Bacteriophage modified SERS substrates integrated with machine learning for trace detection of nitrofurantoin in food and water samples

Nitrofurantoin (NFT) is an antibiotic reported in aquaculture environments due to misuse and tends to bioaccumulate, causing pulmonary and hepatic disorders in humans. HPLC and LC-MS-based detection, although sensitive, is challenging in resource-limited settings, while matrix interference in complex samples limits its accurate quantification. In this work, SERS-active substrate comprising gold nanorods conjugated with bacteriophage T4 was developed and demonstrated for the rapid, specific detection and quantification of NFT in diverse food and water samples. The use of bacteriophage T4 provided an organized spatial network that amplified the localized surface plasmon resonance-induced hotspots of gold nanorods, producing high Raman signal enhancement. An optimized combination of nanorods with an aspect ratio of 3.7 and a bacteriophage titre of 10⁷ PFU/mL yielded the highest sensitivity. The developed SERS substrate was calibrated to detect NFT and showed an exponentially linear response in the range of 0.05–500 ppb (R² = 0.98, RSD ' 15%, n = 3) in wastewater, aquaculture water, milk, and prawn lysate samples. The sensor exhibited excellent selectivity and recovery, establishing its robustness in detecting NFT in biological and environmental samples. However, spectral overlap and matrix effects necessitated multivariate analysis of SERS spectra for reliable quantification. Machine learning-based analysis using Gradient Boosting regression provided a highly accurate NFT quantification (R² = 0.99, RSD ' 8%) and demonstrated a good generalization in the prediction of NFT concentration. The developed sensor thus holds great potential for robust point-of-use assessment of NFT, to minimize chronic human exposure, limiting NFT-induced resistance, monitoring water quality and enabling food security.