KR100836166B1 - Apparatus for prediction of tertiary structure from the protein amino acid sequences and prediction method thereof - Google Patents

Abstract

The present invention relates to an apparatus for predicting tertiary structure from an amino acid sequence of a protein and a method for predicting the same, and specifically, an input means; Means for preprocessing an unknown protein sequence inputted from said input means; Template protein database; Pair comparison means; Protein similarity network database; Global comparison means for comparing an unknown protein and a template protein globally based on the protein similarity network database; Mold selection means; Modeling means; Verification means for verifying the predicted protein structure; And an apparatus for predicting tertiary structure of an unknown protein comprising an output means and a prediction method thereof. The present invention predicts the tertiary structure of proteins more accurately and precisely than conventional protein prediction methods, thereby reducing the cost and time spent experimentally revealing the protein structure, which is useful in the field of research for treating various diseases. Can be used.

Protein, tertiary structure, prediction method, protein similarity network database

Description

Apparatus for prediction of tertiary structure from the protein amino acid sequences and prediction method

1 is a block diagram of an apparatus for predicting the structure of an unknown protein according to one embodiment of the present invention.

2 is a block diagram of a template protein database according to one embodiment of the invention.

3 is a block diagram of a protein similarity network database, according to one embodiment of the invention.

4 is a flowchart for explaining a method for predicting the structure of an unknown protein according to one embodiment of the present invention.

5 is a graph comparing the performance of conventional protein tertiary structure prediction programs.

6 is a graph comparing the performance of the conventional protein tertiary structure prediction program and the prediction device according to the present invention.

7 is a diagram comparing the actual structure and the protein structure predicted according to the conventional method.

8 is a diagram comparing the actual structure and the protein structure predicted according to the method of the present invention.

The present invention relates to an apparatus for predicting tertiary structure from an amino acid sequence of a protein and a method of predicting the same.

Proteins are the foundation of life phenomena and cause and regulate life phenomena. The function of these proteins is determined by their tertiary structure. Therefore, by analyzing the sequence of the protein to find the tertiary structure and predict the function of the protein is the key to understand the life phenomenon, and to treat various diseases caused by the abnormal function of the protein. Therefore, it is very important to identify the tertiary structure of these proteins.

There are two ways to determine the structure of a protein: experimentally revealing the structure and predicting it computationally. Experimental methods using X-ray crystallography or nuclear magnetic resonance (NMR) can reveal the exact protein structure, but it is difficult to apply at the genome scale because it is time and costly. The computational method is to predict tertiary structure from protein amino acid sequence, and to predict structure based on template protein with similar expected structure and ab initio method to predict structure based on protein's physicochemical properties. (Baker, D. and Sali, A. 2001, Science, 294, 93-96). In particular, the template-based prediction method is applicable to a relatively long protein sequence and can be applied on a genome scale because of its wide range of application [McGuffin, LJ, et. al., 2004, Nucleic Acids Res., 2004, D196-D199].

In the method of modeling a structure based on a template, the process of selecting a template protein having the structure most similar to an unknown protein has a great influence on the overall performance.

In general, selecting a template with more than 30% identical sequence predicts a relatively accurate structure, and if a template protein that meets the above criteria cannot be found, the physicochemical properties of the amino acids or the profile of the protein sequence may be additionally used. Altschul, SF, et. al. 1997, Nucleic Acids Res., 25, 3389-3402; Sadreyev, R. and Grishin, N. 2003, J. Mol. Biol., 2003, 317-336.

Many of the conventional protein structure prediction programs used sequence alignment mainly to find the template structure needed to predict the unknown protein structure.

Generally, a tool called BLAST is used as a tool for searching for protein similarity. The means is a program that finds the template structure by comparing the input sequence with known DNA and amino acid sequences. The BLAST first creates a list of character blocks that record scores above a threshold when aligned with a query sequence. In the next step, local alignment is performed by using a table calculated in advance to increase the similar region laterally until there is no similarity between the query sequence and the sequence database. Since the BLAST is sorted in the order of High-scoring Segment Pairs (HSP), which is a local score with high scores, the BLAST is not sorted in order from the top to the end of the gene.

Position-Specific Iterated BLAST (PSI-BLAST), which utilizes specific domains or motifs for pairwise sequence alignment, features local multiple alignments that start with a single sequence and allow gaps. It is widely used together with BLAST.

Korean Patent Laid-Open Publication No. 2005-0064644 discloses a method for predicting the structure of an unknown protein having an amino acid sequence, the method comprising: (a) comparing the sequence with the sequence information of the unknown protein based on a database storing sequence information of the known proteins; Determining a template protein candidate according to the degree of similarity of; (b) determining which group of properties the unknown protein belongs to based on a database in which the property information of known proteins is stored; And it discloses a method and apparatus for predicting the structure of an unknown protein comprising the step of predicting the structure of the unknown protein based on the results of step (a) and (b).

However, since these methods rely only on pairewise comparisons between the unknown and each template protein, they rely only on local information in the region of protein structure, resulting in structure or function failures when sequence similarity is less than 30%. There is a problem that does not effectively find a similar homolog.

In order to solve this problem, a method of introducing a protein network has been developed. Currently, a method of introducing a protein network is Rank Prop. The Rank Prop is based on the existing PSI-BLAST algorithm, and is configured by weighting the sequence similarity between template proteins in a database. However, even in such a case, there is a problem in that a protein having a similar structure or function cannot be effectively found because it depends only on the information of the sequence.

Accordingly, the present inventors have been studying to find a protein having similar structure or function even when the sequence similarity is less than 30%, and using the protein similarity network database after pair comparison between an unknown protein and a template protein, Comparing the structure and function can be found even when the sequence similarity is low, the present invention was completed.

It is an object of the present invention to provide a device for predicting tertiary structure of an unknown protein.

 It is an object of the present invention to provide a method for predicting tertiary structure of an unknown protein.

In order to achieve the above object, the present invention provides an apparatus for predicting tertiary structure from the amino acid sequence of a protein.

The present invention also provides a method for predicting tertiary structure from the amino acid sequence of a protein.

Hereinafter, the present invention will be described in detail.

The present invention

Input means;

Sequence preprocessing means for preprocessing the unknown protein sequence inputted from the input means;

A template protein database in which information of the template protein is stored;

Pair comparison means for comparing the unknown protein with the template protein based on the template protein database;

A protein similarity network database composed and stored of structural similarities between the template proteins;

Global comparison means for comparing an unknown protein and a template protein globally based on the protein similarity network database;

Template selection means for selecting one or more template proteins that are similar in structure to an unknown protein based on the results of the pair comparison means and the global comparison means;

Modeling means for predicting the steric structure of the unknown protein based on the result of the template selection means;

Verification means for verifying the predicted protein structure; And

It provides an apparatus for predicting tertiary structure of an unknown protein comprising an output means for outputting the predicted protein structure.

Hereinafter, the present invention will be described in more detail with reference to FIG. 1 .

1 is a block diagram of an apparatus for predicting the structure of an unknown protein according to an embodiment of the present invention.

As shown in FIG . 1 , an apparatus for predicting the structure of an unknown protein according to an embodiment of the present invention includes an input means 10; a sequence preprocessing means 100; Template protein database 102; Pair comparison means (101) for performing pair comparisons between individual template proteins and unknown proteins in the template protein database (102); Protein similarity network database 104; Global comparison means (103) for performing a global comparison using said protein similarity network database (104); Template selection means 105 for selecting a template protein in consideration of the pair comparison score and the global comparison score; Modeling means 106 for constructing a structural model of the unknown protein using sequence alignment and modeling tools; Model verification means 107 for verifying that the generated model is indeed suitable; And an output means 20 or the like.

The sequence data of the unknown protein can be input to the sequence input window of the protein tertiary structure predicting device according to the present invention by the conventional input means 10. A keyboard is typically used for the input means 10, and in the case of obtaining data of a protein sequence using a file or the Internet, it is possible to input using only a mouse without keyboard manipulation. In addition to the keyboard and / or mouse, tablets, trackballs, electronic pens, scanners, and the like may be used. The sequence data input as described above may be stored in a file form, so that the file may be loaded and operated later.

The sequence pretreatment means 100 extracts and preprocesses information such as secondary structure, solvent exposure, profile, etc. from an unknown protein sequence inputted using its own program or an external program. At this time, Psipred et al. Program can be used to extract the information of the secondary structure from the sequence of the unknown protein [Jones, DT (1999) Protein secondary structure prediction based on position-specific scoring matrices, J Mol Biol , 292 , 195-202], and artificial neural networks can be used to predict solvent exposure [Ahmad, S. and Gromiha, MM (2002) NETASA: neural network based prediction of solvent accessibility, Bioinformatics , 18 , 819-824], a program such as PSI-BLAST can be used for profile generation [Altschul, SF, Madden, TL, Schaffer, AA, Zhang, J., Zhang, Z., Miller, W. and Lipman, DJ ( 1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Nucleic Acids Res , 25 , 3389-3402.

The template protein database 102 is a space for storing data related to the template protein, and the data that can be stored in the template protein database includes the protein sequence 200, the actual structure of the protein 201, and the sequence of the protein. Structure data 202, profile 203, and the like, which are predicted from (see FIG. 2 ). The template protein database 102 is not limited, and for example, a Protein Database (PDB), a Structural Classification of Proteins (SCOP), or the like may be used [Murzin, AG, Brenner, SE, etc.]. , Hubbard, T. and Chothia, C. (1995) SCOP: a structural classification of proteins database for the investigation of sequences and structures, J Mol Biol , 247 , 536-540. The SCOP is one of the commonly used tertiary databases. It is composed of three stages, fold, super family, and family, depending on the similarity and range of proteins. Generally, the same fold or superfamily is the same. Inner protein is used.

The pair comparison means 101 numerically represents the similarity between the unknown protein and the individual proteins in the template protein database based on the sequence alignment, and the pair comparison means used include BLAST [Altschul, SF, Gish, W., Miller, W., Myers, EW and Lipman, DJ (1990) Basic local alignment search tool, J Mol Biol , 215, 403-410], PSI-BLAST [Altschul, SF, Madden, TL, Schaffer, AA, Zhang, J., Zhang, Z., Miller, W. and Lipman, DJ (1997) Gapped BLAST and PSI -BLAST: a new generation of protein database search programs, Nucleic Acids Res , 25, 3389-3402], alignment based on dynamic programming [Needleman, SB and Wunsch, CD (1970) A general method applicable to the search for similarities in the amino acid sequence of two proteins, J. Mol . Biol . , 48, 443-453; Smith, TF and Waterman, MS (1981) Identification of common molecular subsequences, J. Mol . Biol . , 147, 195-197], profile based alignment [Wallner, B., Fang, H., Ohlson, T., Frey-Skott, J., and Elofsson, A. (2004) Using evolutionary information for the query and target improves fold recognition, Proteins , 54, 342-350], threading method [Shi, J., Blundell, TL, and Mizuguchi, K. (2001) J. Mol. Biol., 310, 243-257., Sadreyev, R. and Grishin, N. (2003) COMPASS: A tool for comparison of multiple protein alignments with assessment of statistical significance, J. Mol. Biol., 326, 317-336, and the like, but is not limited thereto.

The protein similarity network database 104 organizes the proteins in the template protein database 102 into nodes, and at least one of the structures and functions between the template proteins so that they can be neighbored if the structures or functions between each node are similar. It is preferable to be connected by a link based on one (see FIG. 3 ).

The global comparison means 103 performs a global comparison by using the protein similarity network database 104, and specifically, a weighted sum of pair comparison scores of neighboring proteins in the protein similarity network database 104 according to the similarity. To calculate the global comparison score. At this time, the weight between neighboring proteins is based on the similarity of the structure or function of the two proteins, the higher the weight, the higher the weight is preferable.

For example, if a global comparison score is calculated using a Z-score (Z _{P , D} ) representing the similarity of the structure between P and D as a weight between template proteins P and D on a protein network, the global comparison score is It can be expressed by the following equation (1).

Where

는 미지 단백질 Q와 주형 단백질 P 사이의 전역 비교 점수이고,

Is the global comparison score between unknown protein Q and template protein P,

는 미지 단백질 Q와 주형 단백질 P 사이의 짝 비교 점수이고,

Is the pair comparison score between unknown protein Q and template protein P,

는 미지 단백질 Q와 주형 단백질 D 사이의 짝 비교 점수이고,

,

는 단백질 네트워크상에서 주형 단백질 P와 D 사이의 가중치를 의미하고,

는 구조의 유사성을 나타내는 Z-스코어로서, 단백질의 구조비교 원점수로부터 얻어지는 통계적인 수치를 의미한다.

Is the pair comparison score between unknown protein Q and template protein D,

,

Is the weight between template proteins P and D on the protein network,

Is a Z-score representing the similarity of the structures, and refers to a statistical value obtained from the structural comparison raw score of the protein.

delete

The template selection means 105 is a means for selecting a template protein most similar to an unknown protein, wherein the template protein preferably selects the one with the largest calculated global comparison score.

The modeling means 106 is a means for generating a structural model by predicting the steric structure of the unknown protein using a sequence alignment method and a modeling tool based on the template protein selected based on the result of the template selection means. At this time, dynamic alignment or threading may be used as the sequence alignment method, and a program such as a modeler or a SWISS-MODEL may be used as a modeling tool. [Sali, A. and Blundell, TL (1993) Comparative protein modeling by satisfaction of spatial restraints, J Mol Biol , 234, 779-815; Arnold, K., Bordoli, L., Kopp, J., and Schwede, T. (2006) The SWISS-MODEL workspace: a web-based environment for protein structure modeling, Bioinformatics , 22, 195-201].

The model verification means 107 is a means for checking whether the model predicted using the verification tool is actually a structure that can exist stably. Available verification tools include Procheck, Whatcheck, etc. [Laskowski, RA, MacArthur, MW, Moss, DS, and Thornton, JM (1993) PROCHECK: a program to check the stereochemical quality of protein structures, J. Appl . Cryst. , 26, 283-291; Hooft, RWW, Vriend, G., Sander, C., and Abola, EE (1996) Errors in protein structures, Nature , 381, 272-272].

The verified protein tertiary structure model is output through the output means 20 of the computer. The most basic output form is output to the screen, and the protein tertiary structure can be output in real time at the same time as the work, and the output result can be printed through a printer, and the file format representing various graphic format files or molecules. Can be stored as

The graphic format file is not particularly limited thereto, but a bitmap format, an image format file, or a vector format file may be applied. The bitmap or image format file is not particularly limited thereto, but JPEG, JPG, GIF, It is preferably selected from the group consisting of TIF, PIC, and BMP, and the file in the vector format is not particularly limited thereto, but is preferably selected from the group consisting of CAD, WMF, DWG, CDR, and AI.

The file format representing the molecule may be any file format used to store the structure of the molecule. The molecular structure file format is not particularly limited, but it is preferable to select a format with PDB or mmCIF.

In addition, the present invention

(a) extracting and preprocessing information from the sequence of the unknown protein input through the input means through sequence preprocessing means (step 1);

(b) pairwise comparing each template protein and the unknown protein stored in the template protein database via pair comparison means to calculate similarity between each protein (step 2);

(c) calculating global similarity of the unknown protein and the template protein on the protein similarity network database through global comparison means (step 3);

(d) selecting a template protein candidate based on the results of steps 2 and 3 through template selection means (step 4);

(e) modeling the structure of the unknown protein based on the template protein selected by the modeling means (step 5); And

and (f) verifying the predicted protein structure through verification means and then outputting the predicted protein structure through output means (step 6).

Hereinafter, the present invention will be described in detail with reference to FIG. 4 .

4 is a flow chart of a method for predicting the structure of an unknown protein according to one embodiment of the invention.

As shown in FIG . 4 , step 1 400 is a step of extracting and preprocessing information from an input unknown protein sequence when an unknown protein sequence is input through an input means. At this time, the information includes secondary structure, solvent exposure degree, profile, and the like, and a program such as Psipred can be used to extract secondary structure information from the sequence of the unknown protein [Jones, DT (1999) Protein secondary structure prediction based on position-specific scoring matrices, J Mol Biol , 292, 195-202], and using neural networks to predict solvent exposure [Ahmad, S. and Gromiha, MM (2002) NETASA: neural network based prediction of solvent accessibility, Bioinformatics , 18, 819-824], and programs such as PSI-BLAST can be used for profile generation [Altschul, SF, Madden, TL, Schaffer, AA, Zhang, J., Zhang, Z., Miller, W. and Lipman, DJ (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Nucleic Acids Res , 25, 3389-3402.

Next, step 2 401 calculates the similarity between each protein by pairing each template protein and the unknown protein stored in the template protein database 407 through pair comparison means by inputting the information obtained through the preprocessing. to be.

In this step, the similarity between the unknown protein and the individual proteins in the template protein database is numerically expressed, and the similarity calculation between the template protein and the unknown protein is preferably performed based on at least one of the sequence and the profile of the unknown protein. . Examples of the protein similarity calculation method may include, but are not limited to, BLAST, PSI-BLAST, alignment based on dynamic programming, alignment based on profile, threading method, and the like. [Altschul, SF, Gish, W., Miller, W., Myers, EW and Lipman, DJ (1990) Basic local alignment search tool, J Mol Biol , 215 , 403-410., Altschul, SF, Madden, TL, Schaffer, AA, Zhang, J., Zhang, Z., Miller, W. and Lipman, DJ (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Nucleic Acids Res , 25 , 3389-3402; Needleman, SB and Wunsch, CD (1970) A general method applicable to the search for similarities in the amino acid sequence of two proteins, J. Mol . Biol . , 48 , 443-453; Smith, TF and Waterman, MS (1981) Identification of common molecular subsequences, J. Mol . Biol . , 147 , 195-197; Wallner, B., Fang, H., Ohlson, T., Frey-Skott, J., and Elofsson, A. (2004) Using evolutionary information for the query and target improves fold recognition, Proteins , 54 , 342-350. , Shi, J., Blundell, TL, and Mizuguchi, K. (2001) J. Mol . Biol . , 310 , 243-257; Sadreyev, R. and Grishin, N. (2003) COMPASS: A tool for comparison of multiple protein alignments with assessment of statistical significance, J. Mol . Biol . , 326 , 317-336.

Step 3 402 then calculates the global similarity of the unknown protein and the template protein on the protein similarity network database 408 via global comparison means.

In this step, a global comparison is performed using the protein similarity network database 408, and specifically, the template protein and the unknown protein are determined according to the similarity between neighboring proteins along the shortest link connected on the protein similarity network database 408. The global comparison score is calculated by taking the weighted sum of the pair comparison results. At this time, the weight between neighboring proteins is based on the similarity of the structure or function of the two proteins, the higher the weight, the higher the weight is preferable.

Step 4 403 is followed by selecting a template protein candidate based on the results of steps 2 and 3 through template selection means.

In this step, the template protein is selected in consideration of the pair comparison similarity and the global comparison similarity, and preferably a template protein having a high global comparison similarity value is selected.

Next, step 5 (404) is a step of modeling the structure of the unknown protein based on the template protein selected through the modeling means.

In this step, a structural model is generated by predicting the conformation of the unknown protein using a sequence alignment method and a modeling tool using the selected template protein as a template. At this time, dynamic alignment or threading may be used as the sequence alignment method, and a program such as a modeler or a SWISS-MODEL may be used as a modeling tool. [Sali, A. and Blundell, TL (1993) Comparative protein modeling by satisfaction of spatial restraints, J Mol Biol , 234, 779-815; Arnold, K., Bordoli, L., Kopp, J., and Schwede, T. (2006) The SWISS-MODEL workspace: a web-based environment for protein structure modeling, Bioinformatics , 22, 195-201].

Next, step 6 405 is a step of verifying the predicted protein structure through the verification means and outputting the predicted protein structure through the output means.

In this step, a verification tool is used to check whether the predicted model is actually a structure that can exist stably. At this time, the available verification tools may be Procheck, Whatcheck, etc. [Laskowski, RA, MacArthur, MW, Moss, DS, and Thornton, JM (1993) PROCHECK: a program to check the stereochemical quality of protein structures, J. Appl. Cryst. , 26 , 283-291; Hooft, RWW, Vriend, G., Sander, C., and Abola, EE (1996) Errors in protein structures, Nature , 381 , 272-272]. The final result of passing the verification process may be output through the output means to predict the tertiary structure of the unknown protein (406).

The present invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium may include all kinds of recording devices that store data that can be read by a computer system. For example, ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage. A device or the like may be used, and may be implemented in the form of a carrier wave (for example, transmission through the Internet). The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

Hereinafter, the present invention will be described in more detail with reference to Examples. However, the following examples are illustrative of the present invention, and the content of the present invention is not limited by the examples.

< Example  1>

In order to determine the accuracy of the tertiary structure predicting apparatus of the unknown protein according to the present invention, the following experiment was performed.

As shown in FIG. 1 , the third protein structure predicting device of the unknown protein was measured, and the performance of the conventional protein protein structure predicting program and the predicting device according to the present invention was measured. The performance measurement is expressed as ROC ₅₀ , where ROC ₅₀ is true-positive if the classification according to the nature of the unknown protein belongs to the correct family, and false-positive if it belongs to another family. -positive), which represents the area of the graph when 50 false positives are accumulated in the relationship graph between false-positive and true-positive. Conventional protein tertiary structure prediction programs and showed the performance of the prediction unit in accordance with the present invention in Figs. 5 and 6 to ₅₀ show the ROC value to the point.

5 is a graph comparing the performance of PSI-BLAST and RankProp of the conventional protein tertiary structure prediction program, Figure 6 is a graph comparing the performance of the PSI-BLAST and the method according to the present invention. As shown in FIG . 5 , ROC ₅₀ of PSI-BLAST and RankProp Although the distribution is generally close to a straight line, the distribution is slightly biased toward the RankProp algorithm in which the protein network is introduced than PSI-BLAST without the protein network.

But,6As shown, the ROC according to the present invention₅₀ The value is ROC of PSI-BLAST₅₀Regardless of the value, it can be seen that a large number is distributed in 0.8 to 1. As a result, the prediction method according to the present invention confirmed that the accuracy of the mold selection is high.

< Example  2>

For the 1bf2_2 domain (database: SCOP b.71.1.1) of which the actual structure is known, a template was selected by the method for predicting the tertiary structure of the unknown protein of FIG. 4 through the tertiary structure predicting apparatus of the unknown protein prepared in Example 1 above . Modeling was performed to predict the structure of the 1bf2_2 domain.

The sequence alignment method used profile-profile alignment and the modeling tool generated the model using a modeler. Procheck was used to evaluate the generated model.

< Comparative example  1>

The 1bf2_2 domain was subjected to tertiary structure prediction using the conventional method, PSI-BLAST program.

<Analysis>

(1) comparison of mold selection

In the template selection, 1m53a1, which is a protein in the same database family (b.71.1.1), was selected as the best template when using the method according to the present invention.However, in the case of using the conventional method, PSI-BLAST, the database of 1bf2_2 was used. The best template was 1epfa1, a protein belonging to (b.1.1.4) different from the family (b.71.1.1). When comparing the list of templates in the top ten among the list of templates converted into scores according to structural similarity, the method according to the present invention selects eight protein templates in the same database family as the database family of 1bf2_2 (b.71.1.1). In contrast, the conventional method selected four protein templates within the same database family and the other six from different folds. Thus, the prediction method according to the present invention confirmed that the accuracy of the mold selection.

(2) Comparison between prediction model and real model

Model the structure based on 1m53a1 or 1epfa1 selected as the best template by the method according to the present invention (Example 1) or the conventional method PSI-BLAST, and compare how the final model is similar to the structure of the actual 1bf2_2 The alignment program CE was used to compare the model with the actual structure. CE expresses structural similarity in Z-scores, which can be said to have biologically meaningful similarities if they have a value above 3.7. The results are shown in Table 1.

template Z-score value Example 1 1m53a1 3.9 PSI-BLAST 1epfa1 2.0

As shown in Table 1, the model predicted by the prediction method according to the present invention shows a Z-score value of 3.9, but there is a significant similarity in comparison with the actual structure, but it is generated by the conventional method PSI-BLAST. The model shows a Z-score of 2.0, which is different from the actual structure.

The predicted models are shown in FIG. 8 and FIG. 9 superimposition with the actual structure of 1bf2_2.

FIG. 7 illustrates the structure (solid line) of 1bf2_2 actually overlapped with the structure (thick line) modeled based on 1epfa1 selected as the best template by PSI-BLAST. As shown in FIG . 7 , it was confirmed that the two structures had different shapes.

FIG. 8 is a diagram of an overlapping structure (solid line) modeled on the basis of the actual structure of 1bf2_2 (solid line) and 1m53a1 selected as the best template by the method according to the present invention. As shown in FIG . 8 , it was confirmed that the secondary structure of the actual structure 1bf2_2 (α-helix structure, β-screen structure, etc.) and the secondary structure of the modeled structure showed similar structures.

Therefore, the prediction method according to the present invention can predict the tertiary structure of a protein more accurately and precisely than conventional protein prediction methods.

As described above, the present invention predicts the tertiary structure of a protein more accurately and precisely than conventional protein prediction methods, thereby reducing the cost and time spent experimentally revealing the protein structure. Therefore, it can be usefully used in the research field for treating various diseases.

Claims (12)

Input means; Sequence preprocessing means for preprocessing the unknown protein sequence inputted from the input means; A template protein database in which information of the template protein is stored; Pair comparison means for comparing the unknown protein with the template protein based on the template protein database; A protein similarity network database composed and stored of structural similarities between the template proteins; Global comparison means for obtaining a global comparison score by taking a weighted sum of pair comparison scores of neighboring proteins in the protein similarity network database according to structural similarities; Template selection means for selecting one or more template proteins having the largest global comparison score based on results of the pair comparison means and the global comparison means; Modeling means for predicting the conformation of the unknown protein using sequencing software and protein structure modeling software, with the selected protein as a template based on the result of the template selecting means; Verification means for examining whether the predicted protein structure is a structure that can be stably present using protein structure analysis software; And An apparatus for predicting tertiary structure of an unknown protein comprising an output means for outputting the predicted protein structure. The apparatus for predicting tertiary structure of an unknown protein according to claim 1, wherein the preprocessing means extracts at least one information selected from the group consisting of secondary structure, solvent exposure degree and profile from the inputted unknown protein sequence. . The tertiary structure prediction of an unknown protein according to claim 1, wherein the template protein database stores at least one information selected from the group consisting of a sequence of a template protein, an actual protein structure, a structure predicted from the sequence, and a profile. Device. The apparatus of claim 1, wherein the protein similarity network database comprises each template protein as a node and forms a link based on at least one of a structure and a function between the template proteins. The tertiary structure of an unknown protein according to claim 1, wherein in the global comparison means, weights between neighboring proteins in the protein similarity network database use Z-scores (Z _{P, D} ) indicating structural similarity. Prediction device. (a) extracting and preprocessing information from the sequence of the unknown protein input through the input means through sequence preprocessing means (step 1); (b) pairwise comparing each template protein and the unknown protein stored in the template protein database via pair comparison means to calculate similarity between each protein (step 2); (c) calculating a global comparison score of the unknown protein and the template protein on the protein similarity network database through global comparison means (step 3); (d) selecting one or more template protein candidates with the highest global comparison scores based on the results of steps 2 and 3 via template selection means (step 4); (e) modeling the structure of the unknown protein using a sequencing software and protein structure modeling software as a template, using modeling means (step 5); And (f) verifying the predicted protein structure through verification means and outputting the predicted protein structure through output means (step 6). The method of claim 6, wherein the pretreatment of step 1 comprises extracting at least one information selected from the group consisting of secondary structure, solvent exposure degree, and profile from an unknown protein sequence inputted through an input means. Method for predicting the structure of an unknown protein. The method of claim 6, wherein the similarity calculation between the template protein and the unknown protein of step 2 is performed based on at least one of a sequence and a profile of the unknown protein. The method of claim 6, wherein the calculation of the global comparison score of the unknown protein and the template protein of step 3 is weighted to the pair comparison result of the template protein and the unknown protein according to the structural similarity between neighboring proteins on the protein similarity network database. A method of predicting the structure of an unknown protein. 10. The method of claim 9, wherein the weights between the template proteins on the protein similarity network database use Z-scores (Z _{P , D} ) _, indicating structure similarity. A computer-readable recording medium for executing the method of any one of claims 6 to 10 on a computer. Records a data structure in which a template protein similarity network database is stored, consisting of a node consisting of proteins in the template protein database and a link numerically representing the similarity based on at least one of the structure or function between the template proteins corresponding to the node. Recording media.

Images