Alternative splicing (AS) is a transcriptional regulation mechanism that participates in multiple aspects of cancer. The present study aimed to identify differential AS events from tumor and non-tumor samples and investigate the potential of AS as a source of candidate cancer diagnostic biomarkers. Deep RNA sequencing of three paired hepatocellular carcinoma (HCC) tumors and adjacent non-tumors was applied to identify AS events. RT-qPCR was performed on 45 HCC clinical samples to validate the splicing differences. The maximal information coefficient was first used to build an association between clinical features and AS changes. We identified 197 significantly differential skipped exon events, of which only 29% overlapped with the differentially expressed genes. The differentially spliced genes were mainly enriched in HCC-characterized biological processes and pathways, clearly separating tumors from non-tumors. We also validated the statistically significant splicing differences of three AS candidates (CEACAM1 exon 7, VPS29 exon 2 and ISOC2 exon 3). Furthermore, a clinicopathological analysis revealed that carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) exon 7 was significantly correlated with the survival time, and VPS29 exon 2 was associated with cell differentiation stages. In conclusion, the findings of the three AS candidates in the present study could be beneficial in HCC prognosis and new treatment strategies.Abbreviations: A3SS, alternative 3'splice site; A5SS, alternative 5'splice site; AS, alternative splicing; DE, differentially expressed; DS, differentially spliced; FDR, false discovery rate; FPKM, fragments per kilobase of transcript per million mapped reads; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; IL, inclusion level; MIC, maximal information coefficient; MXE, mutually exclusive exons; PSI, percent-spliced-in; RI, retained intron; RNA-Seq, RNA sequencing; RT-PCR, reverse transcription quantitative polymerase chain reaction; SE, skipped exon