Construction and validation of a diagnostic model for cholangiocarcinoma based on tumor-educated platelet RNA expression profiles

Abstract

Listen

Translate article

Abstract Objectives We aim to explore the diagnostic value of platelet-based “liquid biopsy” technology for cholangiocarcinoma (CCA), seeking reliable methods for early cancer diagnosis to improve patient prognosis. Methods Bioinformatics methods were utilized to analyze the GEO databases (GSE183635) and (GSE68086), identifying differentially expressed genes and constructing a diagnostic model of CCA using tumor-educated platelet (TEP) RNA expression profiles. GO and KEGG pathway enrichment analysis were performed. Additionally, platelet RNA from CCA patients and controls totaling 60 cases was extracted by qRT-PCR experiments to validate the diagnostic reliability of candidate genes, further confirmed through in vitro experiments. Results A diagnostic model comprising seven platelet genes (CRYM, IFI27, EED, METAP1, RASGRP1, SEC11A, and WDR82) effectively distinguished CCA from controls. Area under curve (AUC) values were 0.862 (training set), 0.875 (internal validation), 0.865 (total internal), and 0.954 (external validation). GO analysis highlighted “non-coding RNA processing,” “nuclear envelope,” and “catalytic activity, acting on RNA.” KEGG pathways included “Ribosome biogenesis in eukaryotes”, “RNA translocation” and “Spliceosome”. qRT-PCR experiments revealed significant differences (p<0.05) in METAP1, SEC11A, WDR82, RASGRP1, and EED gene expression in CCAs, consistent with bioinformatics predictions. CRYM showed significant differences (p<0.001) compared to healthy individuals. WDR82 and CRYM had high diagnostic efficiency (AUC 0.939 and 0.942), surpassing conventional tumor markers (AFP, CEA, and CA19-9). Joint receiver operating characteristics (ROC) analysis yielded an AUC of 0.806, sensitivity of 1.000, and accuracy of 0.833. Conclusions Based on the GEO database, we identified seven TEP RNAs (CRYM, IFI27, METAP1, SEC11A, WDR82, RASGRP1, EED) with strong discriminative ability for CCA, suggesting their potential as reliable non-invasive biomarkers.