503P Self-supervised AI reveals novel determinants of lethality in lung adenocarcinoma
outcome cluster in every dataset is defined by signatures of high immune population, high PD-L1, and low B-cells, all of which were confirmed at the protein level by mIF.It also shows an exclusion of epithelial T-cells visible in routine H&E.Spatially resolved transcriptomics reveals overexpression of stromal extracellular matrix programs, supporting this physical exclusion of T-cells from the tumour regions.Our findings suggest a key role for B-cells in tumour virulence and checkpoint inhibitor sensitivity, and we are now studying their broader influence on the immune microenvironment. Conclusions:We propose a robust new transcriptomic based classification of LUAD which defines 6 LUAD variants, each with key clinicopathological features, and which highlights the defining role of B-cells in tumour virulence.
Authors
- Lucas Farndale (ORCID: https://orcid.org/0009-0003-3667-2001)
- Kai Rakovic (ORCID: https://orcid.org/0000-0002-0676-1973)
- Alexandrina Pancheva (ORCID: https://orcid.org/0000-0002-3026-7078)
- John Le Quesne (ORCID: https://orcid.org/0000-0003-3552-7446)
- Dennis T. Chang
- Adalberto Claudio Quiros (ORCID: https://orcid.org/0000-0003-4804-0741)
- C. Miller
- K. Yuan
Institutions
- Cancer Research UK (GB)
- Cancer Research UK Scotland Institute (GB)
- CRUK Lung Cancer Centre of Excellence (GB)
- University of Glasgow (GB)
Publication Details
- Journal
- ESMO Open
- Published
- 2026-03-17
- DOI
- https://doi.org/10.1016/j.esmoop.2026.106818
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00