Prediction of Dissolution Data Integrated in Tablet Database Using Four-Layered Artificial Neural Networks.

Area of Science:

• Pharmaceutical Sciences
• Computational Chemistry
• Materials Science

Background:

• Tablet formulation design relies on understanding drug dissolution.
• Existing databases lack comprehensive dissolution data for active pharmaceutical ingredients (APIs).
• Predictive models are needed to enhance tablet formulation development.

Purpose of the Study:

• To integrate measured dissolution data into an existing tablet database.
• To develop and evaluate a four-layered artificial neural network (4LNN) for predicting drug dissolution.
• To compare the predictive performance of 4LNN with conventional models.

Main Methods:

• Collected and integrated extensive dissolution data for 14 model APIs.
• Utilized a four-layered artificial neural network (4LNN) implemented in commercial software.
• Employed physicochemical and powder properties of APIs as input features for the 4LNN model.
• Compared 4LNN performance against a three-layered neural network and linear regression models.

Main Results:

• An excellent predictive model for drug dissolution was achieved using the 4LNN method.
• The 4LNN model demonstrated superior performance compared to a three-layered neural network.
• Linear regression models yielded poor predictions for drug dissolution data.

Conclusions:

• The four-layered artificial neural network (4LNN) is highly effective for predicting drug dissolution from API properties.
• 4LNN enhances the value of tablet databases for formulation design.
• Advanced neural network models offer significant advantages over traditional methods in pharmaceutical prediction.