Machine Learning-Based Prediction of Drug Solubility in Lipidic Environments: The Sol_ME Tool for Optimizing...

Area of Science:

• Pharmaceutical Sciences
• Computational Chemistry
• Drug Delivery

Background:

• Lipid-based formulations enhance drug solubility and bioavailability.
• Selecting optimal lipid excipients is a significant challenge in drug development.
• Existing methods often rely on traditional parameters that may not fully capture complex solubility interactions.

Purpose of the Study:

• To introduce Sol_ME, a machine learning model for predicting drug solubility in lipidic environments.
• To streamline the selection of lipid excipients for improved drug formulation.
• To demonstrate the practical application of Sol_ME in optimizing the formulation of a specific drug.

Main Methods:

• Developed Sol_ME using PubChem® fingerprints to correlate drug structures with lipid excipient solubility.
• Trained the model on a dataset of 1,379 drug-solvent entries.
• Applied and validated the model using Apalutamide (BCS Class II drug) and 35 drug-solvent combinations.

Main Results:

• Sol_ME achieved a high predictive accuracy with a correlation coefficient of 0.998.
• Identified Cinnamon oil as the optimal excipient for Apalutamide, with Vanillin for further refinement.
• Reduced formulation volume by 75%, enabling a 240 mg single-unit soft gelatin capsule; experimental validation showed 80% alignment.

Conclusions:

• Sol_ME offers a data-driven approach to optimize lipid-based drug formulation, enhancing efficiency.
• The successful formulation of Apalutamide highlights the model's practical utility and potential for broader application.
• Future work includes expanding the dataset and extending the model to solid lipid systems for advanced drug delivery.