Rapid and quantitative detection of Botryosphaeria dothidea by surface-enhanced Raman spectroscopy with size-controlled spherical metal nanoparticles combined with machine learning

Abstract

Botryosphaeria dothidea infection has become a major factor affecting the quality of postharvest fruits, so detection of B. dothidea infection is very important to control the spread of infection and ensure food safety. In this study, we built a monitoring platform for rapid and quantitative detection of B. dothidea by Surface-enhanced Raman spectroscopy (SERS) with size-controlled spherical metal nanoparticles combined with machine learning, and the platform could also detect a variety of pathogens. With spherical metal nanoparticles as the active substrate, the SERS enhanced effect was significantly size-dependent. 45–60 nm Ag@ICNPs was determined to achieve the maximum SERS signal detection of B. dothidea, and used as the active substrate to construct a SERS platform for rapid and quantitative detection of B. dothidea. This platform had potential practicability in actual sample detection. The platform was used for the detection of Pseudomonas syringae pv. Actinidiae, Pseudomonas aeruginosa, Ralstonia solanacearum and Pseudomonas extremorientalis, and the SERS fingerprint information of these pathogens was successfully captured, and the quantitative analysis ability of these pathogens was also strong. Machine learning analysis was performed on the SERS spectra of pathogens obtained. Based on the differences between the spectral data sets of different pathogens, PCA could effectively distinguish these five pathogens into different groups. The accuracy rates of SVM, Tree, Linear discrimination analysis, Efficient logistic regression, Naive Bayes, K-Nearest Neighbors, Ensemble and Neural network test were 100 %, 96 %, 100 %, 100 %, 100 %, 100 %, 98 % and 100 % respectively, all of which had relatively high accuracy rates. Overall, this study provides a simple, efficient and accurate method for rapid quantitative detection and identification of multiple pathogens, and can be extended to practical agricultural products.