144 PROCEEDINGS of the20-th International Conference SAER-2006 SYSTEM IDENTIFICATION OF BIOLOGICAL PROCESSES Kaloyan Yankov Medical Faculty, Thracian University, Armejska str., 11, Stara Zagora 6000, Bulgaria e-mail:
[email protected] Abstract: This paper discusses some aspects of system identification of biological processes. An iterative optimization method for identification is proposed. The created software aids in appreciation and analysis of effects of new drugs. The major merit is the graphical user interface. It would make the software easy to work for users without specific expertise in system identification despite the sophistication of the theory and algorithms Key words: System identification, Systems biology, Cyclic coordinate descent. 1. INTRODUCTION The development of new drugs requires assessing their safety and efficacy before they can be marketed. The process of assessing consists of two phases: preclinical and clinical. In preclinical phase the experiments are carried out on animals in vivo or on isolated tissues and organs in vitro. The obtained results from the preclinical research phase usually are processed by the methods of statistics or by numeric analysis of their dynamic characteristics [1,2]. The received data are insufficient and it is incorrect to translate the results and conclusions upon the humans. A resources and methods are necessary in order to be able to predict the effects on humans and then to pass to the clinical research phase. To obtain reliable experimental data the use lots of animals is necessary and that is in discrepancy with modern tendencies in biological science. In addition this makes the experiment more expensive and complicates the research conditions. An alternative approach, which proves its effectiveness for system evaluation and determination its quality at reasonable price, is the use of simulation models [3]. The goal of modeling is to create a model of phenomenon using at least minimal number of variables and parameters and correctly to reproduce the main characteristics of phenomenon. Models map the relationships between the variables in the system to be modeled onto mathematical structures like simple algebraic equations, differential equations or even systems of differential equations. In developing a mathematical model two fundamental approaches are possible. The first is based on fundamental understanding of the modeled processes. The other is based on experimental data and is essentially a data-driven approach (black-box model). Experimental modeling is known in the literature as system identification (SI). Hence, nowadays there are many simulators and simulation languages aiding the process of system identification. Аll-purpose or general simulators. Мodels are expressed using differential equations or transfer functions, so they can be used for any kind of system: chemical, boilogical, mechanical, electrical, etc. Examples are Matlab-Simulink and ACSL [4]. The main obstacle in its use is that they require a good knowledge about mathematics and system theory. Specialised simulators.