Heterogeneity characterization of hepatocellular carcinoma based on the sensitivity to 5-fluorouracil and development of a prognostic regression model.

5-Fluorouracil (5-FU) is a widely used chemotherapeutic drug in clinical cancer treatment, including hepatocellular carcinoma (HCC). A correct understanding of the mechanisms leading to a low or lack of sensitivity of HCC to 5-FU-based treatment is a key element in the current personalized medical treatment. Weighted gene co-expression network analysis (WGCNA) was used to analyze the expression profiles of the cancer cell line from GDSC2 to identify 5-FU-related modules and hub genes. According to hub genes, HCC was classified and the machine learning model was developed by ConsensusClusterPlus and five different machine learning algorithms. Furthermore, we performed quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis on the genes in our model. A total of 19 modules of the cancer cell line were divided by WGCNA, and the most negative correlation with 5-FU was the midnight blue module, from which 45 hub genes were identified. HCC was divided into three subgroups (C1, C2, and C3) with significant overall survival (OS) differences. OS of C1 was the shortest, which was characterized by a high clinical grade and later T stage and stage. OS of C3 was the longest. OS of C2 was between the two subtypes, and its immune infiltration was the lowest. Five out of 45 hub genes, namely, , , , , and , were filtered to develop a risk regression model as an independent prognostic indicator for HCC. The qRT-PCR results showed that , , , , and were remarkably highly expressed in hepatocellular carcinoma. The HCC classification based on the sensitivity to 5-FU was in line with the prognostic differences observed in HCC and most of the genomic variation, immune infiltration, and heterogeneity of pathological pathways. The regression model related to 5-FU sensitivity may be of significance in individualized prognostic monitoring of HCC.