False discovery rate control for simultaneously significant features

It is usually of interest to identify genomic features are simultaneously significant in each of two datasets. For example, a user may be interested in whether genes that are differentially expressed in his or her own data are also differentially expressed in a related study publicly available from GEO. Typically this analysis is done by applying FDR corrections separately to the two studies and then taking the intersection of the results, but this procedure does not control the overall false discovery rate of final list of discoveries. Our method provides powerful discovery of simultaneously significant features and theoretically proven control of the overall false discovery rate. It is computationally scalable and can be applied to 10 million features in less than 1 minute.