A multiplex self-referencing detection of pathogens using surface enhanced raman scattering nanoprobes with a nano-DEP microfluidic concentrator

In this dissertation, I successfully developed the multiplex self-referencing SERS pathogen (E.coli O157: H7) detection biosensor platform which offers high sensitivity (10^1 CFU/mL) and strain level discrimination by measuring the superimposed SERS signatures with multiple characteristic peaks. To harvest the effective Raman molecular probes, I developed methods to fabricate anisotropic metallic nanoparticles to serve as SERS enhancers, and more importantly, I developed surface modification methodology to add functionality to the SERS enhancers so that they can bind specifically to their pathogen targets for highly accurate and sensitive detection. Gold nanorods (GNRs) and gold/silver nanocages are successfully fabricated with high particle yield. Three highly effective linker molecules (4-Aminothiophenol (4-ATP), 3-Amino-1,2,4-triazole-5-thiol (ATT), and 3-Mercaptopropionic acid (3-MPA)) are identified and designed to conjugate on gold nanostructures, and then different monoclonal antibody molecules are designed to bond to the different linkers through diazo-histine binding (4-ATP and ATT) and EDC/NHS bonding (3-MPA-antibody). In addition, this platform demonstrated excellent separation and concentration capacities by using DEP microfluidic devices and further improves the sensitivity to 10^0 CFU/mL. The integration of microfluidic devices with SERS detection has yielded simple and miniaturized instrumentation that is suitable for the detection and characterization of small volume of chemical and biological analytes with high sensitivity and specificity. For data analysis, various preprocessing methods are used for spectral background removal, baseline correction, smoothing, and normalization. Principle Component Analysis (PCA) is applied to reduce the variable dimensions. A Support Vector Machine (SVM) discriminant analysis model based on statistical multivariate model is being developed for superimposed spectra classification. The validation of spectral classification model (target binding VS no target binding) is evaluated by the accuracy percentage, which is above 95%