Artificial Neural Networks as Modeling Tools in the Identification of Drugs Release Mechanisms from Hydrodynamically-Balanced Systems Formulated with Various Polymers

The aim of this study was to use artificial neural networks (ANNs) in the attempt to the identification of model drug release mechanisms from hydrodynamically-balanced systems (HBS). As model drug used was ketoprofen (KT) nonsteroid anti-inflammatory drug. The matrices of HBS were formulated with use of hydrophilic polymers, derivatives of cellulose, carrageens and alginates. Cheminformatics software was applied in order to provide numerical representation of HBS qualitative composition. Artificial neural networks were employed as predictive and data-mining tools. Large initial number of input variables (2659) was reduced using subsequent sensitivity analyses to the 17 inputs. A knowledge-based "context-related" sensitivity analysis was applied in order to cope with cut-off point selection procedure. Additional application of inputs correlation analysis resulted in the final reduction to the 8-input vector. Observations of what was remained as the input variables in the above mentioned 8-input vector was the base of theories formulation concerning KT release mechanism.