Early colorectal cancer detection: a serum analysis platform combining SERS and machine learning

Abstract

Colorectal cancer (CRC) is one of the deadliest malignancies globally, with high incidence and mortality rates. Early detection is crucial for improving treatment success rates and patient survival. However, due to the difficulty in detecting early symptoms, many cases are diagnosed at advanced stages, necessitating more sensitive and accurate detection methods. This study proposes a novel approach combining the Principal Component Analysis (PCA)-Dynamic Weighted Nearest Neighbor (DWNN) model with Surface-Enhanced Raman Scattering (SERS) technology to detect the serum of CRC mice at different stages. Establishing the CRC mice model, serum samples were collected for further analysis. An Au Nanocluster (AuNC) substrate was synthesized to ensure optimal SERS enhancement. The PCA-DWNN recognition model was constructed to classify the SERS spectra of CRC at different stages. The synthesized AuNC substrate has high sensitivity, good reproducibility, uniformity, and stability, making it a high-performance nanomaterial. The PCA-DWNN model has significant advantages in identifying high-dimensional and complex SERS spectra, offering excellent classification accuracy and robustness, with an accuracy rate of 97.5%. By analyzing the PCA loading plot, it was observed that as CRC progressed, the content and structure of proteins, lipids, amino acids, and carbohydrates in the serum changed, reflected in different characteristic peaks in the SERS spectra. This study suggests that SERS combined with PCA-DWNN has potential in the early detection of CRC, possibly providing a novel approach for clinical diagnostics.