Abstract
Hepatocellular carcinoma (HCC) remains a global medical problem. Programmed cell death protein 1 (PD-1) is a powerful weapon against many cancers, but it is not sensitive to some patients with HCC. We obtained datasets from the Gene Expression Omnibus (GEO) database on HCC patients and PD-1 immunotherapy to select seven intersecting DEGs. Through Lasso regression, two intersecting genes were acquired as predictors of HCC and PD-1 treatment prognosis, including HAMP and FOS. Logistic regression was performed to build a prediction model. HAMP had a better ability to diagnose HCC and predict PD1 treatment sensitivity. Further, we adapted the support vector machine (SVM) technique using HAMP to predict triple-classified outcomes after PD1 treatment in HCC patients, which had an excellent classification ability. We also performed external validation using TCGA data, which showed that HAMP was elevated in the early stage of HCC. HAMP was positively correlated with the infiltration of 18 major immune cells and the expression of 2 important immune checkpoints, PDCD1 and CTLA4. We discovered a biomarker that can be used for the early diagnosis, prognosis and PD1 immunotherapy efficacy prediction of HCC for the first time and developed a diagnostic model, prognostic model and prediction model of PD1 treatment sensitivity and treatment outcome for HCC patients accordingly.