Data modeling the interplay between single-cell shape, single-cell protein expression, and tissue state.
Clinical Snapshot
PICO Framework
| P — Population | Heterogeneous human tissue samples spanning multiple cell types, diseases (including at least two cancers), and disease states, analysed at single-cell resolution via multiplexed imaging |
| I — Intervention | Computational framework integrating multiplexed imaging-derived single-cell shape features with protein expression data, including machine learning (ML) interpretability and graph neural network (GNN) modelling |
| C — Comparator | Population-averaged or single-marker analyses; baseline GNN disease-state prediction without shape features |
| O — Outcomes | Primary: bidirectional association between single-cell shape and protein expression across cell types and disease states. Secondary: identification of a p53-positive large-cell tumour subpopulation across two cancers; improvement in GNN-based disease-state classification when shape properties are incorporated |
Bottom Line
This computational study uses multiplexed imaging and machine learning to investigate the relationship between single-cell shape and protein expression in human tissues. The authors report a bidirectional, cross-disease association between cell morphology and protein expression, with a specific finding of large, p53-positive tumour cells across two cancer types. While the biological question is important and the methodological approach is innovative, the abstract provides insufficient quantitative detail to fully evaluate the strength or precision of the reported associations. The cross-sectional design precludes causal inference, confounder adjustment is not described, and the claim of universality requires independent external validation. The machine learning interpretability framework generates testable hypotheses rather than mechanistic proof. Clinical translation — including any diagnostic or prognostic application — is premature and would require prospective validation in well-characterised patient cohorts. For Australian clinicians and researchers, this work is of interest as a hypothesis-generating platform study, particularly in the context of precision oncology, but it does not yet support changes to clinical practice or pathological classification systems.
Key Findings
P Value: Not reported in abstract
Effect Size: Not reported in abstract; qualitative description only ('universal', 'amplified')
Primary Outcome: A universal, bidirectional association between single-cell shape and protein expression was identified across multiple cell types, diseases, and disease states, with the association amplified for cell state markers
Nnt Or Sensitivity: GNN disease-state prediction was enhanced by incorporation of shape properties; specific AUC, accuracy, or sensitivity/specificity values not reported in abstract. A subpopulation of large, p53-positive tumour cells was identified across two cancer types — prevalence and effect size not quantified in abstract.
Confidence Interval: Not reported
Clinical Application
Currently limited to research settings with access to multiplexed imaging platforms (e.g., imaging mass cytometry, CODEX, MIBI-TOF) and computational infrastructure for ML/GNN analysis. Clinical translation would require standardisation of imaging protocols, validated segmentation pipelines, and prospective outcome data. Not yet feasible for routine diagnostic or prognostic use. Australia has emerging multiplexed imaging capability at major academic cancer centres (e.g., Peter MacCallum Cancer Centre, Garvan Institute, QIMR Berghofer). The p53-positive large-cell tumour subpopulation finding, if validated, could be relevant to pathological classification and potentially to therapeutic stratification under TGA-approved p53-pathway-targeting agents. RACGP and COSA guidelines do not currently incorporate single-cell morpho-proteomic profiling. PBS listing of any diagnostic or therapeutic application derived from this work is distant and contingent on prospective clinical validation. The computational framework may have future relevance to Australia's national precision oncology initiatives (e.g., Zero Childhood Cancer, PrOSPeCT). Patients with solid tumours (at least two cancer types studied) and potentially other diseases characterised by tissue heterogeneity; specific cancer types, stages, and patient demographics not specified in the abstract
Abstract
While cell shape fundamentally governs tissue function, the underlying links between single-cell shape and protein expression have been difficult to resolve due to limitations in imaging multiplexing and population averaging. Here, we use multiplexed imaging and machine learning to investigate the coupling of cell shape and protein expression in heterogeneous human tissues. Our analysis establishes a universal, bi-directional link between a cell's shape and its protein expression across cell types, diseases, and disease states, amplified for cell state markers. Machine learning interpretability shows that shape features can potentially generate hypotheses of protein functions. Screening all protein-cell type pairs identified a subpopulation of large, p53-positive tumor cells across two cancers. Shape properties further enhanced graph neural network disease state prediction. Our results open the door to unraveling the intricate connections between protein expression, cell shape, tissue organization, and tissue state in a physiological context.
References
- 1.Tamir, Y., Bussi, Y., Owczarek, C., Luque, L., Torrisi, G., Rose, L. A., Kliper-Gross, O., Schumacher, L., Parsons, M., Keren, L., & Zaritsky, A. (2026). Data modeling the interplay between single-cell shape, single-cell protein expression, and tissue state. Cell Reports Methods. https://doi.org/10.1016/j.crmeth.2026.101463
Related Research
The American journal of pathology
An Integrated Machine Learning and Genomic Framework for Precise Detection of Gastric Cancer
3 Aug 2026
Current opinion in chemical biology
Prediction of protein-protein interactions and co-complex models with deep learning
3 Aug 2026
European journal of radiology
A CT-based deep learning model to differentiate between benign and malignant adrenal lesions
3 Aug 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service