Discriminatory Detection of ssDNA by Surface-Enhanced Raman Spectroscopy (SERS) and Tree-Based Support Vector Machine (Tr-SVM)
Seju Kang, Inyoung Kim, Peter J. Vikesland
Abstract
We report label-free detection of 86-base single-stranded DNA (ssDNA) gene segments by surface-enhanced Raman spectroscopy (SERS). The use of a slippery liquid infused porous (SLIP) membrane induced aggregation of 43 nm gold nanoparticles and ssDNA upon pin-free droplet evaporation. The combined SLIPSERS approach generates significant numbers of SERS hot-spots and enabled detection at the 100 nM level of mecA and intI1 gene segments—two genes of interest in the context of antibiotic resistance. Tree-based multiclass support vector machine (Tr-SVM) classifiers were built to discriminate SERS spectra of 12 different gene sequences obtained by SLIPSERS: mecA, intI1, as well as analogues of mecA and intI1, respectively, with 2–10 base mismatches, and two random sequences. The trained predictive Tr-SVM classifiers correctly identified each gene sequence with a prediction accuracy of ∼90%. This study illustrates a novel means for discriminatory label-free SERS detection of ssDNA enabled by Tr-SVM.