We explore noninvasive clinical applications of multiple disease-specific fusion markers recently discovered in prostate cancer to predict the risk of cancer occurrence and aggressiveness at diagnosis. A total of 92 men who were prostate-specific antigen (PSA) screened and scheduled for diagnostic biopsy were enrolled for this study. Prospectively collected urine was blind coded for laboratory tests. RNA from urine sediments was analyzed using a panel of 6 TMPRSS2:ETS fusion markers with a sensitive quantitative PCR platform. The pathology reported 39 biopsy-positive cases from 92 patients (42.4%). In urine test, 10 unique combinations of fusion types were detected in 32 of 92 (34.8%) prebiopsy samples. A novel combination of fusion markers, termed Fx (III, IV, ETS), was identified with a sensitivity of 51.3% and an odds ratio of 10.1 in detecting cancer on biopsy. Incorporating a categorical variable of Fx (III, IV, ETS) with urine PCA3 and serum PSA, a regression model was developed to predict biopsy outcomes with an overall accuracy of 77%. Moreover, the overexpression of Fx (III, IV, or ETS) was shown to be an independent predictor to the high-grade cancer, with a predictive accuracy of 80% when coupled with PSA density. The individualized risk scores further stratified a high-risk group that is composed of 92% high-grade cancers and a low-risk group that harbors mainly clinically insignificant cancers. In conclusion, we have identified a novel combination of fusion types very specific to the clinically significant prostate cancer and developed effective regression models to predict biopsy outcomes and aggressive cancers at diagnosis.