Most cancer drugs are effective in only a fraction of the patients who take them. If this responsive subset could be identified in advance, patients would be spared useless, cytotoxic treatments, and the regulatory pathway for new agents would be simplified. For a handful of drugs, efficacy is known to track closely with the presence of particular mutations, but, overall, the use of tumor genotype data to predict optimal drug regimens is still at an early stage. Two new reports have addressed this problem by pharmacological screening on human cancer cell lines at a larger scale than has been carried out previously. Garnett et al. tested 130 drugs on 639 cell lines, whereas Barretina et al. collected sequence, chromosomal copy number and transcriptional data on 947 cell lines and measured the activity of 24 drugs on 479 of these lines. Both studies uncovered known markers of drug sensitivity as well as novel candidate markers. Moreover, established correlations between drug efficacy and genotype were found to be invalid in some cases. (Nature 483, 570–575 and 603–607, 2012) KA