A microRNA expression profile for vascular invasion can predict overall survival in hepatocellular carcinoma

Background The presence of vascular invasion (VI) in pathology specimens is a well-known unfavorable prognostic factor of hepatocellular carcinoma (HCC) recurrence and overall survival (OS). We investigated the vascular invasion related microRNA (miRNA) expression profiles and potential of prognostic value in HCC. Methods MiRNA and mRNA expression data for HCC were accessed from The Cancer Genome Atlas (TCGA). LASSO logistic regression models were used to develop a miRNA-based classifier for predicting VI. The predictive capability was accessed by area under receiver operating characteristics (AUC). Concordance index (C-index) and time-dependent receiver operating characteristic (td-ROC) were used to determine its prognostic value. We validated the predictive and prognostic accuracy of this classifier in an external independent cohort of 127 patients. Functionally relevant targets of miRNAs were determined using miRNA target prediction, experimental validation and correlation of miRNA and mRNA expression data. Results A 16-miRNA-based classifier was developed which identified VI accurately, with AUC of 0.731 and 0.727 in TCGA set and validation cohort, respectively. C-index and td-ROC showed that the classifier was able to stratify patients into risk groups strongly associated with OS. When stratified by tumor characteristics, the classifier was still a clinically and statistically significant prognostic model. The predictive and prognostic accuracy of the classifier was confirmed in validation cohort. Vascular invasion related miRNA/target pairs were identified by integrating expression patterns of predicted targets, which were validated in cell lines. Conclusions A multi-miRNA-based classifier developed based on the presence of VI, which could effectively predict OS in HCC.