A 14-protein plasma signature identified by Pandya and colleagues allows for the prediction of lung cancer risk more than five years before a clinical diagnosis. Published June 4, 2026, in Cell, the study demonstrates how machine learning can detect molecular changes in the lung microenvironment, potentially enabling targeted prevention strategies before malignancy becomes visible.
How Molecular Signatures Outperform Traditional Screening
Current lung cancer screening relies heavily on age and smoking history, which often leaves light smokers or never-smokers without access to early detection. According to the study published in Cell, the new 14-protein model achieved an area under the curve (AUC) of 0.865. This performance surpasses established models like the Liverpool Lung Project version 3, which scored 0.806, and the LCRAT model, which reached 0.774.
The researchers analyzed data from 48,099 individuals within the UK Biobank. By combining clinical data—such as age, pack-years, and COPD history—with the 14-protein signature, they captured biological shifts occurring two to four years before a cancer diagnosis.
The 14-protein signature includes biomarkers like CXCL17, GDF15, and MMP12, which are linked to inflammation, extracellular matrix remodeling, and pulmonary surfactant biology.
The Link Between Inflammation and Tumor Initiation
This signature does not simply detect an established tumor. Instead, it reflects a perturbed lung environment that may promote cancer initiation. Pandya and colleagues found that the signature is enriched in alveolar type 2 cells and fibroblasts. These findings suggest that the risk of cancer is tied to the inflammatory and regenerative context in which lung cells exist, rather than just the presence of driver mutations.
Environmental factors play a clear role in this process. The study notes that particulate matter and smoking history elevate these specific protein levels. In preclinical models, the researchers observed that inhibiting interleukin-1β (IL-1β) could restrain the expansion of a specific “KAC” (keratin 8+/claudin 4+) alveolar transitional state, which acts as a bottleneck between lung injury and malignant transformation.
Can Biomarkers Improve Preventive Therapy?
The retrospective analysis of the CANTOS trial offers a potential roadmap for future clinical trials. Previously, the CANTOS trial showed that the drug canakinumab could reduce lung cancer incidence, but the high number of patients needed to treat made it difficult to apply to the general population.
Pandya and colleagues re-evaluated 4,651 CANTOS participants and found that the 14-protein signature effectively identified who would benefit from the therapy. In the high-signature group, canakinumab reduced the cumulative lung cancer incidence from 3.88% to 2.06%. In contrast, the low-signature group saw no meaningful reduction. This shift reduced the number needed to treat from 1,516 down to 55 for high-risk individuals.
While these results are promising, the authors emphasize that this blood test is not yet ready for routine clinical use. It serves as a scientific framework for future, prospectively validated prevention trials.
Frequently Asked Questions
Is this blood test available for patients now?
No. According to the study, these findings provide a scientific framework for future research and are not yet ready for routine clinical lung cancer prevention.
Why is this study different from standard liquid biopsies?
Most liquid biopsies look for tumor-derived signals like circulating tumor DNA, which require an established tumor. This signature detects the inflammatory, pre-cancerous microenvironment years before a tumor is clinically visible.
How does particulate matter affect lung cancer risk?
The research shows that particulate matter exposure influences the 14-protein signature, contributing to a lung environment that may favor tumor initiation by triggering inflammatory pathways.
Are you interested in the future of personalized oncology? Subscribe to our newsletter for the latest updates on molecular diagnostics and early detection research.
