JP2019185224A - Identification quality evaluation method and apparatus for endogenous modified peptide - Google Patents

Abstract

To provide a decoy database that can estimate the false discovery rate (FDR) of a database search by PTM-compatible endogenous peptide retrieval method with high accuracy.SOLUTION: When a reader 132 reads data from a target DB12 that contains the sequence of modified proteins, an amino acid distribution statistics creation unit 133 creates amino acid distribution statistics indicating the frequency of occurrence of amino acids, and a PTM distribution statistics creation unit 134 creates PTM distribution statistics showing the frequency of occurrence of modification sites for each type of modification. A decoy sequence generating unit 135 randomly creates the same decoy sequence as the number of sequences of entries in an original target DB12 by weighting, using the amino acid distribution statistics. A decoy PTM information generating unit 136 creates a decoy PTM information corresponding to each decoy sequence by weighting, using the PTM distribution statistics. By collecting the decoy array and decoy PTM information created in this way, the system creates a decoy database with PTM addition information.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a method for evaluating the identification quality of a technique for identifying an endogenous peptide subjected to post-translational modification (endogenous modified peptide) using mass spectrometry, and an apparatus therefor.

  In recent years, protein structures and functions have been rapidly analyzed as post-genomic research. The most widely used data analysis method for comprehensively identifying proteins and peptides using mass spectrometry is a protein sequence database search method (hereinafter simply referred to as “database search method”. (Abbreviated as “DB search method”). As a database search method, MS / MS ion search (see Non-Patent Document 1) included in Mascot, a search engine provided by UK Matrix Science, is widely known. Other than that, free search engines such as X! Tandem have similar functions.

  Since there are various methods (algorithms) for database retrieval methods, comparison of identification quality of these methods is important in selecting an appropriate method. In addition, even if the same technique is used, the identification quality changes if the threshold for judging the score or expected value of the amino acid sequence hit by the search is changed, so it is also important to control the identification quality by appropriately setting such a threshold. . As an objective evaluation index of identification quality used in such a case, a false discovery rate (FDR = False Discovery Rate) is widely used (see Non-Patent Document 2).

  In general, when estimating the false discovery rate in a database search method, a decoy sequence database containing a decoy sequence that is a false sequence that is created to resemble the amino acid sequence of a natural protein but does not exist is used. . When searching for mass spectral information collected by mass spectrometry using a decoy sequence database, the false discovery rate can be estimated by counting the number of decoy sequences identified accidentally.

  In proteome analysis, which analyzes peptides that have been artificially fragmented using a digestive enzyme such as trypsin, the amino acids of each protein recorded in the natural protein sequence database (hereinafter referred to as “target database”) to be searched If the sequence is inverted for each protein or peptide to be fragmented (inversion between the N-terminal side and the C-terminal side) is used as the decoy sequence, the false discovery rate can be estimated easily and with high accuracy. Are known. On the other hand, when comprehensively identifying endogenous peptides produced by unspecific degrading enzymes in the living body, adopting the amino acid sequence reversed at both ends as described above as a decoy sequence would increase the false discovery rate. It is known that there is a risk of evaluation. Therefore, in such a case, a method of generating a decoy sequence by replacing a random sequence based on the amino acid sequence of a protein recorded in a target database is used.

  Many of proteins and peptides existing in the body may be abbreviated as “PTM” as appropriate after the post-translational modification (bonding or dissociation of chemical functional groups or hydrolysis (cleavage) of peptide bonds). The function of which is dynamically and strictly adjusted. It is very important from the viewpoint of applying life science research and its results to industry to analyze which part (which amino acid residue) in the amino acid sequence of protein in the living body has been modified. is there. However, conventional general database search methods cannot identify all modified peptides that can occur in vivo by a single search. This is because the number of theoretical peptides to be searched, that is, the search space expands exponentially in consideration of modification variations, and as a result, peptide candidates are based on the statistical confidence index of the score. This is because it becomes impossible to narrow down uniquely (see Non-Patent Document 2).

  For this reason, a large number of modified peptides are identified with an acceptable degree of reliability, that is, the number of types of modifications to be searched is narrowed down in advance as a limitation to improve the identification sensitivity of modified peptides. It is common to perform a database search above (so-called variable modification method). However, even in such a case, identification sensitivity may decrease if the search is for a modification that occurs in the side chain of an amino acid residue that is highly contained in the protein, such as phosphorylation, that is, the frequency of occurrence is high. Are known. Also, it is known that the search space is greatly increased and the identification sensitivity is lowered when trying to exhaustively search for a cleavage site instead of modification of a functional group such as phosphorylation.

  As another method for suppressing an increase in search space that causes a decrease in identification sensitivity in the modified peptide search described above, a method of performing a search limited to known modification sites is effective. As a method encompassed by this, regarding functional group addition, there is a method in which an amino acid that has undergone a specific functional group modification at a known modification site is associated with a pseudo amino acid symbol that does not exist (patent) Reference 1). For peptide bond cleavage, the combination of cleavage is a sequence detected from the sample to be measured in the past measurement, a sequence reported to have been detected from the sample to be measured in the literature, or the sequence of that sequence. There is a method of performing a search after limiting to a sequence including at least one partial sequence (see Patent Document 2). One of the features of these methods is that the protein sequence database is not used as a search target as it is, but a peptide sequence database of a smaller peptide unit is newly constructed and searched.

  As is well known, an endogenous modified peptide is a peptide that has undergone both functional modification and cleavage. Therefore, as a method for effectively identifying the endogenous modified peptide, an attempt to combine the method described in Patent Document 1 and the method described in Patent Document 2 is one of the possible methods.

However, in the method described in Patent Document 1, since the modified amino acid is symbolized as a pseudo amino acid as described above, for example, phosphorylation is known at three locations on the amino acid sequence of one peptide. If there is a phosphorylation site, it is necessary to develop 2 ³ = 8 peptide sequences as a combination of the presence or absence of phosphorylation and record them in the database. In general, the endogenous peptide is large in size, and it is assumed that a large number of known modification sites are included in one peptide. Therefore, when the method described in Patent Document 1 is applied as it is, The combination becomes enormous. On the other hand, also in the technique described in Patent Document 2, it is necessary to record a huge number of peptides in a database. For this reason, as described above, even if the methods described in Patent Documents 1 and 2 are combined, the size of the database becomes too large to be implemented.

JP 2013-47624 A International Publication No. 2017/0475580 Pamphlet

"Matrix Science-Mascot-MS / MS Ions Search", [online], Matrix Science Ltd., [April 3, 2018 Day search], Internet <URL: http://www.matrixscience.com/cgi/search_form.pl?FORMVER=2&SEARCH=MIS> Akiyasu Yoshizawa, “Which database to use: database search, sequence analysis, misunderstandings, and difficult problems”, Proteomics Letters 2016 1: 63-80, [online], Japan Proteomics Society, [April 2018 3 days search], Internet <URL: https://www.jhupo.org/data/proteomeletters/16008.pdf> Siwy J. and 4 others, “Human urinary peptide database for multiple disease biomarker discovery”, Proteomics Clinical Applications (Proteomics) Clinical Applications), 2011, Vol.5, pp.367-374 "PEFF-PSI Extended Fasta Format", PSI, [online], [April 3, 2018 search], Internet <URL: http://www.psidev.info/node/363>

  In order to reduce the size of the above database, modification information related to functional group modification is handled as additional information for individual amino acid residues on the amino acid sequence, independently of the amino acid sequence, and only the information necessary for the search is moved. Therefore, it is necessary to devise such that it is possible to search by expanding it as a modified peptide on the memory. However, when estimating the false discovery rate in order to evaluate the identification quality in the endogenous modified peptide search method (hereinafter referred to as “PTM-compatible endogenous peptide search method”) using such a technique, the following There's a problem.

  (1) In the PTM-compatible endogenous peptide search method described above, post-translational modification information is required in addition to the amino acid sequence information. However, in the conventional method for creating a decoy database as described above, even if a decoy database containing amino acid decoy sequences can be created, decoy modification information cannot be obtained. Therefore, even if a search using such a database is performed, an appropriate search for a modified peptide is not executed, and the false discovery rate is underestimated. That is, the conventional method for creating a decoy database has a problem that the false discovery rate in the above-described PTM-compatible endogenous peptide search method cannot be estimated with sufficient accuracy to be practically used.

  (2) The protein sequence database or peptide sequence database used in the above-described PTM-compatible endogenous peptide search method is a subset obtained by extracting the amino acid sequence of a protein having sample specificity from the amino acid sequence of a commonly used protein. Or the amino acid sequence of a peptide that is a subset thereof. Therefore, in the PTM-compatible endogenous peptide search method, the number of identifications can be increased while suppressing the false discovery rate to a predetermined value by appropriately integrating the search results obtained by each search using a plurality of different sequence databases. Can do. For example, a peptide identified redundantly among multiple sets of identification results obtained by searching against a plurality of different sequence databases is a highly reliable peptide identified with good reproducibility under different search conditions. Can be estimated. However, the probability that the randomly generated decoy sequences match for a plurality of different target databases is very small. Therefore, when a conventional general method for generating a decoy sequence is used, it is not possible to obtain a false discovery rate for peptides hit commonly among the results identified using a plurality of different databases.

  The present invention has been made to solve the above-mentioned problems, and its main purpose is to provide a fake for evaluating the identification quality of a database search method for identifying a modified endogenous peptide (endogenous modified peptide). An object is to provide an identification quality evaluation method and apparatus for endogenously modified peptides that can accurately determine the discovery rate.

The identification quality evaluation method for endogenous modified peptides according to the present invention made to solve the above problems is a method for evaluating the identification quality when identifying an endogenous modified peptide by database search based on the results of mass spectrometry, In an identification quality evaluation method for endogenous modified peptides that calculates an index value for evaluating identification quality using a decoy sequence,
a) an information reading step for reading the amino acid sequence information and post-translational modification information of the modified protein recorded from the protein sequence database to which post-translational modification information is added;
b) Amino acid distribution statistics acquisition step for obtaining amino acid distribution statistical information indicating the appearance frequency of each amino acid based on the amino acid sequence information obtained by the information reading step;
c) based on the post-translational modification information obtained by the information reading step, a post-translational modification distribution statistics acquisition step for obtaining post-translational modification distribution statistical information indicating the frequency of appearance of each post-translational modification;
d) a decoy sequence generation step for generating a decoy sequence similar to a normal amino acid sequence based on the amino acid sequence information obtained by the information reading step and the amino acid distribution statistical information;
e) Generation of decoy post-translational modification information generating decoy post-translational modification information including the type and site of modification based on the decoy sequence generated in the decoy sequence generation step and the post-translational modification distribution statistical information Steps,
f) By integrating the decoy sequence generated in the decoy sequence generation step and the decoy post-translation modification information generated in the decoy post-translation modification information generation step, a decoy sequence to which post-translation modification information is added is recorded. A decoy database creation step for creating a decorated decoy database;
It is characterized by having.

Moreover, the identification quality evaluation apparatus for endogenous modified peptides according to the present invention is an apparatus for carrying out the identification quality evaluation method for endogenous modified peptides according to the above invention, and the endogenous modified peptides are obtained by database search based on the mass spectrometry results. An apparatus for evaluating identification quality at the time of identification, wherein an identification quality evaluation apparatus for endogenous modified peptides that calculates an index value for evaluating identification quality using a decoy sequence,
a) an information reading unit for reading amino acid sequence information and post-translational modification information of the recorded modified protein from a protein sequence database to which post-translational modification information is added;
b) based on the amino acid sequence information obtained by the information reading unit, amino acid distribution statistics acquisition unit for obtaining amino acid distribution statistical information indicating the appearance frequency of each amino acid;
c) based on the post-translational modification information obtained by the information reading unit, a post-translational modification distribution statistical acquisition unit for obtaining post-translational modification distribution statistical information indicating the appearance frequency of each post-translational modification;
d) a decoy sequence generation unit that generates a decoy sequence similar to a normal amino acid sequence based on the amino acid sequence information obtained by the information reading unit and the amino acid distribution statistical information;
e) Generation of decoy post-translational modification information that generates decoy post-translational modification information including the type and site of modification based on the decoy sequence generated by the decoy sequence generation unit and the post-translational modification distribution statistical information And
f) By integrating the decoy sequence generated by the decoy sequence generation unit and the decoy post-translation modification information generated by the decoy post-translation modification information generation unit, a decoy sequence to which post-translation modification information is added is recorded. A decoy database creation unit for creating a decorated decoy database;
It is characterized by having.

  In the present invention, as the “protein sequence database to which post-translational modification information is added”, for example, UniProtKB provided by Swiss Institute of Bioinformatics (abbreviation: SIB) or the like can be used. Further, instead of the entire existing protein database, a part of the protein database extracted appropriately may be used as the “protein sequence database to which post-translational modification information is added”.

  The present invention is characterized by a procedure for creating a decoy database used when evaluating the identification quality of a database search method for identifying endogenous modified peptides, and a process at the time of creation. That is, in the identification quality evaluation apparatus according to the present invention, when amino acid sequence information is read from the protein sequence database to which post-translational modification information as described above is added, the amino acid distribution statistical information acquisition unit obtains the obtained amino acid sequence information. Based on the above, amino acid distribution statistical information indicating the appearance frequency of each amino acid sequence is created. On the other hand, when the post-translational modification information is read from the protein sequence database to which the post-translational modification information is added, the post-translational modification distribution statistics acquisition unit calculates the appearance frequency of each post-translational modification based on the obtained post-translational modification information. Post-translational modification distribution statistical information indicating the frequency of sites to which each post-translational modification is added is created.

  For example, the decoy sequence generation unit generates a false decoy sequence by randomly replacing amino acids with weighting according to the amino acid distribution statistical information based on the normal amino acid sequence information obtained by the reading. Further, the post-decoy translation modification information generation unit randomly determines a post-translational modification molecule to be added and a site (amino acid residue) to be added to the generated decoy sequence based on post-translational modification distribution statistical information. In the present invention, amino acid distribution statistical information and post-translational modification distribution statistical information are respectively obtained based on a protein sequence database to which post-translational modification information is added, and using these information, a decoy sequence that is a fake amino acid sequence and Decide decoy modifications, which are fake post-translational modifications. Then, the decoy database creation unit creates a decoy database in which the decoy sequence to which post-translation modification information is added is recorded by integrating the generated decoy sequence and post-decoy post-translation modification. The post-translational modification is performed by performing a search using a predetermined database search method for mass spectrum information using the decoy database created in this way, and counting the number of decoy sequence identifications that meet the identification criteria (exceeding the identification threshold). The false discovery rate can also be estimated with high accuracy for the processed protein or peptide.

In the identification quality evaluation method for endogenous modified peptides according to the present invention, as a preferred embodiment,
A database dividing step of dividing the target database, which is a protein sequence database to which post-translational modification information is added, into a plurality of subset databases having no common entry;
For each of the plurality of subset databases, the information reading step, the amino acid distribution statistics acquisition step, the post-translational modification distribution statistics acquisition step, the decoy sequence generation step, the decoy post-translational modification information generation step, and the decoy database A subset decoy database creation step of creating a subset decoy database by performing the processing in the creation step;
A decoy database integration step of creating a whole set decoy database by integrating a plurality of subset decoy databases created in the subset decoy database creation step in correspondence with the target database;
It can have.

  According to this preferable aspect, the decoy sequence and post-translational modification included in each subset decoy database reflect the normal amino acid composition and post-translational modification included in the subset database corresponding to the subset decoy database. It becomes. That is, the subset decoy database is a decoy database that can obtain the false discovery rate with high accuracy. And since the whole set decoy database integrates the information contained in such a subset decoy database, the search results in the search method using a plurality of subset databases are integrated by using this whole set decoy database. It is possible to obtain the false discovery rate with high accuracy.

  According to the present invention, it is possible to accurately obtain a false discovery rate for evaluating identification quality in a database search method for identifying a peptide that has undergone post-translational modification, specifically, a PTM-compatible endogenous peptide search method or the like. it can.

The block block diagram of the modification protein identification evaluation apparatus which is one Embodiment of this invention. The block block diagram of the principal part centering on the decoy database preparation part in FIG. The flowchart which shows the flow of a process at the time of decoy database creation. The schematic diagram which shows an example of the information obtained in the process of decoy database creation. The figure which shows an example of an endogenous peptide database. The graph which shows the relationship between the ratio of the decoy sequence PTM appearance frequency with respect to the target sequence PTM appearance frequency, and the total number of registered PTM sites on the decoy sequence in a specific example of decoy database creation. The figure which shows the estimation result of the relationship between the number of intrinsic identification spectrums, and FDR in the specific example of decoy database preparation. The figure which shows the comparison result of the search result by the variable modification method, and the search result by the PTM corresponding | compatible endogenous peptide search method in the specific example of decoy database creation. FIG. 9 is a Venn diagram showing identification results obtained by the three methods shown in FIG. 8.

An embodiment of an endogenous modified peptide identification and quality evaluation apparatus according to the present invention will be described with reference to the accompanying drawings.
FIG. 1 is a block diagram of the endogenous modified peptide identification quality evaluation apparatus according to this embodiment.

  The endogenous modified peptide identification quality evaluation device 1 is used to identify the quality of identification in the peptide identification unit 11 for identifying the amino acid sequence and post-translational modification of the endogenous modified peptide contained in the sample to be measured, specifically false discovery. The rate is estimated. The endogenous modified peptide identification quality evaluation apparatus 1 includes a target database 12, a decoy database creation unit 13, a decoy database 14, a fake discovery rate estimation unit 15 and the like in addition to the peptide identification unit 11 to be evaluated. Usually, functions other than the databases 12 and 14 can be realized by executing dedicated software installed in the computer on the computer. In this case, the decoy database creation unit 13 and the false discovery rate estimation unit 15 are substantially functions realized by software. The entities of the target database 12 and the decoy database 14 are storage devices that store data in a predetermined format.

Note that the mass analysis unit 2 and the spectrum analysis unit 3 described in FIG. 1 are necessary for inputting MS ² spectrum information to the peptide identification unit 11 when identifying the endogenous modified peptide in the sample. It is a component and is not used when evaluating the identification quality of the peptide identification unit 11.

That is, when identifying the endogenous modified peptide in the sample, the mass spectrometer 2 performs MS ² measurement on the sample containing the target endogenous modified peptide, and the MS ² spectrum (or n is 3 or more). MS ⁿ spectrum). The spectrum analysis unit 3 detects peaks derived from product ions in the MS ² spectrum (or MS ⁿ spectrum), and creates a peak list in which information such as peak mass-to-charge ratio, signal intensity, and valence is aggregated. Then, the peak list is input to the peptide identification unit 11 as mass spectrum information. The peptide identification unit 11 refers to the amino acid sequence of the protein or peptide stored in the target database 12 and the post-translational modification information, and performs a search based on the given measured mass spectrum information for the purpose of endogenous Identify modified peptides.

  In order to obtain the false discovery rate as the identification quality of the peptide identification unit 11, a false decoy sequence and false post-translational modification information (hereinafter referred to as “post-translational modification information”) are used in place of the original peptide sequence and post-translational modification information database. It is necessary to perform a search referring to a decoy database in which “decoy PTM information” is recorded. The decoy database creation unit 13 has a function of creating a decoy database 14 for this purpose.

  2 is a block diagram of the main part centering on the decoy database creation unit 13 in FIG. 1, FIG. 3 is a flowchart showing the flow of processing when creating the decoy database, and FIG. 4 is information obtained in the process of creating the decoy database. It is a schematic diagram which shows an example.

As shown in FIG. 2, the decoy database creation unit 13 includes, as functional blocks, a database division unit, a data reading unit, an amino acid distribution statistical information creation unit, a PTM distribution statistical information creation unit, a decoy sequence generation unit 135, and a decoy PTM information generation unit. 136, a subset decoy database construction unit 137, a whole set decoy database construction unit 138, and the like.
An example of the processing procedure for decoy database creation in the decoy database creation unit 13 is as follows.

  The database dividing unit 131 reads the target database 12 that is a database of modified proteins such as the well-known UniProtKB and the like, such as a human protein database, and does not have a common entry, that is, a plurality of entries without duplication. Are divided into a subset database (step S1). Next, the data reading unit 132 reads data from one of the plurality of subset databases (step S2).

  The amino acid distribution statistical information creating unit 133 examines the appearance frequency for each amino acid based on the amino acid sequence of each modified protein or each modified peptide in the read data, and creates amino acid distribution statistical information indicating the appearance frequency (step) S3). On the other hand, the PTM distribution statistical information creation unit 134 checks the appearance frequency of the modification site (the type of amino acid bound) for each post-translationally modified molecule in the modified protein or modified peptide in the read data, and the appearance frequency Is generated (step S4).

  Next, for each protein entry read by the data reading unit 132 in step S2, the decoy sequence generation unit 135 calculates a decoy sequence having the same sequence length (that is, the number of amino acid residues) as the amino acid sequence corresponding to the entry. It is created by randomly arranging amino acids under weighting according to the amino acid distribution statistical information created in step S3 (step S5). In addition, decoy sequences can be created by randomly replacing amino acids in regular amino acid sequences under weighting according to amino acid distribution statistical information, or by rearranging by permutation (not including sequence inversion). Good. Thereby, a fake decoy sequence corresponding to the appearance frequency of the amino acid appearing in the amino acid sequence of the regular modified protein or modified peptide is generated. However, as amino acid distribution statistical information based on the sequence of the subset database, amino acid distribution statistical information of a protein sequence that is a superset of the target sequence (for example, a registered protein sequence of an analysis target species registered in UniProtKB) may be used. Good.

  Next, the decoy PTM information generation unit 136 randomly modifies the decoy sequence generated in step S5 for each type of modification (amino acid residue) under weighting according to the PTM distribution statistical information generated in step S4. ) Is determined (step S6). Thereby, fake decoy PTM information according to the appearance frequency of the post-translationally modified molecule appearing in the regular modified protein or the modified peptide and the appearance frequency of the modified site is generated. In step S6, the weighting may be corrected so that the distribution of the number of modified sites (or the total number of modified sites of all peptides) matches between the peptide generated from the target sequence and the decoy sequence.

  The subset decoy database construction unit 137 collects the decoy sequence generated in step S5 and the decoy PTM information generated in step S6, and uses the modified protein or modified peptide sequence recorded in the original subset database. And a decoy subset database including fake information is created (step S7). Thereafter, it is determined whether or not the processing of steps S2 to S7 has been executed for all of the plurality of subset databases created by the division in step S1 (step S8). If there is an unprocessed subset database, the process returns from step S8 to S2, and the processes of steps S2 to S7 are executed for the unprocessed subset database. As a result, decoy subset databases are created corresponding to all the subset databases.

Further, the whole set decoy database construction unit 138 integrates the plurality of subset decoy databases generated in step S7, thereby creating a whole set decoy database corresponding to the target database before the division in step S1, and decoying it. Register as the database 14 (step S9).
Through the above processing, a decoy database corresponding to a database of modified proteins such as UniProtKB can be created.

Next, a specific example of creating a decoy database and its evaluation result will be described.
Here, based on the peptide sequence database (see Mosaiques DB, Non-patent Document 3) registered in the urinary peptide sequence database published by Mosaiques, Germany, the method described in Patent Document 2 is used. Peptide variant sequences (10 to 30 residues) created by stretching with reference to the full length sequence of the peptide precursor protein were collected to construct a peptide sequence database for identifying endogenous peptides. For this peptide sequence database, modified amino acids that had undergone oxidation (including hydroxylation), which is a PTM registered in the Mosaiques DB, were registered in the database as known post-translational modification information. Endogenous peptide identification by adopting the PEFF format (see Non-Patent Document 4) proposed by PSI (Proteomics Standards Initiative) as the database format and describing post-translational modification information in the header information of peptide sequence entries Created a target database for

  FIG. 5 is an example of the created endogenous peptide database. Here, using the PEFF format, the accession information includes the peptide precursor protein accession ID (P02641), the peptide N-terminal amino acid residue position (705) and the C-terminal amino acid on the precursor full-length sequence. Residue position (723). Also, post-translational modification information is listed in the ModResUnimod header key. For example, (8 | UNIMOD: 35 | Oxidation) in FIG. 5 means that proline, which is the eighth amino acid residue of the peptide, undergoes oxidation (Oxidation), which is a modification having ID No. 35 of Unimod. is doing.

  As processing corresponding to step S2 in FIG. 3, the amino acid sequence information and post-translational modification information of the registered peptide were read from the target database which is the endogenous peptide database. In addition, amino acid sequence information was read from Swiss-Prot, a superset of this target database. Next, as processing corresponding to step S3 in FIG. 3, with respect to the amino acid information read from Swiss-Prot and the target database, the appearance frequency of amino acids obtained by excluding overlapping amino acids on the protein sequence is calculated as amino acid distribution statistical information. Acquired as.

As a process corresponding to step S5 in FIG. 3, a sequence whose sequence length matches the peptide sequence registered in the target database is randomly generated based on the amino acid distribution statistical information read from the Swiss-Prot database. This was used as a decoy sequence. On the other hand, as a process corresponding to step S4 in FIG. 3, first, based on post-translational modification information obtained as described above, PTM appearance frequency (target sequence) for amino acids excluding overlapping amino acids on the protein sequence. PTM appearance frequency) was determined. The results are shown in Table 1. Here, Pro is proline, Lys is lysine, and Met is methionine. Next, the total number of registered PTM sites on the target database registered peptide sequence including duplicates was determined. The results are shown in Table 2.

Furthermore, the total number of amino acids to be modified on the decoy sequence (with overlap) obtained by generating the decoy sequence was determined. The results are shown in Table 3. In Table 3, the number in parentheses in the total column is a ratio to the number of amino acids to be modified in the target database. Thus, the total number of amino acids to be modified in the case of duplication does not match with a sufficient degree of reliability, which is a little less than 70% of the total number of amino acids to be modified in the target sequence.

Therefore, PTM sites were randomly selected from the amino acids to be modified on the decoy sequence according to the weighting by the PTM appearance frequency proportional to the target sequence PTM appearance frequency, and the total number of registered PTM sites on the decoy sequence (with overlap) was obtained. FIG. 6 is a graph showing the relationship between the ratio of the decoy sequence PTM appearance frequency to the target sequence PTM appearance frequency and the total number of registered PTM sites on the decoy sequence. From FIG. 5, it is estimated that when the ratio is about 4.5, the total number of registered PTM sites on the decoy array is comparable to that of the target database. Table 4 shows the number of decoy PTM sites in the decoy DB when this ratio is 4.5. In Table 4, the numerical value in parentheses in the total column is a ratio to the total number of registered PTM sites on the target database.

  In this way, the type and modification site of the modification obtained by setting the ratio of the decoy sequence PTM appearance frequency to the target sequence PTM appearance frequency as 4.5 was selected as decoy post-translational modification information for the decoy database. Then, by integrating the decoy sequence and decoy post-translation modification information obtained as described above, a decoy database in which the decoy sequence to which post-translation modification information was added was recorded was created.

  In order to evaluate the decoy database thus created, a comparison of FDR accuracy was performed on a peptide database in which the decoy database and the target database were combined. As a database to be compared, a peptide database without decoy post-translational modification information from which decoy post-translational modification information is removed from a decoy sequence, and a ratio of the decoy sequence PTM appearance frequency to the target sequence PTM appearance frequency when acquiring PTM distribution statistical information PTM database without PTM appearance frequency correction using a decoy sequence created by selecting 1.0 (that is, the position on the horizontal axis in FIG. 6 is 1.0).

  For each of the above three types of peptide databases, PTM using the search engine Comet was used for about 7000 MS / MS spectra derived from peptide mixtures in urine obtained by measurement with a liquid chromatograph mass spectrometer (LC-MS). A corresponding endogenous peptide search method was performed. Based on the number of decoy sequences identified thereby, the relationship between the number of intrinsic identification spectra and FDR was estimated based on the combined target decoy database method. The result is shown in FIG. From this result, within the range of 0.01 to 0.05, which is a practical value of FDR, a peptide database without decoy post-translational modification information or a peptide database without PTM appearance frequency correction was used for the peptide database created this time. In some cases, it became clear that the number of intrinsic identification spectra was overestimated. This is a result of underestimating the FDR because the PTM search space for the decoy sequence in the peptide database without modification information after decoy translation and the peptide database without PTM appearance frequency correction is too small compared to the target sequence.

  In addition, a decoy translation in which the decoy sequence is recorded in the same manner as described above, using a sequence database that records only the full-length sequence of the precursor protein (proprotein) of the target sequence registered in the peptide database as described above. A decoy database without post-modification information was created and integrated with the target database to create a proprotein database. This proprotein database is a database that is a superset of the peptide database, and the number of registered peptides was about 10 times that of the peptide database.

  For this proprotein database and peptide database, MS / MS spectra derived from peptide mixtures in urine measured by LC-MS using the search engine Comet were converted to the variable modification method (post-translational modification registered in the database). A comparison was made between the results of a search by a conventional PTM search method that does not use information and the search results of a peptide database by the above-described PTM-compatible endogenous peptide search method. The comparison result is shown in FIG. By performing the PTM-compatible endogenous peptide search method on the peptide database, the FDR is within a practical value range of 0.01 to 0.05 compared to the conventional PTM search by the variable modification method. It was confirmed that the identification sensitivity was improved.

  FIG. 9 summarizes the identification results obtained by the three methods shown in FIG. 8 in a Venn diagram. The set of amino acid sequences in the peptide database is a subset of the proprotein database, and the number of registered peptides is only about 1/10. From FIG. 9, among the peptides identified in the proprotein database, the identification reliability Most of the high results were identified in the peptide database, and it was revealed that the results had sufficient coverage.

  Furthermore, it has been found that the FDR for the result of duplication identified by the three methods compared is smaller than the FDR of the entire identification result. Such an evaluation can be realized by the identification quality evaluation method according to the present invention, in which all decoy-modified peptides in the peptide database searched by the PTM-compatible endogenous peptide search method are converted into the peptide database by the variable modification method. This is because the number of false discovery peptides can be estimated from the results of duplicate identification by each database and PTM search method.

  Next, a more detailed example of creating a decoy database will be described with reference to FIG. In this example, a decoy database is created for each target database having three types of common entries.

  The top-level database 12a including the most information among the three target databases is a human protein database such as UniProtKB. When the sample to be measured is an endogenous modified peptide (urinary peptide) present in human urine, a protein sequence database that is a subset of UniProtKB is used to identify urinary peptides reported in known literature. A collection of proprotein sequences is used as one subset database (protein sequence database). This is the second target database 12b. Furthermore, as described in Patent Document 1, urinary endogenous peptides reported in known literatures, etc., or partial sequences thereof are designated as endogenous peptide sequences generated by degradation in vivo. A record of peptide sequences containing more than the number is used as the lowest subset database (peptide sequence database). This is the third target database 12c.

  The above three protein or peptide sequence databases 12a, 12b, and 12c included in the target database 12 are sequences of UniProtKB or a subset thereof. Therefore, a common entry is included between the three databases. Therefore, when a peptide search based on peak information obtained from an MS / MS spectrum obtained by mass spectrometry (MS / MS analysis) of a peptide obtained by purifying a urine sample is performed, the target database 12 It is estimated that the same peptide is identified redundantly from a plurality of included databases.

  Actually, the amino acid composition of the sequence recorded in the protein sequence database 12b of urinary peptide, which is a subset database, and the amino acid composition of the sequence recorded in the peptide sequence database 12c, which is the subset database. There is little difference between them. Therefore, in this case, the amino acid distribution statistical information and the PTM distribution statistical information are created based on the protein sequence database 12b of urinary peptide, and this is stored in the decoy database corresponding to the protein sequence database 12b of urinary peptide. It is used not only for creating but also for creating a decoy database corresponding to the peptide sequence database 12c.

  On the other hand, the remainder obtained by removing the protein sequence database 12b of the urinary peptide from UniProtKB (the highest protein sequence database 12a) is a set of data that does not have a common entry with the protein sequence database 12b of the urine peptide. That is, this is the same as dividing the top-level protein sequence database 12a into two, one being a protein sequence database 12b of urinary peptides and the other being a difference set database 12d. Is also a subset database. The amino acid composition is different between the protein recorded in the difference set database 12d and the protein recorded in the protein sequence database 12b of the urinary peptide. Therefore, amino acid distribution statistical information and PTM distribution statistical information are created for the difference set database 12d separately from the protein sequence database 12b of urinary peptides.

Thus, after the amino acid distribution statistical information and the PTM distribution statistical information are created based on the protein sequence database 12b and the difference set database 12d of the urine peptide, respectively, the decoy having the same number of sequences as each protein entry of each database is used. The (amino acid) sequence is created according to the procedure described above. In addition, decoy PTM additional information is determined based on PTM distribution statistical information as decoy PTM additional information for each created decoy array. Then, the subset decoy database 14b and the difference set decoy database 14d are respectively created by integrating the created decoy sequence and the decoy PTM additional information with the data in the original database, and further integrated to create the top-level database 12a. The top-level decoy database 14a corresponding to is created.
In this way, a decoy database corresponding to UniProtKB can be created.

  It should be noted that the above embodiment is merely an example of the present invention, and it will be understood that the present invention is encompassed in the scope of the claims of the present application even if appropriate modifications, corrections, additions, etc. are made within the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 ... Endogenous modified peptide identification quality evaluation apparatus 11 ... Peptide identification part 12 ... Target database 12a ... Top-level protein sequence database 12b ... Subset protein sequence database 12c ... Peptide sequence database 12d ... Difference set database 13 ... Decoy database creation part 131 ... database division part 132 ... data reading part 133 ... amino acid distribution statistical information creation part 134 ... PTM distribution statistical information creation part 135 ... decoy sequence generation part 136 ... decoy PTM information generation part 137 ... subset decoy database construction part 138 ... whole set Decoy database construction unit 14 ... Decoy database 14a ... Top-level decoy database 14b ... Subset decoy protein database 14b ... Subset decoy peptide sequence database 14d ... Difference set data Lee database 15 ... false discovery rate estimator 2 ... mass analyzer 3 ... spectrum analysis unit

Claims (3)

A method for evaluating identification quality when identifying endogenous modified peptides by database search based on mass spectrometry results, and calculating an index value for evaluating identification quality using a decoy sequence. In
a) an information reading step for reading the amino acid sequence information and post-translational modification information of the modified protein recorded from the protein sequence database to which post-translational modification information is added;
b) Amino acid distribution statistics acquisition step for obtaining amino acid distribution statistical information indicating the appearance frequency of each amino acid based on the amino acid sequence information obtained by the information reading step;
c) based on the post-translational modification information obtained by the information reading step, a post-translational modification distribution statistics acquisition step for obtaining post-translational modification distribution statistical information indicating the frequency of appearance of each post-translational modification;
d) a decoy sequence generation step for generating a decoy sequence similar to a normal amino acid sequence based on the amino acid sequence information obtained by the information reading step and the amino acid distribution statistical information;
e) Generation of decoy post-translational modification information generating decoy post-translational modification information including the type and site of modification based on the decoy sequence generated in the decoy sequence generation step and the post-translational modification distribution statistical information Steps,
f) By integrating the decoy sequence generated in the decoy sequence generation step and the decoy post-translation modification information generated in the decoy post-translation modification information generation step, a decoy sequence to which post-translation modification information is added is recorded. A decoy database creation step for creating a decorated decoy database;
A method for evaluating the quality of identification of endogenous modified peptides, characterized by comprising: An identification quality evaluation method for an endogenous modified peptide according to claim 1, comprising:
A database dividing step of dividing the target database, which is a protein sequence database to which post-translational modification information is added, into a plurality of subset databases having no common entry;
For each of the plurality of subset databases, the information reading step, the amino acid distribution statistics acquisition step, the post-translational modification distribution statistics acquisition step, the decoy sequence generation step, the decoy post-translational modification information generation step, and the decoy database A subset decoy database creation step of creating a subset decoy database by performing the processing in the creation step;
A decoy database integration step of creating a decoy database by integrating a plurality of subset decoy databases created in the subset decoy database creation step in correspondence with the target database;
A method for evaluating the quality of identification of endogenous modified peptides, characterized by comprising: A device that evaluates the identification quality when identifying endogenous modified peptides by database search based on mass spectrometry results, and calculates the index value for evaluating the identification quality using decoy sequences. In the device
a) an information reading unit for reading amino acid sequence information and post-translational modification information of the recorded modified protein from a protein sequence database to which post-translational modification information is added;
b) based on the amino acid sequence information obtained by the information reading unit, amino acid distribution statistics acquisition unit for obtaining amino acid distribution statistical information indicating the appearance frequency of each amino acid;
c) based on the post-translational modification information obtained by the information reading unit, a post-translational modification distribution statistical acquisition unit for obtaining post-translational modification distribution statistical information indicating the appearance frequency of each post-translational modification;
d) a decoy sequence generation unit that generates a decoy sequence similar to a normal amino acid sequence based on the amino acid sequence information obtained by the information reading unit and the amino acid distribution statistical information;
e) Generation of decoy post-translational modification information that generates decoy post-translational modification information including the type and site of modification based on the decoy sequence generated by the decoy sequence generation unit and the post-translational modification distribution statistical information And
f) By integrating the decoy sequence generated by the decoy sequence generation unit and the decoy post-translation modification information generated by the decoy post-translation modification information generation unit, a decoy sequence to which post-translation modification information is added is recorded. A decoy database creation unit for creating a decorated decoy database;
An identification quality evaluation apparatus for endogenous modified peptides, comprising:

Images