Development of a chemometric-assisted SERS method for simultaneous analysis of HDL and LDL cholesterol in blood serum with silver nanoparticles as substrate.

Cardiovascular diseases (CVD) are becoming a serious threat to human health. These are considered the leading causes of mortality. Abnormal lipid concentrations in the body, such as High-density lipoproteins (HDL) and Low-density lipoproteins (LDL), are major factors that contribute to CVD. Surface-enhanced Raman spectroscopy (SERS) has the potential to be used to compare HDL and LDL cholesterol. In the current study, SERS was employed for the comparative profiling of HDL and LDL cholesterol using clinical blood serum samples along with silver nanoparticles (Ag-NPs) as the SERS substrate. The SERS spectral features of HDL and LDL cholesterol were clearly identified by applying various chemometric statistical tools. The Principal Component Analysis (PCA) was employed for the differentiation of blood serum samples of HDL and LDL cholesterol. Moreover, the support vector machine-Synthetic minority over-sampling technique (SVM-SMOTE) was used to accurately address the different imbalanced concentration of HDL and LDL cholesterol in order to reduce the risk of overfitting as compared to traditional machine learning algorithms. The SMOTE algorithm improves the interpretability of SVM by analyzing the minority classes of data sets. The macro-average Area Under the Curve (AUC) increased slightly from 0.97 to 0.98 with SMOTE, though the test Area Under the Curve was the same as 0.95. These results showed the accuracy and validation of the SMOTE model for the comparison of blood serum samples of HDL and LDL cholesterol.