Heuristic modeling of macromolecule release from PLGA microspheres

Area of Science:

• Biomaterials Science
• Pharmaceutical Sciences
• Computational Modeling

Background:

• Protein dissolution from poly(lactic-co-glycolic acid) (PLGA) particles is complex and poorly understood.
• This lack of understanding hinders the development of predictive models for PLGA-based drug delivery systems, impacting medical applications and toxicity assessments.

Purpose of the Study:

• To develop and compare heuristic models for predicting macromolecule dissolution profiles from PLGA micro- and nanoparticles.
• To identify key variables influencing protein release and establish predictive models with high accuracy.

Main Methods:

• Employed heuristic techniques including artificial neural networks (ANNs), feature selection (fscaret), and genetic programming.
• Reduced 300 input variables to a concise set (11-21 inputs) using feature selection and sensitivity analysis.
• Developed monotone multi-layer perceptron (MON-MLP) neural networks and derived a classical equation using genetic programming.

Main Results:

• The best ANN model (MON-MLP) achieved a root-mean-square error (RMSE) of 15.4 with eleven inputs.
• A classical equation derived from 17 inputs yielded a better generalization error (RMSE) of 14.3 and is feasible for standard nonlinear regression.
• Both heuristic modeling approaches demonstrated good predictive efficiency for macromolecule release from PLGA microspheres.

Conclusions:

• Heuristic modeling, particularly ANNs and genetic programming, provides effective tools for predicting protein dissolution from PLGA particles.
• The derived classical equation offers a practical and predictable model for understanding and optimizing PLGA-based multiparticulate dosage forms.