Development and validation of a six-gene signature for predicting prognosis in prostate cancer patients with lymph node metastasis.
e17524 Background: Prostate cancer (PCa) patients with lymph node metastasis (LNM) always exhibit poor clinical outcomes. A gene signature that could predict survival in these patients would allow for earlier detection of mortality risk and will also guide individualized therapy. Methods: A prediction model was developed using a public cohort consisting of 623 patients with clinicopathologically confirmed PCa. Data were gathered from cBioPortal and UCSC Xena. Genes expressed differentially in patients with lymph node metastasis versus those without lymph node metastasis were identified. Uni-variate Cox regression analysis and LASSO Cox regression were applied to build a prediction model. Time-dependent receiver operating characteristic (ROC) and Kaplan-Meier curves were used to assess the prognostic capacity of the model, followed by external validation using the MSKCC dataset from cBioPortal. Gene Set Enrichment Analysis (GSEA) was performed to further understand the underlying molecular mechanisms. Results: We identified a six-gene signature (covering GSDMB, SSTR1, MX1, CCBE1, MYBPC1, and FAM3D) that could effectively identify a high-risk subset of PCa patients. ROC analysis indicated that the signature had a good performance (AUC > 0.7) in survival prediction in both the training and the testing/validation cohorts. Cox regression analysis showed that the six-gene signature could independently predict disease-free survival (DFS) as well, although with lower predictive power. Subgroup analyses showed that signature-based risk score may serve as a promising marker to predict DFS in different subgroups, including stage T2 (HR = 0.12, p < 0.001), stage T3 (HR = 0.29, p < 0.001), TP53-wild-type (HR = 0.22, p < 0.001), TP53-mutated (HR = 0.07, p < 0001), AR pathways-wild-type (HR = 0.2, p < 0.001) and AR pathways-mutated(HR = 0.16, p = 0.0419). The performance of the six-gene signature in LNM+ was stable for stratifying the patients according to risk of deatch (HR = 0.23, p = 0.0333). Moreover, GSEA revealed distinct pathway enrichment features in the different risk groups, where pathways related to DNA repair were more prominently enriched in the high-risk group while the low-risk group had higher enrichment of androgen response. Conclusions: We developed a robust six-gene signature that can effectively classify PCa patients into groups with low- and high-risk group, which may help select high-risk patients who require more aggressive adjuvant target therapy or immune therapy.