A natural product collection and natural-product-derived combinatorial libraries were virtually screened for potential inhibitors of human 5-lipoxygenase (5-LO) activity. We followed a sequential ligand-based approach in two steps. First, similarity searching with a topological pharmacophore descriptor (CATS 2D method) was performed to enable scaffold-hopping. Eighteen compounds were selected from a virtual hit list of 430 substances, which had mutual pharmacophore features with at least one of 43 known 5-LO inhibitors that served as query structures. Two new chemotypes exhibited significant activity in a cell-based 5-LO activity assay. The two most potent molecules served as seed structures for a second virtual screening round. This time, a focused natural-product-derived combinatorial library was analyzed by different ligand-based virtual screening methods. The best molecules from the final set of screening candidates potently suppressed 5-LO activity in intact cells and may represent a novel class of 5-LO inhibitors. The results demonstrate the potential of natural-product-derived screening libraries for hit and lead structure identification.