Related Experiment Video
Updated: Aug 9, 2026
09:22
Fabricating Superhydrophobic Polymeric Materials for Biomedical Applications
Published on: August 28, 2015
Optimization of Electrospun PCL-PLGA Substrates for Lung Basement Membrane Applications Using Response Surface
Golestan Salimbeigi1, Halima Boutouil1, Tanya Levingstone1
1School of Mechanical & Manufacturing Engineering, Dublin City University, Collins Avenue Extension, Glasnevin, D09 V209, Dublin 9, Ireland.
ACS Omega
|August 8, 2026
Summary
This study optimized poly-(ε-caprolactone) (PCL)-poly-(lactic-co-glycolic acid) (PLGA) electrospun membranes for lung barrier models. Machine learning enhanced the design process, balancing mechanical properties for improved tissue engineering applications.
Area of Science:
- Biomaterials Science
- Tissue Engineering
- Polymer Science
Background:
- Electrospun membranes mimic native tissue extracellular matrix for tissue modeling.
- Developing functional in vitro lung barrier models requires membranes with stability, reproducibility, fine fibers, reduced stiffness, and mechanical strength.
- Poly-(ε-caprolactone) (PCL) and poly-(lactic-co-glycolic acid) (PLGA) blends offer tunable properties for synthetic membrane development.
Purpose of the Study:
- To optimize PCL-PLGA electrospun membranes for in vitro lung barrier models.
- To balance competing structural and mechanical requirements for membrane substrates.
- To integrate Response Surface Methodology (RSM) and Machine Learning (ML) for enhanced membrane design.
Main Methods:
- Utilized Response Surface Methodology (RSM) with a Box-Behnken design to optimize electrospinning parameters.
- Investigated fiber diameter, Young's modulus, ultimate tensile strength (UTS), and strain at break as key responses.
- Applied Machine Learning (ML) techniques, including Gaussian-Copula augmentation, CatBoost, and XGBoost, for inverse prediction and optimization.
Main Results:
- RSM predicted optimized membrane properties: 0.386 μm fiber diameter, 152.156 MPa Young's modulus, 41.850 MPa UTS, and 2.948 strain at break, with <10% deviation in confirmatory experiments.
- ML models, particularly CatBoost, accurately predicted polymer composition (LOO R² of 0.675 for PCL, 0.747 for PLGA).
- XGBoost optimization achieved a desirability of 0.895, surpassing the RSM reference (0.782), indicating superior multiresponse optimization.
Conclusions:
- The developed PCL-PLGA membranes meet essential criteria for lung barrier models.
- The combined RSM and ML workflow effectively optimizes electrospun membranes, balancing complex properties.
- This integrated approach enhances the design of fibrous membranes by prioritizing electrospinning parameters and quantifying uncertainty.
