Fuente: Journal of applied polymer Lugar: RESEARCH ARTICLE Curcumin encapsulation in nanomicelles coupled with a machine learning–guided model for predicting its release profile.

ABSTRACT
This work, leveraging machine-learning algorithms, focuses on the design and in vitro evaluation of oil-in-water surfactant-based nanomicelles as a novel curcumin formulation with comparable cytotoxic effects against MCF-7 human breast cancer cells. Nanomicelles were prepared using Pluronic F-127 as an emulsifying agent and the addition of an ethyl butyrate solution as the organic phase, reaching highly efficient encapsulations of 89.4%, with a particle hydrodynamic size of 534.425 nm. The in vitro drug release studies showed pH-dependent sustained release, with a faster release at acidic pH 5.4 compared to neutral pH 7.4. The cytotoxicity studies showed that curcumin-loaded nanomicelles preserved the anticancer activity of free curcumin against MCF-7 cells while exhibiting a comparable cytotoxicity profile toward normal fibroblast cells, indicating that nanoencapsulation did not compromise efficacy or introduce additional toxicity. Furthermore, this study introduces an ML-based fitting and interpolation framework for modeling pH-responsive curcumin release kinetics from nanocarrier systems. The Random Forest regression model demonstrated outstanding goodness-of-fit (R
2 > 0.999), effectively capturing the nonlinear and complex kinetics of curcumin release across the measured time points. These computational insights provide a useful descriptive tool for the rational design of pH-responsive drug delivery systems with enhanced therapeutic precision.