Background: Recently, microRNAs (miRNAs) have been reported to play essential roles in the pathogenesis of prostate cancer and shown to have the potential as molecule biomarkers for the diagnosis of prostate cancer.However, the knowledge about prostate cancer miRNA biomarkers is quite limited, and the regulation mechanism for these miRNAs remains largely unknown.Methods: We hypothesized that the miRNA activity can be associated with the expression levels of its target genes and developed a novel model for prediction of miRNA biomarkers in prostate cancer, by integrative analyzing of gene expression and miRNA-mRNA interaction data.The predicted microRNAs were then verified by qPCR analysis of microRNA expression difference between prostate cancer and the control tissue samples.Results: With this model, thirty miRNAs were predicted as prostate cancer miRNA biomarkers.Five of them have been previous reported to show aberrant regulation in prostate cancer.We selected eight of the predicted outlier miRNAs for quantitative PCR confirmation.Among them, four microRNAs were shown to be novel prostate cancer miRNA biomarkers.The systematic analyses for the relationship between the outlier miRNAs target genes and prostate cancer further confirmed our finding.We at last applied our model to prostate cancer gene expression data at the early stage and the advanced, metastatic stage ; the results provide putative biomarkers for the prediction of the occurrence or the metastases of prostate cancer.Conclusions: We proposed a novel method for detecting outlier miRNAs biomarkers from gene expression data and then applied the model to prostate cancer miRNA biomarkers identification.Four microRNAs were confirmed with quantitative PCR and can be potentially used as novel biomarkers for the diagnosis of prostate cancer .