Integrative transcriptomic analysis identifies miR-642a-5p as a regulator of POFUT1 expression in colon cancer.

Colorectal cancer (CRC) is one of the most common and deadly cancers worldwide, underscoring the urgent need for novel biomarkers and therapeutic targets. Protein O-fucosyltransferase 1 (POFUT1) is increasingly recognized as an oncogenic driver in CRC, but the upstream regulatory mechanisms responsible for its overexpression remain poorly understood. We implemented an integrative, multi-step bioinformatics framework that combined multiple linear regression with stepwise selection, penalized regression models (LASSO and Elastic Net), Random Forest analysis for non-linear feature evaluation and correlation analysis. Five miRNAs (miR-92b, miR-484, miR-574, miR-642a, miR-3940) showing a consistent inverse association with POFUT1 across these complementary methods, were retained as final candidates. Among them, miR-642a exhibited the strongest negative correlation with POFUT1 and was then chosen to test its impact on POFUT1 expression in colon cancer cell lines. Overexpression of miR-642a-5p, through mimic transfection, significantly reduced POFUT1 transcript levels by approximately 46 % in HCT 116, 50 % in SW480, and 22 % in SW620 cells relative to controls. Dual-luciferase reporter assays confirmed direct binding of miR-642a-5p to two predicted sites within the POFUT1 3'UTR, resulting in a decrease of luciferase activity by 40 or 50 % according to the site. Our study combines bioinformatics screening based on multiple linear regression and machine learning models, with experimental validation, to identify a miRNA related to POFUT1 expression. This approach deserves to be extended to the other candidate miRNAs, which, combined with POFUT1, could serve as a basis for future diagnostic or therapeutic strategies in CRC.