Metastasis is the major cause of hepatocellular carcinoma (HCC) mortality. But the effective biomarkers for HCC metastasis remain underexplored. Here we integrated GEO (Gene Expression Omnibus) and TCGA (The Cancer Genome Atlas) datasets to screen candidate genes for hepatocellular carcinoma metastasis, a consensus metastasis-derived prognostic signature (MDPS) was constructed by machine learning. Based on the risk scores, HCC patients were stratified into high-risk and low-risk groups. Comprehensive analyses were conducted to investigate various aspects including survival outcomes, clinical characteristics, immune cell infiltration, as well as in vitro experiments. Together, we develop a comprehensive machine learning-based program for constructing a consensus MDPS including four genes (SPP1, TYMS, HMMR and MYCN). Our findings revealed that four genes could serve as efficient prognostic biomarkers and therapeutic targets in HCC. In addition, in vitro experiments showed that HMMR overregulation exacerbated tumor progression, including proliferation, migration and invasion.
Keywords: Bioinformatics; Hepatocellular carcinoma; Machine learning; Metastasis; Prognosis.
© 2024. The Author(s).