Purpose: To evaluate the prognostic effect of the integration of genomic and transcriptomic profiles in breast cancer.
Methods: Eight hundred and ten samples from the Cancer Genome Atlas (TCGA) data sets were randomly divided into the training set (540 subjects) and validation set (270 subjects). We first selected single-nucleotide polymorphism (SNPs) and genes associated with breast cancer prognosis in the training set to construct the prognostic prediction model, and then replicated the prediction efficiency in the validation set.
Results: Four SNPs and three genes associated with the prognosis of breast cancer in the training set were included in the prognostic model. Patients were divided into the high-risk group and low-risk group based on the four SNPs and three genes signature-based genetic prognostic index. High-risk patients showed a significant worse overall survival [Hazard Ratio (HR) 9.43, 95% confidence interval (CI) 3.81-23.33, P < 0.001] than the low-risk group. Compared to the model constructed with only gene expression, the C statistics for the signature-based genetic prognostic index [area under curves (AUC) = 0.79, 95% CI 0.72-0.86] showed a significant increase (P < 0.001). Additionally, we further replicated the prognostic prediction model in the validation set as patients in the high-risk group also showed a significantly worse overall survival (HR 4.55, 95% CI 1.50-13.88, P < 0.001), and the C statistics for the signature-based genetic prognostic index was 0.76 (95% CI 0.65-0.86). The following time-dependent ROC revealed that the mean of AUCs were 0.839 and 0.748 in the training set and the validation set, respectively.
Conclusions: Our findings suggested that integrating genomic and transcriptomic profiles could greatly improve the predictive efficiency of the prognosis of breast cancer patients.
Keywords: Breast cancer; Genomic; Prognostic model; Transcriptomic.