AN ENVIRONMENT FOR KNOWLEDGE DISCOVERY IN BIOINFORMATICS APPLICATIONS

AN ENVIRONMENT FOR KNOWLEDGE DISCOVERY IN BIOINFORMATICS APPLICATIONS Junior Barrera, Roberto M Cesar-Jr, João E. Ferreira, Marco D. Gubitoso USP Center for Bioinformatics (BIOINFO-USP) University of São Paulo Rua do Matão, 1010, São Paulo, SP, 05508-900 Brazil Phone: +55 11 3818 6135 http://www.bioinfo.usp.br Abstract The history of biology was strongly changed by two facts: (i) the discovery of basic principles (i.e., genetics and molecular biology phenomena) for understanding all known kinds of life; (ii) the capacity of measuring these phenomena and, consequently, of creating quantitative models. These facts have also a strong impact on biotechnology. New technologies can approach complex problems not manageable before such as early and sharp diagnosis of complex diseases like cancer or refined drug discovery. Furthermore, this unifying approach allows adapting the procedures to quite different contexts such as oncology, virology or agronomy. However, advances in genetics and molecular biology do not make knowledge discovery (i.e., confirmation or rejection of scientists hypotheses) in biology an easy task. Biological phenomena are usually so complex and diverse that require the integration of several sources of knowledge, besides the genetic and molecular information. Typically, studies of biological systems may involve understanding several levels of organization: molecules, genes, cells, tissue, organ, system, organism, population, and the interaction with other populations and the environment. The knowledge available may be qualitative or quantitative and may come from different sources, from structured databases to texts of the literature. Nowadays, researchers that try to understand and to solve these complex problems face a motley collection of tools written in different databases system, running on different operating systems, using various file formats, and based on incompatible data models. This heterogeneity severely affects carrying out productive research. Part of the heterogeneity comes from diverse technological options of measuring the same kind of biological phenomena. However, most of the heterogeneity are due to lack of standardization. In fact, it is common to find databases that store the same kind of information in different formats or software that perform the same operations with completely different data models . Special information systems are required to increase research efficacy. These systems have huge databases and count with sophisticated transformation and mining tools, implemented in high performance hardware. This paper presents a Knowledge Discovering Environment for Biology [1] with such characteristics. The system has a generic kernel that implements the mining functions to be applied to input primary databases, with a warehouse architecture, of biomedical information. Both super-vised and unsupervised classification can be implemented within the kernel and applied to data extracted from the primary database, with the results being suitably stored in a complex object database for knowledge discovery. The kernel also includes a specific high-performance library that allows designing and applying the mining functions in parallel machines. Reference [1] J. Barrera, R.M. Cesar-Jr., J. E. Ferreira, M. D. Gubitoso, An environment for knowledge Discovery in biology, accepted for Computers in Biology and Medicine.