Surface-enhanced Raman scattering (SERS) spectroscopy of single-cell suspensions obtained from fresh specimens of breast cancer tissue and normal breast tissue by mechanical enzymatic digestion was obtained and analysed, which is different from most Raman studies using breast cancer cell lines. Random forest classification was implemented to develop effective diagnostic algorithms for the classification of SERS of different typed cells. We first examined the SERS spectra of the primary breast cancer single cell and normal epithelial single cell obtained by flow sorting cytometry due to their biomarkers of CD326+/CD45-. Comparison analyses on their SERS spectra disclose that the nucleic acid and protein levels of the primary breast cancer single cell are higher than those of the normal epithelial single cell, while the lipids are at a relatively lower level. An important finding is that the cholesterol, palmitic acid, and sphingomyelin in the cancer cell profiles exhibit stronger than those of normal cells, while the glycans are at a relatively lower level. Furthermore, the standard deviation (SD) of the normal epithelial single cell is larger than that of the breast cancer cell, and the SD of the primary breast cancer single cell is more obvious than that of the normal epithelial cells. In addition, the prospective application of an algorithm to the dataset results in an accuracy of 78.2%, a precision of 75.5%, and a recall of 66.7%. The breast cancer diagnostic model laid a solid foundation for judgment of breast-conserving surgical margins and early diagnosis of breast cancer.
Keywords: Flow cytometry; Primary breast cancer cell; Random forest; Surface-enhanced Raman spectroscopy.
Copyright © 2020 Elsevier B.V. All rights reserved.