Abstract
Blood carries some of the most valuable biomarkers for disease screening as it interacts with various tissues and organs in the body. Human blood serum is a reservoir of high molecular weight fraction (HMWF) and low molecular weight fraction (LMWF) proteins. The LMWF proteins are considered disease marker proteins and are often suppressed by HMWF proteins during analysis. This issue is addressed by using a filtration device to isolate the filtrate portion from blood serum samples having biomarker proteins up to the size of the cutoff value of the filtration device. In this research, 100 kDa filter devices are employed to obtain the filtrate portions from blood serum samples of type II diabetes mellitus patients and healthy volunteers, followed by characterization using surface-enhanced Raman spectroscopy (SERS) with silver nanoparticles (Ag NPs) as the SERS substrate. By using this approach, the collected filtrate is expected to contain marker proteins at a size of <100 kDa, which are associated with type II diabetes. These marker proteins are in a narrow size range (cutoff value of 100 kDa). Hence, they may be more easily identified by their characteristic SERS spectral features as compared to their analysis in the respective whole blood serum samples due to the exclusion of larger size proteins. These proteins that are present in the filtrate portions of type II diabetes may include adiponectin, C-reactive protein, insulin, leptin, RBP4, IL-6, TNF-α, Fibroblast Growth Factor 21 (FGF21), albumin, transthyretin, alpha-antitrypsin, transferrin, apolipoprotein A-1 (ApoA-1) and fetuin-A. Some prominent SERS bands are observed at 356 cm-1, 435 cm-1, 490 cm-1, 548 cm-1, 596 cm-1, 729 cm-1, 746 cm-1, 950 cm-1, 1330 cm-1, 1362 cm-1, 1573 cm-1 and 1689 cm-1, which differentiate type II diabetes patients from healthy individuals. Moreover, the SERS spectral data sets of various samples are classified using two chemometric approaches: principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA). The validation of the PLS-DA analysis classification model is indicated with 81% accuracy, 79% specificity, and 85% sensitivity, having a value of AUC = 0.75.