Brusic, Zeleznikow

Knowledge Discovery and Data Mining.

Biological databases continue to grow rapidly. This growth is reflected in increases in both size and complexity of individual databases as well as in the proliferation of new databases. A huge body of data is thus available for the extraction of high level information including the development of new concepts, concept interrelationships and interesting patterns hidden in the databases. Knowledge Discovery in Databases (KDD) is an emerging field combining techniques from Databases, Statistics and Artificial Intelligence, which is concerned with the theoretical and practical issues of extracting high level information (or knowledge) from volumes of low level data. At the core of KDD is Data Mining - the application of specific tools for pattern discovery and extraction. KDD process comprises several data pre-processing steps as well as data mining and knowledge interpretation steps.

Studies of biological data involve access to multiple databases using a variety of query tools. The amounts of biological data are growing faster than the capability to analyse them. KDD offers the capacity to automate complex search and data analysis tasks. We can distinguish two types of goals of KDD systems: verification and discovery. With verification, the system is limited to verifying the users hypothesis. With discovery, the system autonomously finds new patterns. Discovery can be subdivided into prediction and description (explanation) goals. Biological sources are highly heterogeneous, geographically dispersed, constantly evolving, and often high in volume. They represent data from a highly complex domain. Numerous tools suitable for data mining in biology are available, yet the selection of an appropriate tool is a non-trivial task. The KDD process provides for the selection of the appropriate data mining methods by taking into account both domain characteristics and general KDD process requirements.

This tutorial consists of four parts: a) introduction to KDD, b) discussion of the domain concepts from biological data and databases, c) data mining techniques in biology, and d) a case study: application of KDD in immunology.

The goal of the tutorial is to present KDD methodology as an approach which is complementary to laboratory experiments and which can accelerate the process of discovery in biology. This is achieved by both minimisation of the number of necessary experiments and by improved capacity to interpret biological data. The tutorial examples include those demonstrating successful applications of KDD to experiment planning, prediction of biological function and description (postulating hypotheses). The tutorial also contains pointers to the relevant literature.

The target audience includes biologists, medical researchers and computer scientists intersted in biological discovery.

	   Biographical Information

	   Vladimir Brusic, The Walter and Eliza Hall Institute of Medical Research,
	   Melbourne, Australia. (vladimir@wehi.edu.au)

	   Vladimir Brusic is a Bioinformatician at the Walter and Eliza Hall Institute.
	   He received the Masters degree in Biomedical Engineering from the University of
	   Belgrade in 1988 and the degree of Master of Applied Science in Information
           Technology from the Royal Melbourne Institute of Technology in 1997. His
           research interests include computer modelling of biological systems,
           computational immunology and complex systems analysis. He is a creator of the
           MHCPEP, an immunological database. He has developed several data mining tools
           in immunology which have been successfully applied to the prediction and
           determination of vaccine targets in autoimmunity, cancer and malaria research.

           John Zeleznikow, School of Computer Science and Computer Engineering, La Trobe
           University, Melbourne, Australia. (johnz@latcs1.cs.latrobe.edu.au)

           John Zeleznikow is a Senior Lecturer at La Trobe University where he teaches
           Database and Information Systems subjects. He received the PhD degree in
           Mathematics from Monash University, Melbourne in 1980. His research interests
           include building Intelligent Information Systems and Knowledge Representation
           and Reasoning, among others. He has developed several KDD applications in the
           legal domain and more recently expanded his work to the biological domain. He
           has co-authored with Dan Hunter a book on building intelligent legal systems.
           He was a general chair of the Sixth International Conference on Artificial
           Intelligence and Law. He has recently given tutorials at the 7th and 9th
           international Legal Knowledge Based Systems Conferences, the IFIP World
           Computer Congress and the 14th and 15th British Expert Systems conferences.