Bioinspired multi-colored colloidal array for simultaneous visual sensing of bisphenol B and levofloxacin assisted by machine learning.

The integration of multiple optical signals with independent responses within a single sensing architecture is of interest for multiplex small-molecule detection. An anole-inspired multi-colored colloidal array (MCA) probe integrating structural reflection (SR), fluorescein isothiocyanate (FITC), and europium acetylacetonate (EU) emissions was developed for simultaneous visual detection of bisphenol B (BPB) and levofloxacin (LVX). The probe consists of a highly ordered non-close-packed (NCP) colloidal array assembled from FITC- and EU-doped molecularly imprinted particles, enabling the coexistence of structural color and dual-emission luminescence within a single architecture. The periodic array generates an additional SR signal via Bragg diffraction, while the enlarged interparticle spacing suppresses aggregation-induced quenching and non-radiative losses, resulting in enhanced FITC and EU emissions. Upon exposure to BPB and LVX, orthogonal optical responses are triggered, producing concentration-dependent multicolor variations that are directly observable by the naked eye, enabling dual-target discrimination. To quantitatively decode these physico-optical responses, a logic-guided physico-optical decoding (LGPD) framework is employed, enabling simultaneous prediction of BPB and LVX concentrations with high correlation coefficients (range from 0.964 to 0.994). The probe exhibits stable tri-color emission over 20 days, with sensitivity low to 1 ng/mL, a linear range of 1-300 ng/mL, and recoveries of 85.4 %-98.61 % in real samples. This multimodal optical encoding and machine-learning-based decoding strategy enables rapid (<30 min) multiplexed detection and provides a generalizable platform for point-of-care analysis in environmental and biomedical applications. This bioinspired integration of optical encoding and machine-learning-assisted decoding provides an alternative strategy for visual and quantitative multiplex sensing.