A new approach to drug discovery is based on the generation of high diversity libraries of DNA, RNA, peptides or small molecules. Search of such libraries for useful molecules is an optimization problem on high-dimensional molecular fitness landscapes. We utilize a spin-glass-like model, the NK model, to analyze search strategies based on pooling, mutation, recombination and selective hill-climbing. Our results suggest that pooling followed by recombination and/or hill-climbing finds better candidate molecules than pooling alone on most molecular landscapes. Our results point to new experiments to assess the structure of molecular fitness landscapes and improve current models.