTINN--DITLLKLSTAASFSQTVSAGTTCLAGS -+ * + * + * + * + * + * + * LAQTLQQAYLPTVDYAICSSSSYWGSTVKNSMVCAGGNGVRSGCQGDSGGPLHCLVNGQYAVHGVTSFVS::: ::: ::::: ::: ::: ::::::::::::: LQ-- QAS-LPLLSNTNCKK--YWGTKIKDAMICAGASGV-SSCMGDSGGPLVCKKNGAWTLVGIVSWGS + * + * + * + * + * + * + * RLGCNVTRKPTVFTRVSAYISWINNVIASN::: :::::: S-TCS-TSTPGVYARVTALVNWVQA

[0068]

[Table 10] [2] (17-240) 240 elastase porcine ec 3.4.21.36 (10-222) 222 TRYPSINOGEN code name: 1TGN cor.coef. = 0.452 (35.7%) + * + * + * + * + * + * + * SLQYRSGSSWA-HTCGGTLIRQNWVMTAAHCVDRELTFRVVVGEHNLNQNNGTEQYVGVQKIVVHPYWNT:: ::: :: :: :::: ::::: ::: ::::: TVPYQVSLNSGYHFCGGSLINSQWVVSAAHCY * KS + GIVVRNSED + * DDVAAGYDIALLRLAQSVTLNSYVQLGVLPRAGTILANNSPCYITGWGLTRTNGQ-LAQTLQQAYLPTVD :::: :::: :::: ::::::: NTLNN - DIMLIKLKSAASLNSRVASISLP-TSCA-SAGTQCLISGWGNTKSSGTSYPDVLKCLKAPILS + * + * + * + * + * + * + * YAICSSSSYWGSTVKNSMVCAG-GNGVRSGCQGDSGGPLHCLVNGQYAVHGVTSFVSRLGCNVTRKPTVF: ::: :::: ::::::::::::: :: ::: DSSCKSA-Y-PGQITSNMFCAGYLEGGKDSCQGDSGGPVVC--SGK--LQGIVSWGS--GCAQKNKPGVY + * + * + * + * + * + * + * TRVSAYISWINNVIASN::: ::: :::: TKVCNYVSWIKQTIASN

[0069]

[Table 11] [3] (1-240) 240 elastase porcine ec 3.4.21.36 (16-245) 245 CHYMOTRYPSINOGEN * A code name: 2CGA cor. Coef. = 0.502 (39.0%) + * + * + * + * + * + * + * VVGGTEAQRNSWPSQISLQYRSGSSWAHTCGGTLIRQNWVMTAAHCVDRELTFRVVVGEHNLNQNNGTEQ :::::::::::::::::::::::::::: NGNGAV + * + * + * + * + * YVGVQKIVVHPYWNTDDVAAGYDIALLRLAQSVTLNSYVQLGVLPRAGTILANNSPCYITGWGLTR-TNG: :: :::: ::::: ::::::: :: KLKIAKVFKNS--KYNSLTINNDITLLKLSTAASFSQTV + ATS + * QLAQTLQQAYLPTVDYAICSSSSYWGSTVKNSMVCAGGNGVRSGCQGDSGGPLHCLVNGQYAVHGVTSFV :::::::::::::::::::::::::: NTPDRLQQASLPLLSNTNCKK--YWGTKIKDAMICAGASGSGVPLV-SSCMGDSGG + * + * SRLGCNVTRKPTVFTRVSAYISWINNVIASN::::: :::::: SS-TCS-TSTPGVYARVTALVNWVQQTLAAN

[0070]

[Table 12]

[0071]

[Table 13] Protein name: elastase porcine ec 3.4.21.36 Sequence: vvggteaqrn swpsqislqy rsgsswahtc ggtlirqnwv mtaahcvdre ltfrvvvgeh nlnqnngteq yvgvqkivvh pywntddvaa gydiallrla qsvtlnsyvq lgvlpragti lannspcyit gwgltrtngq laqtlqqayl ptvdyaicss ssywgstvkn smvcaggngv rsgcqgdsgg plhclvngqy avhgvtsfvs rlgcnvtrkp tvftrvsayi swinnviasn

[0072]

[Table 14] [Suplementary information] Reference protein: CHYMOTRYPSINOGEN * A 2CGA 4 Sequence: CYS GLY VAL PRO ALA ILE GLN PRO VAL LEU SER GLY LEU SER ARG ILE VAL ASN GLY GLU GLU ALA VAL PRO GLY SER TRP PRO TRP GLN VAL SER LEU GLN ASP LYS THR GLY PHE HIS PHE CYS GLY GLY SER LEU ILE ASN GLU ASN TRP VAL VAL THR ALA ALA HIS CYS GLY VAL THR THR SER ASP VAL VAL VAL ALA GLY GLU PHE ASP GLN GLY SER SER SER GLU LYS GLN LEU LYS ILE ALA LYS VAL PHE LYS ASN SER LYS TYR ASN SER LEU THR ILE ASN ASN ASP ILE THR LEU LEU LYS LEU SER THR ALA ALA SER PHE SER GLN THR VAL SER ALA VAL CYS LEU PRO SER ALA SER ASP PHE ALA GLY THR THR CYS VAL THR THR GLY TRP GLY LEU THR ARG TYR THR ASN ALA ASN THR PRO ASP ARG LEU GLN GLN ALA SER LEU PRO LEU LEU SER ASN THR ASN CYS LYS LYS TYR TRP GLY THR LYS ILE LYS ASP AMET ALA GLY ALA SER GLY VAL SER SER CYS MET GLY ASP SER GLY GLY PRO LEU VAL CYS LYS LYS ASN GLY ALA TRP THR LEU VAL GLY ILE VAL SER TRP GLY SER SER THR CYS SER THR SER THR PRO G LY VAL TYR ALA ARG VAL THR ALA LEU VAL ASN TRP VAL GLN GLN THR LEU ALA ALA ASN Residue pairs: 1-23: 16-38 | 28-53: 40-65 | 55-81: 66-92 | 84-136 : 93-145 | 137-160: 147-170 | 163-180: 171-188 | 182-211: 189-218 | 217-240: 222-245 | Random number generation: 2 Minop has exhausted 300calls of calcfg ** Initial conformation: F = 0.521134E + 07 ** Final conformation: F = 0.323426E + 04

[0073]

[Table 15] HEADER elastase porcine ec 3.4.21.36 COMPND ATOM 1 CA VAL 1 18.851 -2.532 20.987 ATOM 2 CA VAL 2 15.303 -2.327 19.641 ATOM 3 CA GLY 3 14.268 1.317 20.378 ATOM 4 CA GLY 4 16.678 2.285 23.284 ATOM 5 CA THR 5 15.808 2.957 26.986 ATOM 6 CA GLU 6 14.409 5.605 29.237 ATOM 7 CA ALA 7 16.874 8.289 30.295 ATOM 8 CA GLN 8 17.404 9.265 33.926 ATOM 9 CA ARG 9 15.443 12.444 34.524 ATOM 10 CA ASN 10 17.432 15.682 34.242 ATOM 11 CA SER 11 20.388 13.904 32.727 ATOM 12 CA TRP 12 20.195 15.463 29.240 ATOM 13 CA PRO 13 19.813 19.086 30.382 ATOM 14 CA SER 14 20.126 20.787 26.964 ATOM 15 CA GLN 15 17.127 18.902 25.531 ATOM 16 CA ILE 16 14.106 21.076 24.752 ATOM 17 CA SER 17 10.689 20.574 23.259 ATOM 18 CA LEU 18 9.093 22.827 20.699 ATOM 19 CA GLN 19 5.414 23.298 20.964 ATOM 20 CA TYR 20 2.907 25.510 19.275 ATOM 21 CA ARG 21 0.476 28.157 20.613 ATOM 22 CA SER 22 -2.005 25.426 21.582 ATOM 23 CA GLY 23 0.535 22.932 22.760 ATOM 24 CA SER 24 0.467 21.756 19.096 ATOM 25 CA SER 25 2.042 22.804 15. 793

[0074]

[Table 16] ATOM 26 CA TRP 26 3.969 19.750 14.634 ATOM 27 CA ALA 27 2.807 17.611 17.585 ATOM 28 CA HIS 28 4.789 18.927 20.563 ATOM 29 CA THR 29 6.194 19.127 17.020 ATOM 30 CA CYS 30 9.957 19.301 17.210 ATOM 31 CA GLY 31 12.964 19.010 19.458 ATOM 32 CA GLY 32 15.944 21.369 20.022 ATOM 33 CA THR 33 19.182 21.917 22.057 ATOM 34 CA LEU 34 20.207 24.855 24.311 ATOM 35 CA ILE 35 23.569 26.252 23.416 ATOM 36 CA ARG 36 23.218 28.969 26.009 ATOM 37 CA GLN 37 20.595 30.780 ATOM 38 CA ASN 38 19.501 32.806 24.995 ATOM 39 CA TRP 39 19.762 30.486 22.041 ATOM 40 CA VAL 40 18.339 27.129 20.770 ATOM 41 CA MET 41 19.501 25.071 17.786 ATOM 42 CA THR 42 16.988 23.035 15.819 ATOM 43 CA ALA 43 16.242 21.851 12.273 ATOM 44 CA ALA 44 15.217 24.288 9.484 ATOM 45 CA HIS 45 12.578 21.796 8.167 ATOM 46 CA CYS 46 10.856 22.238 11.555 ATOM 47 CA VAL 47 9.278 25.470 10.248 ATOM 48 CA ASP 48 9.424 27.290 13.625 ATOM 49 CA ARG 49 7.681 30.673 14.011 ATOM 50 CA GLU 50 7.681 33.203 16.785

[0075]

[Table 17] ATOM 51 CA LEU 51 4.274 31.867 17.943 ATOM 52 CA THR 52 5.811 28.550 18.988 ATOM 53 CA PHE 53 7.071 28.006 22.591 ATOM 54 CA ARG 54 8.397 24.718 21.106 ATOM 55 CA VAL 55 10.286 26.413 23.928 ATOM 56 CA VAL 56 9.932 24.014 26.848 ATOM 57 CA VAL 57 12.973 23.250 28.996 ATOM 58 CA GLY 58 13.598 20.851 31.890 ATOM 59 CA GLU 59 11.025 18.298 30.786 ATOM 60 CA HIS 60 11.118 14.540 31.293 ATOM 61 CA ASN 61 7.653 13.062 31.184 ATOM 62 CA LEU 62 5.258 14.966 29.043 ATOM 63 CA ASN 63 2.358 13.384 30.844 ATOM 64 CA GLN 64 3.344 15.002 34.151 ATOM 65 CA ASN 65 2.043 18.438 35.406 ATOM 66 CA ASN 66 4.203 18.063 38.430 ATOM 67 CA GLY 67 7.441 18.996 36.718 ATOM 68 CA THR 68 9.139 22.298 37.292 ATOM 69 CA GLU 69 9.719 23.150 33.610 ATOM 70 CA GLN 70 10.445 26.522 31.865 ATOM 71 CA TYR 71 8.236 27.651 29.000 ATOM 72 CA VAL 72 10.232 30.118 26.973 ATOM 73 CA GLY 73 9.090 32.602 24.329 ATOM 74 CA VAL 74 10.871 33.193 21.048 ATOM 75 CAGLN 75 12.145 36.637 19.897

[0076]

[Table 18] ATOM 76 CA LYS 76 13.419 35.909 16.359 ATOM 77 CA ILE 77 13.963 32.885 14.160 ATOM 78 CA VAL 78 17.058 32.573 11.944 ATOM 79 CA VAL 79 16.921 29.883 9.269 ATOM 80 CA HIS 80 20.294 29.135 7.395 ATOM 81 CA PRO 81 19.837 30.674 3.980 ATOM 82 CA TYR 82 16.953 28.351 4.957 ATOM 83 CA TRP 83 20.573 27.735 5.984 ATOM 84 CA ASN 84 21.686 27.864 2.292 ATOM 85 CA THR 85 18.733 25.451 3.256 ATOM 86 CA ASP 86 17.599 23.567 0.387 ATOM 87 CA ASP 87 14.123 22.580 0.848 ATOM 88 CA VAL 88 14.430 20.145 -2.002 ATOM 89 CA ALA 89 17.363 18.113 -0.848 ATOM 90 CA ALA 90 16.890 18.918 2.901 ATOM 91 CA GLY 91 20.567 19.922 2.948 ATOM 92 CA TYR 92 21.984 22.543 5.411 ATOM 93 CA ASP 93 19.056 21.731 7.820 ATOM 94 CA ILE 94 19.470 24.142 10.697 ATOM 95 CA ALA 95 17.722 27.016 12.452 ATOM 96 CA LEU 96 18.585 29.300 15.403 ATOM 97 CA LEU 97 16.026 30.577 17.739 ATOM 98 CA ARG 98 16.756 33.629 19.888 ATOM 99 CA LEU 99 14.620 33.510 23.091 ATOM 100 CA ALA 100 12.579 36.612 24.198

[0077]

[Table 19] ATOM 101 CA GLN 101 12.921 35.234 27.696 ATOM 102 CA SER 102 16.256 33.781 28.839 ATOM 103 CA VAL 103 16.598 30.229 30.219 ATOM 104 CA THR 104 17.429 29.900 33.841 ATOM 105 CA LEU 105 20.290 27.410 34.114 ATOM 106 CA ASN 106 19.975 24.821 36.900 ATOM 107 CA SER 107 20.655 21.250 37.796 ATOM 108 CA TYR 108 18.309 20.142 35.019 ATOM 109 CA VAL 109 18.669 22.991 32.486 ATOM 110 CA GLN 110 22.024 23.750 30.864 ATOM 111 CA LEU 111 23.957 23.834 27.556 ATOM 112 CA GLY 112 25.474 21. ATOM 113 CA VAL 113 28.934 22.164 23.644 ATOM 114 CA LEU 114 29.414 23.002 19.950 ATOM 115 CA PRO 115 32.334 21.340 17.893 ATOM 116 CA ARG 116 35.074 23.456 16.017 ATOM 117 CA ALA 117 34.676 23.489 12.202 ATOM 118 CA GLY 118 37.846 21.310 12.129 ATOM 119 CA THR 119 36.637 18.684 14.544 ATOM 120 CA ILE 120 36.853 15.237 13.204 ATOM 121 CA LEU 121 34.484 12.657 14.608 ATOM 122 CA ALA 122 35.483 9.192 13.197 ATOM 123 CA ASN 123 33.124 6.458 11.904 ATOM 124 CA ASN 124 32.277 3.845 14.454 ATO M 125 CA SER 125 32.462 6.308 17.289

[0078]

[Table 20] ATOM 126 CA PRO 126 29.443 5.603 19.607 ATOM 127 CA CYS 127 27.161 8.681 20.144 ATOM 128 CA TYR 128 23.801 9.364 21.797 ATOM 129 CA ILE 129 20.457 10.524 20.417 ATOM 130 CA THR 130 17.632 11.463 22.732 ATOM 131 CA GLY 131 13.996 12.472 22.429 ATOM 132 CA TRP 132 10.244 11.899 22.536 ATOM 133 CA GLY 133 10.010 10.720 18.922 ATOM 134 CA LEU 134 8.160 7.549 19.898 ATOM 135 CA THR 135 5.145 9.670 20.637 ATOM 136 CA ARG 136 4.723 10.340 16.882 ATOM 137 CA THR 137 -0.358 8.075 17.391 ATOM 138 CA ASN 138 -2.041 4.683 17.584 ATOM 139 CA GLY 139 0.642 3.241 19.679 ATOM 140 CA GLN 140 0.073 6.394 21.641 ATOM 141 CA LEU 141 3.334 5.402 23.381 ATOM 142 CA ALA 142 5.564 6.033 26.461 ATOM 143 CA GLN 143 5.945 9.802 26.947 ATOM 144 CA THR 144 9.266 9.810 28.849 ATOM 145 CA LEU 145 12.605 10.935 27.345 ATOM 146 CA GLN 146 14.358 7.949 25.590 ATOM 147 CA GLN 147 18.018 7.543 24.736 ATOM 148 CA ALA 148 19.880 5.281 22.299 ATOM 149 CA TYR 149 23.601 4.723 21.595 CA LEU 150 24.381 4.597 17.895

[0079]

[Table 21] ATOM 151 CA PRO 151 27.725 4.413 15.936 ATOM 152 CA THR 152 28.655 7.144 13.428 ATOM 153 CA VAL 153 28.940 6.048 9.799 ATOM 154 CA ASP 154 31.047 7.478 6.999 ATOM 155 CA TYR 155 28.950 9.044 4.203 ATOM 156 CA ALA 156 30.379 6.460 1.703 ATOM 157 CA ILE 157 28.888 3.802 3.705 ATOM 158 CA CYS 158 25.671 5.686 4.243 ATOM 159 CA SER 159 25.080 5.897 0.451 ATOM 160 CA SER 160 24.542 2.092 0.480 ATOM 161 CA SER 161 23.473 1.851 4.162 ATOM 162 CA SER 162 20.506 3.770 ATOM 163 CA TYR 163 21.189 2.940 2.015 ATOM 164 CA TRP 164 20.008 6.191 0.847 ATOM 165 CA GLY 165 22.043 6.686 -2.295 ATOM 166 CA SER 166 22.599 10.145 -3.735 ATOM 167 CA THR 167 20.332 11.720 -1.147 ATOM 168 CA VAL 168 23.467 11.834 1.098 ATOM 169 CA LYS 169 25.090 15.334 0.833 ATOM 170 CA ASN 170 28.351 16.462 2.426 ATOM 171 CA SER 171 26.602 18.580 5.035 ATOM 172 CA MET 172 24.920 15.473 6.321 ATOM 173 CA VAL 173 26.021 12.759 8.769 ATOM 174 CA CYS 174 24.491 9.298 9.375 175 CA ALA 175 24.378 7.183 12 .460

[0080]

[Table 22] ATOM 176 CA GLY 176 22.974 3.824 13.593 ATOM 177 CA GLY 177 21.787 0.767 11.650 ATOM 178 CA ASN 178 23.430 -1.515 14.182 ATOM 179 CA GLY 179 20.439 -3.047 15.884 ATOM 180 CA VAL 180 18.859 -0.111 17.378 ATOM 181 CA ARG 181 19.056 3.286 18.939 ATOM 182 CA SER 182 16.589 2.327 16.140 ATOM 183 CA GLY 183 16.885 6.146 15.520 ATOM 184 CA CYS 184 13.181 7.048 15.645 ATOM 185 CA GLN 185 13.352 8.263 19.312 ATOM 186 CA GLY 186 15.002 11.524 18.037 ATOM 187 CA ASP 187 12.469 14.253 17.449 ATOM 188 CA SER 188 13.025 16.375 14.353 ATOM 189 CA GLY 189 15.565 18.979 15.251 ATOM 190 CA GLY 190 16.815 17.051 18.336 ATOM 191 CA PRO 191 20.443 16.249 19.282 ATOM 192 CA LEU 192 23.005 13.724 18.326 ATOM 193 CA HIS 193 25.791 14.108 20.861 ATOM 194 CA CYS 194 29.203 12.596 21.186 ATOM 195 CA LEU 195 31.753 13.081 23.894 ATOM 196 CA VAL 196 34.846 15.124 23.079 ATOM 197 CA ASN 197 37.452 15.597 25.869 ATOM 198 CA GLY 198 34.806 14.392 28.396 ATOM 199 32. 20 0 CA TYR 200 29.016 16.171 25.402

[0081]

[Table 23] ATOM 201 CA ALA 201 29.237 17.882 22.017 ATOM 202 CA VAL 202 26.477 18.407 19.386 ATOM 203 CA HIS 203 27.599 16.495 16.297 ATOM 204 CA GLY 204 24.149 15.989 14.545 ATOM 205 CA VAL 205 20.649 17.525 14.312 ATOM 206 CA THR 206 17.833 14.988 13.502 ATOM 207 CA SER 207 17.004 15.670 9.868 ATOM 208 CA PHE 208 15.361 12.670 8.212 ATOM 209 CA VAL 209 15.198 8.921 7.541 ATOM 210 CA SER 210 12.840 6.012 7.980 ATOM 211 CA ARG 211 9.065 6.721 8.446 ATOM 212 CA LEU 212 9.207 10.363 9.487 ATOM 213 CA GLY 213 10.513 8.176 12.286 ATOM 214 CA CYS 214 14.299 8.540 12.751 ATOM 215 CA ASN 215 14.022 5.057 14.226 ATOM 216 CA VAL 216 16.089 2.589 12.218 ATOM 217 CA THR 217 16.970 -0.988 13.315 ATOM 218 CA ARG 218 18.852 -1.719 10.187 ATOM 219 CA LYS 219 18.625 1.675 8.439 ATOM 220 CA PRO 220 20.773 4.709 9.586 ATOM 221 CA THR 221 19.186 7.973 10.532 ATOM 222 CA VAL 222 20.240 11.136 8.710 ATOM 223 CA PHE 223 21.377 14.250 10.543 ATOM 224 CA THR 224 22.588 17.709 9.840 ATOM 225 CA ARG 225 26 .434 17.861 10.202

[0082]

[Table 24] ATOM 226 CA VAL 226 26.987 20.558 12.851 ATOM 227 CA SER 227 30.732 20.894 12.035 ATOM 228 CA ALA 228 29.767 22.035 8.475 ATOM 229 CA TYR 229 27.315 24.546 9.917 ATOM 230 CA ILE 230 29.176 25.852 12.929 ATOM 231 CA SERSER 28.924 11.237 ATOM 232 CA TRP 232 27.267 30.374 10.371 ATOM 233 CA ILE 233 26.138 29.758 14.035 ATOM 234 CA ASN 234 29.038 31.790 15.523 ATOM 235 CA ASN 235 28.410 34.403 12.987 ATOM 236 CA VAL 236 24.670 34.745 13.693 37.37 37.37 37.3737 ATOM 238 CA ALA 238 28.014 37.384 17.237 ATOM 239 CA SER 239 25.674 39.775 15.407 ATOM 240 CA ASN 240 22.208 39.280 16.899 END

In the above embodiments, the coordinates of the α-carbon atom of the target protein are predicted and calculated. However, the present invention is not limited to this, and similarly, the amino acid residues of all the atoms constituting the protein are similarly calculated. It may be configured to predict and calculate the coordinates of the atoms.

[0084]

As described in detail above, according to the present invention, at least one protein having a high degree of homology with a target protein whose tertiary structure is to be predicted, among proteins whose atomic coordinates of amino acid residues are known. A protein is selected as a reference protein, and the coordinates of the atoms of the amino acid residues of the target protein are predicted by a predetermined method based on the coordinates of the atoms of the amino acid residues of the selected reference protein. A prediction accuracy calculation method and apparatus of the three-dimensional structure of a protein to calculate the accuracy, based on physicochemical parameters of a plurality of types of amino acids contained in the target protein, in the juxtaposed sequence of the reference protein and the target protein, Both the residue number of the reference protein and the residue number of the target protein are set so that the amino acid residues of the reference protein correspond to the amino acid residues of the target protein. The values of the correlation function indicating the correlation with each corresponding amino acid residue are calculated for at least one selected physicochemical parameter pre-selected from the physicochemical parameters of the plurality of amino acids, The average value of the correlation function calculated by averaging the calculated correlation function values is used as the prediction accuracy of the predicted three-dimensional structure of the target protein. Therefore, according to the present invention, it is possible to estimate the local prediction accuracy one-dimensionally along the residue number with respect to the coordinates of the predicted three-dimensional structure of the target protein based only on the amino acid sequence information. By using this method, the primary structures can be compared, and a portion having a high possibility that the three-dimensional structure is similar can be extracted. According to this method, it is possible to examine how long the value of the correlation function is, and how long the chain is expected to be in conformity with the three-dimensional structure.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a prediction calculation and prediction accuracy calculation device for a three-dimensional structure of a protein according to one embodiment of the present invention.

FIG. 2 is a flowchart showing a three-dimensional structure prediction and accuracy calculation process of a protein performed by the apparatus of FIG. 1;

FIG. 3 is a flowchart illustrating a predicted coordinate calculation process as a subroutine of FIG. 2;

FIG. 4 is a view showing a pair of homologous residues when the target protein is elastase and the reference protein is chymotrypsinogen, and the relative distance <d _ij > according to the present invention.

FIG. 5 shows a correlation function Cp (i) between residue number i of target protein X and residue number j of reference protein Y used in the present invention.
It is a figure for demonstrating Formula 5 which shows this.

6 is a stereo view showing a three-dimensional structure (solid line) of the target protein elastase predicted by the apparatus of FIG. 1 and a three-dimensional structure (broken line) of the measurement result by X-ray crystallography.

For 7 residues number of target proteins elastase, a graph showing the deviation l _ii between the three-dimensional structure of the target protein elastase predicted in the apparatus of FIG 1 and the conformation of the measurement result of the X-ray crystallography is there.

8 is a graph showing an average <C (i)> of a correlation function Cp (i) indicating a prediction accuracy coefficient calculated by the apparatus of FIG. 1 with respect to a residue number of a target protein elastase.

[Explanation of symbols]

 Reference numeral 10: Microprocessing unit (MPU), 11: ROM, 12: RAM, 13: Keyboard, 14: CRT display, 15: Printer, 20: Calculation result file, 21: Protein database (PDB) file, RP1 to RP8 ... Residue pairs to be used.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁶ , DB name) C06F 17/30 C07K 1/00 JICST file (JOIS)

Claims (6)

(57) [Claims] At least one protein having a high degree of homology with a target protein whose tertiary structure is to be predicted is selected as a reference protein from among proteins whose coordinates of atoms of amino acid residues are known, and the selected reference protein is selected. Calculating the prediction accuracy of the three-dimensional structure of the target protein, in which the coordinates of the atoms of the amino acid residues of the target protein are predicted by a predetermined method based on the coordinates of the atoms of the amino acid residues of the protein The method, wherein, based on physicochemical parameters of a plurality of types of amino acids contained in the target protein, an amino acid residue of the reference protein and an amino acid residue of the target protein in the juxtaposed sequence of the reference protein and the target protein. By changing both the residue number of the reference protein and the residue number of the target protein so that the groups correspond to each other, The value of the correlation function indicating the correlation is calculated for each of at least one selected physicochemical parameter previously selected from the physicochemical parameters of the plurality of types of amino acids, and the calculated correlation function values are calculated and averaged. A method for calculating the prediction accuracy of the three-dimensional structure of a protein, comprising using the average value of the obtained correlation function as the prediction accuracy of the predicted three-dimensional structure of the target protein. 2. The method according to claim 1, wherein the atom of the amino acid residue is an α-carbon atom. 3. The selected physicochemical parameters include: a degree of polarity, a specific volume, a degree of turn formation, a pK value of an α amino group, a pK value of an α carboxyl group, and a mutation degree. The method according to claim 1, wherein the method includes at least one of the following. 4. At least one protein having a high degree of homology with a target protein whose tertiary structure is to be predicted is selected from among proteins whose coordinates of atoms of amino acid residues are known, and the selected reference protein is selected. Calculating the prediction accuracy of the three-dimensional structure of the target protein, in which the coordinates of the atoms of the amino acid residues of the target protein are predicted by a predetermined method based on the coordinates of the atoms of the amino acid residues of the protein An apparatus, wherein, based on physicochemical parameters of a plurality of types of amino acids contained in the target protein, an amino acid residue of the reference protein and an amino acid residue of the target protein in the juxtaposed sequence of the reference protein and the target protein. By changing both the residue number of the reference protein and the residue number of the target protein so that the groups correspond to each other, First calculating means for calculating the value of the correlation function indicating the correlation for at least one selected physicochemical parameter selected in advance from the physicochemical parameters of the plurality of amino acids; and the first calculating means. A second calculating means for averaging the value of the correlation function calculated by the second calculation means to calculate an average value of the correlation function, and calculating the average value of the correlation function calculated by the second calculation means. An apparatus for predicting the three-dimensional structure of a protein, which is used as the accuracy of predicting the three-dimensional structure of the target protein. 5. The apparatus according to claim 4, wherein the atom of the amino acid residue is an α-carbon atom. 6. The selected physicochemical parameters include a polarity, a partial specificity, a turn formation, a pK value of an α amino group, a pK value of an α carboxyl group, and a mutation degree. The apparatus according to claim 4 or 5, wherein the apparatus further comprises at least one protein.