RULE-BASED MODELING IN BIOUML N. Mandrik* 1,2,3 , E. Kutumova 1,2 , F. Kolpakov 1,2 1 BIOSOFT.RU, LLC, Novosibirsk, Russia 2 Design Technological Institute of Digital Techniques, SB RAS, Novosibirsk, Russia 3 Sobolev Institute of Mathematics, SB RAS, Novosibirsk, Russia e-mail: manikitos@biosoft.ru *Corresponding author Key words: rule-based modeling, BioUML, KaSim, BioNetGen Abstract Motivation and Aim: The traditional approach to mathematical modeling of biological systems involves usage of nonlinear systems of ordinary differential equations (ODEs) with given initial conditions. Talking about the modeling, we emphasize the fact that we consider an abstraction and study the mathematical description of some qualitative and quantitative characteristics of biological processes. The level of detail is dependent on the problem and based on the knowledge of the researcher. On the one hand, many meaningful models consist of few non- linear equations. On the other hand, a detailed study of the biochemical networks leads to development of large-scale models consisting of hundreds of variables and, therefore, equations. Moreover, if we incorporate to the model site-specific details of protein-protein interactions, the number of protein modifications increases dramatically, and complexity of the model becomes combinatorial. For example, a protein comprising n amino acids can be potentially found in 2 n distinct phosphorylation states. Investigation of such models using formalism of differential equations is difficult in view of the fact that we need to analyze thousands of variables whose values are often small. Visualization of the models (graphical representation of the reaction network as diagram) using one of the conventional standards (e.g., SBGN or KEGG) does not simplify the problem, although the diagram is easier to interpret than the corresponding system of equations, and readability of the diagram can be improved. Methods and Algorithms: The main idea to deal with such models is based on representations of protein-protein interactions using rules serving as generators of species and biochemical reactions (or discrete events). This approach is known as «rule-based» modeling. Each rule describes a class of reactions with a common kinetic law and establishes the correspondence between reactant and product patterns defining a set of species with similar chemical compositions and properties. Conclusion: The principles for creation of the «rule-based» models were implemented in several software resources including KaSim (http://dev.executableknowledge.org/) and