CN111710365A - An ontology-based method for building protein/gene synonyms - Google Patents

Abstract

The protein/gene synonym table construction method based on the ontology of the present invention includes: a) acquisition of data sources Uniprot, BioGRID and NCBI Gene; b) segmentation of data files; c) establishment of upper-level ontology; d). The mapping and fusion of Swissprot to the upper ontology; e). The mapping and fusion of BioGRID to the upper ontology; f). The mapping and fusion of NCBI Gene to the upper ontology; g). The deduplication of synonyms. The protein/gene synonym table construction method of the present invention establishes a protein/gene synonym table with more comprehensive synonym scale, more reliable accuracy and more detailed classification information, which provides a prerequisite for more efficient and accurate literature data mining Guarantee is a powerful aid for biomedical experts to make scientific research discoveries.

Description

An ontology-based method for building protein/gene synonyms

technical field

The present invention relates to an ontology-based method for constructing a protein/gene synonym table, more particularly, to a method for constructing an ontology-based protein/gene synonym table using Uniprot-Swissprot, BioGRID and NCBI Gene as data sources.

Background technique

Synonyms, or more academically called synonyms, are a phenomenon that exists in every language in the world. It refers to words that express the same or similar meanings, but different expressions. Decades of biological research practice have resulted in inconsistent usage of some terms, with a large number of "synonymous" terms in datasets. Using different terms to describe the same gene function hinders the search for common features across multiple proteins and species. To make matters worse, the same term is used by scholars in different branches of biology to describe different molecules. Such unorganized data accumulation and acquisition will bring great challenges to the downstream users of the data flow and business flow. In order to ensure interoperability of systems or data, dispersing heterogeneous data requires a unified protein/gene synonym. Especially in the field of protein/gene, on the one hand, due to the complexity of its structure and function and the diversity of source organisms, the same protein may have different names in different functional fields and different species, and even the same protein in the same species. The same physiological process can also be named differently. On the other hand, the nomenclature of proteins/genes is constantly changing with the development of technology, making it more difficult to unify the nomenclature of proteins/genes.

At present, many biomedical synonym tables have been constructed and released in the field of biomedicine, and some institutions have added the attributes of synonyms to protein information on their websites. Common ones such as the information of more than 560,000 proteins included in Uniprot Swiss-Prot include gene synonym information for each protein. For another example, NCBI (National Center for Biotechnology Information, US National Center for Biotechnology Information) has constructed a biomedical subject thesaurus Mesh Terms (Medical Subject Headings Terms), which also includes synonym information for some proteins/genes; The possible synonyms for each gene are specifically indexed and published.

In addition to the special organization of experts in major institutions to collect and organize synonyms, many individual researchers are also constantly trying to automatically identify and extract synonyms from literature through automated means, so as to solve the problem that manual methods require a high professional threshold and cost A lot of manpower and material resources to read massive literature. For example, Morin and Jacquemin used term variants to detect synonyms, but many synonyms are not simple variants, or the two words that make up the variant are not synonyms. This situation is very common in the field of proteins/genes. For example, ABFB and AO090023000001 describe the same protein, but there is no variant characteristic between them; while the characteristics of variants between ABFB and ABFC are obvious, but They are different proteins. Snow et al. used an automatically constructed pattern extractor to obtain synonyms, but this method relies on a grammar parser, and the noise will start from word segmentation, to dependent syntactic parsing, to pattern construction, and until the final synonym discrimination, continuously superimposed and amplified, As a result, the final accuracy and coverage are not ideal.

In the final analysis, both the synonym sets published by various institutions and the published synonym extraction methods have the following shortcomings:

(1) The comprehensiveness of the synonym set is insufficient.

The vocabulary collections compiled by various institutions can be guaranteed to be highly authoritative, but there is still a lack of comprehensiveness. For example, for the protein "P2R3B_HUMAN", "NYREN8" is its synonym, but it is only included in BioGRID; "PPP2R3LY" is also its synonym, but it is only included in NCBI Gene. Due to the different emphases among various data sources, each research institution will focus on the research direction that it is good at when summarizing the synonym list. For example: Uniprot, as a data source focusing on protein sequence and structure, will mine and include more synonyms in protein literature related to sequences, but less attention to the representation of synonyms related to protein functions; BioGRID is used as PPI (Protein- Protein Interactions) data sources, more emphasis is placed on synonym representations in the literature describing protein interactions. The focus of each data source is different, which will inevitably lead to a lack of comprehensiveness in the sorting and collection of synonyms.

(2) The authority of the synonym set is insufficient.

The synonym list obtained by automated means is generally slightly more comprehensive than the existing public vocabulary set, but the accuracy rate is always difficult to match the public vocabulary set verified by the human hand, there will be deviations and errors, and the authority cannot be guaranteed.

(3) No distinction is made by species.

Proteins/genes need to be distinguished by species. When a protein with the same name appears in different species, its synonyms may differ.

One difference is that completely different synonyms appear across species. For example, when the protein/gene named "PUR9" appears in "HUMAN", there are synonyms such as "PURH", "IMPCHASE", "EL-S-70p", "AICARFT" and "ATIC"; but when When appearing in "JANSC" (A genus of the Rhodobacteraceae), there is only one synonym for "purH".

The difference, on the other hand, is that synonyms that are spelled the same will have different capitalization conventions depending on the species. Take "PUR9" as an example, when it appears in "HUMAN", its synonym is "PURH"; when it appears in "JANSC", its synonym is "purH".

Therefore, looking for a method that can effectively summarize, align and organize protein/gene synonyms from various data sources according to protein/gene and species information to form a more complete protein/gene synonym table is a more efficient and The premise and guarantee of accurate literature data mining is a powerful aid for biomedical experts to conduct scientific research and discovery, and is an urgent work in the field of biomedicine.

SUMMARY OF THE INVENTION

In order to overcome the shortcomings of the above technical problems, the present invention provides an ontology-based method for constructing a protein/gene synonym table.

The present invention is realized through the following steps:

a). Acquisition of data sources, to obtain the most authoritative data sources of Uniprot-Swissprot, BioGRID and NCBIGene in the field of biomedicine. The files obtained from the three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene are respectively uniprot-sprot. xml, BIOGRID-ALL-LATSET.psi25.xml and ALL_Date.gene_info;

b). Segmentation of data files. For the xml format file obtained in step a), the segmentation is performed by reading line by line, and the tags identifying the start and end are detected, and each pair of start and end tags and their containing The part of the file is stored as a single entity file; for a single file obtained from the BIOGRID-ALL-LATSET.psi25.xml file with the tags starting and ending with the entity, only the file corresponding to the "Interactor" entity is kept, ignoring "Interaction" and "Interaction" The file corresponding to the Experiment" entity; for ALL_Date.gene_info, since each line stores one entity, it is not divided, and a single entity file is obtained by directly reading line by line;

c). The establishment of the upper-level ontology based on the Uniprot-Swissprot data source, referring to the actual properties of the protein/gene in Uniprot, the construction includes the protein/gene name Name, the TaxID representing the species type id corresponding to the protein/gene, and the species type name Upper ontologies including Specie and protein/gene synonyms Synomym;

d). The mapping and fusion of Uniprot-Swissprot to the upper ontology, mapping the "name" field of the protein in Uniprot to the "Name" field of the upper ontology, and the "NCBI TaxonomyID" field in Uniprot to the "TaxID" field of the upper ontology , the "gene" field in Uniprot is mapped to the "Synomym" field of the upper ontology, until all the proteins and attributes in the Uniprot-Swissprot data source are mapped, so that the protein instances and attributes in the Uniprot data source are all fused into the upper ontology ;

e). For the mapping and fusion of BioGRID to the upper-level ontology, first find the protein corresponding to Uniprot in the reference field of BioGRID, and map it with the accession field in Uniprot, so that with the help of the mapping established in step d), the BioGRID The attribute information of the protein is fused into the upper ontology;

If the corresponding protein in Uniprot cannot be found through the reference field of BioGRID, it indicates that there is no instance of the protein in the upper ontology, then the protein instance needs to be added to the upper ontology, and the "shortLabel" of BioGRID is used to align to the "Name" of the upper ontology 、The "ncbiTaxld" of BioGRID is aligned to the "TaxID" of the upper-level ontology, and the "Synonym" of BioGRID is aligned to the "Synonym" of the upper-level ontology to expand the upper-level ontology;

f). For the mapping and fusion of NCBI Gene to the upper ontology, first convert all the synonyms of proteins in NCBI Gene and proteins in the upper ontology to lowercase or uppercase, and then combine the synonym set of each NCBI Gene protein instance with each synonym in the upper ontology The synonym sets of a protein instance are compared, and the protein instance pairs with co-occurring synonyms are screened out. Finally, according to the species information, the protein instance pairs of the same species are retained, and the attribute information of the NCBI Gene protein instance is fused into the upper ontology;

If there is no synonym between a certain protein in NCBI Gene and the protein in the upper ontology, it means that the protein is not covered in the upper ontology established in step e), then the "Symbol" of NCBI Gene is used to "Name" of the upper ontology. "Alignment, NCBI Gene's "taxID" is aligned to the upper-level ontology's "TaxID", and NCBI Gene's "Synonym" is aligned to the upper-level ontology's "Synonym" to extend the upper-level ontology;

g). De-duplication of synonyms and export of synonym tables. After mapping and merging the protein instances in the three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene, the respective synonym information is extracted. After de-duplication, the obtained The synonym table with species classification information is stored in the synonym table in the upper ontology, that is, the construction of the protein/gene synonym table is realized.

In the ontology-based protein/gene synonym table construction method of the present invention, in step a), there are two ways to obtain data source data. One is to use crawler or API to capture and analyze the protein information on each data source website online. Download; the other is to obtain open source data files from a website or FTP and analyze the file structure locally.

In the ontology-based protein/gene synonym table construction method of the present invention, the fusion or alignment of the BioGRID described in step e) to the upper ontology is achieved through the following steps:

e-1). Determine whether the protein in BioGRID has a corresponding protein in Uniprot, look up the information in the "primaryRef" or "secondaryRef" field in BioGRID, and find the information in the "accession" field of the Uniprot protein contained in the upper ontology. Whether it appears, if it does, it means that the two proteins that appear in BioGRID and Uniprot can be fused, go to step e-2), if not, expand the protein in BioGRID, go to step e-4);

e-2). Fusion of shortLabel values. If the "shortLabel" of the BioGRID protein is different from the "Name" value in the upper body to be fused, keep the "Name" value of the upper body unchanged, and add "shortLabel" to In the "Synonym" of the upper ontology;

e-3). Fusion of Synonym values. If the "Synonym" of the BioGRID protein is exactly the same as the "Synonym" in the upper ontology, keep the "Synonym" in the upper ontology unchanged. "Synonym" is merged with the "Synonym" synonym in the upper ontology to remove duplicates, and is stored in the upper ontology as a new "Synonym" field value;

e-4). The extension of BioGRID to the upper-level ontology protein, using the "shortLabel" of BioGRID to map to the "Name" of the upper-level ontology, the "ncbiTaxld" of BioGRID to map to the "TaxID" of the upper-level ontology, and the "Synonym" of BioGRID to the upper-level ontology The upper-level ontology is expanded by means of the "Synonym" mapping.

In the ontology-based protein/gene synonym table construction method of the present invention, the fusion or alignment of the NCBI Gene described in step f) to the upper ontology is achieved through the following steps:

f-1). Extract the species and synonym information, respectively extract the species information TaxID and synonym information Synonym in the protein in NCBI Gene and the upper body protein, and convert all the synonyms to lowercase;

f-2). Match synonyms, match the synonyms in each NCBI Gene protein to the synonyms in each upper-level ontology protein, if there is an intersection between the lowercase synonym sets between the two, the upper-level ontology protein with the intersection will be used as the upper-level ontology protein. Candidate matching items of NCBI Gene protein; if there is no intersection between the two, the corresponding NCBIGene protein should be expanded to the protein in the upper body, and step f-6);

f-3). Screening of candidate matching items, by judging whether the species information TaxID of the NCBI Gene protein and the species information TaxID in the candidate matching items identify the same species, if they are the same species, it indicates that the NCBI Gene protein and the upper ontology The protein can be fused, and step f-4) is performed; if it is a different species, it means that the NCBI Gene protein cannot be fused with the upper body protein. step f-6);

f-4). Fusion of shortLabel values, if the "Symbol" in the NCBI Gene protein is different from the "Name" value in the upper ontology to be fused, keep the "Name" value of the upper ontology to be fused unchanged, and set the "Symbol" is added to the "Synonym" of the upper ontology to be merged;

f-5). Fusion of Synonym values. If the "Synonym" in the NCBI Gene protein is exactly the same as the "Synonym" in the upper ontology to be fused, keep the "Synonym" in the upper ontology unchanged. Then the "Synonym" in the NCBI Gene protein is merged with the "Synonym" synonym in the upper-level ontology and deduplicated, and then stored in the upper-level ontology as a new "Synonym" field value;

f-6). NCBI Gene protein expansion, using the "Symbol" of NCBI Gene to align to the "Name" of the upper-level ontology, the "taxID" of NCBI Gene to the "TaxID" of the upper-level ontology, and the "Synonym" of NCBI Gene to the upper-level The "Synonym" alignment of the ontology extends the upper ontology.

The beneficial effects of the present invention are as follows: the ontology-based protein/gene synonym table construction method of the present invention uses the most authoritative Uniprot-Swissprot, BioGRID and NCBI Gene in the field of biomedicine as the data sources, Identify a single protein file in each data source; secondly establish an upper-level ontology including the name of the protein/gene, the TaxID of the species type id, the type name Specie, and the Synomym attribute of the thesaurus; again, align the proteins in Uniprot one by one. To the upper ontology, then fuse or align the proteins in BioGRID to the upper ontology; finally, align the proteins in NCBI Gene to the upper ontology, thereby establishing a protein with more comprehensive synonym scale, more reliable accuracy, and more detailed classification information. /Gene Thesaurus provides a more comprehensive and authoritative data query and reference for the specification, storage, reasoning and application of the protein/gene synonym, and provides a prerequisite for more efficient and accurate literature data mining. It is a biomedical A powerful aid for experts to make scientific discoveries.

Description of drawings

Fig. 1 is the overall flow chart of the protein/gene synonym table construction method based on ontology of the present invention;

Fig. 2 is the key attribute map of three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene in the present invention;

Fig. 3 is the attribute alignment mapping diagram of three data sources to upper-level ontology in the present invention;

4 is a schematic diagram of candidate matching items after the synonyms in the NCBI Gene protein and the upper-layer ontology are matched by lowercase synonyms in the present invention;

FIG. 5 is the final protein matching result of the candidate matching items in FIG. 4 after species matching.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and embodiments.

Existing protein/gene synonyms hardly guarantee authoritative and comprehensive questions. For synonym extraction using automated means for biomedical literature, no matter how high the accuracy of the model is, it is difficult to ensure the comprehensiveness of the constructed synonym table without the lack of data benchmarks. In the case of long-term use and verification, it is also difficult to verify the authority of the synonym table.

Data extraction and entity alignment problems. The technical solution used in this paper selects the three most commonly used and authoritative data sources in the field of biomedicine: Uniprot-Swissprot, BioGRID and NCBI Gene as the data base for synonym construction. However, these three data sources have different data fields, different storage formats, different protein/gene names as main names, and different positions and field names of synonym fields. Therefore, it is necessary to align and fuse the protein/gene entities in each data source, so that the proteins in each data source have the same semantic structure, thereby establishing protein mapping.

Synonym classification problem for species. Due to differences in names, capitalization, etc., in the naming of proteins/genes between different species. To establish an effective and usable synonym list, it is necessary to solve the problem of how to effectively classify proteins/genes and their synonyms by species.

The current protein/gene nomenclature lacks mandatory and complete norms. When extracting from text or simply performing synonym matching, there are the following difficulties:

Internationally, the naming of proteins and genes in commonly used species has been standardized, but the remaining species still lack standardization, and the naming is confusing. Species for which nomenclature has been established include: mice, rats, chickens, humans, non-human primates, livestock, flies (Drosophila), fish. The protein and gene names of other species are not clearly defined.

Some academic institutions and journals use their own naming conventions. For example, the Guangdong University of Pharmacy stipulates that the case of gene names and protein names is opposite and the protein name at least the first letter is capitalized, and the existence of exceptions is allowed.

In the naming convention, capitalization, italics, and Greek letters are all important bases for gene/protein naming. The general international nomenclature is shown in Table 1 on the use of upper and lower case, italics and Greek letters:

Table 1

However, in the process of electronic storage of text information, often only the upper and lower case information of the naming is retained, and the main attribute of italics is lost, which makes the alignment of synonym naming and species classification very difficult.

The ontology-based protein/gene synonym table construction method of the present invention is characterized in that, it is realized by the following steps:

a). Acquisition of data sources, to obtain the most authoritative data sources of Uniprot-Swissprot, BioGRID and NCBIGene in the field of biomedicine. The files obtained from the three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene are respectively uniprot-sprot. xml, BIOGRID-ALL-LATSET.psi25.xml and ALL_Date.gene_info;

Sequence aspects: Uniprot-Swissprot. Uniprot is an internationally renowned protein database, and is currently the non-redundant protein sequence database with the most complete sequence data and richest annotation information in the world. His data mainly comes from the protein sequences obtained after the completion of the genome sequencing project. where Swissprot is an expert-verified dataset containing checked, manually annotated entries.

Functional: BioGRID. BioGRID is an old and classic protein interaction database, and it has been continuously updated for a long time. The database screened and sorted out protein interaction information, chimera information and PTM information of post-transcriptional modifications from biomedical literature. Because of its focus on protein interaction, its description of protein entities and synonyms focuses more on the functions and roles of proteins and their transcribed genes.

Genetics: NCBI Gene. The full name of NCBI is the National Center for Biotechnology Information. NCBI Gene is a gene resource collected by NCBI. It integrates all relevant information such as pathway, variation and phenotype with each gene as a unit, involving sequence information from thousands of different organisms, representing chromosomes, organelles, plasmids, viruses , transcription templates and millions of protein profiles.

Ensuring that the obtained data sources are comprehensive and authoritative is the basis for building a synonym table. Therefore, the above three most authoritative data sources in the field of biomedicine are selected to complete synonyms from different emphases.

In this step, there are two ways to obtain data source data. One is to use crawler or API to capture and download entity information on each data source website online; the other is to obtain open source data files from websites or FTP. , which analyzes the file structure locally. Both methods have pros and cons. The online crawling method is slow, seriously affected by network conditions, and has poor stability, but it can obtain the latest entity information without being restricted by website file updates. The method of downloading data files to the local is fast, stable and durable, but on the one hand, it is limited by the update frequency of the data source files, and the obtained data is often not the latest data. On the other hand, the file formats of different data sources are different. The internal organization of the file is also different, and it is often necessary to formulate its own extraction strategy for each data source. This paper adopts the method of downloading data files to the local. The files obtained from the three data sources are uniprot-sprot.xml, BIOGRID-ALL-LATSET.psi25.xml and ALL_Date.gene_info.

The Uniprot website provides the download of Uniprot KB and its sub-database data source files. The download address is: https://www.uniprot.org/downloads. In this article, we choose to download the Swiss-Prot sub-database data verified by experts, and choose the xml file format to obtain more semantic relations between upper and lower levels. The downloaded file is named "uniprot-sprot.xml". BioGRID provides a complete data source file download, the download address is: https://downloads.thebiogrid.org/BioGRID/Latest-Release/, in this article, select the latest release of BIOGRID-ALL-LATEST.psi25.zip with psi25 The data file "BIOGRID-ALL-LATEST.psi25.xml" stored in xml format.

The data files of NCBI can be downloaded from the official FTP server of NCBI. The address of the Gene file is: ftp://ftp.ncbi.nlm.nih.gov/gene/DATA/GENE_INFO/,

Select the ALL_Data.gene_info.gz file to download, decompress it to obtain the ALL_Data.gene_info file, which is essentially a tab-delimited text file. Except the first line is used to describe the field name, each line stores a gene's Entity attribute information.

b). Segmentation of data files. For the xml format file obtained in step a), the segmentation adopts the method of reading line by line, and detects the tags that identify the start and end of the entity, and divides each pair of start and end tags and their Included sections are stored as a single entity file; for a single file obtained from the BIOGRID-ALL-LATSET.psi25.xml file with tags starting and ending with the entity, only the files corresponding to the "Interactor" entity are kept, ignoring "Interaction" and The file corresponding to the "Experiment" entity; for ALL_Date.gene_info, since each line stores one entity, it is not divided, and a single entity file is obtained by directly reading line by line;

Since the downloaded data file is large (Uniprot-sprot is 6.26G, BioGRID is 6.46G, and NCBI Gene is 3.29G), it is difficult to directly process it, so it is necessary to divide the data file and divide it according to entities into small files that store a single entity. document. For the xml file, the segmentation adopts the way of reading line by line, and detects the tags that identify the start and end of the entity, such as "<entry>" and "</entry>" in Uniprot, and converts each pair of start and end tags and the parts it contains are stored as a single file.

c). The establishment of the upper-level ontology based on the Uniprot-Swissprot data source, referring to the actual properties of the protein/gene in Uniprot, including the Name representing the protein/gene, the TaxID representing the species type id corresponding to the protein/gene, the representation The upper-level ontology including the type name Specie of the species corresponding to the protein/gene and the synonym table Synomym representing the protein/gene;

In order to be able to fuse the synonym information of the three data sources, the proteins in the three data sources need to be aligned. The method adopted is to build a higher-level ontology, and fuse the different protein descriptions in the three data sources to the proteins in the ontology. It is required that the ontology schema can establish a semantic description standard for entities and various attribute fields of the entity, and can standardize the storage standard of data values. Requirements:

(1) Ontology can give standardized labels and names of instances (such as P53-HUMAN) by defining classes and adding instances;

(2) The ontology can give the normalized property field name contained in the instance by defining the object property (ObjectProperty) and the numerical property (DataProperty);

(3) Ontology can provide the type, value range, unit of normalized attribute value, and the storage format specification of attribute value by defining numerical attributes;

(4) Ontology can provide a unified definition of semantic relationship, format and representation of relationship.

Indicates the name of the protein/gene Name indicates the common name of the protein/gene or the first name proposed in the historical literature. The name is based on the value of the name field in Uniprot. The format is: <pgso:name rdf:datatype="http://www.w3.org/2001/XMLSchema#string">P53_HUMAN </pgso:name >.

Indicates the TaxID of the species type id corresponding to the protein/gene, which is uniformly represented by the Id in NCBI Taxonomy. The format is: <pgso:taxId rdf:datatype="http://www.w3.org/2001/XMLSchema#string">9606</pgso:taxId>; indicates the type name Specie of the species corresponding to the protein/gene, It is represented by the species abbreviation in Uniprot, and it is matched with NCBI Taxonomy and its taxId one by one. The format is: <pgso:specierdf:datatype="http://www.w3.org/2001/XMLSchema#string">HUMAN</pgso:specie>; and the synonym table Synomym representing proteins/genes, the synonym table should be at least Contains the name of the name field, and each synonym is stored with a separate owl ontology tag. The format is:

<pgso:synonym rdf:datatype="http://www.w3.org/2001/XMLSchema#string">P53</pgso:synonym>

<pgso:synonym rdf:datatype="http://www.w3.org/2001/XMLSchema#string">TP53</pgso:synonym>

<pgso:synonym rdf:datatype="http://www.w3.org/2001/XMLSchema#string">LFS1</pgso:synonym>

...

d). The mapping and fusion of Uniprot-Swissprot to the upper ontology, mapping the "name" field in the protein in Uniprot to the "Name" field of the upper ontology, and the "NCBI TaxonomyID" field in Uniprot to the "TaxID" of the upper ontology field, the "gene" field in Uniprot is mapped to the "Synomym" field of the upper ontology, until all the proteins and attributes in the Uniprot-Swissprot data source are mapped, so that the protein instances and attributes in the Uniprot data source are all fused to the upper ontology middle;

Due to the normative, comprehensive and authoritative nature of the data in Uniprot, the schema of Uniprot is extracted as the main entities and attributes in the original upper-level ontology, and the entities and attributes of other data sources are aligned with this model. For entities or attributes that have not yet been covered, the upper-level ontology is expanded on this basis.

e). For the mapping and fusion of BioGRID to the upper-level ontology, first find the protein corresponding to Uniprot in the reference field of BioGRID, and map it with the accession field in Uniprot, so that with the help of the mapping established in step d), the BioGRID The attribute information of the protein is fused into the upper ontology;

If the corresponding protein in Uniprot cannot be found through the reference field of BioGRID, it indicates that there is no instance of the protein in the upper ontology, then the protein instance needs to be added to the upper ontology, and the "shortLabel" of BioGRID is used to align to the "Name" of the upper ontology , BioGRID's "ncbiTaxld" is aligned to the "TaxID" of the upper-level ontology, and the "Synonym" of BioGRID is aligned to the "Synonym" of the upper-level ontology to extend the upper-level ontology; Figure 2 shows the relationship between the three data sources. Attribute association and mapping relationship. It can be known that the "primaryRef" or "secondaryRef" field in BioGRID records the "accession" field information of the Uniprot referenced by the protein. For example, in the data source BioGRID, the "shortLabel" attribute value of a protein is "HSPA5", and its "secondaryRef" attribute value contains four accession information such as "Q9UK02", "Q2EF78", "B0QZ61", "Q9NPF1", etc. Accession can find the corresponding protein in Uniprot, and the fusion of the protein instance in BioGRID to the protein instance in Uniprot can be established.

It should be noted that a Uniprot protein may have multiple accessions, and a BioGRID protein also corresponds to multiple accession references. Uniprot and BioGRID are actually many-to-many fusion relationships, and each pair of fused relationships needs to be mapped once.

In this step, the fusion or alignment of the BioGRID to the upper-level ontology is achieved through the following steps:

e-1). Determine whether there is a corresponding Uniprot for the BioGRID protein, look for the information in the "primaryRef" or "secondaryRef" field in the BioGRID, and whether it appears in the "accession" field information of the Uniprot protein contained in the upper ontology, if If there is, it means that the two proteins in BioGRID and Uniprot can be fused, and step e-2) is performed. If not, the protein in BioGRID should be expanded, and step e-4) is performed;

e-2). Fusion of shortLabel values, if the "shortLabel" in the BioGRID protein is different from the "Name" value in the upper body to be fused, keep the "Name" value of the upper body unchanged, and add "shortLabel" to the "Synonym" of the upper ontology;

e-3). Fusion of Synonym values. If the "Synonym" in the BioGRID protein is exactly the same as the "Synonym" in the upper ontology, keep the "Synonym" in the upper ontology unchanged. The "Synonym" in the upper-level ontology is combined with the "Synonym" synonym in the upper-level ontology to remove duplicates, and is stored in the upper-level ontology as a new "Synonym" field value;

e-4). The expansion of BioGRID proteins, using the "shortLabel" of BioGRID to map to the "Name" of the upper-level ontology, the "ncbiTaxld" of BioGRID to map to the "TaxID" of the upper-level ontology, and the "Synonym" of BioGRID to the "Synonym" of the upper-level ontology "Mapping to expand the upper ontology.

f). For the mapping and fusion of NCBI Gene to the upper ontology, first convert all the synonyms in the protein in the NCBI Gene and the protein in the upper ontology to lowercase or uppercase, and then combine the synonym set of each NCBI Gene protein instance with the synonyms in the upper ontology The synonym sets of each protein instance are compared, and the protein instance pairs with co-occurring synonyms are screened out. Finally, according to the species information, the protein instance pairs of the same species are retained, and the attribute information of the NCBI Gene protein instance is fused into the upper ontology. ;

If there is no synonym between a protein in NCBI Gene and the upper ontology, it means that the protein is not covered in the upper ontology established in step e), then use the "Symbol" of NCBI Gene to align to the "Name" of the upper ontology, The "taxID" of NCBI Gene is aligned with the "TaxID" of the upper-level ontology, and the "Synonym" of NCBI Gene is aligned with the "Synonym" of the upper-level ontology to extend the upper-level ontology; NCBI Gene itself is difficult to communicate with Uniprot, BioGRID and the upper-level ontology. Directly establish the alignment relationship of proteins. Therefore, it is necessary to start with synonyms and species information, and match NCBIGene with the upper ontology. First, the synonyms of the NCBI Gene protein and the protein in the upper ontology should be converted to lowercase (or uppercase). The upper ontologies can ignore species information for matching. Second, the synonym set of each protein instance in NCBI Gene is compared with the synonym set of each protein in the upper ontology, and the proteins with co-occurring synonyms are screened out. In this way, each NCBI Gene protein can match several proteins in the candidate upper ontology (or may not match at all). Finally, according to the species information of the NCBI Gene protein, the protein of the same species is searched among the proteins of the candidate upper ontology. If present, the two proteins can be aligned.

As can be seen from Figure 2, there is no obvious reference or pointing relationship between NCBI Gene, Uniprot and BioGRID, so it is impossible to directly determine the alignment of proteins through the alignment of key attribute fields as in the case of BioGRID fusion. Therefore, it is necessary to combine the existing attribute values of NCBI Gene and the upper ontology to determine the alignment relationship of proteins, that is, to reflect the matching mapping relationship between proteins by matching the two key information of "TaxId" and "Synonym".

The fusion or alignment of the NCBI Gene to the upper-level ontology described in this step is achieved through the following steps:

f-1). Extract species and synonym information, respectively extract species information TaxID and synonym information Synonym in NCBI Gene protein and upper-level ontology protein, and convert all synonyms to lowercase;

f-2). Match synonyms, match the synonym set of each NCBI Gene protein instance to the protein instance synonym set in each upper ontology, if there is an intersection between the lowercase synonym sets between the two, there will be an upper layer of the intersection. The ontology instance is used as the candidate matching item of the NCBI Gene instance; if there is no intersection between the two, the protein instance in the NCBIGene should be expanded to the protein in the upper ontology, and step f-6);

f-3). Screening of candidate matching items, by judging whether the species information TaxID of the NCBI Gene protein and the species information TaxID in the candidate matching items identify the same species, if they are the same species, it indicates that the NCBI Gene instance and the upper ontology Instances can be fused, and step f-4) is performed; if they are different species, they cannot be fused, and if all candidate matching items are judged, they cannot be fused, then step f-6) is performed;

f-4). Fusion of shortLabel values, if the "Symbol" in the NCBI Gene protein is different from the "Name" value in the upper ontology to be fused, keep the "Name" value of the upper ontology to be fused unchanged, and set the "Symbol" is added to the "Synonym" of the upper ontology to be merged;

f-5). Fusion of Synonym values. If the "Synonym" in the NCBI Gene protein is exactly the same as the "Synonym" in the upper ontology to be fused, keep the "Synonym" in the upper ontology unchanged. Then the "Synonym" in the NCBI Gene protein is merged with the "Synonym" synonym in the upper-level ontology and deduplicated, and then stored in the upper-level ontology as a new "Synonym" field value;

f-6). The extension of NCBI Gene protein to the upper-level ontology, using the "Symbol" of NCBI Gene to align to the "Name" of the upper-level ontology, the "taxID" of NCBI Gene to the "TaxID" of the upper-level ontology, and the "Synonym" of NCBI Gene ” extends the upper-level ontology by aligning the “Synonym” of the upper-level ontology.

As shown in Figure 4, a schematic diagram of candidate matching items after the synonyms in the NCBI Gene protein and the upper-level ontology are matched by lowercase synonyms in the present invention is given, the right side is the synonyms and species in the upper-level ontology, the left side is from the data source The synonyms and species of NCBI Gene, filter out the matching combination of NCBI Gene and the synonym set of the upper-level ontology that have intersections, and then the candidate matching items of the NCBI Gene protein can be obtained, that is, the connecting line part in the figure. It can be seen that for 3 of the NCBI Gene For proteins, there are multiple candidate matches. As shown in Figure 5, the final protein matching results of the candidate matching items in Figure 4 after species matching are given. After the candidate matching items are filtered by species information, the filtering results are shown in the dotted line in Figure 5.

g). Deduplication of synonyms and export of the synonym table. After the proteins in the three data sources of Uniprot-Swissprot, BioGRID and NCBIGene are aligned and fused, the respective synonym information is extracted. There is a synonym table with species classification information, and it is stored in the synonym table in the upper ontology, that is, the construction of the protein/gene synonym table is realized.

visible:

(1) The present invention proposes a method for constructing a gene/protein synonym table, which can effectively integrate synonym information from multiple authoritative data sources, and classify synonyms according to species, so as to achieve a more comprehensive synonym scale and higher accuracy. More reliable and more detailed in classification information.

(2) Different from the traditional synonym construction method, the present invention uses ontology as the semantic specification of multi-source heterogeneous data in the construction of the synonym table, and formulates the semantic mapping of proteins containing synonyms first, and then the fusion of protein synonyms. Then, the semantic standardization construction process of the synonym table is constructed. The use of semantic mapping technology can significantly increase the accuracy of synonym matching; the introduction of ontology can have stronger advantages in the specification, storage, reasoning and application of synonym tables.

(3) The present invention adopts a fusion method of establishing an upper-layer ontology independent of the data source, and establishes a clear fusion direction. The construction of the upper-level ontology means that the synonym information, as knowledge itself, is represented, stored and used in accordance with the international general knowledge representation standard. The existence of the upper-level ontology also enables other data sources to be easily integrated into the existing synonym list through a unified semantic standard, not only limited to the three data sources mentioned in the present invention, and the scalability is greatly enhanced .

Claims (4)

1. a protein/gene synonym table construction method based on ontology, is characterized in that, realizes through the following steps: a). Acquisition of data sources, to obtain the most authoritative data sources of Uniprot-Swissprot, BioGRID and NCBIGene in the field of biomedicine. The files obtained from the three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene are respectively uniprot-sprot. xml, BIOGRID-ALL-LATSET.psi25.xml and ALL_Date.gene_info; b). Segmentation of data files. For the xml format file obtained in step a), the segmentation adopts the method of reading line by line, and detects the tags that identify the start and end of the entity, and divides each pair of start and end tags and their Included sections are stored as a single entity file; for a single file obtained from the BIOGRID-ALL-LATSET.psi25.xml file with tags starting and ending with the entity, only the files corresponding to the "Interactor" entity are kept, ignoring "Interaction" and The file corresponding to the "Experiment" entity; for ALL_Date.gene_info, since each line stores one entity, it is not divided, and a single entity file is obtained by directly reading line by line; c). The establishment of the upper ontology, according to the actual attributes of the protein/gene, including the name of the protein/gene, the TaxID of the species type id corresponding to the protein/gene, and the type name of the species corresponding to the protein/gene Specie And the upper-level ontology including the synonym Synomym representing proteins/genes; d). The mapping and fusion of Uniprot-Swissprot to the upper ontology, mapping the "name" field of the protein in Uniprot to the "Name" field of the upper ontology, and the "NCBI TaxonomyID" field in Uniprot to the "TaxID" field of the upper ontology , the "gene" field in Uniprot is mapped to the "Synomym" field of the upper ontology, until all the proteins and attributes in the Uniprot-Swissprot data source are mapped, so that all protein instances in the Uniprot data source are merged into the upper ontology; e). The mapping and fusion of BioGRID to the upper-level ontology. First, look for the protein corresponding to Uniprot in the reference field of BioGRID, and map it with the accession field in Uniprot, so that with the help of the mapping established in step d), the BioGRID The attribute information of the protein is fused into the upper ontology; If the corresponding protein in the corresponding Uniprot cannot be found through the reference field of BioGRID, it indicates that there is no instance of the protein in the upper ontology, then the protein instance needs to be added to the upper ontology, and the "shortLabel" of BioGRID is used to "Name" of the upper ontology. ” alignment, the “ncbiTaxld” of BioGRID is aligned to the “TaxID” of the upper-level ontology, and the “Synonym” of BioGRID is aligned to the “Synonym” of the upper-level ontology to extend the upper-level ontology; f). For the mapping and fusion of NCBI Gene to the upper ontology, first convert all the synonyms of proteins in NCBI Gene and proteins in the upper ontology to lowercase or uppercase, and then combine the synonym set of each NCBI Gene protein with each of the upper ontology The synonym sets of proteins are compared, and the protein pairs with co-occurring synonyms are screened out. Finally, according to the species information of the protein, the protein pairs of the same species are retained, and the attribute information of the NCBI Gene protein is fused into the upper ontology; If there is no synonym between a protein in NCBI Gene and the upper ontology, it means that the protein is not covered in the upper ontology established in step e), then use the "Symbol" of NCBI Gene to align to the "Name" of the upper ontology, The "taxID" of NCBI Gene is aligned to the "TaxID" of the upper-level ontology, and the "Synonym" of NCBI Gene is aligned to the "Synonym" of the upper-level ontology to expand the upper-level ontology; g). Deduplication of synonyms and export of synonym tables. After mapping and fusion of the proteins in the three data sources of Uniprot-Swissprot, BioGRID and NCBI Gene, the respective synonym information is extracted. There is a synonym table with species classification information, and it is stored in the synonym table in the upper ontology, that is, the construction of the protein/gene synonym table is realized. 2. the protein/gene synonym table construction method based on ontology according to claim 1, is characterized in that: in step a), the mode that obtains data source data has two kinds, and a kind of is to utilize crawler or API, online to each The protein information on the data source website is captured and downloaded; the other is to obtain open source data files from the website or FTP, and analyze the file structure locally. 3. ontology-based protein/gene synonym table construction method according to claim 1 and 2 is characterized in that, the fusion or alignment of the BioGRID described in step e) to the upper ontology is realized by the following steps: e-1). Determine whether the protein in BioGRID has a corresponding protein in Uniprot, look up the information in the "primaryRef" or "secondaryRef" field in BioGRID, and find the information in the "accession" field of the Uniprot protein contained in the upper ontology. Whether it appears, if it does, it means that the two proteins that appear in BioGRID and Uniprot can be fused, go to step e-2), if not, expand the protein in BioGRID, go to step e-4); e-2). Fusion of shortLabel values, if the "shortLabel" in the BioGRID protein is different from the "Name" value in the upper body to be fused, keep the "Name" value of the upper body unchanged, and add "shortLabel" to the "Synonym" of the upper ontology; e-3). Fusion of Synonym values. If the "Synonym" in the BioGRID protein is exactly the same as the "Synonym" in the upper ontology, keep the "Synonym" in the upper ontology unchanged. The "Synonym" in the upper-level ontology is combined with the "Synonym" synonym in the upper-level ontology to remove duplicates, and is stored in the upper-level ontology as a new "Synonym" field value; e-4). The expansion of the upper ontology by the BioGRID protein instance uses the "shortLabel" of BioGRID to map to the "Name" of the upper ontology, the "ncbiTaxld" of BioGRID to the "TaxID" of the upper ontology, and the "Synonym" of BioGRID to the upper ontology The "Synonym" mapping of the ontology extends the upper-level ontology. 4. the protein/gene synonym table construction method based on ontology according to claim 1 and 2 is characterized in that, the fusion or alignment of the NCBI Gene described in step f) to the upper ontology is realized by the following steps: f-1). Extract species and synonym information, respectively extract species information TaxID and synonym information Synonym in NCBI Gene protein and upper-level ontology protein, and convert all synonyms to lowercase; f-2). To match synonyms, match the synonym set of each NCBI Gene protein to the synonym set of each upper-level ontology protein. If there is an intersection between the lowercase synonym sets between the two, the upper-level ontology protein with the intersection will be used as the upper-level ontology protein. Candidate matching items of NCBI Gene protein; if there is no intersection between the two, the protein in NCBI Gene should be expanded to the upper ontology, and step f-6); f-3). Screening of candidate matching items, by judging whether the species information TaxID of the NCBI Gene protein and the species information TaxID in the candidate matching items identify the same species, if they are the same species, it indicates that the NCBI Gene protein and the upper ontology The protein can be fused, and step f-4) is performed; if it is a different species, it cannot be fused, and if all candidate matching items are judged, they cannot be fused, then step f-6) is performed; f-4). Fusion of shortLabel values, if the "Symbol" in the NCBI Gene protein is different from the "Name" value in the upper ontology to be fused, keep the "Name" value of the upper ontology to be fused unchanged, and set the "Symbol" is added to the "Synonym" of the upper ontology to be merged; f-5). Fusion of Synonym values. If the "Synonym" in the NCBI Gene protein is exactly the same as the "Synonym" in the upper ontology to be fused, keep the "Synonym" in the upper ontology unchanged. Then the "Synonym" in the NCBI Gene protein is merged with the "Synonym" synonym in the upper-level ontology and deduplicated, and then stored in the upper-level ontology as a new "Synonym" field value; f-6). The extension of NCBI Gene protein to the upper-level ontology, using the "Symbol" of NCBI Gene to align to the "Name" of the upper-level ontology, the "taxID" of NCBI Gene to the "TaxID" of the upper-level ontology, and the "Synonym" of NCBI Gene ” extends the upper-level ontology in a way of aligning with the “Synonym” of the upper-level ontology.

Images