Optimization of Savitzky-Golay smoothing filter for salivary surface enhanced Raman spectra of non structural protein 1

Abstract

Raman spectroscopy is a vibration based spectroscopic technique for identifying chemical constituent. The Raman signal can be enhanced by binding with metal nanostructures, resulting in so-called surface-enhanced Raman scattering (SERS). Non-structural protein 1 (NS1), a biomarker for flavivirus origin diseases, is found to present in saliva in low concentration. SERS is promising as an early detection method for NS1 in saliva. However, unwanted features are found to embed in the SERS spectrum, making the characteristic features undetected with visual or automated means. In particular, at low concentration of NS1, the intensity at characteristic peak (Imax) and its signal-to-noise ratio (SNR) is so low that the noisy features actually bury the fingerprint characteristics. In addition to background subtraction and baseline removal, spectrum smoothing using suitable filter is found to improve the spectra SNR. In this study, NS1 dilution dataset from the UiTM-NMRR-12868-NS1-DENV spectral database is used. It is found that SG smoothing filter optimized with a span of 13 and a polynomial degree of order 3 produces the most intense smoothing effect while preserving most of the characteristics of the original spectral waveform, with SNR of 3.76, %RMSE of 33.49% and maximum intensity (Imax) of 98%.