ICEO: a biological ontology for representing and analyzing the bacterial integrative and conjugative element Meng Liua,b, Hong-Yu Oua, Yongqun Heb a State Key Laboratory of Microbial Metabolism, School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China b University of Michigan Medical School, Ann Arbor, MI 48109, USA Abstract (T4SS). The regulation module refers to those genes and sequence contributing to stabilization and maintenance of ICEs. Besides, Bacterial integrative and conjugative elements (ICEs) are mobile virulence factors (VFs) and acquired antibiotic resistance genes genetic elements critical to horizontal gene transfer and organism (ARGs) often exist inside ICEs as the cargo genes (also called the evolution. To better understand and analyze ICEs, it is critical to accessory module of ICEs) and can confer the hosts with selective systematically represent, integrate and classify gene components, advantages, which make ICEs a vital role in the process of functional modules and related information of available ICEs bacterial adaptation and evolution (2). Based on two different archived in the ICEberg database. Toward this goal, we conjugal manners, ICEs can be categorized as T4SS-type ICEs developed a community-driven ICE ontology (ICEO). ICEO is and actinomycete ICEs (AICEs). T4SS-type ICEs are widely aligned with the Basic Formal Ontology (BFO) to allow the distributed both in Gram-negative and Gram-positive bacteria, integration with other ontologies. ICEO reused the existing while actinomycete ICEs (AICEs) only have been found in reliable ontologies, such as Ontology of Gene and Genome, Actinobacteria, mainly in Streptomyces. And T4SS-type ICEs are Protein Ontology and NCBITaxon. ICEO now represents the transferred as linear single-stranded DNA (ssDNA) typically information about over 270 experimentally verified ICEs from depended on a relaxase and a conjugative type IV secretion 235 bacterial strains. Two query use cases were provided, system (T4SS). AICEs are delivered as double-stranded DNA including a DL query of ICE-contained genes that are also (dsDNA) relied on proteins for replications and translocation. virulence factors and a SPARQL query of ICEs under an upper level taxonomy type of Gammaproteobacteria. Our study The information about thousands of experimentally validated or demonstrated that ICEO supports computer-assisted reasoning computationally predicted bacterial ICEs is freely accessible in and efficient SPARQL query. ICEberg (http://db-mml.sjtu.edu.cn/ICEberg/), a comprehensive web-based ICE database that is developed by our group at the Keywords: Shanghai Jiao Tong University (3). However, to make the best ICE; ontology; mobile genetic elements use of these available data and the ongoing increase of information, and to facilitate more effective and accurate Introduction identification and annotation of ICEs from single bacterial genomes or even metagenomes, a knowledge base about available bacterial ICEs in a format compliant for computer analysis is Integrative and conjugative elements (ICEs), also called desired (4, 5). Ontology, a hierarchical and interconnected conjugative transposons before, are a large family of the bacterial controlled vocabulary that emphasizes the logical organization mobile genetic elements (MGEs) (1). ICEs are integrative to the and representation of complex data and knowledge, provides such bacterial chromosome and encode functional conjugation a platform to achieve this goal. In this big data and IT era, machinery for the self-transmission between bacterial cells. structured ontology has been widely used in biological data and Similar to all other bacterial MGEs, typically, ICEs have a highly metadata standardization, integration, sharing, and analysis (6). modular structure with three core genetic modules: (i) For example, one of the most successful and widely-used recombination (integration and excision) module; (ii) conjugation ontology, Gene Ontology (GO; http://www.geneontology.org/) module; and (iii) regulation module. The recombination module (7), which represents the information of cellular components, refers to those genes and sequence within the ICE responsible for biological processes, and molecular functions, is often used as the site-specific integration and excision of the element from the the standard to describe the function of gene and gene products host chromosome, including genes encoding the integrase and or across different databases and to conduct various gene expression recombination directionality factor (also known as excisionase, analyses. The usage of ontology supports better representation, which influences the direction of recombination mediated by the integration, and analysis of big data. integrase to favor excision). The conjugation module denotes those gene and sequence involved in the conjugal process, such In this study, we report the development strategy of a community- as genes encoding relaxase and the type IV secretion system driven Ontology of the Integrative and Conjugative Element Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). (ICEO), which is aimed to ontologically represent and integrate application of well-established ontology design patterns (ODPs), the ICE gene information and functional modules to support and involvement of community efforts for new ontology develop- computer-assisted reasoning. There are 260 experimentally ment (9). verified T4SS-type ICEs (113 with the entire nucleotide 2. ICE-related ontology term reuse sequences) and 11 experimentally validated AICEs (7 with the entire nucleotide sequences) in ICEberg. The current ICEO is To support ontology interoperability and avoid reinventing the focused on ontological organization of the information about wheel, related existing terms from reliable ontologies were im- these 271 experimentally validated ICEs for now here. ICEO is ported into ICEO via an Ontofox (http://ontofox.hegroup.org) im- developed by reusing many terms from existing ontologies and port strategy (12). The external ontologies used here include On- using the state-of-the-art ontology engineering technologies (8, tology of Genes and Genomes (OGG) (13), PRotein Ontology 9). A systematic analysis of the ICEO-represented knowledge (PR) (14), Gene Ontology (GO) (7) and a taxonomy ontology of base allows us to generate new insights about these widely NCBI organismal classification (NCBITaxon) (15). distributed integrative genetic elements. 3. New ICEO term generation Methods Based on the available information, an ontology design pattern (ODP) was developed. Many new annotations and relations between different entities were added by utilizing the Ontorat 1. ICEO ontology development strategy (http://ontorat.hegroup.org/) (16), an online program designed to The development of ICEO follows the Open Biological and Bio- support ODP-based creation of new ontology terms, hierarchies, medical Ontologies (OBO) Foundry principles (10), such as open- annotations, and logical axioms. ness, collaboration, use of a common shared syntax and so on. The Protégé-OWL editor (version 5.2) Therefore, the ICEO information can be easily integrated and pro- (http://protege.stanford.edu/) was used for the ICEO manual cessed with other ontologies in the OBO library. To support the processing and editing, ontology term merging and visualization. data FAIRness (Findable, Accessible, Interoperable and Reusa- ICEO-specific terms were generated using new ICEO identifiers ble) (11), the eXtensible Ontology Development (XOD) strategy with the prefix “ICEO_” followed by auto-generated 7 digits. The (9) was also applied for the ontology development of ICEO. Ba- Hermit reasoner (http://hermit-reasoner.com/) was applied for sically, the XOD strategy recommends the reuse of existing terms semantic consistency checking and inferencing. and semantic relations from reliable ontologies, development and Fig. 1. (A) Classical ICE conserved modules. ICEs typically contain three core modules: (i) a recombination (integration and excision) module, (ii) a conjugation module and (iii) a regulation module. In addition to three core modules, most ICEs possess conserved accessory regions. (B) ICEO top-level hierarchy. Terms with ontology abbreviations inside parentheses are imported from external ontologies, while terms without an identified source are ICEO terms. Some intermediate terms such as those terms in between different layers are not shown to make the relations simple and clear. All the arrows indicate the ‘is a’ relation. SPARQL query, ICEO was stored in the Resource Description 4. ICEO format, source code, and access Framework (RDF; https://www.w3.org/RDF/) triples in the On- The ICEO is developed using the format of W3C standard Web tobee RDF triple store (17, 18). The Ontobee SPARQL query in- Ontology Language (OWL2) (https://www.w3.org/TR/owl- terface (http://www.ontobee.org/sparql) was then used for ICEO guide/) (10). The source code of ICEO is open and available for specific SPARQL query. public view and download on the GitHub website: https://github.com/ontoice/ICEO. The ICEO source code is freely Results available under the Creative Commons 4.0 License (http://creativecommons.org/licenses/by/4.0/), which allows ICEO users to freely distribute and use ICEO. The latest version 1. ICEO top-level design and development of ICEO is also accessible for visualization and downloading from Ontobee (17, 18) ontology repository website: a. ICEO is aligned with BFO and OBO foundry ontologies http://www.ontobee.org/ontology/ICEO, or NCBO’s BioPortal Fig. 1A illustrates the genetic functional modules of the ICEs website: https://bioportal.bioontology.org/ontologies/ICEO. abundant in both Gram-positive and Gram-negative bacteria. Basically, a typical ICE includes three core modules: 5. ICEO knowledge query and analysis recombination (integration and excision) module, conjugation The knowledge stored in the ICEO ontology can be queried module, and regulation module, which relate to the ICE life cycle. through different approaches. In this study, we used the Descrip- In addition to these three core modules, most ICEs carry some tion Logic (DL) query and SPARQL (a recursive acronym for cargo genes (also called accessory genes) like virulence factors SPARQL Protocol and RDF Query Language) query. The DL (VFs) genes and acquired antibiotic resistance genes (ARGs) query was performed using the Protégé OWL editor. For the conferring adaptive phenotypes to the hosts of ICEs. Fig. 2. Gene and protein label naming strategy of ICEO. The ‘ybtE’ gene archived in the NCBI GenBank database is assigned with gene label ‘KP1_3592(ybtE)’ in ICEO, and its corresponding protein label is ‘BAH64185.1(YbtE)’. Fig. 1B represents the basic top-level ICEO hierarchical structure stands for time-independent entities (e.g., material entity and their in accordance with the ICE genetic functional modules (Fig. 1A). quality and roles), while the ‘occurrent’ branch represents time- Specifically, ICEO is aligned to the upper-level Basic Formal related entities (e.g. process and time). Since BFO has been used Ontology (BFO) 2.0 version (19, 20). BFO consists of as the upper-level ontology by over 100 ontologies, the alignment ‘continuant’ and ‘occurrent’ branches. The ‘continuant’ branch of ICEO with BFO facilitates the effective integration of ICEO organisms but with the same names archived in the NCBI with many other ontologies. GenBank database is inevitable. To avoid name conflicts, the original OGG designed a special scheme to automatically assign To enable the reusability of existing ontologies, ICEO imports gene IDs by mapping ontology ID with NCBITaxon IDs and many related terms and relations from OBO library ontologies. NCBI Gene IDs (13). However, the integer sequence identifiers As shown in Fig. 1B, ICEO imports OGG and PR to represent the known as “GIs” and Gene ID are no longer provided and used by genes and proteins of ICEs. NCBITaxon terms are imported to NCBI for the sequence records in non-reference strains since represent various ICE-containing organisms in the taxonomic September 2016 (21). Furthermore, for many organisms organism hierarchy. GO terms are imported to represent the harboring ICEs, for example, Escherichia coli strain ECOR31 processes in the whole life cycle of ICEs. that carries ICE gene components, there are no available b. Modified and extensive gene and protein ID assignments NCBITaxon IDs. In addition, OGG still faces the gene label and label naming strategy redundancy since OGG only used the gene name or locus tag as the ontology label of genes. Given the ever-growing number of genes sequenced and annotated, the phenomena of having genes or proteins in different Fig. 3. ICEO design pattern and an example. (A) Generic ontology design pattern for relations among terms in ICEO. (B) An example of ICEKp1 representation by applying ICEO design pattern with extended information obtained from the ICEberg database. Due to these reasons, we have worked with the OGG development top-level hierarchy (Fig. 1) which shows the hierarchical team and developed an OGG-based extended strategy of relationships among different terms, Fig. 3 shows the logical generating new OGG IDs for gene assignments for ICE-related relations of related terms across different hierarchical structures genes. Simply put, this strategy assigns OGG gene IDs using in ICEO. Together, the combination of Fig. 1 and Fig. 3 presents NCBI locus_tag identifiers commonly seen in GenBank gene us a general framework of the ontological design of ICEO. records. Generally, if gene name is available for a gene, then it’s As shown in Fig. 3A, the basic ICEO design pattern is to represent gene label will be assigned as ‘locus_tag(gene_name)’; if not, the ICE from the view of typical function modules. ICE ‘has part’ gene label will be ‘locus_tag’. In addition, ‘product’ information integration, conjugation, regulation and accessory module will be added to ‘dc:description’ property of the corresponding components, and these components and or their encoding proteins protein. Such a naming strategy allows us to develop and design ‘participants in’ specific ICE life process. computer programs to automatically generate readable and nonredundant ICEO gene label. For example, the yetE gene in An example of applying this general design pattern to a concrete Klebsiella pneumoniae strain NTUH-K2044 has a locus tag of example is shown in Fig. 3B where the design pattern is used to KP1_5092. Accordingly, we assign this yetE gene as represent ICEKp1, a virulence-associated ICE found in Klebsiella ‘KP1_5092(yetE)’ and assign its ID as “OGG_KP1_5092” (Fig. pneumoniae subsp. pneumoniae NTUH-K2044 causing a primary 2). liver abscess (22, 23). Basically, ICEKp1 contains all essential genes necessary to the whole life cycle of ICEKp1 (that is, the ICE is essentially a genetic feature so that the gene representation process of excision, conjugation, regulation, and integration) is our priority. Since the Protein Ontology (PR) does not include (Fig. 3B). Besides, a virulence factor gene cluster within all the proteins included in ICEO, we applied a similar strategy to ICEKp1 responsible for the synthesis, regulation, and transport represent protein names in ICEO (Fig. 2). We have also contacted of the siderophore yersiniabactin confers the high virulence to the Protein Ontology (PR) team, and will request new classes in the K. pneumoniae NTUH-K2044 (22, 23). Fig. 4 is the more spe- the PR to improve ICEO. cific demonstration of such an example visualized by Protégé. 2. ICEO ontology design pattern Fig. 3 illustrates the ICEO ontology design pattern which logically links different types of entities. Compared to the ICEO Fig. 4. Demonstration of ICEO linkage of different entities and representation of the virulence factor gene ybtE of the accessory module in ICEKp1. 3. ICEO statistics 4.ICEO applications The latest release of ICEO contains a total of 7738 terms, ICEO is formatted in the machine-interpretable OWL format, including 7604 classes, 52 object properties, and 78 annotation which is easily understood by computer programs and can support properties. Among these terms, 1713 terms have ICEO_ various advanced queries and analyses. Therefore, ICEO can be namespace. The full ontology statistics of ICEO is accessible in used for various applications, such as DL query and SPARQL the Ontobee ICEO statistics page query, which is designed for RDF triple and cannot be done (http://www.ontobee.org/ontostat/ICEO). directly in ICEberg. Two ICEO use cases are provided in the study as follows: Use Case 1: Use DL Query to query the specified group of ('has role' some 'virulence factor gene role') and ('part genes of an ICE of' some ICEKp1) ICEO supports OWL-based automated reasoning using reasoning As shown in Fig. 5, our query identified 15 genes that are part of programs within OWL editors. As an example, we designed the the specific ICEKp1 and also encode for virulence factors in the following question for query the ontology: host bacterium of the ICEKp1. The result indicates that ICEKp1 is responsible for transporting this set of virulence factor genes to What genes in the ICEKp1 encode virulence factors? Klebsiella pneumoniae strain NTUH-K2044, the bacterium that In ICEO, a virulence factor gene is logically defined as hosts the ICE. ICEKp1 is indeed critical to make the bacterium “something that has role some virulence factor gene role”. To virulent (22, 23). answer this question, the following query was simply performed using the DL Query function in ProtégéOWL editor version 5.2 (Fig. 5): Fig. 5. DL Query window in Protégé5. The query of all the virulence factor genes of ICEKp1 was performed using the DL Query of OWL editor Protégé5.2. The query code is shown on the top, and the query results are displayed at the bottom. Use Case 2: Use SPARQL query for advanced analysis Fig. 6 demonstrates a SPARQL query over ICEO. This example includes only a few lines of SPARQL query code. However, it As the OWL-formatted ICEO is stored in the Ontobee RDF triple enabled the identification of the organisms under the taxonomic store (17, 18), the ICEO information can be also queried and class of ‘Gammaproteobacteria’ (NCBITaxon_1236) that include analyzed using the RDF query language, SPARQL experimentally verified ICEKp1 family ICEs. A variety of queries (https://www.w3.org/TR/rdf-sparql-query-protocol/). can be achieved with new SPARQL query scripts for more practical and advanced analysis. Fig. 6. SPARQL query of all the ICEKp1 family ICEs in the taxonomy rank of Gammaproteobacteria. The ICEO term ICEO_0000141 is ‘ICEKp1 family ICE’ class, ICEO_0000020 refers to an object property ‘is ICE of organism’, and the NCBITaxon term NCBITaxon_1236 points to ‘Gammaproteobacteria’ class. The query was performed using the Ontobee SPARQL query interface (http://www.ontobee.org/sparql/). information for the advanced data sharing and organization. Discussion Second, we are going to combine and integrate the structured ICEO ontology into the ICEfinder (3), an ICE prediction tool developed also by our group with both web server and standalone In this study, we developed a community-driven Integrative and versions available. With the use of ICEO, the enhanced ICEfinder Conjugative Element Ontology (ICEO). ICEO will facilitate more effective, accurate prediction of ICE and ontologically represents the complex hierarchical structure of powerful sequence analysis from the raw genome sequence. ICEs, ICE components, and the relations among ICE and ICE Furthermore, ICEO may also facilitate ontology-based literature components. As demonstrated in two use cases, the ICEO mining, which has been shown in many other ontology-based representation of the experimentally verified ICE knowledge research domains (28, 29). supports computer-assisted data integration, efficient query, and reasoning. Conclusions ICEO now is built by standardizing and integrating the rich information from ICEberg database. And ICEO can perform tasks that cannot be done in current ICEberg. For example, in our use To conclude, ICEO is a biological ontology and a knowledge- case 1, we were able to easily query any virulence factors for any centric platform of the bacterial integrative and conjugative level of MGEs. Currently, ICEberg will label VF for those MGEs element. ICEO can serve as an ICE knowledgebase and facilitate that are virulence factors. However, it is still impossible for users the systematical representation, integration and automatical to query all VFs for a specific bacterial group. In our use case 2, computer-assisted reasoning of ICE data. we further illustrate an efficient way of using ICEO to query ICEs under any specific bacterial taxon level like class, species or Acknowledgments family. ICEberg can only query based on species level. However, rather than being only a complement or translation of ICEberg National Key R&D Program of China [2017YFC1600100 to database, ICEO and ICEO-based features will be explored to be H.Y.O.]; ML was supported by a jointly funded Ph.D.- integrated into ICEberg in the future. studentship of the China Scholarship Council and University of ICEO is the first BFO-based ICE ontology. Toussaint et al. Michigan Medical School (Grant No. 201806230209). developed the MeGO, a Gene Ontology dedicated to the functions of mobile genetic elements, and used it in the ACALME database (A CLAssification of Mobile genetic Elements, Address for correspondence http://aclame.ulb.ac.be/) (24–26). MeGO is a non-OBO ontology expanded from the Phage Ontology (PhiGO). MeGO contains 375 YH and HYO are the co-corresponding authors. Their email classes, a single object property (which is “part of”), and 22 addresses are yongqunh@med.umich.edu and hyou@sjtu.edu.cn annotation properties. Most of MeGO terms are related to phages, respectively. Any suggestions or questions are highly welcomed. GO, and sequences. Only a few terms directly related to ICE are included in MeGO. MeGO does not include any specific ICEs and References ICE gene components. In addition, MeGO terms are poorly aligned. It is also noted that the MeGO and ACALME database have not been updated in the past six years. In comparison, ICEO 1. Johnson,C.M. and Grossman,A.D. (2015) Integrative and is systematically developed by aligning with the widely used BFO Conjugative Elements (ICEs): What They Do and How They upper level ontology and following the OBO Foundry principles. Work. Annu Rev Genet, 49, 577–601. ICEO represents the complicated gene components, functional 2. Delavat,F., Miyazaki,R., Carraro,N., Pradervand,N. and van modules and related information about T4SS-type ICEs and der Meer,J.R. (2017) The hidden life of integrative and AICEs, making it possible to perform the automatically conjugative elements. FEMS microbiology reviews, 41, 512–537. computer-assisted reasoning, query, and advanced analysis of ICE. 3. Liu,M., Li,X., Xie,Y., Bi,D., Sun,J., Li,J., Tai,C., Deng,Z. and Ou,H.-Y. ICEberg 2.0: an updated database of bacterial However, ICEO is currently at its early development stage and integrative and conjugative elements. Nucleic Acids Res, will be further developed in the future. We will represent more 10.1093/nar/gky1123. known information about T4SS-type ICEs. Due to differences in cell membrane structure, Gram-positive and Gram-negative 4. Frost,L.S., Leplae,R., Summers,A.O. and Toussaint,A. (2005) bacteria have different T4SS organization and features and are Mobile genetic elements: the agents of open source evolution. associated with different T4SS-type ICEs information. Such Nature Reviews Microbiology, 3, 722–732. differential characteristics will be further categorized, modeled, 5. Toussaint,A. and Chandler,M. (2012) Prokaryote Genome and represented in ICEO. Besides, only 11 experimentally Fluidity: Toward a System Approach of the Mobilome. In van verified AICE are included in ICEO. Compared with T4SS-type Helden,J., Toussaint,A., Thieffry,D. (eds), Bacterial Molecular ICEs, AICE is less commonly seen in bacteria. However, AICE Networks: Methods and Protocols, Methods in Molecular is also important in terms of developing useful tools for genetic Biology. Springer New York, New York, NY, pp. 57–80. engineering of Actinobacteria (27). In the future, we plan to more systematically represent and analyze AICE information in ICEO. 6. Hoehndorf,R., Schofield,P.N. and Gkoutos,G.V. (2015) The role of ontologies in biological and biomedical research: a ICEO will be also be used for more applications. A major functional perspective. Brief Bioinform, 16, 1069–1080. application is to apply ICEO as the base of the next version of the ICEberg database to standardize and integrate the rich ICE 7. The Gene Ontology Consortium (2019) The Gene Ontology ICEKp1-Associated Genomic Heterogeneity in a Klebsiella Resource: 20 years and still GOing strong. Nucleic Acids pneumoniae Strain Isolated from a Primary Liver Abscess. J Research, 47, D330–D338. Bacteriol, 190, 515–526. 8. Smith,B., Ashburner,M., Rosse,C., Bard,J., Bug,W., 23. Wu,K.-M., Li,L.-H., Yan,J.-J., Tsao,N., Liao,T.-L., Tsai,H.- Ceusters,W., Goldberg,L.J., Eilbeck,K., Ireland,A., Mungall,C.J., C., Fung,C.-P., Chen,H.-J., Liu,Y.-M., Wang,J.-T., et al. (2009) et al. (2007) The OBO Foundry: coordinated evolution of Genome sequencing and comparative analysis of Klebsiella ontologies to support biomedical data integration. Nat pneumoniae NTUH-K2044, a strain causing liver abscess and Biotechnol, 25, 1251–5. meningitis. J. Bacteriol., 191, 4492–4501. 9. He,Y., Xiang,Z., Zheng,J., Lin,Y., Overton,J.A. and Ong,E. 24. Leplae,R., Hebrant,A., Wodak,S.J. and Toussaint,A. (2004) (2018) The eXtensible ontology development (XOD) principles ACLAME: A CLAssification of Mobile genetic Elements. and tool implementation to support ontology interoperability. J Nucleic Acids Res, 32, D45–D49. Biomed Semantics, 9, 3. 25. Toussaint,A., Lima-Mendez,G. and Leplae,R. (2007) PhiGO, 10. Group,O.W. and others (2009) OWL 2 Web Ontology a phage ontology associated with the ACLAME database. Language Document Overview: W3C Recommendation 27 Research in microbiology, 158, 567–71. October 2009. 26. Leplae,R., Lima-Mendez,G. and Toussaint,A. (2010) 11. Wilkinson,M.D., Dumontier,M., Aalbersberg,Ij.J., ACLAME: A CLAssification of Mobile genetic Elements, update Appleton,G., Axton,M., Baak,A., Blomberg,N., Boiten,J.-W., da 2010. Nucleic Acids Research, 38, D57–D61. Silva Santos,L.B., Bourne,P.E., et al. (2016) The FAIR Guiding 27. Raynal,A., Karray,F., Tuphile,K., Darbon-Rongère,E. and Principles for scientific data management and stewardship. Pernodet,J.-L. (2006) Excisable Cassettes: New Tools for Scientific Data, 3, 160018. Functional Analysis of Streptomyces Genomes. Appl. Environ. 12. Xiang,Z., Courtot,M., Brinkman,R.R., Ruttenberg,A. and Microbiol., 72, 4839–4844. He,Y. (2010) OntoFox: web-based support for ontology reuse. 28. Hur,J., Ozgur,A., Xiang,Z. and He,Y. (2012) Identification of BMC research notes, 3, 175. fever and vaccine-associated gene interaction networks using 13. He,Y., Liu,Y. and Zhao,B. (2014) OGG: a Biological ontology-based literature mining. Journal of biomedical Ontology for Representing Genes and Genomes in Specific semantics, 3, 18. Organisms. In ICBO. Citeseer, pp. 13–20. 29. Hur,J., Özgür,A., Xiang,Z. and He,Y. (2015) Development 14. Natale,D.A., Arighi,C.N., Blake,J.A., Bult,C.J., Christie,K.R., and application of an interaction network ontology for literature Cowart,J., D’Eustachio,P., Diehl,A.D., Drabkin,H.J., Helfer,O., mining of vaccine-associated gene-gene interactions. Journal of et al. (2014) Protein Ontology: a controlled structured network of Biomedical Semantics, 6, 2. protein entities. Nucleic Acids Res, 42, D415–D421. 15. NCBITaxon: An ontology representation of the NCBI organismal taxonomy. http://obofoundry.org/ontology/ncbitaxon.html. 16. Xiang,Z., Zheng,J., Lin,Y. and He,Y. (2015) Ontorat: automatic generation of new ontology terms, annotations, and axioms based on ontology design patterns. Journal of biomedical semantics, 6, 4. 17. Xiang,Z., Mungall,C., Ruttenberg,A. and He,Y. (2011) Ontobee: A linked data server and browser for ontology terms. In ICBO. 18. Ong,E., Xiang,Z., Zhao,B., Liu,Y., Lin,Y., Zheng,J., Mungall,C., Courtot,M., Ruttenberg,A. and He,Y. (2016) Ontobee: A linked ontology data server to support ontology term dereferencing, linkage, query and integration. Nucleic acids research, 10.1093/nar/gkw918. 19. Grenon,P. (2003) Spatio-temporality in basic formal ontology Ifomis. 20. Arp,R., Smith,B. and Spear,A. (2015) Building Ontologies using basic formal ontology. Cambridge, MA, USA, 2015. 21. NCBI is phasing out sequence GIs - use Accession.Version instead! (2012) https://www.ncbi.nlm.nih.gov/books/NBK431010/#news_03- 02-2016-phase-out-of-GI-numbers. 22. Lin,T.-L., Lee,C.-Z., Hsieh,P.-F., Tsai,S.-F. and Wang,J.-T. (2008) Characterization of Integrative and Conjugative Element Responses to the Reviewers: The greatest shortcoming of this paper is that it does not clearly ----------------------------------------------------------------- demonstrate the benefits of the ontological translation. Are the We appreciate the time and efforts of the reviewers who re- DL and SPARQL queries doing work that could not be done di- viewed our manuscript. Their suggestions and comments are rectly in ICEberg? constructive and helpful. We have revised our manuscript by in- corporating these comments accordingly as described below. It Reply: This is a good point. We have added a new paragraph in is noted that the italicized paragraphs below are the reviewer’s the Discussion part to address the reviewer’s comment: comments and our followed replies are in regular font style. “ICEO now is built by standardizing and integrating the rich information from ICEberg database. And ICEO can ----------------------- REVIEW 1 --------------------- perform tasks that cannot be done in current ICEberg. SUBMISSION: 23 For example, in our use case 1, we were able to easily TITLE: ICEO: a biological ontology for representing and ana- query any virulence factors for any level of MGEs. lyzing the bacterial integrative and conjugative element Currently, ICEberg will label VF for those MGEs that AUTHORS: Meng Liu, Hong-Yu Ou and Yongqun He are virulence factors. However, it is still impossible for users to query all VFs for a specific bacterial group. In ----------- Overall evaluation ----------- our use case 2, we further illustrate an efficient way of SCORE: 2 (accept) using ICEO to query ICEs under any specific bacterial taxon level like class, species or family. ICEberg can ----- TEXT: only query based on species level. However, rather than Nice, clearly written paper on the development of an application being only a complement or translation of ICEberg ontology for bacterial mobile genetic elements. Good strong database, ICEO and ICEO-based features will be use-case and I liked the inclusion of example queries that the on- explored to be integrated into ICEberg in the future.” tology could be used for. (Discussion part, column 1, paragraph 2) The gene-naming issue seems like it might be an issue long-term, but this is largely out of the authors control and they seem to ad- I reviewed the paper and looked at the project's GitHub reposi- dress it in a pragmatic way. tory and OWL files. The `iceo_merged.owl` file loaded in Protege and reasoned under HermiT. ICEO claims to be devel- Reply: Thanks for your kind and positive comments. oped according to OBO principles. It uses an OBO namespace , however no OBO ID The authors mention their use of the "eXtensive Ontology Devel- has been requested for ICEO. The paper claims a CC-BY 4.0 li- opment (XOD) strategy" which includes the involvement of com- cense, but the GitHub repository and OWL files contain no li- munity efforts to develop ontologies, but how they in fact involve cense information. I did not see any textual definitions for ICEO the community is not discussed. The paper would be improved terms. by the authors discussing how they had approached their inter- action with the community - e.g. by providing the ontology to ex- Reply: We have submitted the OBO ID request for ICEO and perts for review. have been refining the ICEO according to the community’s com- ments. CC-BY 4.0 license and all the textual definitions have Reply: We have submitted the ICEO to the OBO Foundry com- been added to the ICEO. munity and have been checking and refining the ICEO according to the community’s comments. Minor points: ----------------------- REVIEW 2 --------------------- - p2 "eXtensive Ontology Development" should be "eXtensible SUBMISSION: 23 Ontology Development" TITLE: ICEO: a biological ontology for representing and ana- - Figure 3 "particpants in" should be "participates in". lyzing the bacterial integrative and conjugative element AUTHORS: Meng Liu, Hong-Yu Ou and Yongqun He Reply: All the above points have been corrected in the latest version manuscript. ----------- Overall evaluation ----------- SCORE: 1 (weak accept) ----- TEXT: This paper describes the translation of a database (ICEberg) of ----------------------- REVIEW 3 --------------------- bacterial integrative and conjugative elements (ICEs) into an SUBMISSION: 23 ontology (ICEO). ICEO builds on existing OBO ontologies such TITLE: ICEO: a biological ontology for representing and ana- as the Basic Formal Ontology, the Protein Ontology, and the lyzing the bacterial integrative and conjugative element Ontology of Genes and Genomes. The authors explain the de- AUTHORS: Meng Liu, Hong-Yu Ou and Yongqun He sign patterns used to build ICEO and demonstrate how the re- sults can be queried. The presentation is clear and the subject is ----------- Overall evaluation ----------- in scope for ICBO, however the novelty is limited. SCORE: 2 (accept) ----- TEXT: Reply: We appreciate the reviewer’s summary and positive com- Liu et al. present the ICEO ontology for the representation of ments. bacterial integrative and conjugative element, ICE. In general the presentation is clear and the paper is fairly well written with just a few errors. Positive points: 1) The ontology is grounded in BFO and OBO-Foundry princi- ples quite well, and the authors make an appropriate nod to the FAIR principles. 2) The design patterns presented seem appropriate for repre- senting the domain of ICE. 3) The authors present the use of ICEO for querying of ICE via DL and SPARQL, and presumably the results would be useful to researchers in this domain. Reply: Thanks for your positive comments and advice. Of interest: The Authors state: "Since the Protein Ontology (PR) does not include all the proteins included in ICEO. We applied a similar strategy to represent protein names in ICEO (Fig. 2)." The should be a single sentence, but more importantly, the authors should request classes in the PR to cover the protein entities of interest, and then revise their ontology appropriately. This is part of working within the OBO Foundry community. Reply: Thanks for your advice. We have added a new paragraph in the Results part to address the reviewer’s comment: “ICE is essentially a genetic feature so that the gene representation is our priority. Since the Protein Ontol- ogy (PR) does not include all the proteins included in ICEO, we applied a similar strategy to represent protein names in ICEO (Fig. 2). We have also contacted the Protein Ontology (PR) team, and will request new clas- ses in the PR to improve ICEO.” (Results part, column 1, paragraph 4)