TITLE: “P-value-free solution to fix exaggerated false positives by popular differential expression methods.”
ABSTRACT: We report a surprising phenomenon that popular bioinformatics methods for identifying differentially expressed genes (DEG) between two conditions have unexpectedly high false discovery rates (FDRs) on large-sample-size RNA-seq datasets. Failed FDR control is likely due to the invalid p-values which rely on unrealistic assumptions. To address this issue, we use a general statistical framework Clipper to control the FDR in DEG analysis without relying on p-values.
