Machine learning-based SERS label-free detection of plasma and exosome binding in early-stage lung cancer

Lung cancer is extremely lethal and early screening can be an effective treatment for lung cancer. Liquid biopsies allow for rapid, non-destructive testing and are promising for early screening. Plasma is rich in biomarkers and has been widely used for cancer diagnosis. Exosomes derived from plasma are considered a more promising biomarker. This paper reports a label-free detection strategy for plasma and exosomes of cancer-free (HC), adenocarcinoma in situ (AIS) and early-stage invasive adenocarcinoma (IAC) using surface-enhanced Raman spectroscopy (SERS). SERS spectra of plasma and exosomes were trained one to each other using four machine learning algorithms. Notably, this work breaks new ground by proposing a SERS “Combined Spectrum” of plasma combined with exosomes. The SERS “Combined Spectrum” synthesizes SERS spectra of plasma and exosomes from the same case sample into a completely new spectrum for training purposes. Calculated by each of the four machine learning algorithms, the plasma and exosome SERS “Combined Spectrum” were calculated with an accuracy higher than 97 % and an AUC value higher than 0.9500. Based on these results, the combined SERS “Combined Spectrum” of plasma and exosome can be used for the early detection of lung cancer, and it also has a promising future for application in the detection of other diseases.