UC Irvine

Optical bio-nanosensors are powerful alternatives to conventional analytical techniques because they enable direct, real-time and label-free detection of many biological and chemical substances. Surface enhanced Raman scattering (SERS) spectroscopy is a label-free optical biosensing technique that provides target specific information, enables real-time monitoring of continuous flows and multiplex detection of small volatile molecules. For example, the detection of small molecule metabolites has proved to be crucial in study of microbial activity and early detection of diseases. In the first part of this research, I prepare SERS substrates designed to improve the analytical capabilities of SERS. First, synthesis of uniform Au nanopsheres via a kinetically controlled seeded growth method is performed. Second, chemically assembled Au nanoparticle clusters on self-organized templates are characterized as SERS substrates. The enhancement of SERS signals by controlling nanometer gap spacings between plasmonic nanospheres using the length of the chemical cross linker enables detection of various metabolites down to 1 pg.mL-1.

In the second part of this work, I discuss the advantages of using deep learning algorithms for quantification of metabolites. Feed-forward neural networks and 1 D- Convolutional neural networks are trained on indole and 2-aminoacetophenone SERS spectral data to get robust quantification of concentrations ranging from 1 pg.mL-1 to 1 mg.mL-1. The models are further assessed for future applications of SERS sensors for multianalyte detection in biological fluids.