PCA-TLNN-based SERS analysis platform for label-free detection and identification of cisplatin-treated gastric cancer

Serum analysis is crucial for favourable prognosis of gastric cancer (GC) and for improving patient survival rates. However, it remains a challenge to develop an effective strategy to accurately identify differences in gastric cancer before and after treatment to guide efficacy evaluation. In this study, we combined surface-enhanced Raman scattering (SERS) with principal component analysis (PCA)-two-layer nearest neighbour (TLNN) to propose a promising serum analytical platform for label-free detection of cisplatin-treated GC mice. A microarray chip fabricated from Au nano-hexagon (AuNH) substrates was employed to measure the SERS spectra of the serum of GC mice at different treatment stages, and then a model for recognition of SERS spectra was constructed using a PCA-TLNN algorithm. The results revealed that the microarray chip exhibited superior portability, SERS activity, stability, and uniformity. Through PCA-TLNN, the GC mice at different treatment stages were successfully segregated, and several key spectral features for distinguishing different treatment stages were captured. The established PCA-TLNN model achieved satisfactory results, with an accuracy of over 97.5%, a sensitivity of over 90%, and a specificity of over 96.7%. Label-free serum SERS in combination with multivariate analysis could serve as a potential technique for the clinical diagnosis and staging of treatments. • The novel microarray chip can realize rapid, sensitive, label-free and high-throughput detection of SERS spectra of serum. • PCA-TLNN successfully differentiated the SERS spectra of serum from cisplatin-treated GC mice at different stages. • The most prominent spectral features for distinguishing different treatment stages • were captured in PCs loading plots. • PCA-TLNN was superior to traditional multivariate algorithm in accuracy, sensitivity and specificity.