Background: PROSTOX is a germline genetic signature previously found to predict late Grade > 2 genito-urinary (GU) toxicity after stereotactic body radiotherapy (SBRT). MIRAGE is a clinical trial evaluating toxicity in patients treated with CT or MRI-guided SBRT. We evaluated the ability of PROSTOX to predict toxicity in MIRAGE at 2-year follow-up. Methods: We evaluated PROSTOX’s performance in MIRAGE using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1 score, and AUC. We also evaluated the predictive utility of PROSTOX as well as defined a new signature of acute toxicity. We performed a gene ontology (GO) analysis to assess the biological relevance of the biomarkers involved in PROSTOX. Variants were mapped to their corresponding genes, and GO analysis was conducted with an adjusted p-value cutoff of 0.05. We set the genomic background to all the measured variants to account for and remove bias from the pre-selection of variants. All analyses were conducted in R using biomaRt and clusterProfiler packages. Results: The validation study included 62 and 57 patients receiving MRI and CT-SBRT treatment, respectively. There were 10 patients in the MRI arm and 16 patients in the CT arm with late GU toxicity. Ages of the patients ranged from 51-86 with a mean of 71 years old. PROSTOX predicted late GU toxicity with an AUC of 0.77 and 0.75 in the MRI and CT treated groups, respectively. Compared to the original training LOOCV metrics, in the Mirage MRI and CT groups PROSTOX resulted in higher sensitivity (0.942, 0.927) and F1 score (0.933, 0.884), but a lower specificity (0.600, 0.563) and NPV (0.667, 0.750). PROSTOX did not predict acute toxicity, but a unique genetic signature for acute toxicity was identified in the MRI-treated cohort (AUC=0.764). However, this signature was not as predictive in the CT-treated cohort (AUC=0.63). The GO analysis revealed 14 pathways enriched in the PROSTOX signature (adjusted p-value < 0.05). Notably, all pathways related to RNA, which appeared to fall into three main categories: RNA regulation, RNA processes, and transcription. Of the 24 unique genes involved in PROSTOX, 15 were conserved across all enriched pathways. Conclusions: Our study successfully validated PROSTOX to accurately predict late GU toxicity in patients treated with SBRT, regardless of delivery method. PROSTOX does not predict acute GU toxicity, which appears to have a unique signature but may be impacted by treatment approach. Our GO analysis identified RNA pathways, related to the regulatory mechanisms that control RNA synthesis within the cell, including the biochemical processes necessary for the synthesis, modification, degradation, and turnover of RNA molecules in predicting late GU toxicity. These findings provide insight into potential molecular mechanisms contributing to radiation-induced late toxicity and could lead to therapeutic strategies in the future. Clinical trial information: NCT04384770.