Integration of protein and metabolomic blood biomarkers enhances classification of radiation exposure in X-irradiated humanized mice

The general population is at risk of exposure to ionizing radiation due to nuclear warfare, terrorism, or radiological accidents. Such exposures can lead to acute radiation syndrome (ARS), a set of debilitating and often fatal health effects that start within the first week after irradiation. ARS is largely defined by injuries to the bone marrow and intestinal tract, two organ systems that are highly sensitive to the deleterious effects of ionizing radiation. There is currently no FDA-approved, minimally invasive assay to identify absorbed radiation dose in subjects suspected to have been exposed to ionizing radiation. Moreover, no methods are available for the early detection of radiation injuries, such that intervention strategies can be implemented before injuries become life-threatening. The primary objective of this study was to develop a multiparameter targeted biomarker approach to improve dose and injury assessment using X-irradiated humanized mice as a model.
Metabolomic and proteomic analyses were performed on plasma and leukocytes derived from blood collected 2 and 7 days after X-ray exposure (doses 0, 1, 2, and 3 Gy). The dose-dependent depletion of human blood cell lymphocytes (T and B cells) was used as a surrogate for bone marrow injury after radiation exposure. Multiple machine learning models were used for predictive modeling.
Multimodal integration of proteomic, lipidomic, and metabolomic analytes in blood cells (leukocytes) improved the performance metrics for binary classification (AUC > 0.95) on day 7 after exposure. Moreover, this integrated approach was able to distinguish between high-dose (2 and 3 Gy) and low-dose (1 Gy) exposure (AUC > 0.9), demonstrating its potential utility for biomarker-based assessment.
The findings highlight the potential for a multiparameter biomarker approach in human leukocytes for radiation exposure. With further validation studies, this approach could be leveraged for novel dosimetry algorithms for human use.