The Public Multi-Omics Data Landscape for Immune Disease

A multi-omics program is only as good as the data it can reach, and for immune-mediated disease the public landscape is richer — and more uneven — than most researchers assume.

The large cancer consortia (TCGA, CPTAC, ICGC/PCAWG, HTAN) remain the backbone for developing integration methods: they offer matched genome, transcriptome, proteome, and methylation at scale. But the resources that matter most for autoimmune work are the immune-specific cohorts — ImmPort (over 58,000 subjects of human immunology data with linked clinical metadata), the TEDDY birth cohort (a 15-year prospective multi-omic study of type 1 diabetes onset), and immune-cell eQTL maps such as DICE and OneK1K (1.27 million PBMCs across 982 donors) that let you interpret an autoimmune GWAS hit in terms of a specific cell type.

Atlases and population cohorts

On top of these sit the reference atlases — the Human Cell Atlas (~62M cells), CZ CELLxGENE Discover (a standardized Census of ~217M cells with a programmatic API), and HuBMAP's multi-modal spatial atlas — and the population multi-omics of UK Biobank, which now pairs whole-genome sequencing with Olink proteomics and NMR metabolomics across hundreds of thousands of participants. Open Targets ties much of this together into harmonized target–disease associations spanning genetics, expression, and drugs.

The honest asymmetry

Maturity is uneven, and it is worth being candid about it. Common, tissue-accessible, GWAS-powered diseases — RA, SLE, IBD, MS, T1D, psoriasis, atopic dermatitis — have cell-resolution atlases and consortium infrastructure (the AMP program, IMSGC, IIBDGC). Rarer or less tissue-accessible diseases — systemic sclerosis, Sjögren's, myasthenia gravis, CIDP, ankylosing spondylitis — remain largely single-institution and study-poor, with limited independent replication. Knowing which bucket your disease is in should shape the study you design.

How to reach it

Open processed data lives in GEO and ArrayExpress/BioStudies; controlled human data (raw sequence, germline, identifiable cohorts) is gated behind EGA and dbGaP data-access committees. Increasingly, standardized portals — the CELLxGENE Census API chief among them — make cross-dataset single-cell integration practical without re-harmonizing every study by hand. A maintained catalog of these resources, with scales and access terms, is the first asset any serious multi-omics program should build.