KR100889940B1 - Method to predict protein secondary structure using NMR spectroscopy - Google Patents

Abstract

The present invention relates to a method for predicting protein secondary structure using nuclear magnetic spectroscopy.

That is, the present invention builds a separate database in order to increase the accuracy of protein secondary structure prediction, and builds a GetSBY program, which is a protein secondary structure prediction program using this database, and compares the protein secondary structure with the conventional prediction method. It can be used to determine the tertiary structure of proteins, with the advantages of accurate and superior prediction ability, hydration or dihedral prediction, and many options, conveniences, and speeds, and library of extensive protein and NMR data. The aim of the present invention is to provide a method for predicting protein secondary structure using nuclear magnetic spectroscopy, which can help to clarify the relationship between NMR information and protein structure.

Nuclear Magnetic Spectroscopy, Proteins, Secondary Structure, GetSBY Program, YDB Database, Protein Database Bank, Bio Magnetic Retention Bank

Description

Method to predict protein secondary structure using NMR spectroscopy

1 is a flow chart illustrating a method for building a YDB database according to the present invention;

2 is a capture screen showing the basic interface of the GetSBY program according to the present invention;

3 is a calculation result screen of the GetSBY program according to the present invention;

4 is a flowchart illustrating the operation of the GetSBY program according to the present invention;

5 is a schematic diagram showing a Ramachandran plot for protein secondary structure determination.

The present invention relates to a method for predicting protein secondary structure using nuclear magnetic spectroscopy, and more specifically, reference (YDB), which is a database linking protein structure and NMR experimental values, and chemical movement through NMR experiment ( The present invention relates to a method for predicting protein secondary structure using nuclear magnetic spectroscopy to accurately predict the secondary structure of a protein through a special matching system using a chemical shift value.

The function of the proteins that make up the human body is closely related to the structure of the protein, and thus, many efforts have been made to determine the structure of the protein to determine the function of the protein.

The structure of proteins can be largely divided into 1, 2, 3, and 4 structures.

The primary structure refers to the amino acid sequence of a protein, and is composed of 20 amino acids, for example, MVSTAGIKLMN .........

Secondary structure refers to the formation of a series of amino acid bases through hydrogen bonding, especially in the form of spiral helix or beta sheet, and secondary structure to predict tertiary structure. Prediction should be preceded by PSI-PRED and Talos program.

In the case of PSI-PRED, the secondary structure is predicted only by input of a sequence (primary structure), and in the case of Talos, C, CA, CB, which are backbone atoms of peptide bonds of proteins The chemical shifts and sequences of N and H are input together.

The PSI-PRED is known to approach the results from PSI-BLAST (see http://www.ncbi.nlm.nih.gov/blast/) in a two feed-forward neural network. have.

In the case of Talos, the available range of PHI and PSI dihedral angle is compared with the inputted information by referring to the structural information of 78 proteins and the chemical shift information of backbone atoms of the proteins. In this case, the secondary structure is predicted by mapping the PHI and PSI angles with the range of secondary structures that can be derived from the Ramachandran plot based on the PHI and PSI dihedral angle information.

In the process of determining the structure of protein through NMR experiment, the secondary structure is confirmed using Talos most frequently, and in the case of PSI-PRED, the sequence before and after the sequence to know the secondary structure The secondary structure is determined by measuring the frequency, including the sequence, in a data file that is made up of a weight matrix.

These methods are used in many papers because they are often used and have good results, but they also have problems. In the case of PSI-PRED, the probability of second-order prediction is simple, although it is simple to input only a sequence. Is relatively low, and the case of Talos is not predicting the structure of the entire sequence, but rather the reference database is small (78 proteins), thus predicting only a specific region in the whole sequence. The prediction part can also be divided into "GOOD" and "NEW", which shows quite high accuracy for the sequence marked "GOOD", but the prediction reliability is low for the part marked "NEW". And, for parts where no chemical shift is given, secondary structure prediction itself is not.

Also, in real-time, protein structure files and chemical shift files are searched for similar sequence patterns (for example, for three consecutive sequences) according to system specifications. It's different, but it's about 30 minutes on a server-class computer to make a second-order prediction for about 100 sequences of proteins.

In addition, Talos also supports Talos input format, which is inconvenient to convert the input file.

The present invention is a result of research and development to solve the above problems, and to build a separate database (hereinafter referred to as YDB (Yonsei Database Bank) database) to improve the accuracy of protein secondary structure prediction using this database By constructing a protein secondary structure prediction program (hereinafter referred to as GetSBY (Get Secondary structure By YDB) program), it is more accurate and superior prediction ability, hydration degree or dihedral prediction of protein secondary structure than the conventional prediction method. In addition, it provides many options, conveniences, and speeds, and can be used for the determination of tertiary structure of proteins.It is also possible to fully understand the relationship between NMR information and protein structure through the library of vast protein and NMR data. To provide a method for predicting protein secondary structure using nuclear magnetic spectroscopy to help Never.

The present invention for achieving the above object: a chemical shift by the NMR experiment of the PDB files and protein atoms of the Protein Database Bank (PDB) that contains the coordinate information of the protein (chemical atoms) Constructing a YDB database, which is a relational database based on chemical shift data of a Bio Magnetic Resonance Bank (BMRB) having information; And a step of constructing a GetSBY program for predicting a secondary structure only by assignment of six backbone atoms to a protein that has been subjected to NMR experiments to predict a protein secondary structure based on the YDB database; Reading a secondary structure, a back angle, and a sign language database index file which processes the YDB database with the operation of the GetSBY program so as to be loaded into a memory to have a fast processing speed of a search; The GetSBY program receives inputs in BMRB's Star file format and Sparky's shifts file format, and includes a BMRB containing data indicating NMR backbone chemical shifts of proteins to be predicted for secondary structure. a calculation step for predicting the protein secondary structure of the star or Sparky assignment table file path and selecting the desired option; It provides a protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that it comprises a.

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

First, main terms used below will be defined in order to clarify the contents while helping the understanding according to the description of the present invention.

1. YDB database

PDB files in the Protein Database Bank (PDB) containing the coordinates of the protein, and biomagnetic retention banks containing chemical shift information by NMR experiments of protein atoms. Bio Magnetic Resonance Bank (BMRB)) is a relational database based on chemical shift data.

2. GetSBY Program

As a micro window compatible program, it is a program for predicting the protein secondary structure according to the present invention based on the above YDB database.

3. Protein Database Bank (PDB)

A database containing PDB files containing protein coordinate information, hereinafter referred to as PDB.

4. Bio Magnetic Resonance Bank (BMRB)

A database containing chemical shift information by NMR experiments of protein atoms, hereinafter referred to as BMRB.

First, the method of constructing the YDB database will be described with reference to the flowchart of FIG. 1.

To create a YDB database, a series of programs called YDB_Maker are used. The roles of this program are as follows.

First, import the dbmatch.csv file provided by the ftp site of the BMRB (Bio Magnetic Resonance Bank) (S101).

The dbmatch.csv file consists of four fields. The first field shows what kind of protein structure database the BMRB file is linked to, and the second field contains the entry ID and the third field. Indicates whether the result is a direct experiment or speculative information in Y / N, and the fourth field indicates Entry ID in the BMRB.

The dbmatch.csv file matches the chemical shift of the protein when there is a structure for the same protein, even if the information is not actually tested. These are the parts that are cleaned up when purifying.

In this way, the dbmatch.csv file is downloaded and parsed into the TMatchParse class (S102).

The TMatchParse parses a dbmatch.csv file and returns an entry id of a BMRB when an entry id of the PDB database is input, and returns an entry id of a PDB when an entry id of the BMRB is input to indicate which files are matched. Is a class.

As such, it is possible to know which PDB file and BMRB file are used through the dbmatch.csv file.

Next, while checking the existence of the PDB file and the STAR file of the BMRB while looping and downloading through the HTTP protocol (S103 and S104).

Once you have downloaded all the necessary files, you will actually match the PDB file with the BMRB file and create a SQL statement that will contain the information you want to enter into the database.

TPdbParse and TStarParse are the classes for parsing the PDB file and the BMRB file, respectively. The main role of the TPdbParse is to obtain the sequence of proteins shown in the PDB file and the coordinates of the atoms in each sequence. Creates an empty list for each of them and puts the chemical shift information there.

In addition, the TStarParse is the main role of reading the chemical shift (chemical shift) information in the BMRB STAR file, the STAR file contains the chemical shift information of the atoms of a particular sequence, one STAR file in the PDB It may or may not contain chemical shift information for every atom in every sequence in the file, it may contain just a portion of a sequence, or just some atoms.

And because the list of sequences can be different from the PDB, and the indexes are different, the general matching of these is not so easy.

Therefore, several BMRB files may be matched to one PDB, and one BMRB file may cover several PDBs. Therefore, a class that implements this function is implemented separately, which is TPdb2Bmrb (S105).

The TPdb2Bmrb class receives instances of TPdbParse and TStarParse and matches them.The TPdb2Bmrb class inputs chemical shift information to an empty list in the sequence structure of TPdbParse to input several atoms of a sequence. Enter chemical shift information into the atoms.

In this case, a special algorithm for matching the index between the PDB file and the STAR file is used. The case where a sequence is shown in PDB is obtained in ten consecutive sequences and the sequence in STAR is obtained in ten consecutive sequences is checked for correctness. Enter it.

After the chemical shift information is entered into TPdbParse, a file database is first created (S106).

The file databases are designed to have the following information.

Protein structure information (coordinates, interatomic distance, dihedral angle), chemical shift, secondary structure, degree of hydration, SCOP FOLD classification, etc. are all entered into each record, and sequence information, 2 Vehicle structure pattern information is also stored to form a large pool.

Thereafter, a SQL file is generated using the file database, and then uploaded to a corresponding SQL server (S107 and S108).

Accordingly, the user can obtain the protein structure data and chemical shift information corresponding to the condition by inputting the condition.

The table structure of the YDB database thus constructed is divided into five types, as shown in Tables 1 to 6.

Users can access the above five kinds of tables to get the desired data and find new trends as needed. There are 20 _X and _cs_X tables, 400 _XX and 1 table each, and the YDB database has a total of 443. It consists of two tables.

In general, the YDB database is distinguished from the databases used for protein research. It is built as a relational database.

Being built as a relational database means that you can access the database in real time and query and query for various conditions to obtain and process the data.

For example, the average value of chemical shift of carbon alpha of Phenylalanin whose secondary structure is determined as strand among all beta protein types on SCOP FOLD. If you want to get the standard deviation value, you can get a simple query like the following example 1.

If you want to get the data you want, but there is no relational database like the YDB database, then check the proteins classified in SCOP and chemical shift to BMRB among all beta proteins among them. dl registered proteins, and among them, the carbon alpha values of Phenylalanine should be inputted to Excel and processed, but the data obtained by this invisible operation can be obtained. The YDB database can be obtained in less than a second.

Accordingly, the YDB database according to the present invention makes the time consuming process simple, fast, and programmable, and thus will be very useful in the protein structure research process based thereon.

In addition to the advantages of building as a relational database, the YDB database has other advantages.

In other words, it can be said that the YDB database of the present invention can be used as a powerful research tool compared to other databases in that it provides both a sequence database, a fragment database, and a classification database.

The data that can be obtained by accessing existing databases such as PDB, BMRB, and SCOP is limited, and the only way to check the linkage between these databases is to go to the site where the user directly provides the database.

However, as the YDB database of the present invention significantly improves the connectivity between the desired databases, it is possible to obtain the desired data at high speed with only one line of query, eliminating all the manual work of searching and checking each other database individually. .

In addition, the YDB database of the present invention is a database that can study the relationship between several factors, not only to obtain the chemical shift (chemical shift) under the structural conditions, but also to the structure within the chemical shift (chemical shift) range You can also get

In other words, the structure information can be obtained from the chemical shift as shown in Example 2, in contrast to the Example 1 obtained from the chemical shift as described above.

In Example 2 above, we trace back the structure information at the chemical shift, and we have the commonality that both proteins are present at the end of the secondary structure. This is not a good idea, but it is a good example to show that a traceback from chemical shift to structure is possible with the YDB database.

Meanwhile, in the construction of the YDB database of the present invention, various purification methods were selected to deal with the fact that the conditions for the various nmr experiments were different, and that there was a possibility of error in the results of the initial NMR experiment and the protein structure. There will be six types of databases with different characteristics, and the six databases are shown in Table 7 below.

As above, the GetSBY program is built with the advantage of enabling backtracking from the chemical shift of the YDB database to the structure.

The GetSBY program can predict the secondary structure very accurately with the assignment of six backbone atoms to the protein after nmr experiment.

In addition, it is possible to predict the optimal secondary structure according to the situation by applying the six databases (see Table 6) of the YDB database, in order to run the GetSBY program of the present invention six databases included in the YDB database (Table 6 18 data sets are obtained for each database, and the average and standard deviation of the secondary structure, the degree of hydration, and the dihedral angle for each database. Will have

All three types of files listed above have a similar format, followed by amino acid type, atomic type, index, mean, and standard deviation.

In the case of the index, the secondary structure is composed of H, E, T, G, and C B, and the remainder is composed of numbers from 1 to 10.

In the above secondary structure, H stands for Alpha Helix, E stands for Beta Strand, T stands for Turn, G stands for 310 Helix, C stands for Random Coil, and B stands for Bridge.

In the general secondary structure prediction program, all coils are processed except for helix and strands, which is also followed by the GetSBY program of the present invention.

In the case of the degree of hydration, 1 is represented by 0% to 10%, 2 is represented by 10% to 20%, and as a representative value up to 10, and the dihedral angle divides PHI and PSI into sizes of 9 × 9, respectively, and PHI Both PSIs range from -180 to +180 degrees, which is represented by a representative value of 1 to 81 in the GetSBY program. For example, 1 ranges from -180 to -140 for both PHI and PSI. .

The GetSBY program according to the present invention receives two types of files as input files, and receives them in the BMRB Star file format and the Sparky shifts file format.

The above two file formats are very useful, and they have the advantage that they can be applied immediately during the work because they are files that occur during normal nmr backbone assignment.

The GetSBY program is a Microsoft Windows compatible program and is designed to be easily used in a general desktop environment. The attached FIG. 2 shows a basic interface of the GetSBY program.

Set the input file in the Input file field and press the Calculate button to firstly predict the secondary structure, the degree of hydration and back angle, and press the Post Process button. It will smooth the predicted result.

Smoothing is a kind of artificial intelligence. In the protein structure, helix should be 4 or more in the case of Helix, or when the accuracy of prediction is low, the secondary is connected to the YDB database in real time. This function is to re-determine the secondary structure by selecting a large number of frequencies.

This predicted result can be saved in csv file format by pressing Save button, and also Rama. If you click the plot button, you can see which residue is set in which area and how the degree of hydration is displayed on the graph with the PHI angle on the x-axis and the PSI angle on the y-axis. As shown in 3.

Here, the prediction algorithm of the GetSBY program according to the present invention is applied in the same manner to the secondary structure, the degree of sign language, and the backside angle. Referring to the flowchart of FIG. 4 to which the operation of the GetSBY program is attached, it is as follows. .

The first time you run the GetSBY program, it reads the secondary structure, backside angle, and sign language database index file that processed the YDB database.

These are loaded into memory and searched for faster processing speeds.

Then, give the path of the BMRB star or Sparky assignment table file containing the data assigned to the NMR backbone chemical shifts of the protein to be predicted for secondary structure, select the desired option, and then calculate Press the () button to predict the secondary structure.

In this case, the options to choose are whether to use the standard deviation or only the difference between the means when calculating the six database types and scores, and the chemical movement of the six types of backbone atoms ( Because chemical shift values are represented by a distribution of different values, dividing each difference by a standard deviation will set the standard for the degree of that difference.

Therefore, if this option is selected and the secondary structure is predicted, the following process is performed.

1) Read the input file.

At this time, a list for storing chemical shift information is generated. A PEAKDATA structure representing a sequence is put in the list, and a list called pAtomList is created in the structure to convert chemical shift data. Let's put a structure called ATOMDATA.

The amino acid type is described in PEAKDATA, and ATOMDATA contains atomic type and chemical shift values, and PEAKDATA and ATOMDATA structures are as follows.

2) Calculate the score by looping the number of PEAKDATA in the list containing PEAKDATA, taking ATOMDATA in PEAKDATA.

At this time, the method of calculating the score is slightly different depending on whether or not to apply the standard deviation.

Basically, when calculating scores for a sequence, six nuclides are tested and summed. The scores of each of the six nuclides can be expressed as follows.

Score = absolute value (observed c.s.- c.s.average for each option) / standard deviation

Here, observed cs (chemical shift) refers to the measured chemical shift, and the difference between it and the average chemical shift in the reference file divided by the standard deviation for the six nuclides. Each is obtained and summed.

Of course, the smaller the score means that the closer to the option (option), the option (option) refers to the average value of the secondary structure shown in the YDB database in the amino acid sequence (amino acid sequence).

In the same way, you can get the sign and back angle.

3) In this way, the lowest score for the three structures H, E, and C is selected. At this time, the reliability can be obtained.

Since reliability means that the score of the selected structure is smaller than that of other structures, the reliability can be expressed as follows.

Reliability = 2nd lowest score / (1st low score + 2nd low score) * 100

The reliability is closer to 100 as the first lower score becomes much smaller than the second lowest score, and the greater the reliability, the higher the probability of being relatively right.

 4) Click on the Process button for smoothing. This process handles structural states that cannot be displayed, very low reliability, or no chemical shift.

First, in the case of the secondary structure, Helix or Strand has a characteristic of being connected to some extent. Therefore, if one appears, it is necessary to look around and confirm it.

First, the number of secondary structures shown in the next two sequences, the present self, and the subsequent two sequences is selected, and the second most frequently appeared secondary structure is selected as its own secondary structure. This will give you a first look at the bouncing secondary structure.

Secondly, when the front and back secondary structures are identified as Helix or Strand and are different from their current secondary structure, the reliability of the front and back secondary structures is 90%. In this case, the secondary structure of oneself should follow the secondary structure before and after to have connectivity.

5) Finally, if the confidence level is 50.5% or no chemical shift is given, it is not reliable for the currently selected secondary structure.

This allows you to search the tables of type _XX for yourself and immediately before and after the sequence of three sequences so that you can select the second most popular secondary structure, in which there is no chemical shift. The most likely secondary structure can also be chosen for the case.

Basically, proline induces breakage of the secondary structure, so if proline appears, the sequence and the previous one must be treated as a random random coil.

Here, as an experimental example of the present invention, to test how efficient it is to predict the secondary structure in the GetSBY program with a database file created by processing the YDB database, the PDBs not referenced in the YDB database are SCOP FOLD-specific. We tested a few of them to see how good they were and how effective they were.

For comparison, we used PSI-PRED and Talos programs, which are the most commonly used for secondary structure prediction. They are different from GetSBY program, but they are the most widely used tools for protein structure prediction and determination.

Since the GetSBY program of the present invention can directly receive the BMRB star file, the PDB and the corresponding star file are directly downloaded from the BMRB site and run directly, and the PSI-PRED used for comparison was 2.5 version, and the Talos 2003.027 The version was .13.05, and the list of proteins used for testing is shown in Table 8 below.

Run the GetSBY, PSI-PRED, and Talos programs for the list in Table 8 above, but in the case of Talos, the torsion angle is actually given, and the result is not given for every sequence. Among the results shown in NEW and GOOD, the secondary structure was determined by matching the Ramachandran plot as shown in FIG. 5.

As shown in FIG. 3, the range of Phi and Psi angle which appear according to a secondary structure is determined.

Therefore, it is possible to determine the secondary structure with Talos, and the secondary structure of Talos was determined under the following conditions.

The conditions for determining Alpha Helix are when Phi is between -180 and -40 and Psi is between -80 and -30, or Phi is between 40 and 60 and Psi is between 20 and 100.

Further, the condition for determining the beta sheet is when Phi is between -180 and -40 and Psi is between 20 and 180 or between -180 and -165. For the rest of the cases, a random coil was determined.

On the other hand, the PSI-PRED was used to download and install the latest version 2.5 version from the PSI-PRED official website http://bioinf.cs.ucl.ac.uk/psipred/.

Results for the GetSBY, PSI-PRED, and Talos programs are shown in Tables 9 to 13 below.

In Tables 9 to 13, if _v is added next to the ID of the PDB, it means that the database used is using a database that has been validated.

In addition, G denotes ydb_genuine, P denotes ydb_purified, and R denotes ydb_refdb, and a case where N is attached is omitted to divide the standard deviation, and a case denoted by S is divided by standard deviation. will be.

In addition, PP is the result of PSI-PRED and Talos is the result obtained by driving Talos.

The values in each column represent the secondary structure predicted hit rates for the entire sequence, assuming that 1 is a 100% probability.

Looking at the secondary structure predictions, it can be seen that the overall method using the GetSBY program shows better results than the results of PSI-PRED or Talos.

In the case of validation, when the chemical shift does not include the outer 5%, on average, structural information appears to be unfavorable for secondary structure determination, because the specificity of the information is out of the average. In addition, it seems that too much data is removed by applying validation from ydb_purified.

In general, the prediction of the secondary structure is obtained by applying a standard deviation using ydb_refdb or ydb_purified, and in the case of hydration, a standard deviation is applied to ydb_genuine. Applying standard deviation has been shown to be appropriate.

Of course, selecting a different database generally yields better results than other methods, and predicting results by looking at alternative trends and reliability by predicting with multiple databases.

As described above, according to the method of predicting protein secondary structure using nuclear magnetic spectroscopy, the GetSBY program is a method known in the past by reinterpreting and processing a database called YDB, which stores protein and NMR test results in a special way. It can more accurately predict secondary structure and provide hydration and dihedral angles.

In the case of conventional Talos, the chemical shift values and five consecutive sequences of five backbone atoms, except for HA, within 78 limited protein databases that depend only on chemical shifts. The range of torsion angles is obtained with the sequence, and the range of torsion angles is not constant, and there is no guarantee that the matching of the ramachandran plot is accurate. It shows the lowest secondary structure prediction hit rate among the methods.

In addition, Talos is very slow and does not predict chemical shifts and structures that cannot be found in the reference database. On the other hand, PSI-PRED is less CPU occupied. The reprocessed weight matrix is used to predict the frequency of the secondary structure in sequence only.

However, due to the advantage of receiving only sequence and giving a good prediction result, it also shows a relatively low predictive hit rate compared to the GetSBY program of the present invention due to the limitation of only inputting the sequence.

The GetSBY program of the present invention uses fast speed and accuracy, and results from multiple database reprocessing, and various options (choose backbone atom to use), torsion angle, and hydrophobicity that are not predicted by other tools. It has the advantage of giving results.

Accurate prediction of the secondary structure region of a protein means that the tertiary structure can be predicted by the spatial arrangement of the determined secondary structures, and the prediction of the tertiary structure is equivalent to the prediction of the function of the protein.

Therefore, the accurate prediction of secondary structure is an important task that many scientists have been working on in the past, and the tertiary structure is, after all, knowing the secondary structure, backside angle between amino acids, and the degree of external exposure of amino acids (hydration degree). Because it is predictable, it is necessary to accurately predict the underlying secondary structure. In view of this, the GetSBY program of the present invention has superior prediction capability and sign language compared to the secondary structure prediction tools known to date. It is a tool that has many advantages such as degree or dihedral prediction, and many options, conveniences, and speeds.

In addition, the YDB database, which enables the GetSBY program, can also play a huge role in fully clarifying the relationship between NMR information and protein structure and eventually uncovering protein structure secrets through extensive library of protein and NMR data.

Claims (10)

PDB files in the Protein Database Bank (PDB) containing the coordinates of the protein, and biomagnetic retention banks containing chemical shift information by NMR experiments of protein atoms. Constructing a YDB database, which is a relational database based on chemical shift data of a Bio Magnetic Resonance Bank (BMRB); And A step of constructing a GetSBY program for predicting a secondary structure only by assignment of six backbone atoms to the NMR-tested protein to predict the protein secondary structure based on the YDB database; Reading a secondary structure, a back angle, and a sign language database index file which processes the YDB database with the operation of the GetSBY program so as to be loaded into a memory to have a fast processing speed of a search; The GetSBY program receives inputs in BMRB's Star file format and Sparky's shifts file format, and includes a BMRB containing data indicating NMR backbone chemical shifts of proteins to be predicted for secondary structure. A method for predicting protein secondary structure using nuclear magnetic spectroscopy, comprising calculating a step for predicting a protein secondary structure in which a star or Sparky assignment table file path is selected and a desired option is selected. The method of claim 1, wherein in the calculating step, the desired option is to apply six databases of the YDB database, so that standard deviations or only differences from the mean are used when calculating the six database types and scores. Method for predicting protein secondary structure using nuclear magnetic spectroscopy, characterized in that the selected one. The method of claim 1, wherein building the YDB database is: Importing a dbmatch.csv file provided at an ftp site of a BMRB (Bio Magnetic Resonance Bank); Download the dbmatch.csv file, find the sequence of proteins in the PDB file and the coordinates of the atoms in each sequence, and create an empty list for each sequence parsing with a TMatchParse class that makes the information available for shift; Checking the existence of the PDB file and the BMRB STAR file in a loop and downloading them by the HTTP protocol, and actually matching the PDB file and the BMRB file to make SQL information into a database; Separately implementing a TPdb2Bmrb class in which several BMRB files may match a single PDB and one BMRB file may cover several PDBs; Inputting chemical shift information into TPdbParse to create a file database firstly; Generating an SQL file using the file database and uploading the SQL file to a corresponding SQL server; Protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that comprises a. The method of claim 3, wherein the TMatchParse parses a dbmatch.csv file and returns an entry id of a BMRB when the entry id of the PDB database is input. Protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that the class that tells the match. The method of claim 3, wherein a special algorithm for matching the index between the PDB file and the STAR file is used, wherein the sequence shown in the PDB is taken in ten consecutive sequences, and the sequence shown in the STAR is taken in ten consecutive sequences to be checked for correctness. Method for predicting protein secondary structure using nuclear magnetic spectroscopy, characterized in that the input to the case. The method of claim 3, wherein the file database is a protein structure information (coordinates, interatomic distance, back angle), and chemical shift (secondary structure), hydration degree, SCOP FOLD classification, etc. A method of predicting protein secondary structure using nuclear magnetic spectroscopy, which is performed by inputting a record of and storing sequence information and secondary structure pattern information to form a large pool. The method of claim 1, wherein the driving of the GetSBY program is: Generating a list for storing chemical shift information by receiving an input in a BMRB Star file format and a Sparky shifts file format; Calculating a score by receiving ATOMDATA in PEAKDATA by looping the number of PEAKDATA of the list containing the PEAKDATA; Determining reliability by selecting the least score for the three structures H, E, and C; A smoothing step of processing the unrepresented structural state of the protein, a case where the reliability is very low or when no chemical shift is given at all; Storing the final result in csv format; Protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that it comprises a. The nuclear magnetic field of claim 7, wherein a PEAKDATA structure representing a sequence is put in the list, and a list called pAtomList is generated in the structure to contain a structure called ATOMDATA having chemical shift data. Protein secondary structure prediction method using spectroscopy. The method of claim 7, wherein each score of the six nuclides is [Score = absolute value (observed c.s.-c.s. average for each option) / standard deviation] where the observed c.s. (chemical shift) is the measured chemical shift. Protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that obtained. The method of claim 7, wherein the reliability is [Reliability = 2nd Low Score / (1st Low Score + 2nd Low Score) * 100] Protein secondary structure prediction method using nuclear magnetic spectroscopy, characterized in that obtained.

Images