Hepatocellular carcinoma (HCC) is the sixth most common cancer and the fourth lethal cancer worldwide. We performed proteomic and metabolomic profiling of sera from 67 HCC patients who received pembrolizumab therapy and 15 healthy individuals. We then trained a machine learning model using proteomic and metabolomic measurements from a training cohort containing 26 patients with resistance to PD-1 treatment and 23 patients with response to PD-1 treatment. The model was validated using a testing cohort containing 7 patients with resistance to PD-1 treatment and 6 patients with response to PD-1 treatment, 12 of which were correctly classified. Our analysis reveals the proteomic and metabolic features of HCC patients which correlated with PD1 antibody therapy and provide serum predictive markers for selection of patients with potential response to PD1 therapy.