Machine Learning-Guided Design of High-Performance Antifungal Polymers against Candida albicans

Area of Science:

• Polymer Chemistry
• Computational Chemistry
• Microbiology

Background:

• Developing effective antifungal agents is crucial to combat rising fungal infections.
• Synthetic polymers offer tunable properties for antimicrobial applications.
• Predictive modeling can accelerate the discovery of potent antifungal materials.

Purpose of the Study:

• To utilize machine learning (ML) to identify critical features governing antifungal activity in synthetic polymers.
• To design and synthesize novel antifungal polymers based on ML-identified characteristics.
• To optimize polymer properties for enhanced antifungal potency and biocompatibility.

Main Methods:

• Trained a Random Forest binary classification model on a curated dataset of antifungal polymers.
• Identified five key features predicting antifungal activity against Candida albicans.
• Synthesized polymer libraries using reversible addition-fragmentation chain transfer (RAFT) polymerization.
• Evaluated antifungal activity (MIC90) and biocompatibility (HC50).

Main Results:

• The ML model identified optimal ranges for hydrophilic, hydrophobic, partition coefficient (cLogP), degree of polymerization (DP), and cationic composition.
• A synthesized polymer meeting all five criteria achieved an MIC90 of 32 μg/mL.
• Further optimization using a different RAFT agent (DTPA) yielded a polymer with MIC90 of 16 μg/mL and HC50 > 2000 μg/mL.
• The optimized polymer demonstrated a selectivity index > 125, indicating high efficacy and low toxicity.

Conclusions:

• ML-guided design is an efficient strategy for developing advanced antifungal polymers.
• The identified polymer features enable the creation of potent and biocompatible antifungal materials.
• This approach accelerates the discovery pipeline for novel antimicrobial therapeutics.