Authors
Liu, K., Zhong, H., Elangovan, R. S., Sun, Y., Wang, L., Williams, A. P., Valerio, T. I., Furrer, C., Adams, J. P., Hoover, A. R., Chen, W. R.
Abstract
Most pancreatic ductal adenocarcinoma (PDAC) single-cell and spatial studies analyze one cohort or platform, obscuring recurrent biology. We assembled a human pancreas single-cell and single-nucleus reference of 1,186,130 cells from 19 studies and interpreted 176 Visium sections comprising 458,877 spots across non-diseased pancreas, chronic pancreatitis, PanIN, IPMN, primary PDAC and metastasis. Marker-supported labels were used after two RNA-based copy-number callers failed known-diploid controls. BANKSY domains, two reference-mapping methods and sample-level analyses resolved a cross-sectional epithelial axis extending from acinar-rich to malignant tissue. Five trajectory algorithms recovered similar ordering on a shared embedding; their consensus was interpreted as transformation-associated, not temporal or clonal. Malignant regions were globally segregated from fibroblast and myeloid compartments. Signed-distance analysis refined this pattern into a malignant core, a CAF/myeloid surround beginning at the tumour boundary and a more distal lymphoid compartment. Candidate extracellular-matrix communication, led by COLLAGEN, LAMININ and FN1, concentrated at the interface. Changes were reproduced in six patient-matched Normal-tumour pairs using exact patient-level tests. Visium HD resolved the same organization at single-cell resolution and showed that 8-um bins distorted immune-adjacency estimates. Xenium also revealed recurrent neighbourhoods but sample-specific stromal boundaries. We provide a confound-aware framework for identifying recurrent epithelial and microenvironmental organization in PDAC.
Preprint server:
bioRxiv
The authors list and abstract were imported from bioRxiv on 04 Aug 2026.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 28
- Comments 0