5-Fluorouracil (5-FU) is a commonly employed chemotherapeutic agent in multiple cancers, including hepatocellular carcinoma (HCC). Understanding why some HCC cases respond poorly or are resistant to 5-FU is essential for advancing precision oncology and optimizing treatment strategies. We applied Weighted Gene Co-expression Network Analysis (WGCNA) on gene expression data from the GDSC2 cancer cell line collection to detect 5-FU-related co-expression modules and hub genes. Based on these hub genes, HCC samples were classified into subgroups, and predictive models were developed using ConsensusClusterPlus combined with five machine learning algorithms. Additionally, the expression of key genes in the model was validated via quantitative reverse transcription-polymerase chain reaction (qRT-PCR).
WGCNA identified 19 distinct gene modules in the cancer cell lines, with the midnight blue module showing the strongest inverse association with 5-FU response. Within this module, 45 hub genes were selected. HCC patients were categorized into three subtypes: C1, C2, and C3. C1 had the poorest overall survival (OS) and was marked by a higher clinical grade and advanced T stage and stage, while C3 exhibited the most favorable OS. C2 had intermediate OS and displayed the lowest immune cell infiltration. From the 45 hub genes, five—TOMM40L, SNRPA, ILF3, CPSF6, and NUP205—were chosen to construct a prognostic regression model for HCC. qRT-PCR confirmed that these genes were markedly overexpressed in HCC tissue samples. Stratification of HCC according to 5-FU sensitivity aligns with prognostic differences and reflects heterogeneity in genomic features, immune infiltration, and signaling pathways. The derived 5-FU-related risk model may serve as a valuable tool for individualized prognosis monitoring in HCC patients.
