Machine Learning-Assisted SERS Reveals the Biochemical Signature of Enhanced Protein Secretion from Surface-Modified Magnetic Nanoparticles.

This study introduces a novel investigation of the interaction between cells and iron oxide-based magnetic nanoparticles (FeO MNPs) via protein secretion and machine learning (ML)-assisted surface-enhanced Raman scattering (SERS). For the first time, we produced FeO, FeO@PEG, FeO@PEI, and FeO@PEI MNPs by a one-pot coprecipitation reaction. The addition of polymers to the reaction conditions significantly affected the shape, surface charge, size, and size distribution of the MNPs. The surface modification of MNPs is effectively accomplished using polyethylenimine (PEI), and the ζ-potential values of the MNPs exceed +25 mV under the NHOH control. The homogeneity of MNPs synthesized with NHOH is more pronounced according to transmission electron microscopy (TEM) pictures. All MNPs exhibited excellent immobilization efficiency (>92%) when we used 250 ppm Fe-containing MNP solutions. Smaller MNPs uniformly encapsulated the surface of cells, whereas larger MNPs exhibited irregular accumulation. cells exhibited excellent viability in all MNP solutions at up to 1000 ppm of Fe concentrations. Finally, the highest recombinant azurin protein secretion rate was obtained in FeO@PEI MNP-immobilized cells (about 1.3 times). The ML-assisted SERS analysis revealed that MNP interactions with cells were mediated by proteins such as mannoproteins and membrane transporter proteins as well as N-acetylglucosamine (i.e., chitin). These findings revealed the effect of the size and surface properties of MNPs on the immobilization of cells and the enormous potential of magnetic immobilization for protein secretion.