Abstract
BACKGROUND: Pancreatic cancer (PaCa) is a deadly malignancy that is often diagnosed at an advanced stage, limiting treatment and reducing survival. There is an urgent need for convenient and accurate diagnostic markers for the early detection of PaCa. METHODS: In this multicenter case-control study, we performed transcriptome analysis of 673 platelet samples from different in-house and public cohorts. RNA sequencing and RT-qPCR were used to discover and validate potential platelet biomarkers. A multi-gene signature was developed using binomial generalized linear model and independently validated in multicenter cohorts. RESULTS: Two platelet RNAs, SCN1B and MAGOHB, consistently showed robust altered expression patterns between PaCa and healthy controls across cohorts, as confirmed by both RNA sequencing and RT-qPCR. The diagnostic two-RNA signature, PLA2Sig, demonstrated remarkable performance in detecting PaCa, with area under the receiver operating characteristic curve (AUC) values of 0.808, 0.900, 0.783, and 0.830 across multicenter cohorts. Furthermore, PLA2Sig effectively identified resectable stage I&II PaCa cases with an AUC of 0.812. Notably, PLA2Sig outperformed the traditional serum markers carcinoembryonic antigen and carbohydrate antigen 19-9 in distinguishing PaCa from healthy controls, and is complementary to established blood-based screening biomarkers. CONCLUSION: These findings provide preliminary but promising evidence for the potential utility of platelet RNAs as an alternative non-invasive liquid biopsy tool for the early detection of PaCa.