Ontology-Driven Information Extraction and Knowledge Acquisition from Heterogeneous, Distributed, Autonomous Biological Data Sources
Vasant Honavar, Carson M. Andorf, Doina Caragea, Adrian Silvescu · 2002
Scientific discovery in data rich domains (e.g., biological sciences, atmospheric sciences) presents several challenges in information extraction and knowledge acquisition from heterogeneous, distributed, autonomously operated, dynamic data sources. This paper describes these problems and outlines the key elements of algorithmic and systems solutions for computer assisted scientific discovery in such domains. These include: ontology-assisted approaches to customizable data integration and information extraction from heterogeneous, distributed data sources; distributed data mining algorithms for knowledge acquisition from large, distributed data sets which obviate the need for transmitting large volumes of data across the network; ontology-driven approaches to exploratory data analysis from alternative ontological perspectives; and modular and extensible agent-based implementations of the algorithms within a platform-independent agent infrastructure. Prototype implementations of the proposed system are being used for discovery of macromolecular structure-function relationships in computational biology and distributed coordinated intrusion detection in computer networks. Challenges in Integration and Analysis of Heterogeneous Distributed Data Development of high throughput data acquisition technologies in biological sciences, together with advances in digital storage, computing, and communications technologies have resulted in unprecedented opportunities for large scale, computer assisted, data-driven scientific discovery [Baxevanis et al., 1999]. Data sets of interest to computational biologists are often heterogeneous in structure, content, and semantics. Examples include sequence data (DNA, RNA, and protein sequences, expressed sequence tags) [Benson et al., 1997; Boguski et al., 1997]; numeric measurements (e.g., gene expression data); symbolic data describing relations among entities; structured or semi-structured text (e.g., annotations associated with DNA sequences, protein structures, and gene expression data); temporal data (e.g., gene expression time series); structures containing numeric as well as symbolic information (e.g., 3-dimensional protein structures); and results of various types of analysis [Baxevanis, 2000; Discala et al., 2000]. They currently include data stored in flat files, relational databases, and object-oriented databases. The term biological database is used loosely to refer to a biological data collection in any of these forms. How best to organize genome data is still a matter of debate [Frenkel, 1991; Gelbart, 1998] although several objectoriented databases and have been proposed in recent years [Gray, 1990; Goodman, 1995; Ghosh, 1999; Durbin, 1991]. Applications such as characterization of macromolecular structure function relationships and inference of genetic regulatory pathways require selection and extraction of relevant information from such data (e.g., features from sequences, counts and statistical summaries from measurements, structured representation of relevant information from textual annotations). They also call for data integration from multiple sources into a coherent form that lends itself to further analysis (e.g., data mining) by bridging syntactic and semantic gaps among them. Typical data analysis tasks that arise in computational biology are difficult to express using standard query languages and thus application programs have to be constructed using program libraries. While queries expressed in declarative languages like SQL are still useful in biological databases, the use of programming interfaces is unavoidable for many types of data analysis (e.g., data mining). This follows from the fact that the same set of data may have to be analyzed in different ways depending on the information extraction and knowledge acquisition objectives of the user. It is impossible to foresee all the potential uses of data when designing data repositories or data analysis services. The data sources of interest in computational molecular biology are large, diverse in structure and content, and typically autonomously maintained [Fasman, 1994]. Transforming these data into useful knowledge (e.g., inference of genetic networks from gene expression data, building predictive models of protein function from protein sequence) calls for algorithmic and systems solutions for computer assisted knowledge acquisition and data and knowledge visualization. Machine learning algorithms [Mitchell, 1997] currently offer one of the most cost effective approaches to data-driven knowledge acquisition (discovery of features, correlations, and other complex relationships and hypotheses that describe potentially interesting regularities from large data sets) in increasingly data rich domains such as computational biology [Baldi and Brunak, 1998]. However, application of machine learning algorithms to large scale knowledge discovery from