Abstract
Triple-negative breast cancer (TNBC) is a highly heterogeneous disease with an aggressive clinical course. Prognostic models are needed to chart potential patient outcomes. To address this, we used alternative 3'UTR patterns to improve postoperative risk stratification. We collected 327 publicly available microarrays and generated the 3'UTR landscape based on expression ratios of alternative 3'UTR. After initial feature filtering, we built a 17-3'UTR-based classifier using an elastic net model. Time-dependent ROC comparisons and Kaplan-Meier analyses confirmed an outstanding discriminating power of our prognostic model for TNBC patients. In the training cohort, 5-year event-free survival (EFS) was 78.6% (95% CI 71.2-86.0) for the low-risk group, and 16.3% (95% CI 2.3-30.4) for the high-risk group (log-rank p<0.0001; hazard ratio [HR] 8.29, 95% CI 4.78-14.4), In the validation set, 5-year EFS was 75.6% (95% CI 68.0-83.2) for the low-risk group, and 33.2% (95% CI 17.1-49.3) for the high-risk group (log-rank p<0.0001; HR 3.17, 95% CI 1.66-5.42). In conclusion, the 17-3'UTR-based classifier provides a superior prognostic performance for estimating disease recurrence and metastasis in TNBC patients and it may permit personalized management strategies.