Deep Learning-Enabled Plasmonic-Catalytic Au@PBA Array for Simultaneous SERS/Colorimetric Quantification of Exhaled Benzene-Related VOCs

Abstract

Abstract Accurate determination of benzene-related volatile organic compounds (B-VOCs) in exhaled breath is of considerable importance for disease screening and occupational exposure assessment. However, the simultaneous identification and quantification of structurally similar B-VOCs remain a challenge due to severe cross-interference and nonlinear signal coupling in multicomponent matrices. Herein, we report a deep learning-enabled plasmonic-catalytic sensor array based on core−shell Au@PBA(M) (M = Co, Ni, Mn) nanozymes featuring an ultrathin PBA shell for the multiplexed analysis of five B-VOCs (toluene, ethylbenzene, xylene, styrene, and phenol). The dual-response core−shell Au@PBA nanoparticles provide both strong plasmonic enhancement and peroxidase-mimicking activity, enabling the catalytic oxidation of TMB with concurrent surface-enhanced Raman scattering (SERS) and colorimetric readouts. In the presence of different B-VOCs, the oxidation process is inhibited to different extents due to the differential binding affinities of the B-VOCs to the Au@PBA nanoparticles, generating highly characteristic dual-mode fingerprints. Based on this mechanism, linear discriminant analysis enabled reliable discrimination of the five B-VOCs. Notably, quantitative analysis of single-component systems demonstrated broad quantitative ranges of 0.001−500 ppm for ethylbenzene and phenol and 0.005−500 ppm for toluene, xylene, and styrene. More importantly, the integration of multichannel SERS fingerprints with an artificial neural network enabled simultaneous quantification of five B-VOCs in 25 complex breath-relevant samples, yielding excellent predictive accuracy (R2 > 0.97) and low root-mean-square errors (0.27−0.30 in log10 scale). These results demonstrate that the proposed strategy effectively resolves the cross-responsiveness of complex gas mixtures, providing a robust and noninvasive platform for intelligent breath analysis and environmental monitoring.