New atlas maps tumour microenvironment across 258 cancer datasets in humans and mice
Researchers built a large single-cell RNA sequencing atlas combining 2,855 human samples from 190 studies and 526 mouse samples from 68 studies, drawn from public repositories GEO and ArrayExpress spanning 2014 to mid-2025. The team applied a multi-step pipeline—quality control, batch-effect correction via Harmony, FastMNN or scVI, and supervised cell-type annotation using marker genes, copy number inference and classification tools—to integrate the datasets into unified atlases of the tumour microenvironment.