- Open Access
Structuring and extracting knowledge for the support of hypothesis generation in molecular biology
© Roos et al; licensee BioMed Central Ltd. 2009
- Published: 01 October 2009
Hypothesis generation in molecular and cellular biology is an empirical process in which knowledge derived from prior experiments is distilled into a comprehensible model. The requirement of automated support is exemplified by the difficulty of considering all relevant facts that are contained in the millions of documents available from PubMed. Semantic Web provides tools for sharing prior knowledge, while information retrieval and information extraction techniques enable its extraction from literature. Their combination makes prior knowledge available for computational analysis and inference. While some tools provide complete solutions that limit the control over the modeling and extraction processes, we seek a methodology that supports control by the experimenter over these critical processes.
We describe progress towards automated support for the generation of biomolecular hypotheses. Semantic Web technologies are used to structure and store knowledge, while a workflow extracts knowledge from text. We designed minimal proto-ontologies in OWL for capturing different aspects of a text mining experiment: the biological hypothesis, text and documents, text mining, and workflow provenance. The models fit a methodology that allows focus on the requirements of a single experiment while supporting reuse and posterior analysis of extracted knowledge from multiple experiments. Our workflow is composed of services from the 'Adaptive Information Disclosure Application' (AIDA) toolkit as well as a few others. The output is a semantic model with putative biological relations, with each relation linked to the corresponding evidence.
We demonstrated a 'do-it-yourself' approach for structuring and extracting knowledge in the context of experimental research on biomolecular mechanisms. The methodology can be used to bootstrap the construction of semantically rich biological models using the results of knowledge extraction processes. Models specific to particular experiments can be constructed that, in turn, link with other semantic models, creating a web of knowledge that spans experiments. Mapping mechanisms can link to other knowledge resources such as OBO ontologies or SKOS vocabularies. AIDA Web Services can be used to design personalized knowledge extraction procedures. In our example experiment, we found three proteins (NF-Kappa B, p21, and Bax) potentially playing a role in the interplay between nutrients and epigenetic gene regulation.
- Text Mining
- Biological Model
- Conditional Random Field
- Entity Recognition
- Knowledge Extraction
We present the methodology in the following order: 1) a description of representing prior knowledge through proto-ontologies; 2) extension of the proto-ontologies by a workflow that adds instances to a semantic repository preloaded with the proto-ontologies; 3) a description of how to query the knowledge base; 4) a description of the toolkit that we use for knowledge extraction and knowledge management. Data and references are accessible from pack 58 on myExperiment.org .
Model representation in OWL
Different types of knowledge
Step one of our methodology is to define machine readable 'proto-ontologies' to represent our biological hypothesis within the scope of an experiment. The experiment in this case is a procedure to extract protein relations from literature. Our approach is based on the assumption that knowledge models can grow with each experiment that we or others perform. Therefore, we created a minimal OWL ontology of the relevant biological domain entities and their biological relations for our knowledge extraction experiment. The purpose of the experiment is to populate (enrich) the proto-ontologies with instances derived from literature. We also modeled the evidence that led to these instances. For instance, the process by which a protein name was found and in which document it was found. We find a clash between our intention of enriching a biological model, and the factual observations of a text mining procedure such as 'term', 'interaction assertion', or 'term collocation'. For example, it is obvious that collocation of the terms 'HDAC1' and 'p53' in one abstract does not necessarily imply collocation of the referred proteins in a cell. In order to avoid conflation of knowledge from the different stages of our knowledge extraction process, we purposefully kept distinct OWL models. This lead to the creation of the following models that will be treated in detail below:
❑ Biological knowledge for our hypothesis (Protein, Association)
❑ Text (Terms, Document references)
❑ Knowledge extraction process (Steps of the procedure)
❑ Extraction procedure implementation (Web Service and Workflow runs)
❑ Mapping model to integrate the above through references.
❑ Results (Instances of extracted terms and relations)
Text mining model
For more complete knowledge provenance, we also created a model representing the implementation of the text mining process as a workflow of (AIDA) Web Services. Example instances are (references to) the AIDA Web Services, and runs of these services. Following the properties of these instances we can retrace a particular run of the workflow.
Knowledge extraction experiment
The proto-ontologies form the basis of our knowledge base. They provide the initial templates for the knowledge that we wish to be able to interrogate in search of new hypotheses. The next step is to populate the knowledge base with instances. At the modeling stage we already anticipated that our first source of knowledge would be literature, and that we would obtain instances by text mining. An element of our approach is to regard knowledge extraction procedures as 'computational experiments' analogous to a wet laboratory experiments. We therefore used the workflow paradigm to design the protocol of our text mining experiment, here with the workflow design and enactment tool Taverna [13, 21]. A basic text mining workflow consists of the following steps: (i) Retrieve relevant documents from MedLine, in particular their abstracts, (ii) Extract protein names from the retrieved abstracts, and (iii) Present the results for inspection. We implemented the text mining process as a workflow (Figure 6). We added an additional sub-workflow to process the input query in order to extract known protein names from the input query and expand the query with synonyms for known protein names. For this, we employed a Web Service that provides UniProt identifiers and synonyms for human, rat and mouse gene names. These were derived from a combination of several public databases . The query is first split into its individual terms with a service from the AIDA Toolkit that wraps the Lucene tokenizer, and then all the terms (tokens) from the original query are checked for having a UniProt identifier by which they are identified as referring to a known protein. The sub-workflow makes the synonyms, UniProt identifiers, and the expanded query available for the rest of the workflow. The expanded query is the input for the next sub-workflow: document retrieval. We applied the document search service from the AIDA Toolkit, parameterized to use the regularly updated MedLine index that is stored on our AIDA server and updated daily. The output of this retrieval service is an xml document that contains elements of the retrieved documents, such as the PubMed identifier, title, abstract, and the authors. We then extract titles and abstracts for the next sub-workflow: i.e. protein name recognition. Sub-workflow 3 employs the AIDA Web Service 'applyCRF' to recognize protein (or gene) names in text. This service wraps a machine learning method based on the 'conditional random fields' approach . In this case it uses a recognition model trained on protein/gene names. We added the aforementioned UniProt service again to mark the extracted results as genuine human, rat, or mouse protein/gene names. In a number of cases the workflow produced more than one identifier for a single protein name. This is due to the ambiguity in gene and protein names. For instance, Tuason et al. reported 6.6% ambiguous occurrences of mouse gene names in text, and percentages ranging from 2.4% to 32.9% depending on the organism . The final step of our text mining procedure was to calculate a likelihood score for the extracted proteins to be found in documents retrieved through the expanded input query. We used a statistical method where the likelihood of finding a document with input query (q) and discovered protein name (d) is calculated by: , in which Q, D, and QD are the frequencies of documents containing q, d, and q and d, respectively; QD exp is the expected frequency of documents containing q and d assuming that their co-occurrence is a random event; N is the total number of documents in MedLine.
Querying the knowledge base
The AIDA Toolkit for knowledge extraction and knowledge management
The methodology that we propose enables a 'do-it-yourself' approach to extracting knowledge that can support hypothesis generation. To support this approach, we are developing an open source toolkit called Adaptive Information Disclosure Application (AIDA). AIDA is a generic set of components that can perform a variety of tasks related to knowledge extraction and knowledge management, such as perform specialized search on resource collections, learn new pattern recognition models, and store knowledge in a repository. W3C standards are used to make data accessible and manageable with Semantic Web technologies such as OWL, RDF(S), and SKOS. AIDA is also based on Lucene and Sesame. Most components are available as web services and are open source under an Apache license. AIDA is composed of three main modules: Search, Learning, and Storage.
Search – the information retrieval module
AIDA provides components which enable retrieval from a set of documents given a query, similar to popular search engines such as Google, Yahoo!, or PubMed. To make a set of documents (a corpus) searchable, an 'index' needs to be created first . For this the AIDA's configurable Indexer can be used. The Indexer and Search components are built upon Apache Lucene, version 2.1.0 , and, hence, indexes or other systems based on Lucene can easily be integrated with AIDA. The Indexer component takes care of the preprocessing (the conversion, tokenization, and possibly normalization) of the text of each document as well as the subsequent index generation. Different fields can be made retrievable such as title, document name, authors, or the entire contents. The currently supported document encodings are Microsoft Word, Portable Document Format (PDF), MedLine, XML, and plain text. The so-called "DocumentHandlers" which handle the actual conversion of each source file are loaded at runtime, so a handler for any other proprietary document encoding can be created and used instantly. Because Lucene is used as a basis, a plethora of options and/or languages are available for stemming, tokenization, normalization, or stop word removal which may all be set on a per-field, per-document type, or per-index basis using the configuration. An index can currently be constructed using either the command-line, a SOAP webservice (with the limitation of 1 document per call), or using a Taverna plugin.
Learning – the machine learning module
AIDA includes several components which enable information extraction from text data in the Learning module. These components are referred to as learning tools. The large community working on the information extraction task has already produced numerous data sets and tools to work with. To be able to use existing solutions, we incorporated some of the models trained on the large corpora into the named entity recognition web service NERecognizerService. These models are provided by LingPipe and range from the very general named entity recognition (detecting locations, person and organization names) to the specific models in the biomedical field created to recognize protein names and other bio-entities. We specified several options for input/output, which gives us an opportunity to work with either text data or the output of the search engine Lucene. We also offer the LearnModel web service whose aim is to produce a model from annotated text data. A model is based on the contextual information and uses learning methods provided by Weka  libraries. Once such a model is created, it can be used by the TestModel web service to annotate texts in the same domain. In this paper we use an AIDA service that applies a service for an algorithm that uses sequential models, such as conditional random fields (CRFs)/CRFs have an advantage over Hiddem Markov Models because of their ability to relax the independence assumption by defining a conditional probability distribution over label sequences given an observation sequence. We used CRFs to detect named entities in several domains like acids of various lengths in the food informatics field or protein names in the biomedical field .
Named entity recognition constitutes only one subtask in information extraction. Relation extraction can be viewed as the logical next step after the named entity recognition is carried out . This task can be decomposed into the detection of named entities, followed by the verification of a given relation among them. For example, given extracted protein names, it should possible to infer whether there is any interaction between two proteins. This task is accomplished by the RelationLearner web service. It uses an annotated corpus of relations to induce a model, which consequently can be applied to the test data with already detected named entities. The RelationLearner focuses on extraction of binary relations given the sentential context. Its output is a list of the named entities pairs, where the given relation holds.
The other relevant area for information extraction is detection of the collocations (or n-grams in the broader sense). This functionality is provided by the CollocationService which, given a folder with text documents, outputs the n-grams of the desired frequency and length.
Storage – the metadata storage module
AIDA includes components for the storage and processing of ontologies, vocabularies, and other structured metadata in the Storage module. The main component, also for the work described in this paper, is RepositoryWS, a service wrapper for Sesame – an open source framework for storage, inferencing and querying of RDF data on which most of this module's implementation is based [30, 31]. ThesaurusRepositoryWS is an extension of RepositoryWS that provides convenient access methods for SKOS thesauri. The Sesame RDF repository offers an HTTP interface and a Java API. In order to be able to integrate Sesame into workflows we created a SOAP service that gives access to the Sesame Java API. We accommodate for extensions to other RDF repositories, such as the HP Jena, Virtuoso, Allegrograph repositories or future versions of Sesame, by implementing the Factory design pattern.
Complementary services from BioSemantics applications
One of the advantages of a workflow approach is the ability to include services created elsewhere in the scientific community ('collaboration by Web Services'). For instance, in our BioAID workflows operations are used for query expansion and validation of protein names by UniProt identifiers. AIDA is therefore complemented by services derived from text mining applications such as Anni 2.0 from the BioSemantics group . The 'BioSemantics' group is particularly strong in disambiguation of the names of biological entities such as genes/proteins, intelligent biological query expansion (manuscript in preparation), and provision of several well known identifiers for biological entities through carefully compiled sets of names and identifiers around a biological concept.
User interfaces for AIDA
In addition to RDF manipulation within workflows as described in this document, several examples of user interactions have been made available in AIDA clients such as HTML web forms, AJAX web applications, and a Firefox toolbar. The clients access RepositoryWS for querying RDF through the provided Java Servlets. The web services in Storage have recently been updated from the Sesame 1.2 Java API to the Sesame 2.0 Java API. Some of the new features that Sesame 2.0 provides, such as SPARQL support and named graphs, are now being added to our web service API's and incorporated into our applications.
Our methodology for supporting the generation of a hypothesis about a biomolecular mechanism is based on a combination of tools and expertise from the fields of Semantic Web, e-Science, information retrieval, and information extraction. This novel combination has a number of benefits. First, the use of RDF and OWL removes the technical obstacle for making models interoperable with other knowledge resources on the Semantic Web although semantic interoperability will often require an alignment process to take place for more far reaching compatibility. The modeling approach that we propose is complementary to the efforts of communities such as the Open Biomedical Ontology (OBO) community. This community's stated purpose is to create an 'accurate representation of biological reality' by developing comprehensive domain ontologies and reconciling existing ontologies according to a number of governing principles . Our ambitions are more modest. We start with a minimal model to represent a hypothesis, i.e. a particular model of reality. We define our own classes and properties within the scope of a knowledge extraction experiment, but because of the modularity supported by OWL this does not exclude integration with other ontologies. In fact, integration with existing knowledge resources enables a complementary approach for finding facts potentially relevant to a hypothesis. Clearly, in order to scale up our methodology to represent knowledge beyond the experiments of a small group of researchers, alignment with standards would have to be considered. Upper ontologies can facilitate integration (for an example see ), and we can benefit from the OBO guidelines and the tools that have been developed to convert OBO ontologies to OWL [33–35]. Another interesting possibility is the integration with thesauri based on the SKOS framework . Relations between SKOS concepts (terms) are defined by simple 'narrower' and 'broader' relations that turn out to be effective for human computer interfaces, and may be the best option for labeling the elements in our semantic models. Instead of providing a text string as a human readable label, we could associate an element with an entry in a SKOS thesaurus, which is a valuable knowledge resource in itself. The SKOS format is useful as an approach for 'light-weight' knowledge integration that avoids the problems of ontological over-commitment associated with more powerful logics like OWL DL .
A second benefit of our methodology comes from the implementation of the knowledge extraction procedure as a workflow. The procedure for populating an ontology is similar to the one previously described by Witte et al. , but our implementation allows the accumulation of knowledge by repeatedly running the same workflow or adaptations of it. This enables us to perform posterior analyses over the results from several experiments by querying the knowledge base, for instance in a new workflow that uses the AIDA semantic repository service. Moreover, the approach is not limited to text mining. If one considers text documents as a particular form of data, we can generalize the principle to any computational experiment in which the output can be related to a qualitative biological model. As such, this work extends previous work on integration of genome data via semantic annotation . In this case the annotation is carried out by a workflow. Considering that there are thousands of Web Services and hundreds of workflows available for bioinformaticians , numerous extensions to our workflow can be explored. In addition, the combination with a semantic model allows us to collect evidence information as a type of knowledge provenance during workflow execution. In this way, we were able to address the issue of keeping a proper log of what has happened to our data during computational experimentation, analogous to the lab journal typically required in wet labs . Ideally, the knowledge provenance captured in our approach would be more directly supported by existing workflow systems. However, this is not yet possible. There seems to be a knowledge gap between workflow investigators and the users from a particular application domain with regard to provenance. We propose that workflow systems take care of execution level provenance and provide an RDF interface on which users can build their own provenance model. In this context, it will be interesting to see if we will be able to replace our workflow model and link directly to the light weight provenance model that is being implemented for Taverna 2 . A third benefit is that the application of Semantic Web, Web Services, and workflows stored on myExperiment.org, allow all resources relevant to an experiment to be shared on the web, making our results more reproducible. We would like to increase the 'liquidity' of knowledge so that knowledge extracted from computational experiments can eventually fit into frameworks for scientific discourse (hypotheses, research statements and questions, etc.) such as Semantic Web Applications in Neuromedicine (SWAN) . If it is to be global, interoperability across modes of discourse would require large scale consensus on how to express knowledge provenance, not only about knowledge produced from computational experiments but also from manual or human assertions. Some groups are attempting to address various aspects of this problem, such as the Scientific Discourse task force  in the W3C Semantic Web Health Care and Life Sciences Interest Group , the Concept Web Alliance  and the Shared Names initiative .
In this paper we demonstrate a methodology for a 'do it yourself' approach for the extraction and management of knowledge in support of generating hypotheses about biomolecular mechanisms. Our approach describes how one can create a personal model for a specific hypothesis and how a personal 'computational experiment' can be designed and executed to extract knowledge from literature and populate a knowledge base. A significant advantage of the methodology is the possibility it creates to perform analyses across the results of several of these knowledge extraction experiments. Moreover, the principle of semantic disclosure of results from a computational experiment is not limited to text mining. In principle, it can be applied to any kind of experiment of which the (interpretations of) results can be converted to semantic models, almost as a 'side effect' of the experiment at hand. Experimental data is automatically semantically annotated which makes it manageable within the context of its purpose: biological study. We consider this an intuitive and flexible way of enabling the reuse of data. With the use of Web Services from the AIDA Toolkit and others, we also demonstrated the exploitation of the expertise of computational scientists with diverse backgrounds, i.e. where knowledge sharing takes place at the level of services and qualitative models. We consider the demonstration of e-Science and Semantic Web tools for a personalized approach in the context of scientific communities to be one of the main contributions of our methodology. In summary, the methodology provides a basis for automated support for hypothesis formation in the context of experimental science. Future extensions will be driven by biological studies on specific biomolecular mechanisms such as the role of histone modifications in transcription. We also plan to evaluate general strategies for extracting novel ideas from a growing repository of structured knowledge.
We thank the myGrid team and OMII-UK for their support in applying their e-Science tools, and Machiel Jansen for his contribution to the early development of AIDA. This work was carried out in the context of the Virtual Laboratory for e-Science program (VL-e) and the BioRange program. These programs are supported by BSIK grants from the Dutch Ministry of Education, Culture and Science (OC&W). Special thanks go to Bob Hertzberger who made the VL-e project a reality.
This article has been published as part of BMC Bioinformatics Volume 10 Supplement 10, 2009: Semantic Web Applications and Tools for Life Sciences, 2008. The full contents of the supplement are available online at http://www.biomedcentral.com/1471-2105/10?issue=S10.
- Antoniou G, van Harmelen F: A Semantic Web Primer. MIT Press; 2004.Google Scholar
- Neumann E: A life science Semantic Web: are we there yet? Sci STKE 2005, 2005(283):pe22. 10.1126/stke.2832005pe22PubMedGoogle Scholar
- Rubin DL, Shah NH, Noy NF: Biomedical ontologies: a functional perspective. Brief Bioinform 2008, 9(1):75–90. 10.1093/bib/bbm059View ArticlePubMedGoogle Scholar
- Smith B, Ashburner M, Rosse C, Bard J, Bug W, Ceusters W, Goldberg LJ, Eilbeck K, Ireland A, Mungall CJ, et al.: The OBO Foundry: coordinated evolution of ontologies to support biomedical data integration. Nature biotechnology 2007, 25(11):1251–1255. 10.1038/nbt1346PubMed CentralView ArticlePubMedGoogle Scholar
- Stein LD: Towards a cyberinfrastructure for the biological sciences: progress, visions and challenges. Nature reviews 2008, 9(9):678–688.View ArticlePubMedGoogle Scholar
- Ashburner M, Mungall CJ, Lewis SE: Ontologies for biologists: a community model for the annotation of genomic data. Cold Spring Harbor symposia on quantitative biology 2003, 68: 227–235. 10.1101/sqb.2003.68.227View ArticlePubMedGoogle Scholar
- Spasic I, Ananiadou S, McNaught J, Kumar A: Text mining and ontologies in biomedicine: making sense of raw text. Brief Bioinform 2005, 6(3):239–251. 10.1093/bib/6.3.239View ArticlePubMedGoogle Scholar
- Weeber M, Kors JA, Mons B: Online tools to support literature-based discovery in the life sciences. Brief Bioinform 2005, 6(3):277–286. 10.1093/bib/6.3.277View ArticlePubMedGoogle Scholar
- Katrenko S, Adriaans P: Using Semi-Supervised Techniques to Detect Gene Mentions. Second BioCreative Challenge Workshop: 2007 2007.Google Scholar
- Gomez-Perez A, Manzano-Macho D: An overview of methods and tools for ontology learning from texts. Knowledge Engineering Review 2004, 19(3):187–212.View ArticleGoogle Scholar
- Missikoff M, Velardi P, Fabriani P: Text mining techniques to automatically enrich a domain ontology. Applied Intelligence 2003, 18(3):323–340. 10.1023/A:1023254205945View ArticleGoogle Scholar
- Goble C, Stevens R, Hull D, Wolstencroft K, Lopez R: Data curation + process curation = data integration + science. Brief Bioinform 2008, 9(6):506–517. 10.1093/bib/bbn034View ArticlePubMedGoogle Scholar
- Hull D, Wolstencroft K, Stevens R, Goble C, Pocock MR, Li P, Oinn T: Taverna: a tool for building and running workflows of services. Nucl Acids Res 2006, (34 Web Server):W729-W732. 10.1093/nar/gkl320Google Scholar
- Inda MA, van Batenburg MF, Roos M, Belloum AS, Vasunin D, Wibisono A, van Kampen AH, Breit TM: SigWin-detector: a Grid-enabled workflow for discovering enriched windows of genomic features related to DNA sequences. BMC research notes 2008, 1: 63. 10.1186/1756-0500-1-63PubMed CentralView ArticlePubMedGoogle Scholar
- Romano P: Automation of in-silico data analysis processes through workflow management systems. Brief Bioinform 2008, 9(1):57–68. 10.1093/bib/bbm056View ArticlePubMedGoogle Scholar
- Fisher P, Hedeler C, Wolstencroft K, Hulme H, Noyes H, Kemp S, Stevens R, Brass A: A systematic strategy for large-scale analysis of genotype phenotype correlations: identification of candidate genes involved in African trypanosomiasis. Nucleic Acids Res 2007, 35(16):5625–5633. 10.1093/nar/gkm623PubMed CentralView ArticlePubMedGoogle Scholar
- Goble CA, De Roure DC: myExperiment: social networking for workflow-using e-scientists. In 2nd workshop on Workflows in support of large-scale science. Monterey, California, USA: ACM Press; 2007:1–2.View ArticleGoogle Scholar
- Web Services Description Language (WSDL) 1.1[http://www.w3.org/TR/wsdl]
- Meij EJ, IJzereef LHL, Azzopardi LA, Kamps J, de Rijke M, Voorhees EM, P BL: Combining Thesauri-based Methods for Biomedical Retrieval. In The Fourteenth Text REtrieval Conference (TREC 2005). National Institute of Standards and Technology. NIST Special Publication; 2006.Google Scholar
- Supporting material for this paper on myExperiment (pack 58)[http://www.myexperiment.org/packs/58]
- The Taverna workbench[http://www.mygrid.org.uk/tools/taverna/]
- Kors JA, Schuemie MJ, Schijvenaars BJA, Weeber M, Mons B: Combination of Genetic Databases for Improving Identification of Genes and Proteins in Text. BioLINK: 2005; Detroit, Michigan, USA 2005.Google Scholar
- Katrenko S, Adriaans PW: Using Semi-Supervised Techniques to Detect Gene Mentions. Second BioCreative Challenge Workshop: 2007; Madrid, Spain 2007.Google Scholar
- Tuason O, Chen L, Liu H, Blake JA, Friedman C: Biological Nomenclatures: A source of Lexical Knowledge and Ambiguity. Pacific Symposium on Biocomputing. Fairmont Orchid, Hawaii 2004.Google Scholar
- Manning CD, Raghavan P, Schütze H: Introduction to Information Retrieval. Cambridge University Press; 2004.Google Scholar
- The Apache Lucene project[http://lucene.apache.org]
- Witten IH, Frank E: Data Mining: Practical machine learning tools and techniques. 2nd edition. San Francisco: Morgan Kaufmann; 2005.Google Scholar
- Katrenko S, Adriaans P: Learning Relations from Biomedical Corpora Using Dependency Trees. KDECB (Knowledge Discovery and Emergent Complexity in BioInformatics): 2006 2006.Google Scholar
- Sesame Open Source community web site[http://www.openrdf.org/]
- Broekstra J, Kampman A, van Harmelen F: Sesame: A Generic Architecture for Storing and Querying RDF and RDF Schema. In The Semantic Web – ISWC 2002: First International Semantic Web Conference. Volume 2342/2002. Sardinia, Italy: Springer Berlin/Heidelberg; 2002:54.View ArticleGoogle Scholar
- Jelier R, Schuemie MJ, Veldhoven A, Dorssers LC, Jenster G, Kors JA: Anni 2.0: a multipurpose text-mining tool for the life sciences. Genome Biology 2008, 9(6):R96. 10.1186/gb-2008-9-6-r96PubMed CentralView ArticlePubMedGoogle Scholar
- Hoehndorf R, Loebe F, Kelso J, Herre H: Representing default knowledge in biomedical ontologies: application to the integration of anatomy and phenotype ontologies. BMC Bioinformatics 2007, 8: 377. 10.1186/1471-2105-8-377PubMed CentralView ArticlePubMedGoogle Scholar
- Moreira DA, Musen MA: OBO to OWL: a protege OWL tab to read/save OBO ontologies. Bioinformatics 2007, 23(14):1868–1870. 10.1093/bioinformatics/btm258View ArticlePubMedGoogle Scholar
- Mungall CJ: Obol: integrating language and meaning in bio-ontologies. Comparative and functional genomics 2004, 5(6–7):509–520. 10.1002/cfg.435PubMed CentralView ArticlePubMedGoogle Scholar
- SKOS Simple Knowledge Organization System Reference Google Scholar
- Jupp S, Stevens R, Bechhofer S, Yesilada Y, Kostkova P: Knowledge Representation for Web Navigation. Semantic Web Applications and Tools for the Life Sciences (SWAT4LS 2008). Edinburgh 2008.Google Scholar
- Witte R, Kappler T, Baker CJO: Ontology Design for Biomedical Text Mining. In Semantic Web: Revolutionizing Knowledge Discovery in the Life Sciences. Edited by: Baker CJO, Cheung K-H. New York: Springer Science+Business Media; 2007:281–313.View ArticleGoogle Scholar
- Post LJ, Roos M, Marshall MS, van Driel R, Breit TM: A semantic web approach applied to integrative bioinformatics experimentation: a biological use case with genomics data. Bioinformatics 2007, 23(22):3080–3087. 10.1093/bioinformatics/btm461View ArticlePubMedGoogle Scholar
- Stevens R, Zhao J, Goble C: Using provenance to manage knowledge of in silico experiments. Brief Bioinform 2007, 8(3):183–194. 10.1093/bib/bbm015View ArticlePubMedGoogle Scholar
- Missier P, Belhajjame K, Zhao J, Goble C: Data lineage model for Taverna workflows with lightweight annotation requirements. IPAW'08. Salt Lake City, Utah 2008.Google Scholar
- Clark T, Kinoshita J: Alzforum and SWAN: the present and future of scientific web communities. Brief Bioinform 2007, 8(3):163–171. 10.1093/bib/bbm012View ArticlePubMedGoogle Scholar
- Scientific Discourse task group of W3C Health Care and Life Science Interest Group[http://esw.w3.org/topic/HCLSIG/SWANSIOC]
- W3C Semantic Web Health Care and Life Science Interest Group[http://www.w3.org/2001/sw/hcls/]
- Concept Web Alliance[http://conceptweblog.wordpress.com/about/]
- Shared Names Initiative[http://sharedname.org]
- Verschure PJ: Chromosome organization and gene control: it is difficult to see the picture when you are inside the frame. Journal of cellular biochemistry 2006, 99(1):23–34. 10.1002/jcb.20957View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.