Rapid detection of beer spoilage bacteria based on label-free SERS technology

Abstract

Beer spoilage bacteria have been a headache for major breweries. In order to rapidly identify spoilage bacteria and improve the sensitivity and signal-to-noise ratio of bacterial SERS detection, the label-free SERS technique was used as a starting point, and we found eight bacteria species that led to beer spoilage. The impact of AgNP concentration and AgNP and bacterial binding time on the final results were thoroughly investigated. To maximize the increase in the SERS signal, an aluminized chip was created. We merged the t-SNE reduced dimensional analysis algorithm, and SVM, KNN, and LDA machine learning algorithms to further investigate the effect of the approach on the final identification rate. The results demonstrate that SERS spectra had an increased intensity and signal-to-noise ratio. The machine learning classification accuracy rates were all above 90%, indicating that the bacteria were correctly classified and identified.