Classification analyses for prostate cancer, benign prostate hyperplasia and healthy subjects by SERS-based immunoassay of multiple tumour markers.

Prostate cancer (PCa) is a leading cause of cancer-related death among males globally. To date, prostate-specific antigen (PSA), as a typical tumour marker, has been widely used in the early diagnosis of PCa. However, in practical clinical tests, high serum levels of PSA show a high probability for false-positive results, leading to misdiagnoses. In this study, we developed a new classification system for PCa, benign prostate hyperplasia (BPH) and healthy subjects by using a surface-enhanced Raman scattering (SERS)-based immunoassay of multiple tumour markers along with a support vector machine (SVM) algorithm. Silver nanoparticles (AgNPs) as immune probes and SiC@Ag@Ag-NPs SERS as immune substrates were constructed into a sandwich structure to serve as an ultrasensitive SERS-based immunoassay platform of tumour markers. With this assay, the limits of detection for PSA, prostate-specific membrane antigen (PSMA) and human kallikrein 2 (hK2) were as low as 0.46 fg mL-1, 1.05 fg mL-1 and 0.67 fg mL-1, respectively. Furthermore, the serum levels of PSA, PSMA and hK2 in clinical samples were successfully detected using the SERS-based immunoassay platform, and correct classifications of PCa, BPH and healthy subjects were feasible with help of the linear SVM algorithm. These results demonstrate the potential for improving the diagnostic accuracy of PCa. Overall, the linear SVM classification model with multiple tumour markers exhibited good classifications of PCa, BPH and healthy subjects, with a PCa diagnostic accuracy of 70% that was significantly superior to that of the linear SVM classification model based only on the serum level of PSA (50%). Therefore, combining the SERS-based immunoassay with pattern recognition technology can allow for comprehensive analyses of the serum levels of multiple tumour markers to effectively improve the diagnostic accuracy of cancer with potential applications in point-of-care testing.