A harmonized cross-disease compendium of 3,676 public blood transcriptomes for type I interferon scoring in lupus, Sjögren's disease, systemic sclerosis and myositis against shared healthy controls

Harmonized compendium of public whole-blood and PBMC transcriptomes from systemic lupus erythematosus, Sjögren's disease, systemic sclerosis and idiopathic inflammatory myopathy, with rheumatoid arthritis and pulmonary arterial hypertension comparators and healthy controls, built from 16 GEO and ArrayExpress series (Affymetrix 3 prime, HuEx, Clariom S, HTA 2.0, Illumina and RNA-seq). 3,680 samples were curated; 4 failed quality control, leaving 3,530 whole-blood samples (harmonized on 9,190 genes) and 146 PBMC samples (15,332 genes). The record contains per-series gene matrices (GSE88884 in three frozen-RMA parts plus reference), uncorrected and ComBat matrices by compartment, the curated annotation with quality statistics, inferred sex, module scores and interferon-high calls on within-series and harmonized healthy-control references, and validation outputs. Every series was normalized separately, within its own platform, and series were never normalized jointly. The ComBat (harmonized) matrices are provided for exploration and visualization only. For inference, analyze within each series or batch (per-series or per-batch matrices, or uncorrected matrices with series as a factor) and combine series by random-effects meta-analysis of within-series effect sizes. Do not pool differently normalized batches (for example, correlating genes or scores across the pooled samples), because between-series differences in normalization and platform can create or hide associations.