Distribution of immunodeficiency fact files with XML--from Web to WAP.

BACKGROUND: Although biomedical information is growing rapidly, it is difficult to find and retrieve validated data especially for rare hereditary diseases. There is an increased need for services capable of integrating and validating information as well as proving it in a logically organized structure. A XML-based language enables creation of open source databases for storage, maintenance and delivery for different platforms.
METHODS: Here we present a new data model called fact file and an XML-based specification Inherited Disease Markup Language (IDML), that were developed to facilitate disease information integration, storage and exchange. The data model was applied to primary immunodeficiencies, but it can be used for any hereditary disease. Fact files integrate biomedical, genetic and clinical information related to hereditary diseases.
RESULTS: IDML and fact files were used to build a comprehensive Web and WAP accessible knowledge base ImmunoDeficiency Resource (IDR) available at http://bioinf.uta.fi/idr/. A fact file is a user oriented user interface, which serves as a starting point to explore information on hereditary diseases.
CONCLUSION: The IDML enables the seamless integration and presentation of genetic and disease information resources in the Internet. IDML can be used to build information services for all kinds of inherited diseases. The open source specification and related programs are available at http://bioinf.uta.fi/idml/.

Background

Biomedical information is often very complex. Deciphering the roles of genes in human health and disease is a grand challenge for many reasons, including impediments to defining phenotypes, difficulties in identifying and quantifying environmental effects, technical problems in generating genotypic information, and the difficulties of studying humans [1]. The completion of the draft sequence of the human genome [2,3] and advances in molecular biology provide new opportunities to increase our understanding of the role of genetic factors in human health and disease [1]. The number of identified genetic diseases has increased exponentially [4]. The new knowledge can be applied to the prevention, diagnosis and treatment of diseases. This far, the knowledge of genetics has had a large role in the health care of only a few patients and a small role in the health care of many [5]. The biomedical informatics holds great promise for developing informatics methods that will be crucial in the development of genomic medicine [6].

Most hereditary diseases are rare and the diagnosed patients for a condition are often randomly spread out in the world. One doctor usually has only a few patients with a disease. It is often difficult to find comprehensive and validated biomedical information related to rare diseases. In addition, it is more and more difficult to publish results in scientific journals only from a few cases even when they are interesting [7]. Still, all these pieces of information can contain clues to understanding the fundamental defects at molecular level and can help to develop targeted treatments. The scattering of the disease-related information to literature and Internet is a big obstacle especially for those interested in rare diseases. First of all, there may not be that much data for these diseases and secondly it may be very difficult to find and collect. Further, the user has often difficulties in assessing the quality of data.

There is an increasing need for tools and services capable of integrating information from a variety of sources. Clinicians and researchers could benefit from a more consolidated and unified view of the available biomedical data. Systems biology researchers need to integrate disparate information from multiple public sources to merge with their own experimental data to generate models of processes. Biomedical data mining attempts to extract information from biomedical databases by using e.g. automated natural language processing (NLP) techniques [8]. Processing of biomedical texts presents many challenges such as in the areas of terminology or ontology building, information extraction from texts, knowledge discovery from collections of documents, as well as sharing and integrating knowledge from factual and textual data bases, semantic annotation, etc. Without standardized nomenclature the information extraction (IE) about a particular subject from various resources is difficult. Due to ambiguity of terms, a search for a particular term often retrieves results for unrelated entities. Since there are also some technical problems arising from the diversity of computer hardware and software, there is a need for such a data form, that can be handled by any computer and which can be easily presented on any platform.

The Extensible Markup Language (XML) is a standard created by the World Wide Web Consortium (W3C) for characterizing the content and structure of documents [9]. It is designed to improve the functionality of the Web by enabling more flexible and adaptable information identification and presentation. XML allows to define tags and document structures for own context-specific use. It was derived from SGML (Standard Generalized Markup Language), the international standard for defining descriptions of the structure and content of different types of electronic documents [10]. XML is simpler than SGML, but it allows the use of richly structured documents over the Internet. Information encoded in XML is easy to read and understand, and easy to process by computers. In XML files, structured data are bounded by tags and attributes. XML tags, attributes and element structure provide context information that facilitates the interpretation of the meaning of content, thereby making it feasible to develop efficient search engines and agents and perform intelligent data mining, etc. The XML allows the separation of content, logic and presentation.

Beyond XML there are a number of additional specifications such as Document Object Model (DOM) [11], XML Schemas [12], XSL Transformations [13], and Resource Description Framework (RDF) [14]. XML will have a big role in integration and interoperation of biological databases. Some biomedical information models have been implemented using XML specifications [15,16], many of them being clinical models for electronic healthcare documents [17-19].

A unified data format of resources is required for comparison between similar diseases and reutilization of information. Here we present a new data model called fact file, which integrates biomedical information related to hereditary diseases into a Web and WAP accessible knowledge base. Our scope is wider than e.g. in gene oriented knowledge bases such as GeneCards [20], UniGene [21], or LocusLink [22]. The disease information sources are even more diverse than those for genetic information. The fact files concentrate on sharing and integrating biomedical knowledge from different sources. The presented data model can be applied to any hereditary disease.

The fact files were applied to build a comprehensive, validated knowledge base for primary immunodeficiencies (PIDs) called ImmunoDeficiency Resource (IDR) [23,24]. It is designed for different user groups such as researchers, physicians and nurses as well as patients and their families and the general public. The IDR is the major information source to immunodeficiencies in the Web. Fact files serve as the core of the IDR knowledge base.

Methods

The fact file data model and the Inherited Disease Markup Language (IDML) were developed to facilitate disease information integration, storage and exchange in the first place for immunodeficiencies, but in principle for any hereditary disease. The IDML is an XML specification and container for bioinformatical data on hereditary diseases. The fact file data model schema was defined according to W3C XML specification [12, see Additional files 1, 2, 3]. The fact file data model can be depicted as a tree structure graph where a element is a root (Figure 1). Fact files make use of the following specifications, standards and databases: HUGO nomenclature [25], RefSeq [22], Swiss-Prot [26] and SOURCE [27].

Figure 1

The tree diagram of main IDML elements of fact file data model. The operators "?" (optional), "*" (zero or more) and "+" (one or more) are used to denote cardinality indicating how many instances of an element type are permitted.

Stand-alone IDML fact files have been generated for each PID. The fact files are uniquely identified by an id attribute of FactFile root element. The major concepts in the first tier of the fact file hierarchy below the root level are general information, clinical information and molecular biology (Table 1). In addition, there are other resources, which contain links to related information providers. Each of these elements, in turn, comprises one or more additional levels of guideline constructs.

Table 1

The description of high-level concepts in the fact file document model

Element	Description	Content	Content modela
FactFile	The root element for IDML-based fact file document	Elements	(GeneralInformation, ClinicalInformation, MolecularBiology, Other)
GeneralInformation	Describes the disease in general terms	Elements	(DiseaseName, Abbreviation*, AlternativeNames?, Description, Classification?, Omim*, CrossReferences?, Incidence?)
ClinicalInformation	The short overview of characteristic clinical features	Elements	(ClinicalDescription?, Diagnosis?, TherapeuticOptions?, ResearchPrograms?)
MolecularBiology	Molecular genetic elements	Elements	(GeneInformation?, AnimalModels?, ProteinInformation?, ExpressionPattern?)
Other	Other related information	Elements	(Publications?, Societies?, OtherSites?)

a Content model consists of a set of parenthesis enclosing some combination of child element names and operators. The order operators "," (strict sequence) and "|" (choice) indicate how elements may be combined. The operators "?" (optional), "*" (zero or more) and "+" (one or more) are used to denote cardinality indicating how many instances of an element type are permitted?

The components of the fact file model are defined as IDML elements. According to XML, elements have distinct names and they are delimited with start and end tag, e.g. X-linked agammaglobulinemia. Elements may contain other elements or attributes, they may store text, or they may be empty. Elements may appear as often as required. Many IDML elements contain href attribute for hyperlinking to more detailed information by using globally unique idenfier URL (Unified Resource Locator). The element naming convention follows the approach used by Electronic Business XML (ebXML) core components [28]. The IDML specified element names are in upper camel case (UpperCamelCase) and attribute names are in lower camel case (lowerCamelCase) notations. The usage of acronyms has been avoided, but when they are used the capitalization remains (example: ReferenceDNA).

General information elements identify a particular genetic disease and describe pattern of heritance and frequency in general terms (Table 2). The element includes commonly used abbreviations and the element lists known aliases and synonymous names for the disease. The element provides a short overview in general terms. The description text may contain several Glink -tagged words that can act as links to a glossary, which is an integral part of the IDR service. The element is used to classify a disease explicitly to a group of related diseases. It exploits the hierarchic structure of XML documents by nesting elements. Each element contains a unique identifier in level attribute and a class name in