A field-wide assessment of differential RNAseq reveals ubiquitous bias

We analyzed the field of expression profiling by high throughput sequencing, or RNA-seq, in terms of replicability and reproducibility, using data from the NCBI GEO (Gene Expression Omnibus) repository. Our work puts an upper bound of 56% to field-wide reproducibility, based on the types of files submitted to GEO. Archived dataset contains following files: - output/parsed_suppfiles.csv, p-value histograms, histogram classes, estimated number of true null hypotheses (pi0). - output/document_summaries.csv, document summaries of GEO series - output/publications.csv, publication info of GEO series - output/scopus_citedbycount.csv, Scopus citation info of GEO series - output/single-cell.csv, single cell experiments - spots.csv, sequencing run metadata: number of spots and bases - suppfilenames.txt, list of all supplementary file names of GEO submissions. One filename per row. - suppfilenames_filtered.txt, list of supplementary file names used for downloading files from NCBI GEO. One filename per row.