Background: Immune-related genes possess promising prognostic potential in multiple
cancer types. Here, we describe the development of an immune-related prognostic signature for
predicting prostate cancer recurrence.
Methods: Prostate cancer gene expression profiles for 477 prostate cases, as well as accompanying
follow-up information were downloaded from The Cancer Genome Atlas (TCGA) and GEO. The
samples were divided into 3 groups and immune gene sets significantly associated with prognosis
were identified by evaluating the relationship between the expression of 1039 immune genes and
prognosis in the training set. Relative expression levels of these genes were used to identify
prognostic gene pairs. LASSO was used for feature selection and robust biomarkers selected.
Finally, the identified immune prognostic markers were validated using dataset and GEO
validation dataset and their performance compared with existing prognostic models.
Results: In total, 87 immune genes, significantly associated with prognosis, were identified and
2447 immune gene pairs (IRGPs) established. Univariate survival analysis identified 641
prognosis-associated immune gene pairs. 8-IRGPs were obtained via LASSO feature selection and
an 8-IRGPs signature established. The 8-IRGPs signature exhibited an independent prognosis
value in prostate cancer of the training set, test set, and external validation set (p = <0.001). The 5-
year survival AUC in both the training set and the validation set was >0.7. The 8-IRGPs
outperformed clinical tumor classification features, including T, N, radiation therapy (RT) and
targeted molecular therapy (TMT) (p <0.01). In addition, we compared the prognostic
characteristics of 8-IRGPs with 3 reported prostate cancers and found that 8-IRGPs achieved a
high C index (0.85) and had the highest predictive performance within 10 years of follow-up (HR:
10.5). Finally, we integrated T, N, RT, TMT, and 8-IRGPs and generated a novel alignment chart
to aid the prediction of prostate cancer recurrence in individual patients (p <0.01).
Conclusion: Here, we identified an 8-IRGP novel prognostic signature for the prediction of
prostate cancer recurrence.