Label-Free Multiplex Profiling of Exosomal Proteins with a Deep Learning-Driven 3D Surround-Enhancing SERS Platform for Early Cancer Diagnosis.

Abstract

Identification of protein profiling on plasma exosomes by SERS can be a promising strategy for early cancer diagnosis. However, it is still challenging to detect multiple exosomal proteins simultaneously by SERS since the Raman signals of exosomes detected by conventional colloidal nanocrystals or two-dimensional SERS substrates are incomplete and complex. Herein, we develop a novel three-dimensional (3D) surround-enhancing SERS platform, named 3D se-SERS, for the multiplex detection of exosomal proteins. In this 3D se-SERS, proteins and exosomes are covered with "hotspots" generated by the gold nanoparticles, which surround the analytes densely and three-dimensionally, providing sensitive and comprehensive SERS signals. Combining this 3D se-SERS with a deep learning model, we successfully quantitatively profiled seven proteins including CD63, CD81, CD9, CD151, CD171, TSPAN8, and PD-L1 on the surface of plasma exosomes from patients, which can predict the occurrence and advancement of lung cancer. This 3D se-SERS integrating deep learning technique benefits from high sensitivity and significant multiplexing ability for comprehensive analysis of proteins and exosomes, demonstrating the potential of deep learning-driven 3D se-SERS technology for plasma exosome-based early cancer diagnosis.