Comprehensive Characterization of RNA-Binding Proteins in Colon Adenocarcinoma Identifies a Novel Prognostic Signature for Predicting Clinical Outcomes and Immunotherapy Responses Based on Machine Learning

Background: RNA-binding proteins (RBPs) are crucial factors that function in the posttranscriptional modification process and are significant in cancer. Objective: This research aimed for a multigene signature to predict the prognosis and immunotherapy response of patients with colon adenocarcinoma (COAD) based on the expression profile of RNA-binding proteins (RBPs). Methods: COAD samples retrieved from the TCGA and GEO datasets were utilized for a training dataset and a validation dataset. Totally, 14 shared RBP genes with prognostic significance were identified. Non-negative matrix factorization clusters defined by these RBPs could stratify COAD patients into two molecular subtypes. Cox regression analysis and identification of 8-gene signature categorized COAD patients into high- and low-risk populations with significantly different prognosis and immunotherapy responses. Results: Our prediction signature was superior to another five well-established prediction models. A nomogram was generated to quantificationally predict the overall survival (OS) rate, validated by calibration curves. Our findings also indicated that high-risk populations possessed an enhanced immune evasion capacity and low-risk populations might benefit immunotherapy, especially for the joint combination of PD-1 and CTLA4 immunosuppressants. DHX15 and LARS2 were detected with significantly different expressions in both datasets, which were further confirmed by qRTPCR and immunohistochemical staining. Conclusion: Our observations supported an eight-RBP-related signature that could be applied for survival prediction and immunotherapy response of patients with COAD.