Screening Gene Expression-Related Alternative Splicing Event Signature for Colon Cancer Prognostic Prediction

Abstract

Colon cancer is a kind of common intestinal disease, and early diagnosis of colon cancer is crucial for patient’s prognosis. RNA alternative splicing (AS) is an RNA modification that affects cancer occurrence. RNA AS detection is promising to improve the in-depth understanding of the pathological mechanisms in colon cancer. In this study, differential analysis was performed to determine colon cancer-related AS events and the corresponding parental genes. Subsequently, GO functional annotation analysis was carried out on the parental genes, which revealed that these AS events might affect cell adhesion and cell growth. Besides, protein-protein interaction (PPI) network was established with the parental genes, in which MCODE was utilized to identify major functional modules. Enrichment analysis for the major functional module was implemented again, which demonstrated that these genes were mainly concentrated in the ribosome, protein ubiquitination, cell adhesion molecule binding, and other relevant biological functions. Next, differentially expressed genes (DEGs) were screened from colon cancer and normal tissues and overlapped with the parental genes, by which 55 gene expression-associated AS and the corresponding 45 genes were obtained. Moreover, a correlation analysis between splicing factors (SFs) and AS was done to identify interactions. On this basis, an SF-AS network was constructed. The univariate Cox regression analysis was employed to screen prognostic AS signature and establish a risk model. To assess the model, K-M and ROC analyses were done for model assessment, indicating the effective prediction performance. Combined with common clinicopathological features, the multivariate Cox regression analysis was conducted to confirm whether the risk model could be considered as an independent prognostic indicator. Finally, the expression status of the parental genes for the prognostic AS was evaluated between normal and colon cancer cells using qRT-PCR. In summary, TCGA SpliceSeq data were comprehensively analyzed, and a 5-AS prognostic model was constructed for colon cancer.