Modelling formulations using gene expression programming--a comparative analysis with artificial neural networks.

Area of Science:

• Computational intelligence
• Pharmaceutical sciences
• Data modeling

Background:

• Pharmaceutical formulation development relies on understanding complex data relationships.
• Neural networks are widely used but often act as black boxes.
• Gene expression programming (GEP) is an evolutionary computing technique for automatic equation generation.

Purpose of the Study:

• To investigate the utility and advantages of GEP for pharmaceutical formulation data modeling.
• To compare GEP-generated models with those from neural networks.
• To assess the interpretability of models generated by both techniques.

Main Methods:

• Applied GEP to four types of pharmaceutical formulation data.
• Utilized neural networks for comparison in formulation development.
• Evaluated models based on their ability to discover non-linear relationships and interpretability.

Main Results:

• Both GEP and neural networks identified subtle, non-linear data relationships without predefined functional forms.
• Neural networks developed models faster with higher ANOVA R(2) values but lacked interpretability.
• GEP, though slower, generated simpler, interpretable equations describing key relationships.

Conclusions:

• GEP is an effective and efficient modeling technique for pharmaceutical formulation data.
• GEP provides greater insight into cause-and-effect relationships compared to neural networks.
• GEP's interpretable models are valuable for formulation development and understanding.