Polygenic risk score for prediction of radiotherapy efficacy and radiosensitivity in patients with non-metastatic breast cancer

Objective

To construct a novel polygenic risk scoring model, in order to predict the benefits of radiosensitivity in patients with non-metastatic breast cancer (NMBC).

Methods

A total of 450 NMBC patients from The Cancer Genome Atlas (TCGA) were enrolled and randomly assigned 6:4 (training vs. validation). The empirical Bayes differential analysis was used to perform differential expression analysis, univariate Cox regression and Kaplan-Meier analysis were used to screen for prognosis-related genes. Finally, LASSO regression and stepwise regression were used to select key prognostic-related genes. We constructed a multivariate Cox proportional risk regression model using key genes. The pRRophetic function was used to predict drug sensitivity of radiosensitivity (RS) and radioresistance (RR) groups for adjuvant therapy.

Results

Eight genes (AMH, H2BU1, HOXB13, TMEM132A, TMEM270, ODF3L1, RIIAD1 and RIMBP2) were screened to build a polygenic risk scoring model. The region of characteristic (ROC) curves were drawn based on the 3-, 5- and 10-year overall survival (OS), with area under curves (AUCs) of 0.816, 0.822 and 0.806, respectively. RS and RR can be effectively distinguished according to the risk score of 2.004. Gene set enrichment analysis (GSEA) showed that necroptosis was significantly enriched in RS, while complement and coagulation cascade, JAK-STAT and PPAR signaling pathways were significantly enriched in RR. Alternatively, for those radioresistant patients, the chemotherapy drugs that might be more helpful are Cisplatin, Docetaxel, Methotrexate and Vinblastine with higher drug sensitivity.

Conclusion

The polygenic risk scoring model showed prediction for the benefit of radiotherapy in NMBC patients, which might be used to guide clinical practice.