This study analysed data from two recent large-scale pharmacogenomic studies: the Cancer Genome Project and the Cancer Cell line Encyclopedia. Overlapping data from these studies (15 drugs and 471 cell lines were tested in both studies) enabled the authors to show that genomic data were well correlated between studies, but the drug response data were very divergent. The authors note that this creates an obstacle to the use of data from individual studies to make or validate predictive models of drug response.