Highly Sensitive Detection and Molecular Subtyping of Breast Cancer Cells Using Machine Learning-assisted SERS Technology

Abstract

Breast cancer has always been a research hotspot in the medical field due to its highest incidence and mortality rates among women worldwide. However, the significant molecular heterogeneity of breast cancer presents major challenges for its diagnosis and treatment. Surface-enhanced Raman spectroscopy (SERS) has gained considerable attention for its capability in trace detection and molecular analysis. To accurately identify different breast cancer cell subtypes, constructing reliable SERS bioprobes is essential. Therefore, a specific highly expressed receptor, human epidermal growth factor receptor 2 (HER-2), was employed to explore SERS bioprobes in this study. Two bioprobes capable of targeting breast cancer cells, Au NPs@4-MBA@PDA@aHER-2 and Au NPs@4-MPY@PDA@aHER-2, were synthesized. SERS performance testing indicated that the Au NPs were able to detect and trace molecules at concentrations as low as 2 × 10^–9 mol/L. Additionally, the two bioprobes exhibited good spectral stability with a relative standard deviation (RSD) of 9.58%. Moreover, by constructing a “symphonic SERS spectra” of the two bioprobes with prominent component analysis-linear discriminant analysis (PCA-LDA), the classification accuracy of distinguishing white blood cells (WBCs) and two breast cancer cell subtypes (SK-BR-3 and MDA-MB-231) reached up to 97.33%. The integration of machine learning with SERS detection provides a novel technological pathway for the early diagnosis and personalized treatment of breast cancer.