Machine Learning-Assisted “Shrink-Restricted” SERS Strategy for Classification of Environmental Nanoplastic-Induced Cell Death

Abstract

The biotoxicity of nanoplastics (NPs), especially from environmental sources, and “NPs carrier effect” are in the early stages of research. This study presents a machine learning-assisted “shrink-restricted” SERS strategy (SRSS) to monitor molecular changes in the cellular secretome exposure to six types of NPs. Utilizing three-dimensional (3D) Ag@hydrogel-based SRSS, active targeting of molecules within adjustable nanogaps was achieved to track information. Machine learning was employed to analyze the overall spectral profiles, biochemical signatures, and time-dependent changes. Results indicate that environmentally derived NPs exhibited higher toxicity to BEAS-2B and L02 cells. Notably, the “NPs carrier effect,” resulting from pollutant adsorption, proved to be more harmful. This effect altered the death pathway of BEAS-2B cells from a combination of apoptosis and ferroptosis to primarily ferroptosis. Furthermore, L02 cells demonstrated greater metabolic vulnerability to NPs exposure than that of BEAS-2B cells, especially concerning the “NPs carrier effect.” Traditional detection methods for cell death often rely on end point assays, which limit temporal resolution and focus on single or multiple markers. In contrast, our study pioneers a machine learning-assisted SERS approach for monitoring overall metabolic levels post-NPs exposure at both cellular and molecular levels. This endeavor has significantly advanced our understanding of the risks associated with plastic pollution.