ExactFDR: exact computation of false discovery rate estimate in case-control association studies

Abstract Summary: Genome-wide association studies require accurate and fast statistical methods to identify relevant signals from the background noise generated by a huge number of simultaneously tested hypotheses. It is now commonly accepted that exact computations of association probability value (P-value) are preferred to χ2 and permutation-based approximations. Following the same principle, the ExactFDR software package improves speed and accuracy of the permutation-based false discovery rate (FDR) estimation method by replacing the permutation-based estimation of the null distribution by the generalization of the algorithm used for computing individual exact P-values. It provides a quick and accurate non-conservative estimator of the proportion of false positives in a given selection of markers, and is therefore an efficient and pragmatic tool for the analysis of genome-wide association studies. Availability: A Java 1.6 (1.5-compatible) version is available on SourceForge: http://sourceforge.net/projects/exactfdr. Contact: Jerome.wojcik@merckserono.net Supplementary information: Supplementary data are available at Bioinformatics online