Researchers analysed more than 15 million cells from 3,381 samples in a single atlas covering 42 types of solid tumour in humans and mice. Led by Michael Longaker at Stanford University, the study was published in Nature on 18 November 2026. It finds recurring patterns in the cells surrounding tumours and brings together approximately 15.4 million cells.
The team assembled publicly available single-cell RNA-sequencing studies published or deposited between 2014 and 2025. Earlier research often examined each cancer separately, making it harder to distinguish features shared across organs from those specific to one tumour type. The atlas compares cellular states across tissues and species. It documents broadly comparable supporting-cell populations, while also letting researchers examine tissue-specific features.
The central finding is that many tumours contain similar populations of supporting cells. Héctor Peinado, a researcher at the Spanish National Centre for Biotechnology (CNB-CSIC) in Madrid who was not involved, estimates that tumour cells make up 45% to 55% of a tumour. The remainder belongs to its microenvironment.
That surrounding ecosystem includes fibroblasts, which maintain the structural framework of organs. Endothelial cells form abnormal blood vessels, and the study also identifies immune-cell populations associated with tumour progression. Rather than treating fibroblasts as a single group, the researchers classified cancer-associated fibroblasts into seven states. These included inflammatory cells marked by CXCL8 or ISG15 and myofibroblastic cells marked by LRRC15 or COL4A1. They also identified specialised stromal populations marked by PI16 or CXCL14, alongside antigen-presenting fibroblasts.
One fibroblast subtype stood out: ISG15-positive inflammatory cancer-associated fibroblasts were strongly enriched in tumours and detected across stages of neoplastic growth. Their gene-expression patterns were consistent with interferon-gamma signalling from T cells. In the clinical datasets analysed, their presence correlated with significantly shorter survival. The authors say this population could become a therapeutic target. That remains a possibility raised by the findings, not an established treatment; the survival association does not show that the cells cause poorer outcomes.
The shared patterns fit a broader role for the local environment in determining whether altered cells develop into a tumour. Toni Gabaldón, an ICREA researcher at IRB Barcelona and the Barcelona Supercomputing Center who did not take part in the study, says some cells with tumour-related alterations never form tumours when their surroundings inhibit growth. Limited access to oxygen or nutrients can be one constraint.
The study links some recurring cell populations with poorer patient prognoses, but Longaker cautions that the analysis establishes correlation rather than cause. To test whether the atlas's predictions could be found in tissue, the team examined pancreatic and colon tumour samples for a fibroblast type capable of presenting antigens to immune cells. Thyroid tumour samples were also examined. In laboratory work, the team exposed pancreatic tumour cells to inflammatory signals to see whether they adopted predicted profiles. These analyses add experimental observations to the comparative atlas, but do not establish that the identified cell populations cause tumour progression.
Similarities between human and mouse tumours may help researchers interpret studies using mice. The Stanford team hopes the atlas can inform combined strategies aimed at tumour cells and the surrounding tissue. RUSSPAIN has also covered a different technology story involving Craviotto's electric car.