The high sequence conservation of druggable pockets of closely related proteins can make it challenging to develop selective inhibitors. We designed a new drug discovery approach that exploits both the static and dynamic differences of two orthologues. We applied it, as a proof of concept, to identify compounds that discriminate between the molecular chaperone Hsp90 of the protozoan pathogen Plasmodium falciparum (Pf) and that of its human host. We found that the ATP-binding pocket has a Pf-specific extension, whose sequence lining is identical in human Hsp90 but which differs by tertiary structure and dynamics. Using these insights for a structure-based drug screen, we discovered novel 7-azaindole compounds that exclusively bind the recombinant N-terminal domain of PfHsp90 but not of human Hsp90 nor of a PfHsp90 mutant with "human-like" dynamics. Moreover, these compounds preferentially inhibit the growth of yeast complemented by PfHsp90 and block the growth of Pf in culture.