Drug discovery strategies include from broad random screening to focussed target-based approaches. Structure and substrate information greatly enables target-based design, but this is limited to relatively few targets; cell-based screening can identify new targets but often suffers from low hit rates and difficult hit optimization. Thus, newer approaches are needed that can improve the efficiency of screening and hit optimization. Here, we describe an efficient approach for hit generation, which may be called "biofocussed chemoprospecting." With bio-likeness and ease of synthesis as priority criteria, libraries may be constructed with good optimization potential, physicochemical diversity, drug likeness and low cost. Following this approach, two libraries based on linear and cyclic dipeptide scaffolds were designed, first as virtual libraries comprising of more than 30000 compounds, and after subsequent filtering, as a small library of a total of 51 compounds. These provided good diversity at low cost, and were tested for bioactivities. The discovery of six active compounds demonstrates a hit rate greater than 10%. This is comparable to target-based approaches, but the "chemoprospecting" method described here has the additional potential to identify new targets and mechanisms.