Data Card: Sepsis Thrombocytopenia Whole-Blood RNA-seq Dataset (GSE273700)

Summary

Expression + sample metadata + feature metadata for GSE273700, a prospective RNA-seq study investigating transcriptomic differences between ICU sepsis patients with and without thrombocytopenia, including T-cell receptor (TCR) repertoire diversity assessed from RNA-Seq data via MiXCR and VDJTools.

Source accession

Accession	N (samples)	N (patients)	Alignment	Retrieval source
GSE273700	104	52 (22 thrombocytopenia, 30 without)	GRCh38, Ensembl gene IDs	NCBI GEO

All 52 patients have exactly 2 samples (day 1 and day 8) — a fully balanced paired design, no missing timepoints.

Files

• sample_metadata.parquet — one row per sample: sample_id, patient_id, thrombocytopenia (logical), timepoint (Day1/Day8)
• feature_metadata.parquet — one row per gene (GENEID, Ensembl gene ID): annotation from Ensembl v113 (SYMBOL, ENTREZID, GENEBIOTYPE, SEQNAME, GENESEQSTART, GENESEQEND, SEQSTRAND)
• expression.parquet — single file, long format: feature_id, sample_id, value

Feature annotation notes

Gene IDs are standard Ensembl format (ENSG..., no version suffix), 60,675 total. No specific information in the paper or acession on source or version of the annotation. I used Ensembl v113, the most recent from bioconductor AnnotationHub.

Expression value processing

Raw, un-normalized gene-level counts. No NCBI reprocessing exists for this accession — the counts matrix was downloaded directly from GEO's supplementary files (GSE273700_All_read_count_matrix.txt.gz). No filtering or normalization has been applied.

Provenance / reproducibility

See pull_gse273700.R.