Machine learning-based exosome profiling of multi-receptor SERS sensors for differentiating adenocarcinoma in situ from early-stage invasive adenocarcinoma.

Exosomes, extracellular vesicles released by cells, hold potential as diagnostic markers for the early detection of lung cancer. Despite their clinical promise, current technologies lack rapid and effective means to discriminate between exosomes derived from adenocarcinoma in situ (AIS) and early-stage invasive adenocarcinoma (IAC). This challenge arises from the intrinsic structural heterogeneity of exosomes, necessitating the development of advanced methodologies for precise differentiation. Here, we demonstrate a novel approach for plasma exosome detection utilizing multi-receptor surface-enhanced Raman spectroscopy (SERS) technology to differentiate between AIS and early-stage IAC. To accomplish this, we synthesized a stable and uniform two-dimensional SERS substrate (BC/Au NPs film) by fabricating gold nanoparticles onto bacterial cellulose. We then enhanced its capabilities by introducing multi-receptor SERS functionality via modifying the substrate with both low-specificity and physicochemical-selective molecules. Furthermore, by strategically combining all capturer-exosome SERS spectra, comprehensive "combined-SERS spectra" are reconstructed to enhance spectral variations of the exosome. Combining these features with partial least squares regression-discriminant analysis (PLS-DA) modeling significantly improved discriminatory accuracy, achieving 90% sensitivity and 95% specificity in distinguishing AIS from early-stage IAC. Our developed SERS sensor provides an effective method for early detection of lung cancer, thereby paving a new way for innovative advancements in diagnosing lung cancer.