Integration of relational and textual biomedical sources. A pilot experiment using a semi-automated method for logical schema acquisition.

OBJECTIVES: Bringing together structured and text-based sources is an exciting challenge for biomedical informaticians, since most relevant biomedical sources belong to one of these categories. In this paper we evaluate the feasibility of integrating relational and text-based biomedical sources using: i) an original logical schema acquisition method for textual databases developed by the authors, and ii) OntoFusion, a system originally designed by the authors for the integration of relational sources. METHODS: We conducted an integration experiment involving a test set of seven differently structured sources covering the domain of genetic diseases. We used our logical schema acquisition method to generate schemas for all textual sources. The sources were integrated using the methods and tools provided by OntoFusion. The integration was validated using a test set of 500 queries. RESULTS: A panel of experts answered a questionnaire to evaluate i) the quality of the extracted schemas, ii) the query processing performance of the integrated set of sources, and iii) the relevance of the retrieved results. The results of the survey show that our method extracts coherent and representative logical schemas. Experts' feedback on the performance of the integrated system and the relevance of the retrieved results was also positive. Regarding the validation of the integration, the system successfully provided correct results for all queries in the test set. CONCLUSIONS: The results of the experiment suggest that text-based sources including a logical schema can be regarded as equivalent to structured databases. Using our method, previous research and existing tools designed for the integration of structured databases can be reused - possibly subject to minor modifications - to integrate differently structured sources.

An agent- and ontology-based system for integrating public gene, protein, and disease databases.

In this paper, we describe OntoFusion, a database integration system. This system has been designed to provide unified access to multiple, heterogeneous biological and medical data sources that are publicly available over Internet. Many of these databases do not offer a direct connection, and inquiries must be made via Web forms, returning results as HTML pages. A special module in the OntoFusion system is needed to integrate these public 'Web-based' databases. Domain ontologies are used to do this and provide database mapping and unification. We have used the system to integrate seven significant and widely used public biomedical databases: OMIM, PubMed, Enzyme, Prosite and Prosite documentation, PDB, SNP, and InterPro. A case study is detailed in depth, showing system performance. We analyze the system's architecture and methods and discuss its use as a tool for biomedical researchers.

Designing new methodologies for integrating biomedical information in clinical trials.

OBJECTIVES: To propose a modification to current methodologies for clinical trials, improving data collection and cost-efficiency. To describe a system to integrate distributed and heterogeneous medical and genetic databases for improving information access, retrieval and analysis of biomedical information. METHODS: Data for clinical trials can be collected from remote, distributed and heterogeneous data sources. In this distributed scenario, we propose an ontologybased approach, with two basic operations: mapping and unification. Mapping outputs the semantic model of a virtual repository with the information model of a specific database. Unification provides a single schema for two or more previously available virtual repositories. In both processes, domain ontologies can improve other traditional approaches. RESULTS: Private clinical databases and public genomic and disease databases (e.g., OMIM, Prosite and others) were integrated. We successfully tested the system using thirteen databases containing clinical and biological information and biomedical vocabularies. CONCLUSIONS: We present a domain-independent approach to biomedical database integration, used in this paper as a reference for the design of future models of clinico-genomic trials where information will be integrated, retrieved and analyzed. Such an approach to biomedical data integration has been one of the goals of the IST INFOBIOMED Network of Excellence in Biomedical Informatics, funded by the European Commission, and the new ACGT (Advanced Clinico-Genomic Trials on Cancer) project, where the authors will apply these methods to research experiments.

ONTOFUSION: ontology-based integration of genomic and clinical databases.

ONTOFUSION is an ontology-based system designed for biomedical database integration. It is based on two processes: mapping and unification. Mapping is a semi-automated process that uses ontologies to link a database schema with a conceptual framework-named virtual schema. There are three methodologies for creating virtual schemas, according to the origin of the domain ontology used: (1) top-down--e.g. using an existing ontology, such as the UMLS or Gene Ontology--, (2) bottom-up--building a new domain ontology-- and (3) a hybrid combination. Unification is an automated process for integrating ontologies and hence the database to which they are linked. Using these methods, we employed ONTOFUSION to integrate a large number of public genomic and clinical databases, as well as biomedical ontologies.

Establishing an agenda for biomedical informatics.

OBJECTIVE: To describe potential areas of collaboration between Medical Informatics (BI) and Bioinformatics (BI) and their effects on planning future work in both disciplines. METHODS: Some reflections on the objectives and rationale underpinning MI and BI are given, and preliminary results from the BIOINFOMED workgroup, supported by the European Commission, are introduced. RESULTS: Applications from both subfields suggest topics for sharing and exchange between the subfields within the emerging field of Biomedical Informatics. CONCLUSIONS: We suggest how the nature and degree of collaboration between the sub-disciplines can impact future work in molecular medicine.

Theory, abstraction and design in medical informatics.

OBJECTIVE: To analyze the scientific and engineering components of Medical Informatics. A clear characterization of these components should be undertaken to categorize different areas of Medical Informatics and create a research agenda for the future. METHODS: We have adapted a classical ACM and IEEE report on computing to analyze Medical Informatics from three different viewpoints: Theory, Abstraction, and Design. RESULTS: We suggest that Medical Informatics can be considered from these three perspectives: (1) Theory, from which medical informaticians formally characterize the properties of the objects of study, creating new theories or using and adapting existing theories (e.g., from mathematics), (2) Abstraction, from which medical informaticians deal with all aspects of medical information and create new abstractions, methods, and technology-independent models, which can be experimentally verified, and (3) Design, from which medical informaticians develop systems or act as information brokers or advisors between medical and technology professionals, to improve the quality of computer applications in medicine. CONCLUSION: Based on this framework, we suggest that Medical Informatics has an independent scientific character, different from other applied informatics areas. Finally, we analyze these three perspectives using data mining in medicine.

A comparison of machine learning methods for the diagnosis of pigmented skin lesions.

We analyze the discriminatory power of k-nearest neighbors, logistic regression, artificial neural networks (ANNs), decision tress, and support vector machines (SVMs) on the task of classifying pigmented skin lesions as common nevi, dysplastic nevi, or melanoma. Three different classification tasks were used as benchmarks: the dichotomous problem of distinguishing common nevi from dysplastic nevi and melanoma, the dichotomous problem of distinguishing melanoma from common and dysplastic nevi, and the trichotomous problem of correctly distinguishing all three classes. Using ROC analysis to measure the discriminatory power of the methods shows that excellent results for specific classification problems in the domain of pigmented skin lesions can be achieved with machine-learning methods. On both dichotomous and trichotomous tasks, logistic regression, ANNs, and SVMs performed on about the same level, with k-nearest neighbors and decision trees performing worse.