Longitudinal Plasma Proteomics Reveals an Immuno-thrombotic Signature That Predicts Radiation Pneumonitis in Lung...

Radiation pneumonitis (RP) is a dose-limiting toxicity in lung cancer radiation therapy, often poorly predicted by static clinical and dosimetric models. We aimed to identify a robust, blood-based proteomic signature grounded in the longitudinal biological response to radiation to enable accurate, early risk stratification.

We conducted a prospective study using high-throughput plasma proteomics on 267 longitudinal samples from a discovery cohort of 57 patients with lung cancer (non-small cell lung cancer and small cell lung cancer). To ensure statistical rigor and eliminate data leakage, feature selection was performed strictly within a 70% training partition. Linear mixed-effects models and Weighted Gene Co-expression Network Analysis were used to identify protein trajectories associated with symptomatic RP (Common Terminology Criteria for Adverse Events v5.0 grade ≥ 2). An Elastic Net machine learning model was developed and evaluated on an independent test set. The 10-protein signature was subsequently validated via enzyme-linked immunosorbent assay in a large, independent external cohort (n = 320).

Longitudinal analysis identified a dysregulated immuno-thrombotic axis as a central driver of RP, characterized by significant enrichment in "Platelet Activation" and "Serpin" pathways. In patients experiencing grade ≥ 2 RP, we observed a "systemic exhaustion" profile marked by the progressive decline of key anticoagulant and anti-inflammatory proteins. The 10-protein model achieved an unbiased area under the curve of 0.744 (95% CI: 0.622-0.861), outperforming a baseline clinical model (area under the curve = 0.679). External validation in 320 patients confirmed the robustness of a 4-protein clinical core (PROZ, SERPINA7, SERPINA6, and HAGH), with Pvalues ranging from 10^-8 to 10^-33. This signature remained a significant independent predictor (P < .05) after adjusting for mean lung dose, V20, and immunotherapy status, and was distinct from markers of overall survival.

A systemic immuno-thrombotic signature accurately predicts RP across different histologies and treatment regimens. This validated 4-protein core provides a clinically transportable tool for early risk stratification and personalized toxicity mitigation.