How can we discover new bioactive compounds from complex mixtures, e.g. medicinal plant extracts, prior to any isolation? Natural products are an important source of drugs. However, their discovery is hindered by many challenges including the high rediscovery rate and time consuming work. Methods used so far fail to consider the structural novelty. This work describes comparative metabolomic approaches based on three independent analytical methods (1H NMR, 2D NMR and LC-MS) to detect, prior to any isolation, extracts containing novel compounds as revealed by multivariate data analysis and discriminatory MS/NMR signals. These new holistic methods were applied to compare 17 Hypericum species. Novel natural products from selected extracts were isolated guided by their NMR/MS signals. The investigation of 3 prioritized species resulted in the isolation and characterization of 23 new natural products.The structures of isolated compounds were established by NMR and HR-MS data supported by chemical synthesis and ECD calculations for absolute configuration determination. Notably, novel and significantly anti-HIV active compounds were found from H. roeperianum.