The alternative splicing landscape of hepatocellular carcinoma and its potential for HCC detection

Background: Pre-mRNA alternative splicing contributes to oncogenic gene expression in hepatocellular carcinoma (HCC), and some oncogenic isoforms escape the tumor into circulation. This study aimed to characterize the alternative splicing landscape of HCC and investigate its potential for early detection by identifying HCC-specific alternative splicing events (ASEs) in peripheral fluids. Methods: We analyzed RNA-Seq data from 787 HCC tumor samples across 16 studies to identify commonly mis-spliced ASEs. ASEs were examined in relation to HCC etiology, progression, and functional domain changes. We then identified HCC-tumor-associated ASEs in samples isolated from various peripheral fluids, including circulating epithelial cells (CECs), tumor-educated platelets, serum, and plasma. Machine-learning models were trained to classify HCC versus non-HCC based on ASEs in CECs and plasma. Results: We identified 3416 ASEs commonly mis-spliced in HCC across studies (FDR p <0.05, |ΔPSI|>0.1). We made these ASEs widely available through a new web application, HCC-INSIGHT ( IN vestigation of S pliced I soforms and G enes in Human Tissues ) . Ten ASEs in CECs distinguished HCC from non-HCC in an independent test cohort with a balanced accuracy of 74%, 66% sensitivity, and 81% specificity. When combined with alpha-fetoprotein (AFP), classification performance improved to a balanced accuracy of 78%, 66% sensitivity, and 90% specificity. Conclusions: This study provides a comprehensive characterization of ASEs in HCC and highlights their potential as circulating biomarkers. Our findings support the utility of ASEs in CECs for HCC detection, particularly in combination with AFP, although additional studies are needed to validate these ASEs in an independent prospective cohort. Finally, we introduce HCC-INSIGHT as a resource for further research into HCC splicing alterations.