Identification of a prognostic gene signature based on invasion-related genes in breast cancer

Purpose: This study focused on establishing an invasion-related prognosis prediction model for breast cancer (BC). Methods: : mRNA expression profiles and corresponding clinical information were collected from BC patients in The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. Thereafter, we identified invasion-related genes based on from Cancer SEA database. Invasion-related differentially expressed genes (DEGs) were identified through differential expression analysis. In addition, a risk model was built on the basis of univariate Cox regression and least absolute shrinkage and selection operator (LASSO) analysis for the TCGA cohort. Moreover, GEO-derived BC patients were used in verification. Besides, relation of risk model with tumor microenvironment was analyzed using the ESTIMATE and CIBERSORT algorithms. This was verified using Quantitative Real-time PCR (RT-qPCR). Results: : Altogether75 invasion-related DEGs were identified in BC versus control samples. KRT19, PSME2, HMGB3, MRPL13, and SHCBP1 were identified as prognostic signatures for the risk model. In line with the signature-based risk scores, we classified all patients as low- or high-risk group. In training and validation sets, Kaplan-Meier survival and receiver operating characteristic (ROC)analyses verified that our as-constructed 5-gene signature performed well in prediction. MRPL13 and KRT-19 showed significantly increased expression, whereas SHCBP1 showed decreased expression in BC samples compared to that in normal samples. The ESTIMATE and CIBERSORT algorithms revealed different immune statuses of both riskgroups. Conclusion: According to our findings, the prognosis prediction model constructed by incorporating 5 invasion-related genes is feasible in predicting BC prognosis.