Predicting liposome formulations by the integrated machine learning and molecular modeling approaches

Area of Science:

• Pharmaceutical Sciences
• Biotechnology
• Computational Chemistry

Background:

• Liposomes are widely used drug carriers due to biocompatibility and biodegradability.
• Current liposome formulation screening relies on inefficient trial-and-error methods.
• Machine learning (ML) offers a promising approach to optimize liposome formulation design.

Purpose of the Study:

• To develop ML models for predicting key liposome formulation parameters.
• To rank formulation features for improved liposome design guidance.
• To validate ML model predictions with experimental data for drug delivery systems.

Main Methods:

• Development of individual ML models for predicting liposome size, PDI, zeta potential, and encapsulation.
• Ranking of formulation features to identify critical parameters for liposome design.
• Experimental preparation and validation of liposome formulations for naproxen (NAP) and palmatine HCl (PAL).
• Coarse-grained molecular dynamics simulations to investigate drug-liposome interactions.

Main Results:

• ML models accurately predicted liposome parameters, validating their efficacy.
• Drug properties like logS, molecular complexity, and XLogP3 were identified as crucial for high encapsulation.
• Naproxen distributed within the lipid layer, while palmatine HCl aggregated in the aqueous core, highlighting the impact of drug properties.

Conclusions:

• ML models provide an efficient and rational approach for predicting liposome formulations.
• Understanding drug properties is critical for successful liposome-based drug delivery system design.
• These intelligent prediction systems can significantly advance future liposome formulation development.