Abstract
Molecular targeted therapy has shown potential in hepatocellular carcinoma (HCC) patients, and immunotherapy applications are developing rapidly. However, clinical guidance for making individualized therapy decisions for HCC patients remains lacking. MDH (Medication Decision in HCC) gene signatures comprising 70 genes were screened using transcriptomic data from multikinase inhibitor (TKI)-resistant HCC cells and HCC patient-derived xenograft model (PDX) models. Four MDH subtypes with distinct biological and clinical characteristics were defined by unsupervised cluster analysis of HCC data from The Cancer Genome Atlas (TCGA) database. To facilitate individualized and reasonable clinical guidance for each HCC patient, we constructed the MDH score. Comprehensive analysis suggested high MDH scores were associated with TKI resistance, a high proportion of stromal cell infiltration and poor survival outcomes. We recommend concomitant stromal activity intervention and immunotherapy for this type of HCC. Moreover, low MDH scores indicate TKI sensitivity, and a combination of targeted and immunotherapy is recommended. The nomogram constructed by iteration least absolute shrinkage and selection operator (LASSO) Cox regression analysis successfully predicted 3- or 5-year survival outcomes and mortality risks of HCC patients. In conclusion, TKI resistance model-based MDH gene signatures provide novel insight into potential mechanisms of drug resistance and heterogeneity in HCC. Integrative analysis plus a simplified decision model may aid personalized treatment and prognostic assessment among HCC patients.