Multiplexed Surface Protein Detection and Cancer Classification Using Gap-Enhanced Magnetic–Plasmonic Core–Shell Raman Nanotags and Machine Learning Algorithm

• Published in: ACS applied materials & interfaces Vol. 16; no. 2; pp. 2041 - 2057
• Main Authors: Rodriguez-Nieves, Alberto Luis, Taylor, Mitchell Lee, Wilson, Raymond, Eldridge, Brinton King, Nawalage, Samadhi, Annamer, Assam, Miller, Hailey Grace, Alle, Madhusudhan Reddy, Gomrok, Saghar, Zhang, Dongmao, Wang, Yongmei, Huang, Xiaohua
• Format: Journal Article
• Language: English
• Published: United States American Chemical Society 17.01.2024
• Subjects: Biological and Medical Applications of Materials and Interfaces; cancer detection; multiplex; SERS; machine learning; DDSCAT; magnetic–plasmonic nanoparticles; FDTD
• ISSN: 1944-8244; 1944-8252
• DOI: 10.1021/acsami.3c13921

Cancer is the second leading cause of death attributed to disease worldwide. Current standard detection methods often rely on a single cancer marker, which can lead to inaccurate results, including false negatives, and an inability to detect multiple cancers simultaneously. Here, we developed a multiplex method that can effectively detect and classify surface proteins associated with three distinct types of breast cancer by utilizing gap-enhanced Raman scattering nanotags and machine learning algorithm. We synthesized anisotropic magnetic core–gold shell gap-enhanced Raman nanotags incorporating three different Raman reporters. These multicolor Raman nanotags were employed to distinguish specific surface protein markers in breast cancer cells. The acquired signals were deconvoluted and analyzed using classical least-squares regression to generate a surface protein profile and characterize the breast cancer cells. Furthermore, computational data obtained via finite-difference time-domain and discrete dipole approximation showed the amplification of the electric fields within the gap region due to plasmonic coupling between the two gold layers. Finally, a random forest classifier achieved an impressive classification and profiling accuracy of 93.9%, enabling effective distinguishing between the three different types of breast cancer cell lines in a mixed solution. With the combination of immunomagnetic multiplex target specificity and separation, gap-enhancement Raman nanotags, and machine learning, our method provides an accurate and integrated platform to profile and classify different cancer cells, giving implications for identification of the origin of circulating tumor cells in the blood system.