Ontologized knowledge semantic annotation method for protein domain and online analysis engine system

Abstract

An ontologized knowledge semantic annotation method for a protein domain and an online analysis engine system. By integrating differently defined protein domain data types, a protein domain semantic annotation map is provided for a user. The proposed semantic annotation method comprises: using protein domain information and protein ontology annotations to construct a mapping matrix between domains / superdomains and ontology terms; respectively performing overall statistical inference and relative statistical inference on the mapping matrix to retain the most relevant ontology terms; and obtaining domain-centric ontology annotations according to a real path rule, and for each domain / superdomain, screening for an association relationship between terms according to the obtained overall statistical inference result and relative statistical inference result, so as to form an ontology annotation configuration file. A user is supported to perform semantic annotation on a corresponding protein domain, and enrichment analysis of ontologized knowledge is also supported, thereby providing a strong support for exploration in the field of protein domain semantic annotation.

Description

Ontology knowledge semantic annotation method and online analysis engine system for protein domain TECHNICAL FIELD

[0001] The present application relates to the technical field of ontology knowledge analysis of protein domains, and in particular to a data and ontology knowledge semantic annotation method and analysis engine system for protein domains. BACKGROUND

[0002] In the vast field of molecular biology, accurate annotation of protein function is undoubtedly the key to understanding the nature of life, and has great biomedical and pharmaceutical significance. Although molecular experiments can provide the most reliable protein function annotation, due to the relatively low throughput and limited scope, computational prediction has gradually emerged in the field of function annotation. Accurate biological macromolecule function annotation generated by computers provides a powerful tool for biologists to rapidly identify the potential role of proteins in biological processes or pathways. In recent years, with the continuous emergence of new technologies, computational prediction of protein structure has become feasible, especially after the advent of AI tool AlphaFold2, which quickly resolved almost all known proteins on Earth, generating a large amount of protein data. And the recently released AlphaFold3 successfully predicted protein structure and interaction with unprecedented accuracy. At the same time, with the rapid development of sequencing technology, the demand for genome analysis is growing rapidly, leading to an urgent need to use computers to annotate most proteins. However, the field of protein function annotation still faces great challenges. How to further improve the current computing resources to accurately predict protein function has become a frontier problem in the field of computing and molecular biology. In order to address this challenge, the Critical Assessment of Functional Annotations (CAFA) challenge was born, which is committed to evaluating and improving computational annotation methods of protein function to narrow the gap between the ever-expanding molecular database and limited understanding resources. As the basic functional unit of proteins, domains are more directly and meaningfully associated with individual domains than with ontology terms associated with full-length proteins. However, most protein sequences currently lack biological annotations, especially in the field of ontology knowledge annotation, full-length protein research dominates, and protein domain research is relatively lagging behind, which has led to a lack of corresponding research tools and data resources in this field. In functional genomics, domain-centered functional annotation has obviously lagged behind protein-level annotation. Given that more and more ontologies are being created and used for full-length protein annotation, and protein domains often serve as functional and evolutionary units of proteins, they play an important role in protein function prediction and phenotype prediction, it is particularly necessary to conduct comprehensive semantic annotation of protein domains. Therefore, how to effectively conduct comprehensive biological annotation of protein domains and deeply mine ontology and domain resources has become an urgent problem to be solved.

[0003] Protein domains, as indispensable functional units of proteins, have been proven to play an important role in protein function prediction and phenotype prediction. However, there is currently a lack of a database integrating various protein domains and their ontology knowledge, and a system that can further analyze and mine data in depth. 1) Most known protein sequences lack biological annotations. Compared with full-length proteins, protein domains have received less attention in ontology annotation, which to some extent limits the in-depth understanding of their functions; 2) In current research, there is no systematic work to effectively integrate various protein domains and their related ontology knowledge databases. This fragmented state of information not only makes the research work complex, but also hinders researchers' in-depth study in the field of protein domains; 3) Ensuring the accuracy of protein prediction calculation has always been a great challenge. Due to the complexity and diversity of protein functions, we need more advanced and accurate computational methods to support our prediction work; 4) Although there are some databases for storing and querying information related to protein domains, their performance still needs to be further improved and perfected. SUMMARY

[0004] The purpose of the present application is to overcome the defects of the prior art and provide a protein domain data and ontology knowledge semantic annotation method and analysis engine system.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] As a first aspect of the present application, a protein domain ontology knowledge semantic annotation method is provided, and the method steps include:

[0007] A mapping matrix between domain / superdomain and ontology terms is constructed using protein domain information and protein ontology annotations;

[0008] Based on the obtained mapping matrix, overall and relative statistical inferences are respectively performed in the case of respecting the ontology directed acyclic graph, and the most relevant ontology terms are retained;

[0009] A complete domain-centered ontology annotation is obtained by following the real path rule, and for each domain / superdomain, the association between terms is filtered according to the obtained overall and relative statistical inference results to construct an ontology annotation configuration file.

[0010] As a preferred technical solution, the overall and relative statistical inference has the following specific steps:

[0011] Fisher's test is used to infer the association between the row ontology term annotation information and the specific domain / superdomain in the mapping matrix, and the ontology terms are presented in the form of a directed acyclic graph;

[0012] determine the best association using the real path rule of ontology directed acyclic graph;

[0013] calculate P-value measuring the probability of the association between the ontology and the domain / super domain, evaluate the significance of the association of the statistical inference results between the mapping matrices using the false discovery rate of multiple hypothesis testing, and quantify the association strength using the hypergeometric distribution.

[0014] As a preferred technical solution, the real path rule of ontology directed acyclic graph is used to determine the best association, which is specifically to compare the importance of each term in two different contexts, one context containing all UniProt proteins available for analysis, and the other context containing only UniProt proteins directly annotated by the parent term.

[0015] If a certain ontology term and its parent term are significantly associated with the domain / super domain in the first context, i.e., the overall P-value P overall is greater than the threshold value; but in the second context, the association with the parent term is not significant, i.e., the relative P-value P relative is less than the threshold value, only the parent term is associated with the domain / super domain.

[0016] As a preferred technical solution, the P-value measuring the probability of the association between the ontology and the domain / super domain is calculated as follows:

[0017] P = max{P overall , P relative}

[0018] Where P represents the maximum P-value in the overall P-value P overall and the relative P-value P relative of the overall enrichment test statistical inference and the relative enrichment test statistical inference, representing the statistical significance of the prediction results.

[0019] The overall P-value of the overall enrichment test statistical inference is calculated as follows:

[0020] Where N represents the number of protein domains containing at least one ontology and one domain, M represents the number of protein domains containing at least one domain, K represents the number of protein domains annotated to the ontology, X represents the number of observed protein domains annotated to the ontology and containing the domain, P overall represents the expected probability of observing X or more protein domains under the first context hypergeometric distribution.

[0021] The relative P-value of the relative enrichment test statistical inference is calculated as follows:

[0022] Where N paM is the number of all direct parent annotations of the protein domain in the ontology of the directed acyclic graph structure, K represents the number of annotations to the ontology in the protein domain sequence, X represents the number of observed protein domain annotations to the ontology and the number of domains containing the domains, P pa represents the number of protein domains annotated to the ontology in the protein domain sequence, K represents the number of annotations to the ontology in the protein domain sequence, X represents the number of observed protein domain annotations to the ontology and the number of domains containing the domains, P pa represents the number of protein domains annotated to the ontology in the protein domain sequence, K represents the number of annotations to the ontology in the protein domain sequence, X represents the number of observed protein domain annotations to the ontology and the number of domains containing the domains, P relative represents the expected probability of observing X or more protein domains under the second background hypergeometric distribution.

[0023] As a preferred technical solution, when a parent node has multiple child node annotation information, the associated information with the smallest FDR and the largest hypergeometric score is selected, and the associated ontology term is expanded to all parent term to constitute a complete ontology annotation profile.

[0024] As a preferred technical solution, the method step is to further optimize the annotation profile, and the directed acyclic graph structure of the ontology is divided according to the information content of the ontology term, that is, the negative logarithmic transformation of the domain / superdomain frequency annotated to the term, each partition is located in a different path but reflects the same or similar specificity:

[0025] A GO slim is generated by multiple information content increasing seeds, that is, a subset created by folding the terms upwards to more general and higher-level terms related to a specific background;

[0026] The specificity of the ontology is divided into multiple granularity-increasing ontology term levels by using IC, and the specific steps are as follows:

[0027] All ontology terms in the directed acyclic graph are unmarked, and the IC value is initialized;

[0028] The term closest to the initialized IC is identified, the identified term and its parent node and child node are marked, and are excluded from further analysis, the first two steps are repeated, and the unmarked GO term is iteratively identified until all terms are marked and correspond to an IC value.

[0029] As a preferred technical solution, the method provides two versions of ontology and protein domain association information: high-quality association information and high-coverage association information; the high-quality association information is based on single-domain proteins with known functions, and has causal domain evidence support; the high-coverage association information further includes the association supported by known multi-domain proteins through statistical disambiguation.

[0030] As a second aspect of the present application, an online analysis engine system for protein domain ontology knowledge is provided, and the analysis engine system comprises:

[0031] Protein domain and ontology knowledge knowledge base, including different definition protein domain, protein domain enrichment related ontology knowledge semantic annotation dataset;

[0032] Protein domain and ontology knowledge online analysis main engine for providing ontology knowledge and protein domain online webpage data mining, performing the ontology knowledge semantic annotation method of protein domain as described above;

[0033] Large language model machine learning protein domain prediction main engine, obtaining user input query problem, extracting main information in text and performing learning and reasoning task, generating text to answer user's question;

[0034] Human-computer interaction module, converting search terms into data recognizable by each engine, and visualizing the results.

[0035] As a preferred technical solution, the protein domain and ontology knowledge online analysis main engine comprises:

[0036] Protein domain and ontology knowledge page search sub-engine for receiving user input keywords from human-computer interaction module to query related protein domain and ontology knowledge resources; according to the data provided by the protein domain enrichment related ontology knowledge sub-library, the user input keywords are queried and matched; the query results of protein domain or ontology knowledge are returned to the human-computer interaction module in the form of visualized display page;

[0037] Ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine for retrieving user selected ontology knowledge type from human-computer interaction module, presenting and data mining the user selected ontology knowledge type according to different definition protein domain knowledge sub-library and protein domain enrichment related ontology knowledge sub-library, returning the selected results to the human-computer interaction module in the form of visualized display data;

[0038] Domain ontology knowledge annotation enrichment analysis sub-engine for receiving user input protein domain list from human-computer interaction module; according to the protein domain enrichment related ontology knowledge sub-library, the data of the user input protein domain list is retrieved and ontologically enriched; the retrieval results and ontologically enriched analysis results are returned to the human-computer interaction module in the form of visualized display data.

[0039] As a preferred technical solution, the large language model machine learning protein domain prediction main engine is provided with an Evoformer module and a Diffusion Network diffusion network, starting from original atomic coordinates, learning and training a diffusion model, simulating the interaction of biomolecules, and expanding the cycle of disordered regions, and finally realizing protein function prediction by integrating the above module network.

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] 1) The present application provides a semantic annotation method for protein domain biomedical ontology, which aims to annotate the function of protein domains, including superdomains in multi-domain. In the case of respecting the directed acyclic graph of gene ontology, two types of statistical inference, overall and relative, are carried out, and only the parent term is associated with the domain / superdomain. This double constraint ensures that only the gene ontology terms most significantly related to the domain / superdomain are retained.

[0042] 2) The present application follows the true path rule to obtain a complete domain-centered gene ontology annotation significance measure. For each domain / superdomain, the associated gene ontology terms (i.e. direct annotations) will be extended to all parent ancestor terms (i.e. inherited annotations), which together constitute a complete gene ontology annotation profile.

[0043] 3) In order to further optimize the annotation profile, the present application divides the directed acyclic graph structure of the gene ontology according to the information content of the gene ontology term (i.e. the negative logarithmic transformation of the frequency of the domain / superdomain annotated by the term), each partition is located in a different path but can reflect the same or similar specificity. Not only rely on the depth of the ontology graph to define the specificity of the term, but also fully consider the actual use of the term when determining the four-level depth classification of the domain / superdomain.

[0044] 4) In addition to individual domains, the present application further associates gene ontology terms with superdomains. Superdomains are a combination of two or more continuous domains that can work together, and in the present application, only those completely assigned superdomains without significant gaps between domains are included, excluding superdomains not assigned to known domain region domains, thereby ensuring the accuracy of the annotation. The association of gene ontology with superdomains not only helps us to deeply understand how domain combination promotes functional diversification, but also provides strong support for predicting multi-domain protein function.

[0045] 5) The large language model machine learning protein domain prediction main engine of the application will be based on machine learning, extract the main information in the text, automatically perform learning and reasoning tasks, use the previously learned patterns and structures to generate new text to answer the user's questions, and thus timely feedback the results to the user; it can provide a convenient, fast and simple protein domain information mining tool for relevant researchers. And add Evoformer module and Diffusion Network diffusion network, starting from the original atomic coordinates, learn and train the diffusion model, simulate the interaction of biological molecules, and extend the cycle in the disordered region to better discover the flexibility and variability of these regions. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a schematic diagram of the logical architecture of the protein domain biomedical ontology semantic annotation engine system of the application;

[0047] Figure 2 is a flowchart of the specific operation process of the protein domain and ontology knowledge page search sub-engine of the application;

[0048] Figure 3 is a flowchart of the specific operation process of the ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine of the application;

[0049] Figure 4 is a flowchart of the specific operation process of the domain ontology knowledge annotation enrichment analysis sub-engine of the application. DETAILED DESCRIPTION

[0050] The application will be described in detail below in conjunction with the drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the application, and gives a detailed implementation and specific operation process, but the protection scope of the application is not limited to the following examples.

[0051] Example 1

[0052] As one of the embodiments of the application, the present embodiment provides a protein domain and ontology knowledge online analysis engine system, which integrates different definitions of protein domain data types (such as SCOP, Pfam and InterPro) to provide users with corrected protein domain semantic annotation atlas. The protein domain and ontology knowledge knowledge base of the application contains different definitions of protein domain data and protein domain enrichment related ontology knowledge data. The enrichment relationship atlas constructed by multiple types of annotation data supports users to perform semantic annotation on the corresponding protein domain, and also supports ontology enrichment analysis.

[0053] As shown in Figure 1, the protein domain biomedical ontology semantic annotation engine system includes:

[0054] Protein domain and ontology knowledge knowledge base;

[0055] Protein domain and ontology knowledge online analysis main engine, relying on protein domain and ontology knowledge, the first analysis main engine accesses the protein domain and ontology knowledge knowledge base;

[0056] Large language model machine learning protein domain prediction main engine, relying on large language model machine learning knowledge, used for providing ontology knowledge and protein domain online webpage data mining; the second prediction main engine accesses the protein domain and ontology knowledge knowledge base;

[0057] Human-computer interaction module, receiving search terms, and inputting the search terms to the first annotation main engine or the second annotation main engine, receiving and displaying the interpretation results of each annotation main engine.

[0058] Among them, the protein domain and ontology knowledge knowledge base includes different definition protein domain knowledge sub-library, protein domain enrichment related ontology knowledge, etc. Semantic annotation dataset, the semantic annotation knowledge base specifically includes: different definition protein domain knowledge sub-library, protein domain enrichment related ontology knowledge sub-library;

[0059] The different definition protein domain knowledge sub-library collects protein domain data from SCOP superfamily domain, Pfam domain and InterPro domain databases, integrates multiple protein domain databases to comprehensively and completely query protein domain, and is suitable for data type retrieval and data mining functions of protein domain.

[0060] The protein domain enrichment related ontology knowledge sub-library is used for providing ontology knowledge of protein domain semantic annotation, and these ontology knowledge includes function, pathway, transcription regulation factor, molecular marker, phenotype, disease, drug, etc. The ontology supported by the protein domain enrichment related ontology knowledge sub-library currently includes:

[0061] (1) Function: Gene Ontology GO (June 2024 version), including Biological Process Ontology (GOBP), Molecular Function Ontology (GOMF) and Cellular Component Ontology (GOCC).

[0062] (2) Pathway: KEGG (110.1 version), REACTOME (V89 version), PANTHER (19.0 version), WikiPathways (January 2024 version) and MitoCarta (MitoCarta3.0 version).

[0063] (3) Transcription regulation factor: ENRICHR Consensus TFs (2023 month version) and TRRUST (November 23, 2021 version).

[0064] (4) Molecular markers: MSigDB (October 2023 release).

[0065] (5) Phenotypes: Human Phenotype Ontology HPO (June 2024 release), Mammalian Phenotype Ontology MPO (June 2024 release), and other phenotypic and anatomical ontologies for model organisms such as WormBase (WS292 release), FlyBase (FB2024_03), ZFIN (April 2024 release), and TAIR (May 2024 release).

[0066] (6) Diseases: Mondo Disease Ontology (June 4, 2024 release), Experimental Factor Ontology EFO (3.67.0 release) for GWAS disease traits.

[0067] (7) Drugs: DGIdb Druggable Categories (June 2024 release) and Open Targets Tractability Buckets (June 2024 release).

[0068] The ontology knowledge sub-library related to protein domain enrichment can be applied to the ontology knowledge semantic annotation of the data type of the protein domain.

[0069] The protein domain and ontology knowledge online analysis main engine can be used to provide online webpage data mining functions of ontology knowledge and protein domains, and specifically includes: a protein domain and ontology knowledge page search sub-engine, an ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine, and a domain ontology knowledge annotation enrichment analysis sub-engine; the above sub-engines complement each other to provide a comprehensive, fast, and effective query analysis function for users.

[0070] The protein domain and ontology knowledge page search sub-engine is used to receive the keywords input by the user from the human-computer interaction module to query related protein domain and ontology knowledge resources. The user inputs the keywords of interest, such as protein domain information, SCOP / Pfam / InterPro ID, or ontology terms. According to the data provided by the ontology knowledge sub-library related to protein domain enrichment, the keywords input by the user are queried and matched, and finally the query results of the protein domain or ontology knowledge are returned to the human-computer interaction module in a visual display page.

[0071] The ontology knowledge has the directed acyclic graph annotation information and the annotated protein domain browsing sub-engine for retrieving the ontology knowledge type selected by the user from the human-computer interaction module, presenting and data mining the ontology knowledge type selected by the user according to the different defined protein domain knowledge sub-library and the ontology knowledge sub-library related to the protein domain enrichment, and finally returning the selected result to the human-computer interaction module in the form of visual display data.

[0072] The ontology knowledge annotation enrichment analysis sub-engine of the domain is used for receiving the protein domain list input by the user from the human-computer interaction module, retrieving and ontology enrichment analysis of the data of the protein domain list input by the user according to the ontology knowledge sub-library related to the protein domain enrichment, and finally returning the retrieval result and the ontology enrichment analysis result to the human-computer interaction module in the form of visual display data.

[0073] The human-computer interaction module is a web interactive system, which converts the search terms into data suitable for each engine to recognize, unifies the data format for data analysis, and finally visualizes the results in the form of HTML web page. Further, the human-computer interaction module is composed of an interaction layer, a front-end UI and an access layer, supports all browsers of PC and mobile terminals, and supports Window, Linux and MacOS systems.

[0074] The interaction layer is used to convert the search terms into data suitable for each engine to recognize, unify the data format, and input the converted search terms into the protein domain and ontology knowledge online analysis main engine or large language model machine learning protein domain prediction main engine located in the back end connected with the interaction layer; the front-end UI layer is used to recognize the input search terms.

[0075] The large language model machine learning protein domain prediction main engine uses machine learning as the principle, collects artificially annotated protein domains as the training set, improves the performance by systematically learning data, finds out the patterns and rules hidden in the data, thereby realizes the prediction of new data, and trains a large amount of text data to deepen the model's understanding of human language, so as to carry out effective dialogue and question answering. The main engine is a generative question answering, the user inputs the question to be queried to the main engine, the model can extract the main information in the text, automatically perform learning and reasoning tasks, use the previously learned patterns and structures to generate new text to answer the user's question, so as to feedback the result to the user in time. The main engine can provide a convenient, fast and simple protein domain information mining tool for relevant researchers, realize the function prediction of protein domain, and is very important for studying molecular mechanism and structure-based drug design.

[0076] The large language model machine learning protein domain prediction main engine will add an Evoformer module (a protein language model) and a Diffusion Network diffusion network, starting from the original atomic coordinates, learning and training a diffusion model, simulating the interaction of biomolecules, and extending the cycle of disordered regions to better discover the flexibility and variability of these regions, and by integrating the above functions, ultimately achieving protein function prediction.

[0077] 1) The protein domain semantic annotation database constructed by the present application is not only large in scale, but also detailed in content. It covers various types of protein domains such as SCOP, Pfam and InterPro, and integrates rich ontological knowledge data. This comprehensive and detailed data coverage provides a solid foundation for mapping from protein domains to ontological knowledge, ensuring the completeness and comprehensiveness of protein domain semantic annotation. In particular, the knowledge base of the present application has the characteristics of open source and scalability, meaning that its data resources are not only rich, but also can be continuously expanded and optimized according to actual needs, further meeting the in-depth exploration needs of researchers on protein domain semantic annotation.

[0078] 2) In addition to providing data retrieval and browsing functions, the present application also gives users the ability to explore domain data functions and phenotypes in depth. Through the platform of the present application, users can examine and analyze their data from a more comprehensive perspective, thereby more accurately grasping the mechanism of protein domains in biological processes. This in-depth data mining function not only provides strong data support for their research work, but also helps users discover new scientific laws.

[0079] 3) In terms of resource mining, the present application provides a more integrated and user-friendly way. Through intuitive and easy-to-use interface design and detailed user manuals, users can easily get started and efficiently use the functions of the present application. Whether they are beginners or experienced users, they can easily browse and effectively use the resources of the present application through step-by-step instructions and clear guidance. This user-friendly design not only improves the user experience, but also promotes the integration and efficient discovery of knowledge, providing strong support for researchers in the field of protein domain semantic annotation.

[0080] Example 2

[0081] As a second aspect of the present application, the protein domain biomedical ontological semantic annotation enrichment analysis sub-engine also develops a protein domain biomedical ontological semantic annotation method. This algorithm aims to annotate the function of protein domains (including superdomains in multi-domain). Taking gene ontology as an example, the application of other ontologies such as pathway ontology, phenotype ontology and disease ontology in the knowledge sub-library is the same in principle.

[0082] Step 1: Using protein domain information from SUPERFAMILY and protein Gene Ontology annotation from UniProtKB-GOA, a mapping matrix between domain / superdomain and Gene Ontology terms is constructed: columns are specific domain / superdomain and rows are Gene Ontology annotation information corresponding to protein domain and ontology knowledge repository.

[0083] Step 2: Based on the mapping matrix obtained in Step 1, overall and relative types of statistical inference are performed in respect of the directed acyclic graph of Gene Ontology, which ensures that only the most relevant Gene Ontology terms are retained. The specific steps are as follows: first, Fisher's exact test is used to infer the association between rows (Gene Ontology annotation information) and columns (specific domain / superdomain). The terms in the gene are presented in the form of a directed acyclic graph, in which a single item is a node, and its relationship with the parent term (allowing multiple parent terms) is regarded as a directed edge. Then, the true path rule of the directed acyclic graph of Gene Ontology is used to determine the best association, i.e. the importance of each term in two different contexts is compared: one context contains all UniProt proteins available for analysis (which can be annotated by Gene Ontology), and the other context contains only UniProt proteins directly annotated by the parent term. If a Gene Ontology term and its parent term are significantly associated with the domain / superdomain in the first context (overall P value, formula 1), but not significantly associated with the parent term in the second context (relative P value, formula 2), we only associate the parent term with the domain / superdomain. This double constraint ensures that only the most significantly associated Gene Ontology terms with the domain / superdomain are retained. Then, the P value is calculated to measure the probability of the association between the Gene Ontology and the domain / superdomain, and the higher the P value, the stronger the prediction ability. The false discovery rate (FDR) of multiple hypothesis testing is used to screen the threshold less than 10 -3 , and the FDR is used to evaluate the significance of the above statistical inference results between matrices (formula 3), and the h-score is used to quantify the association strength (formula 4).

[0084] where N represents the number of protein domains containing at least one Gene Ontology and one domain, M represents the number of protein domains containing at least one domain, K represents the number of protein domains annotated to Gene Ontology, X represents the number of observed protein domains annotated to Gene Ontology and containing domains, P overall represents the expected probability of observing X or more protein domains under the first background hypergeometric distribution.

[0085] where N paM is the number of all direct parent annotations of protein domains of the gene ontology in the directed acyclic graph structure, K pa represents the number of protein domains that intersect with N pa represents the number of protein domains that intersect with N relative represents the expected probability of observing X or more protein domains under the second background hypergeometric distribution.

[0086] P = max{P overall ,P relative} (3)

[0087] wherein P represents the maximum P value of the overall enrichment test and the relative enrichment test. And calculate its corresponding FDR, representing the statistical significance of the prediction results.

[0088] h-score = min{S overall ,S relative} (4)

[0089] wherein h-score represents the association strength, used to evaluate the prediction results. S overall and S relative represent the scores calculated by the z-score, and the minimum values are selected.

[0090] Step 3: Follow the true path rule to obtain a complete domain-centered gene ontology annotation. For each domain / super domain, according to the association relationship between the FDR and the h-score of the terms screened in step 2, for example, when a parent node has multiple child node annotation information, the association information with the minimum FDR and the maximum h-score is selected, and the finally associated gene ontology term (i.e. direct annotation) will be extended to all parent term (i.e. inherited annotation), together constituting a complete gene ontology annotation configuration file.

[0091] In addition, to further optimize this annotation profile, the DAG structure of the gene ontology is partitioned according to the information content of the gene ontology terms (i.e. the negative log-transform of the frequency of the term in the domain / GO-slim annotated), each partition is located in a different path but reflects the same or similar specificity. By four seeds of increasing information content, we generated a "GO slim" (a subset of terms created by folding terms upwards to more general, higher-level terms relevant to a specific context for improved maneuverability and efficiency of use, terms are human subjective decisions), using IC (Information content) to measure the specificity of the gene ontology, it is divided into four levels of increasing granularity of gene ontology terms: highly general, general, specific and highly specific. The specific steps are: first, all gene ontology terms in the DAG are unmarked, initialize the IC value (such as 0.5, 1, 1.5 or 2), then identify the term closest to the initialized IC, mark the identified term and its parent node, child node, and exclude it from further analysis, repeat the first two steps, iteratively identify unmarked GO terms, until all terms are marked and correspond to an IC value. Where IC = 0.5 is defined as the highly general level, IC = 1 is defined as the general level, IC = 1.5 is defined as the specific level, and IC = 2 is defined as the highly specific level. Since the gene ontology is created for annotating proteins, some parts of its structure may not be as effective as others in annotating domains / GO-slims. Therefore, the algorithm does not rely solely on the depth of the ontology graph to define the specificity of the term, but also takes into account the actual use of the term when determining the four-level depth classification of the domain / GO-slim.

[0092] Based on the above algorithm, the present application provides two versions of gene ontology and protein domain association information for users: high-quality association information and high-coverage association information. Among them, the high-quality association information is based on single-domain proteins with known functions, and has clear causal domain evidence support. While the high-coverage association information further incorporates known multi-domain proteins, and statistics supported by disambiguation, although its reliability is slightly inferior to the former, but for large-scale research, it provides more extensive functional coverage. From the perspective of domain center, high-quality association information is undoubtedly more reliable, but high-coverage association information is also reliable enough for extensive functional exploration and analysis, and provides users with more abundant functional annotation. In practical application, considering the positive impact of annotation coverage on enrichment analysis results, we strongly recommend using high-coverage association information for gene ontology annotation in large-scale research. For domain-based enrichment analysis, it is strongly recommended to limit the annotation to "GO slim", for example, the ontology annotation enrichment analysis sub-engine of the domain involved in this engine will follow this rule.

[0093] In addition to individual domains, the present application further correlates GO terms with superdomains. Superdomains are combinations of two or more contiguous domains that work synergistically, and in the present application only those superdomains are included that are completely assigned without significant gaps between domains, excluding superdomains that are not assigned to known domain regions, thus ensuring the accuracy of the annotation. The correlation of GO terms with superdomains not only helps us to understand how domain combinations contribute to functional diversification, but also provides strong support for predicting the functions of multidomain proteins.

[0094] Example 3

[0095] Figure 2 shows a flow chart of the specific operation process of the embodiment of the protein domain and ontology knowledge facet search sub-engine of the present application, and the implementation steps are as follows:

[0096] Step 3.1: The user inputs the keyword on the webpage (human-computer interaction module), and the keyword type can be the protein domain name (such as SCOP, Pfam and InterPro domain) or the ontology knowledge;

[0097] Step 3.2: According to the input keyword type, the protein domain and ontology knowledge facet search sub-engine starts to perform the search function, and feeds back the search results to the front-end layer;

[0098] Step 3.3: The user can browse the search results in the access layer. An expansion box will be generated below the search box, which will provide a hyperlink to the specific page of this search result. Clicking the hyperlink will automatically jump to the webpage that not only provides the keyword information, but also provides the protein domain and ontology knowledge table, the ontology knowledge annotation information and the hyperlink information of the specific page;

[0099] Example 4

[0100] Figure 3 shows a flow chart of the specific operation process of the embodiment of the ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine of the present application, and the implementation steps are as follows:

[0101] Step 4.1: The user selects the ontology knowledge category on the webpage (human-computer interaction module);

[0102] Step 4.2: The ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine starts to analyze the ontology knowledge selected by the user, and feeds back the search results to the front-end layer;

[0103] Step 4.3: The user can browse and further select and browse the hierarchical structure page of the sub-class name in the access layer. The SCOP, Pfam and InterPro information are displayed respectively;

[0104] Example 5

[0105] As shown in Figure 4, the specific operation process flow chart of the implementation case of the ontology knowledge annotation enrichment analysis sub-engine of the protein domain is as follows:

[0106] Step 5.1: The user inputs the protein domain data on the webpage (human-computer interaction module);

[0107] Step 5.2: The user selects the corresponding domain database according to the needs, please ensure that the identifier of the input domain is consistent with the level of the domain;

[0108] Step 5.3: The user selects the appropriate knowledge sub-library type from the provided protein domain and ontology knowledge knowledge base;

[0109] Step 5.4: The user selects the parameters required for ontology enrichment analysis, and the ontology knowledge sub-library related to protein domain enrichment;

[0110] Step 5.5: The user clicks the SUBMIT button on the webpage to call the ontology knowledge annotation enrichment analysis sub-engine of the domain to perform analysis.

[0111] Step 5.6: The ontology knowledge annotation enrichment analysis sub-engine of the domain starts to perform enrichment analysis on the data provided by the user, and feeds back the enrichment analysis results to the front-end layer.

[0112] Step 5.7: The user can conveniently browse and download detailed analysis results in the access layer. The analysis results include three interactive tables and high-resolution images of enrichment results: ① The Input domain information module provides an interactive table that lists the protein domain information input by the user in detail. This table not only helps users perform data mining, but also clearly marks the running time in the table header, making it easy for users to understand the analysis situation. ② The Enriched ontology terms module provides an enriched ontology knowledge table. Each ontology knowledge provides a hyperlink to a specific information page. The table provides annotation information of the protein domain, including ontology knowledge type information and response regulation interaction information. ③ The Dotplot for enriched ontology terms module provides a dotplot for enrichment results, which globally displays the enrichment of ontology knowledge, providing users with intuitive visual analysis results. In addition, users can also download high-quality PDF files for further analysis and display locally. Through these three modules, users can comprehensively and deeply understand the function and enrichment of the protein domain, providing strong support for scientific research work.

[0113] The preferred embodiments of the present application have been described above in detail. It should be understood that modifications and variations to the preferred embodiments could be made by those skilled in the art in light of the teachings above. It is therefore contemplated that the application can encompass other variations and modifications that fall within the scope of the claims.

Claims

1. A method of ontology-based knowledge semantic annotation of protein domains, characterized in that, The method steps include: Using protein domain information and protein ontology annotation to construct a mapping matrix between domain / super domain and ontology terms; Based on the resulting mapping matrix, make overall and relative statistical inferences respectively, while respecting the ontology directed acyclic graph, and retain the most relevant ontology terms; Follow the true path rule to obtain a complete domain-centered ontology annotation, and for each domain / super domain, filter the association between terms according to the resulting overall and relative statistical inference results to construct an ontology annotation profile.

2. The method of claim 1, wherein the protein domain ontology knowledge semantic annotation is characterized by, The overall and relative statistical inference has the following specific steps: Use Fisher's test to infer the association between the ontology term annotation information in the mapping matrix and a specific domain / super domain, and the ontology terms are presented in the form of a directed acyclic graph; Determine the best association using the true path rule of the ontology directed acyclic graph; Calculate the P value that measures the probability of the association between the ontology and the domain / super domain, use the false discovery rate of multiple hypothesis testing to evaluate the significance of the statistical inference results associated with the mapping matrix, and use the hypergeometric distribution to quantify the strength of the association.

3. The method of claim 2, wherein the method further comprises: The specific method for determining the best association using the true path rule of the ontology directed acyclic graph is to compare the importance of each term in two different contexts, one context containing all UniProt proteins available for analysis, and the other context containing only UniProt proteins directly annotated by the parent term; If a certain ontology term and its parent term are significantly associated with a domain / superdomain in the first context, i.e. the overall P-value P overall is greater than a threshold; but are not significantly associated with the parent term in the second context, i.e. the relative P-value P relative is less than a threshold, then only the parent term is associated with the domain / superdomain.

4. The method of claim 3, wherein the protein domain ontology knowledge semantic annotation is characterized by, The P-value calculation of the probability of the association between the measure ontology and the domain / superdomain is as follows: P = max{P overall ,P relative} where P represents the overall P-value P of the overall enrichment test statistic inference overall and the relative P-value P of the relative enrichment test statistic inference relative the maximum P-value among P and P, representing the statistical significance of the prediction result; The overall enrichment test statistics overall P-value is calculated as follows: where N represents the number of protein domains that contain at least one ontology and one domain, M represents the number of protein domains that contain at least one domain, K represents the number of protein domains that are annotated to the ontology, X represents the number of protein domains observed that are annotated to the ontology and contain a domain, P overall represents the expected probability of observing X or more protein domains under the first background hypergeometric distribution; The relative P-value for the relative enrichment test statistic is computed as follows: where N pa is the number of protein domains annotated to the ontology that are direct parents of the ontology in the directed acyclic graph structure, M pa represents the number of protein domains that contain the domain after intersecting the protein domain sequence in N pa , K represents the number of annotations to the ontology in the protein domain, and X represents the number of protein domains observed to be annotated to the ontology and contain the domain, P relative represents the expected probability of observing X or more protein domains under the second background hypergeometric distribution.

5. The method of claim 2, wherein the method further comprises: When a parent node has multiple child node annotation information, select the association information with the smallest FDR and the largest hypergeometric score, and the associated ontology term will be expanded to all parent terms to form a complete ontology annotation profile.

6. The method of claim 5, wherein the method further comprises: The method steps are further optimized to annotate the profile, and the directed acyclic graph structure of the ontology is divided according to the information content of the ontology term, i.e. the negative logarithmic transformation of the frequency of the domain / super domain annotated by the term, each partition is located in a different path but reflects the same or similar specificity: Generate a GO slim by multiple information content increasing seeds, i.e. create a subset by folding terms upwards to more general and higher-level terms related to a specific context; Use IC to divide the specificity of the ontology into multiple ontology term levels with increasing granularity, and the specific steps are as follows: All ontology terms in the directed acyclic graph are unmarked, and the IC value is initialized; Identify the term closest to the initialized IC, mark the identified term and its parent and child nodes, and exclude them from further analysis, repeat the first two steps, and iteratively identify unmarked GO terms until all terms are marked and correspond to a unique IC value.

7. The method of claim 1, wherein the method further comprises: determining a semantic annotation of the protein domain based on the ontology of the protein domain. The method provides two versions of ontology and protein domain association information: high-quality association information and high-coverage association information; the high-quality association information is based on single-domain proteins with known functions and supported by causal domain evidence; the high-coverage association information further includes associations supported by statistical disambiguation through known multi-domain proteins.

8. An online analysis engine system for protein domain ontology knowledge, characterized in that, The analysis engine system includes: The protein domain and ontology knowledge knowledge base includes different definitions of protein domains and ontology knowledge semantic annotation data sets related to protein domain enrichment; The protein domain and ontology knowledge online analysis main engine is used to provide online webpage data mining of ontology knowledge and protein domains, and perform the ontology knowledge semantic annotation method of the protein domain as claimed in any one of claims 1-7; The large language model machine learning protein domain prediction main engine obtains user input query questions, extracts main information in the text and performs learning and reasoning tasks, and generates a text to answer the user's question; The human-computer interaction module converts the search terms into data recognizable by each engine and visualizes the results.

9. The online analysis engine system of protein domain ontologization knowledge according to claim 8, characterized in that, The protein domain and ontology knowledge online analysis main engine includes: The protein domain and ontology knowledge page-by-page search sub-engine receives user input keywords from the human-computer interaction module to query related protein domain and ontology knowledge resources, performs resource query and matching on the user input keywords based on the data provided by the protein domain enrichment related ontology knowledge sub-library, and returns the protein domain or ontology knowledge query results to the human-computer interaction module in a visual display page; The ontology knowledge directed acyclic graph annotation information and annotated protein domain browsing sub-engine retrieves the user-selected ontology knowledge type from the human-computer interaction module, presents and data mines the user-selected ontology knowledge type based on the different definitions of protein domain knowledge sub-library and protein domain enrichment related ontology knowledge sub-library, and returns the selected results to the human-computer interaction module in a visual display data form; The domain ontology knowledge annotation enrichment analysis sub-engine receives a user input protein domain list from the human-computer interaction module, retrieves and ontology enriches the data of the user input protein domain list based on the protein domain enrichment related ontology knowledge sub-library, and returns the retrieval results and ontology enrichment analysis results to the human-computer interaction module in a visual display data form.

10. The online analysis engine system of protein domain ontology knowledge according to claim 8, characterized in that, The large language model machine learning protein domain prediction main engine is provided with an Evoformer module and a Diffusion Network diffusion network, starting from the original atomic coordinates, learning and training the diffusion model, simulating the interaction of biological molecules, and extending the cycle of disordered regions, and finally realizing protein function prediction by integrating the above module network.

Images