Eradication of malaria requires a novel type of drug that blocks transmission from the human to the mosquito host, but selection of such a drug is hampered by a lack of translational models. Experimental mosquito infections yield infection intensities that are substantially higher than observed in natural infections and, as a consequence, underestimate the drug effect on the proportion of mosquitoes that become infected. Here we introduce a novel experimental and computational method to adequately describe drug efficacy at natural parasite densities. Parameters of a beta-binomial infection model were established and validated using a large number of experimental mosquito infections at different parasite densities. Analyses of 15 experimental and marketed drugs revealed a class-specific ability to block parasite transmission. Our results highlight the parasite's elongation factor EF2, PI4 kinase and the ATP4 sodium channel as key targets for interruption of transmission, and compounds DDD107498 and KAE609 as most advanced drug candidates.