Abstract
Colorectal cancer (CRC) is one of the most common malignant tumors with complex molecular regulatory mechanisms. Alternative splicing (AS), a fundamental regulatory process of gene expression, plays an important role in the occurrence and development of CRC. This study analyzed AS Percent Spliced In (PSI) values from 49 pairs of CRC and normal samples in the TCGA SpliceSeq database. Using Lasso and SVM, AS features that can differentiate colorectal cancer from normal were screened. Univariate COX regression analysis identified prognosis-related AS events. A risk model was constructed and validated using machine learning, Kaplan-Meier analysis, and Decision Curve Analysis. The regulatory effect of protein arginine methyltransferase 5 (PRMT5) on poly(RC) binding protein 1 (PCBP1) was verified by immunoprecipitation experiments, and the effect of PCBP1 on the AS of Obscurin (OBSCN) was verified by PCR. Five AS events, including HNF4A.59461.AP and HNF4A.59462.AP, were identified, which can distinguish CRC from normal tissue. A machine learning model using 21 key AS events accurately predicted CRC prognosis. High-risk patients had significantly shorter survival times. PRMT5 was found to regulate PCBP1 function and then influence OBSCN AS, which may drive CRC progression. The study concluded that some AS events is significantly different in CRC and normal tissues, and some of these AS events are related to the prognosis of CRC. In addition, PRMT family-driven arginine modifications play an important role in CRC-specific AS events.