Neural network-based expert system was created for improvement of pharmaceutical formulation development process. The knowledge database for this system was created from original, experimental data as well as from literature data. Different models for soluble and poorly soluble drugs were trained and tested. Best results were found when specific, pharmaceutical criterions were used to split base dataset in order to create more specialized neural models predicting behavior of some, well defined class of excipients. ANN function as a PharmCAD system (Pharmaceutical Computer Aided Design) was confirmed.
展开▼