Background
Breast cancer (BC) is the most common malignancy among women in the world. Alternative splicing (AS) is an important mechanism for regulating gene expression and producing proteome diversity, which is closely related to tumorigenesis. Understanding the role of AS in BC may be helpful to reveal new therapeutic targets for clinical interventions.
Methods
RNA-seq, clinical and AS data of TCGA-BRCA were downloaded from TCGA and TCGA SpliceSeq databases. AS events associated with prognosis were filtered by univariate Cox regression. The AS risk model of BC was built by Lasso regression, random forest and multivariate Cox regression. The accuracy of the AS risk model and clinicopathological factors were evaluated by time-dependent receiver operating characteristic (ROC) curves. The significant factors were used to construct the nomogram model. Tumor microenvironment analysis, immune infiltration and immune checkpoint analysis were performed to show the differences between the high and low AS risk groups. The expression differences of genes of AS events constituting the risk model in tumor tissues and normal tissues were analyzed, the genes with significant differences were screened, and their relationship with prognosis, tumor microenvironment, immune infiltration and immune checkpoint were analyzed. Finally, Pearson correlation analysis was used to calculate the correlation coefficient between splicing factors (SF) and prognostic AS events in TCGA-BRCA. The results were imported into Cytoscape, and the associated network was constructed.
Results
A total of 21,232 genes had 45,421 AS events occurring in TCGA-BRCA, while 1604 AS events were found to be significantly correlated with survival. The BRCA risk model consisted of 5 AS events, (TTC39C|44853|AT*− 2.67) + (HSPBP1|52052|AP*− 4.28) + (MAZ|35942|ES*2.34) + (ANK3|11845|AP*1.18) + (ZC3HAV1|81940|AT*1.59), which were confirmed to be valuable for predicting BRCA prognosis to a certain degree, including ROC curve, survival analysis, tumor microenvironment analysis, immune infiltration and immune checkpoint analysis. Based on this, we constructed a nomogram prediction model composed of clinicopathological features and the AS risk signature. Furthermore, we found that MAZ was a core gene indicating the connection of tumor prognosis and AS events. Ultimately, a network of SF-AS regulation was established to reveal the relationship between them.
Conclusions
We constructed a nomogram model combined with clinicopathological features and AS risk score to predict the prognosis of BC. The detailed analysis of tumor microenvironment and immune infiltration in the AS risk model may further reveal the potential mechanisms of BC recurrence and development.