Intelligence computational analysis of letrozole solubility in supercritical solvent via machine learning models

Area of Science:

• Pharmaceutical Science
• Chemical Engineering
• Computational Chemistry

Background:

• Supercritical fluids (SCFs), particularly supercritical carbon dioxide (SCCO2), offer an environmentally friendly alternative to organic solvents for nanoparticle preparation.
• SCFs possess unique properties combining gas-like transport with liquid-like solvation, making them advantageous for improving drug solubility.
• Accurate models are crucial for estimating the solubility of chemotherapeutic agents in SCFs to advance pharmaceutical applications.

Purpose of the Study:

• To employ machine learning (ML) approaches for estimating the solubility of the chemotherapeutic drug Letrozole (LET) in SCCO2.
• To correlate LET solubility with varying temperature and pressure conditions using predictive models.
• To identify the most suitable ML model for accurate solubility prediction in this context.

Main Methods:

• Utilized three machine learning models: Passive Aggressive Regression (PAR), Random Forest (RF), and Support Vector Machine with RBF kernel (RBF-SVM).
• Optimized model hyperparameters using a Genetic Algorithm (GA).
• Evaluated model performance using metrics such as coefficient of determination (R-squared) and Mean Squared Error (MSE).

Main Results:

• Optimized PAR, RF, and RBF-SVM models achieved R-squared values of 0.8277, 0.9534, and 0.9947, respectively.
• Corresponding MSE error rates were 0.1342 for PAR, 0.0305 for RF, and 0.0045 for RBF-SVM.
• The optimized RBF-SVM model demonstrated the highest accuracy, with a maximum prediction error of 0.1289.

Conclusions:

• The optimized RBF-SVM model is highly effective for predicting Letrozole solubility in supercritical carbon dioxide across a range of temperatures and pressures.
• Machine learning offers a powerful tool for estimating drug solubility in SCFs, facilitating the development of advanced drug delivery systems.
• This study highlights the potential of SCCO2 and ML for greener and more efficient pharmaceutical manufacturing processes.