Detecting multiple replicating signals using adaptive filtering procedures

Replicability is a fundamental quality of scientific discoveries: we are interested in those signals that detectable different laboratories, populations, across time etc. Unlike meta-analysis which accounts for experimental variability but does not guarantee replicability, testing partial conjunction (PC) null aims specifically to identify the discovered multiple studies. In many contemporary applications, example, comparing high-throughput genetic experiments, large number M PC nulls need be tested simultaneously, calling comparisons correction. However, standard adjustments on p-values can severely conservative, especially when and sparse. We introduce AdaFilter, new procedure increases power by adaptively filtering out unlikely candidates nulls. prove AdaFilter control FWER FDR as long data studies independent, has much higher than other existing methods. illustrate application with three examples: microarray Duchenne muscular dystrophy, single-cell RNA sequencing T cells lung cancer tumors GWAS metabolomics.