Biomedical ontologies in action: role in knowledge management, data integration and decision support.

OBJECTIVES: To provide typical examples of biomedical ontologies in action, emphasizing the role played by biomedical ontologies in knowledge management, data integration and decision support. METHODS: Biomedical ontologies selected for their practical impact are examined from a functional perspective. Examples of applications are taken from operational systems and the biomedical literature, with a bias towards recent journal articles. RESULTS: The ontologies under investigation in this survey include SNOMED CT, the Logical Observation Identifiers, Names, and Codes (LOINC), the Foundational Model of Anatomy, the Gene Ontology, RxNorm, the National Cancer Institute Thesaurus, the International Classification of Diseases, the Medical Subject Headings (MeSH) and the Unified Medical Language System (UMLS). The roles played by biomedical ontologies are classified into three major categories: knowledge management (indexing and retrieval of data and information, access to information, mapping among ontologies); data integration, exchange and semantic interoperability; and decision support and reasoning (data selection and aggregation, decision support, natural language processing applications, knowledge discovery). CONCLUSIONS: Ontologies play an important role in biomedical research through a variety of applications. While ontologies are used primarily as a source of vocabulary for standardization and integration purposes, many applications also use them as a source of computable knowledge. Barriers to the use of ontologies in biomedical applications are discussed.

Accessing and integrating data and knowledge for biomedical research.

OBJECTIVES: To review the issues that have arisen with the advent of translational research in terms of integration of data and knowledge, and survey current efforts to address these issues. METHODS: Using examples form the biomedical literature, we identified new trends in biomedical research and their impact on bioinformatics. We analyzed the requirements for effective knowledge repositories and studied issues in the integration of biomedical knowledge. RESULTS: New diagnostic and therapeutic approaches based on gene expression patterns have brought about new issues in the statistical analysis of data, and new workflows are needed are needed to support translational research. Interoperable data repositories based on standard annotations, infrastructures and services are needed to support the pooling and meta-analysis of data, as well as their comparison to earlier experiments. High-quality, integrated ontologies and knowledge bases serve as a source of prior knowledge used in combination with traditional data mining techniques and contribute to the development of more effective data analysis strategies. CONCLUSION: As biomedical research evolves from traditional clinical and biological investigations towards omics sciences and translational research, specific needs have emerged, including integrating data collected in research studies with patient clinical data, linking omics knowledge with medical knowledge, modeling the molecular basis of diseases, and developing tools that support in-depth analysis of research data. As such, translational research illustrates the need to bridge the gap between bioinformatics and medical informatics, and opens new avenues for biomedical informatics research.

Investigating implicit knowledge in ontologies with application to the anatomical domain.

Knowledge in biomedical ontologies can be explicitly represented (often by means of semantic relations), but may also be implicit, i.e., embedded in the concept names and inferable from various combinations of semantic relations. This paper investigates implicit knowledge in two ontologies of anatomy: the Foundational Model of Anatomy and GALEN. The methods consist of extracting the knowledge explicitly represented, acquiring the implicit knowledge through augmentation and inference techniques, and identifying the origin of each semantic relation. The number of relations (12 million in FMA and 4.6 million in GALEN), broken down by source, is presented. Major findings include: each technique provides specific relations; and many relations can be generated by more than one technique. The application of these findings to ontology auditing, validation, and maintenance is discussed, as well as the application to ontology integration.

An evaluation of hybrid methods for matching biomedical terminologies: mapping the gene ontology to the UMLS.

Integration of disparate biomedical terminologies is becoming increasingly important as links between biological science and clinical medicine grow. Mapping concepts in the Gene Ontology (GO) to the UMLS may help further this integration and allow for more efficient information exchange among researchers. Using a gold standard of GO term--UMLS concept mappings provided by the NCI, we examined the performance of various published and combined mapping techniques, in order to maximize precision and recall. We found that for the previously published techniques precision varied between (0.61-0.95), and recall varied from (0.65-0.90), whereas for the hybrid techniques, precision varied between (0.66-0.97), and recall from (0.59-0.93). Our study reveals the benefits of using mapping techniques that incorporate domain knowledge, and provides a basis for future approaches to mapping between distinct biomedical vocabularies.

From phenotype to genotype: issues in navigating the available information resources.

OBJECTIVES: As part of an investigation of connecting health professionals and the lay public to both disease and genomic information, we assessed the availability and nature of the data from the Human Genome Project relating to human genetic diseases. METHODS: We focused on a set of single gene diseases selected from main topics in MEDLINEplus, the NLM's principal resource focused on consumers. We used publicly available websites to investigate specific questions about the genes and gene products associated with the diseases. We also investigated questions of knowledge and data representation for the information resources and navigational issues. RESULTS: Many online resources are available but they are complex and technical. The major challenges encountered when navigating from phenotype to genotype were (1) complexity of the data, (2) dynamic nature of the data, (3) diversity of foci and number of information resources, and (4) lack of use of standard data and knowledge representation methods. CONCLUSIONS: Three major informatics issues arise from the navigational challenges. First, the official gene names are insufficient for navigation of these web resources. Second, navigational inconsistencies arise from difficulties in determining the number and function of alternate forms of the gene or gene product and maintaining currency with this information. Third, synonymy and polysemy cause much confusion. These are severe obstacles to computational navigation from phenotype to genotype, especially for individuals who are novices in the underlying science. Tools and standards to facilitate this navigation are sorely needed.

An object-oriented model for representing semantic locality in the UMLS.

Several information models have been developed for the Unified Medical Language System (UMLS). While some models are term-oriented, a knowledge-oriented model is needed for representing semantic locality, i.e. the various semantic links among concepts. We propose an object-oriented model in which the semantic features of the UMLS are made available through four major classes for representing Metathesaurus concepts, semantic types, inter-concept relationships and Semantic Network relationships. Additional semantic methods for reducing the complexity of the hierarchical relationships represented in the UMLS are proposed. Implementation details are presented, as well as examples of use. The interest of this approach is discussed.

Methods for exploring the semantics of the relationships between co-occurring UMLS concepts.

OBJECTIVE: To characterize the relationships among UMLS concepts that co-occur as MeSH descriptors in MEDLINE citations (1990-1999). DESIGN: 18,485 UMLS concepts involved in 7,928,608 directed pairs of co-occurring concepts were studied. For each directed pair of concepts C1-C2: (i) the "family" of C1 was built, using the UMLS Metathesaurus, and we tested whether or not C2 belonged to C1's family; (ii) we used the semantic categorization of Metathesaurus concepts through the UMLS Semantic Network and Semantic Groups to represent the semantics of the relationships between C1 and C2. RESULTS: In 6.5% of the directed pairs, the co-occurring concept C2 was found within the "family" of C1. Detailed results are given. The most frequent co-occurrences involved "Chemicals and Drugs" and "Chemicals and Drugs", as well as "Disorders" and "Chemicals and Drugs". DISCUSSION: This work takes advantage of both symbolic and statistical information represented in the UMLS, and analyzes their overlap. Further research is suggested.

Aggregating UMLS semantic types for reducing conceptual complexity.

The conceptual complexity of a domain can make it difficult for users of information systems to comprehend and interact with the knowledge embedded in those systems. The Unified Medical Language System (UMLS) currently integrates over 730,000 biomedical concepts from more than fifty biomedical vocabularies. The UMLS semantic network reduces the complexity of this construct by grouping concepts according to the semantic types that have been assigned to them. For certain purposes, however, an even smaller and coarser-grained set of semantic type groupings may be desirable. In this paper, we discuss our approach to creating such a set. We present six basic principles, and then apply those principles in aggregating the existing 134 semantic types into a set of 15 groupings. We present some of the difficulties we encountered and the consequences of the decisions we have made. We discuss some possible uses of the semantic groups, and we conclude with implications for future work.

Mapping the UMLS Semantic Network into general ontologies.

In this study, we analyzed the compatibility between an ontology of the biomedical domain (the UMLS Semantic Network) and two other ontologies: the Upper Cyc Ontology (UCO) and WordNet. 1) We manually mapped UMLS Semantic Types to UCO. One fifth of the UMLS Semantic Types had exact mapping to UCO types. UCO provides generic concepts and a structure that relies on a larger number of categories, despite its lack of depth in the biomedical domain. 2) We compared semantic classes in the UMLS and WordNet. 2% of the UMLS concepts from the Health Disorder class were present in WordNet, and compatibility between classes was 48%. WordNet, as a general language-oriented ontology is a source of lay knowledge, particularly important for consumer health applications.

Circular hierarchical relationships in the UMLS: etiology, diagnosis, treatment, complications and prevention.

The Unified Medical Language System (UMLS) is a large repository of some 800,000 concepts for the biomedical domain, organized by several millions of inter-concept relationships, either inherited from the source vocabularies, or specifically generated. This paper focuses on hierarchical relationships in the UMLS Metathesaurus, and especially, on circular hierarchical relationships. Using the metaphor of a disease, we first analyze the causal mechanisms for circular hierarchical relationships. Then, we discuss methods to identify and remove these relationships. Finally, we briefly discuss the consequences of these relationships for applications based on the UMLS, and we propose some prevention measures.

Evaluating UMLS strings for natural language processing.

The National Library of Medicine's Unified Medical Language System (UMLS) is a rich source of knowledge in the biomedical domain. The UMLS is used for research and development in a range of different applications, including natural language processing (NLP). In this paper we investigate the nature of the strings found in the UMLS Metathesaurus and evaluate them for their usefulness in NLP. We begin by identifying a number of properties that might allow us to predict the likelihood of a given string being found or not found in a corpus. We use a statistical model to test these predictors against our corpus, which is derived from the MEDLINE database. For one set of properties the model correctly predicted 77% of the strings that do not belong to the corpus, and 85% of the strings that do belong to the corpus. For another set of properties the model correctly predicted 96% of the strings that do not belong to the corpus and 29% of the strings that do belong to the corpus.

The NLM Indexing Initiative.

The objective of NLM's Indexing Initiative (IND) is to investigate methods whereby automated indexing methods partially or completely substitute for current indexing practices. The project will be considered a success if methods can be designed and implemented that result in retrieval performance that is equal to or better than the retrieval performance of systems based principally on humanly assigned index terms. We describe the current state of the project and discuss our plans for the future.

Using UMLS semantics for classification purposes.

The Unified Medical Language System (UMLS) contains semantic information about terms from various sources; each concept can be understood and located by its relationships to other concepts. We describe a method in which the semantic relationships between UMLS concepts are exploited for the purpose of classification. This method combines three existing components: 1) Mapping terms to UMLS concepts; 2) Restricting UMLS concepts to MeSH; and 3) Mapping MeSH terms to disease categories. When applied to the automatic classification of condition terms into broad disease categories in the Clinical Trials database, this method assigned relevant categories to 92% of the 1823 condition terms encountered. 135 (7%) failed to be classified and 14 (.77%) were misclassified. The limits of this method are discussed, as well as the reuse of existing components, and the tuning required to achieve automatic classification.

Identifying proper names in parallel medical terminologies.

We propose several criteria to identify proper names in biomedical terminologies. Traditional, pattern-based methods that rely on the immediate context of a proper name are not applicable. However, the availability of translations of some terminologies supports methods based on invariant words instead. A combination of five criteria achieved 86% precision and 88% recall on the 16,401 word forms of the International Classification of Diseases.

Periosteum and bone marrow in bone lengthening: a DEXA quantitative evaluation in rabbits.

We quantitatively studied the role of periosteum and bone marrow-endosteum during lengthening in 18 growing rabbits, comparing four surgical procedures: 1) periosteum and bone marrow preservation, 2) periosteum preservation, bone marrow destruction, 3) periosteum destruction, bone marrow preservation, 4) periosteum and bone marrow destruction. An external fixator was set on one femur, the other serving as a control. Distraction began on day 5 and stopped on day 25 (0.25 mm/12 hours). On day 30, femora were harvested with a layer of muscle. Area, bone mineral content and density were measured by dual-energy x-ray absorptiometry. Procedure 2 showed the highest increase in bone mineral content around the elongated callus (127%) compared to procedures: 1 (81%), 3 (25%) and 4 (-8%, i.e., resorption of bone ends). A statistically significant effect on bone formation was observed when preserving (vs. destroying): 1) periosteum, 2) bone marrow (effect observed only around the distraction gap), 3) periosteum and bone marrow in combination. Periosteum alone forms a larger callus, with more mineral content than bone marrow alone, and destruction of both results in the absence of bone formation around the distraction area. Careful preservation of periosteum is essential to bone healing. Formation of bone with a large mineral content does not require bone marrow preservation, but there is an interaction effect on healing between bone marrow and periosteum.

Evaluation of the Unified Medical Language System as a medical knowledge source.

OBJECTIVE: The authors evaluated the use of the Unified Medical Language System (UMLS) as a medical knowledge source for the representation of medical procedures in the MAOUSSC system. DESIGN: MAOUSSC, a multiaxial coding system, was used for the representation of 1500 procedures from 15 clinical specialties, using UMLS concepts (augmented by full sources for three new vocabularies being added to the UMLS) and relationships whenever possible. Evaluation criteria for the UMLS included (1) completeness of representation of concepts and of inter-concept relationships, (2) consistency in the categorization of both concepts and inter-concept relationships, and (3) usability, including adaptability of the UMLS to a foreign language (French), its suitability to a geographic region with different medical practices than the USA, and issues relative to the annual update changes in the test vocabularies. RESULTS: During the MAOUSSC trial, the number of missing concepts or relationships identified in the augmented UMLS sources was deemed to be inconsequential relative to overall project goals. "Missing" UMLS inter-concept relationships were identified, although they were small in number. Some inconsistencies in the UMLS were noted, especially in the area of hierarchic relationships. CONCLUSION: After UMLS was used for five years as a knowledge source for representing 1500 complex medical procedures in MAOUSSC, its value is considered significant. Future editions of the UMLS are expected to improve representation of inter-concept relationships and global consistency.

Beyond synonymy: exploiting the UMLS semantics in mapping vocabularies.

The Unified Medical Language System (UMLS) contains semantic information about terms from various sources, each concept can be understood and located by its relationships to other concepts: this is a result of the organizing principle of semantic locality. We describe a method in which the semantic relationships between concepts are used to map concepts from different vocabularies in the UMLS. Applied to mapping concepts to MeSH, this method is able to map 50 to 65% of the non-MeSH concepts to MeSH. A manual review of the mapping shows a relevance rate of 61%. Causes of failure include a lack of consistently represented relationships in the UMLS, and some inconsistencies in the categorization of the concepts. The limits of this method are discussed, as well as possible adaptations for other uses.

Integration of the analytical and alphabetical ICD10 in a coding help system. Proposal of a theoretical model for the ICD representation.

UNLABELLED: In French hospitals, medical diagnosis coding with the ICD10 is commonly performed and the use of effective tools would help coders in their task. AIM OF THIS WORK: To ameliorate an existing coding help system. This system, which already consists of the ICD10 analytical index, would be increased with the terms of the alphabetical index that includes lexical variants and additional terms as well. The addition of the second volume of the ICD would allow the coding of more terms and would lessen documentary silence. METHODS: The first step of this work was a careful study of a theoretical model of the ICD content. Then the alphabetical index file was submitted to a lexical analysis, and it was automatically transformed to be integrated into the existing coding help system. RESULTS: Compromise had to be made between a theoretical model and between what could be obtained in practice by an automatic processing of the file. Finally the alphabetical index was added to the initial thesaurus, which represents 42,000 terms and 4,000 additional words. Links between words and codes were also considerably increased, which has enhanced the searching possibilities of the tool and lessen documentary silence. Conversely the research time has been increased. CONCLUSION: Difficulties have to be encountered when trying to turn a manual tool into an automatic research tool.

A collaborative approach to building a terminology for medical procedures using a Web-based application: from specifications to daily use.

The MAOUSSC (Model for Assistance in the Orientation of a User within Coding Systems) Web server supports a collaborative work on the description of medical procedures. The specifications for the MAOUSSC application are conceptual modeling, definition of semantically fully described procedures, re-use of an existing vocabulary, the UMLS, and sharability. This paper reports on some difficulties in applying those principles in a networked building and updating of the terminology. The users are physicians who have to represent procedure terms in the MAOUSSC formalism. They must apply the constraints of the underlying model, and re-use the representation of the UMLS knowledge base. In our experience, we found that the implementation of syntactic and semantic constraints was not sufficient. Guidelines for pragmatical aspects in representation are required to make a collaborative approach in terminology building more operational.