Data sources

We retrospectively analyzed publicly available transcriptome profiling data (blood samples) and related clinical parameters from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Two public datasets, {"type":"entrez-geo","attrs":{"text":"GSE65682","term_id":"65682"}}GSE65682 (n=760, only 479 sepsis patients with complete data for survival status) and {"type":"entrez-geo","attrs":{"text":"GSE95233","term_id":"95233"}}GSE95233 (n=51), including data from 811 sepsis patients, were selected. Raw transcriptional expression profiles with clinical data were used as screening criteria, and the mean value was used when more than one probe was for a gene.

The Biological Information Database of Sepsis (BIDOS, http://www.swmubidos.com/), a database of transcriptome profiling data of patients with sepsis, was used to explore the diagnostic and prognostic roles of individual ERGs in sepsis. ERGs were acquired from the GeneCards Database (https://www.genecards.org/) [18], and 1,116 ERGs with a relevance score ≥ 7 were selected.