An aptamer-based SERS method for rapid screening and identification of pathogens assisted by machine learning technique with robustness evaluation

Abstract

Pathogen testing is one of the compulsory items for the quality control of pharmaceutical products. Up to now, rapid screening and identification of low-concentration pathogen contaminations is still full of challenge. Herein, an aptamer-coated magnetic beads (Apt-MBs)-based isolation strategy combined with culture medium turbidity observing and label-free SERS detection technique has been developed for rapid screening and detection of Escherichia coli, Salmonella Typhimurium and Staphylococcus aureus. Apt-MBs were employed for capturing and isolation of target pathogens. Then, with a simple enrichment process, the potential pathogens could be screened out by observing the turbidity of liquid culture medium. The presumptive bacteria species was furtherly identified via a silicon-based SERS chip couple with an optimized machine learning model, which was selected by directly evaluating the model’s generalization ability on test data. The results demonstrated that SoftMax classifier had the optimal performance with the recognition accuracy of 100 % for all samples. Finally, the method was successfully applied for identification of pathogens at ∼1 CFU spiked level in drug samples within 24 h, which is much shorter than that of the standard culture-based method (3–5 d). Therefore, the developed method showed a promising application in monitoring pathogen levels in pharmaceutical industries.