largeQvalue: A program for calculating FDR estimates with large datasets

Abstract

This is an implementation of the R statistical software qvalue package [Dabney et al., 2014], designed for use with large datasets where memory or computation time is limiting. In addition to estimating p values adjusted for multiple testing, the software outputs a script which can be pasted into R to produce diagnostic plots and parameter estimates. This program runs almost 30 times faster and requests substantially less memory than the qvalue package when analysing 10 million p values on a high performance cluster. The software has been used to control for the multiple testing of 390 million tests when analysing a full cis scan of RNA-seq exon level gene expression from the Eurobats project [Brown et al., 2014]. The source code and links to executable files for linux and Mac OSX can be found here: https://github.com/abrown25/qvalue. Help for the package can be found by running ./largeQvalue --help.