|
1F - QSAR (TUP/52) Automated building of metabolic simulators. Mekenyan, Ovanes1, Pavlov, Todor1, Dimitrov, Sabcho1, Veith, Gilman2, 1 Laboratory of Mathematical Chemistry, University “Prof. As. Zlatarov”, Bourgas, Bulgaria, Bulgaria2 U.S.EPA, Mid-Continent Ecology Division, 6201 Congdon Blvd, Duluth, MN 55804, USA ABSTRACT- A large amount of observed metabolic transformations and maps is stored in the regulatory agencies and proprietary databases. Each of these individual transformations and metabolic maps is associated with organ or tissue specific transformation environment. This data could be very helpful in identifying already observed molecular transformations, however, it is very limited for predicting metabolism of new chemicals not existing in documented databases. Recently, for that purpose we have introduced simulators of metabolic transformations based on library of transformations organized in hierarchy. A probabilistic scheme was used to determine this hierarchy and substructure-matching engine - to apply transformations. As a result of application of metabolic simulator metabolic maps are generated. The probabilistic approach to determine transformation hierarchy, however, could be applied if consistent data for rate disappearance and/or identified metabolic products is available. Otherwise, the process of hierarchical ordering of known transformations is a time-consuming process given the combinatorial difficulties due to large number of transformations that need to be ordered. In the present work, an engine for automated building of molecular simulators is proposed based on evolutionary algorithms, which optimizes the hierarchy of user-defined list of transformations to maximize the reproduction of documented metabolic pathways. The system has been used to build metabolic simulators for liver and skin metabolism. Key words: metabolic simulator, metabolism |
