Classification of hepatocellular carcinoma (HCC) into clinical subtypes is in high demand for the selection of proper treatment. For patients with HCC who are not or no longer candidates for locoregional therapy, oral multi-kinase inhibitors are the only established systemic alternative. Yet, the response rate of these FDA-approved drugs against HCC are less than two percent, necessitating in-depth, personalized molecular characterization of the disease. Here, we adopted quantitative proteomic approaches in an attempt to characterize HCC subtypes in proteome level, and link these subtypes to drug response. We developed a liver reference proteome, a mixture of equal amounts of total proteins from seven liver cancer cell lines metabolically labeled with stable isotopes. This labeled reference proteome was utilized to cross-compare protein abundances of ten liver cancer cell lines that are widely used in research. Shotgun proteomic analysis on a hybrid quadrupole-orbitrap mass spectrometer with high mass accuracy at the MS and MS/MS levels yielded a liver proteome composed of 8,883 identified proteins, spanning 7 orders of magnitude in signal intensity. High-accuracy quantification allowed robust differentiation of HCC cell lines into three subtypes. Among these subtypes, one subtype displayed similar protein expression pattern as HCC tissues correlated with poor outcome, vascular invasion, high alpha-fetoprotein levels, and hepatitis B virus infection. The proteome profile of the other two subtypes did not mimic any of the catalogued proteome profile of HCC tissues. HCC characterization by protein expression may facilitate the establishment of clinical guidelines for subtype-specific systematic therapy.