fdrci: FDR confidence interval selection and adjustment for large-scale hypothesis testing

Abstract

Abstract Motivation Approaches that control error by applying a priori fixed discovery thresholds such as 0.05 limit the ability of investigators to identify and publish weak effects even when evidence suggests that such effects exist. However, current false discovery rate (FDR) estimation methods lack a principled approach for post hoc identification of discovery thresholds other than 0.05. Results We describe a flexible approach that hinges on the precision of a permutation-based FDR estimator. A series of discovery thresholds are proposed, and an FDR confidence interval selection and adjustment technique is used to identify intervals that do not cover one, implying that some discoveries are expected to be true. We report an application to a transcriptome-wide association study of the MAVERICC clinical trial involving patients with metastatic colorectal cancer. Several genes are identified whose predicted expression is associated with progression-free or overall survival. Availability and implementation Software is provided via the CRAN repository (https://cran.r-project.org/web/packages/fdrci/index.html). Supplementary information Supplementary data are available at Bioinformatics Advances online.