Development of fucoidan/polyethyleneimine based sorafenib-loaded self-assembled nanoparticles with machine learning and DoE-ANN implementation: Optimization, characterization, and in-vitro assessment for the anticancer drug delivery.

This study aims to develop sorafenib-loaded self-assembled nanoparticles (SFB-SANPs) using the combined approach of artificial neural network and design of experiments (ANN-DoE) and to compare it with other machine learning (ML) models. The central composite design (CCD) and ML algorithms were used to screen the effects of concentrations of both the polymers (polyethyleneimine and fucoidan) on the outcome responses, i.e., particle size and entrapment efficiency with defined constraints. The prediction from different ML models (bootstrap forest, K-nearest neighbors, artificial neural network, generalized regression-lasso and support vector machines) were compared with ANN-DoE model. The ANN-DoE model showed better accuracy and predictability and outperformed all the other models. This depicted that the concept of using ANN and DoE combination approach provided the best, uncomplicated and cost-effective way to optimized the nanoformulations. The optimized formulation generated from the ANN-DoE combined model was further evaluated for characterization and anticancer activity. The optimized SFB-SANPs were prepared using the polyelectrolyte complexation method with Polyethyleneimine (PEI) as a cationic polymer and fucoidan (FCD) as an anionic. The SFB-SANPs were nanometric in size (280.4 ± 0.089 nm) and slightly anionic in nature (zeta potential = -6.03 ± 0.92 mV) with an encapsulation efficiency of 95.56 ± 0.30 %. The drug release from SFB-SANPs was controlled and sustained in the cancer microenvironment (pH 5.0). The SFB-SANPs were compatible with red blood cells (RBCs), and the % hemolysis was found to be <5.0 %. The anticancer activity of the SFB-SANPs exhibited an IC50 at 2.017 ± 0.516 μM against MDMB-231 cells, showing a significantly high inhibitory effect on breast cancer cell lines. Therefore, the nanocarriers developed using various ML tools inherit a huge promise in anticancer drug delivery.