Machine learning-enhanced SERS detection of melamine and its analogues in non-pretreated milk via filter-pressing assembled polytetrafluoroethylene-AgNPs substrate

Melamine contamination from illegal additives, packaging contaminants, and pesticide residues threatens dairy product safety, demanding rapid detection. Traditional methods such as chromatography or mass spectrometry are precise but lack field applicability due to complexity, time consumption, and cost. Surface-enhanced Raman spectroscopy (SERS) is a promising alternative for sensitive, rapid, and label-free analysis. However, current SERS implementations face challenges like complex substrate synthesis, environmentally harmful sample processing, and lack of discrimination of analogues. Thus, developing a simple SERS-based method for detecting melamine and its analogues without pretreatment remains urgent. In this paper, a straightforward SERS detection method is proposed to achieve accurate and rapid pretreatment-free detection of melamine in milk. Polytetrafluoroethylene‑silver nanospheres (PTFE-AgNPs) SERS substrate is fabricated by mixing silver colloid, sodium hydroxide and sodium chloride solution, followed by deposition onto a PTFE filter membrane by filter-pressing assembly. Additionally, diluted milk is subjected to SERS testing directly without any pretreatment. Furthermore, both qualitative and quantitative analyses were performed using RF, PCA-SVM and CNN. Among the three algorithms, CNN classification model achieved the best accuracy 99.25 % for distinguishing melamine, ammeline, ammelide, and blank controls, while the CNN regression model yielded a coefficient of determination (R 2 ) of 0.9999 for melamine quantification. The limit of detection (LOD) for melamine in milk was 3.32 × 10 −6 M, lower than the World Health Organization (WHO) recommended limit. This method, featuring simple SERS substrate preparation and non-pretreatment, enables rapid and efficient detection of melamine and its analogues, promoting broader applications of SERS in food safety monitoring.