A pathway-guided strategy identifies a metabolic signature for prognosis prediction and precision therapy for...

Hepatocellular carcinoma (HCC) is a highly lethal and heterogeneous disease with a poor prognosis and no effective treatments. Herein, we presented a pathway-guided computational framework to establish a metabolic signature with the capacity for HCC prognosis prediction. By using the TCGA dataset as a training cohort (n = 365), we built an eight-gene (ACADS, ALDH1A2, FTCD, GOT2, GPX7, HADHA, LDHA and UGT2A1) risk score called the MGP score from the 20 metabolic pathways downregulated in HCC. The robustness of the MGP model was successfully validated in seven other independent cohorts (LIRI-JP, n = 231; Chinese, n = 159; GSE148355, n = 33; GSE14520, n = 225; GSE54236, n = 81; E-TABM-36, n = 41; and qPCR, n = 126). Moreover, three subtypes, L, H1 and H2, with distinct clinical outcomes were further stratified by using 761 HCC patients in the combined RNA-Seq cohort. Further analysis identified strong negative associations between metabolic pathways and other molecular features, including immune infiltration, expression of immune checkpoint genes, and hypoxic conditions, among the three subtypes. In 81 liver cancer cell lines, the MGP score indicated sensitivity to three preclinical agents (erastin, piperlongumine and PI-103), which may have potential therapeutic implications for the high-MGP score subtypes H1 and H2. Overall, our analysis highlights the potential of applying the MGP score for prognosis prediction and precision therapy for HCC.