Engineering and supercritical systems for improving the solubility and nanoparticles by development of computational...

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Pharmaceutical Sciences

Background:

• Accurate pharmaceutical solubility estimation is vital for supercritical processing.
• Supercritical carbon dioxide is a key solvent for enhancing drug solubility.
• Developing robust predictive models for solubility is an ongoing challenge.

Purpose of the Study:

• To accurately estimate drug solubility in supercritical carbon dioxide.
• To develop and compare artificial intelligence models for solubility prediction.
• To identify key operational parameters influencing drug solubility.

Main Methods:

• Developed and optimized Artificial Intelligence Models (AIMs): XGBoost, Gradient Boosting, and Random Forest.
• Utilized a comprehensive dataset of 1619 data points for 58 drugs under varying temperature and pressure.
• Employed logarithmic transformation of solubility values for enhanced model robustness.

Main Results:

• XGBoost and Gradient Boosting models demonstrated superior performance in predicting drug solubility.
• Optimized XGBoost model achieved high accuracy with R² of 0.998 (training) and 0.9864 (testing).
• Sensitivity analysis confirmed pressure and temperature as the most influential variables for solubility prediction.

Conclusions:

• Artificial intelligence models, particularly XGBoost and Gradient Boosting, are effective for estimating drug solubility in supercritical CO₂.
• The developed models provide reliable predictions crucial for optimizing supercritical fluid applications in pharmaceuticals.
• Understanding the impact of pressure and temperature is key to controlling and predicting drug solubility.