Accessing the rich source of compounds from natural herbs for use in the pharmaceutical industry using conventional bioassay-based screening platforms has low efficiency and is cost-prohibitive. In this study, we developed a new method involving traditional Chinese medicine (TCM) molecular networking and virtual screening coupled with affinity mass spectrometry (MN/VS-AM) for the efficient discovery of herb-derived ligands. The in silico MS/MS fragmentation database (ISDB) generated by molecular networking of TCM can rapidly identify compounds in complex herb extracts and perform compound activity mapping. Additionally, the pre-virtual screening conveniently includes candidate herbs with potential bioactivity, while affinity MS screening completely eliminates the requirement for a tedious pure compound preparation at the initial screening phase. After applying this approach, two types of compounds, isoamylene flavanonols and 20(s)-protopanoxadio saponins, which were confirmed to interact with the small GTPase of Ras, were successfully identified from a dozen anti-cancer TCM herbs. The results demonstrate that the modified screening strategy dramatically improved the accuracy and throughput sensitivity of ligand screening from herbal extracts.