M. Sproull, Y. Fan, Q. Chen, D. Meerzaman, K. Camphausen
Radiation Research 202 (4), 697-705, (3 September 2024) https://doi.org/10.1667/RADE-24-00092.1
In future mass casualty medical management scenarios involving radiation injury, medical diagnostics to both identify those who have been exposed and the level of exposure will be needed. As almost all exposures in the field are heterogeneous, determination of degree of exposure and which vital organs have been exposed will be essential for effective medical management. In the current study we sought to characterize novel proteomic biomarkers of radiation exposure and develop exposure and dose prediction algorithms for a variety of exposure paradigms to include uniform total-body exposures, and organ-specific partial-body exposures to only the brain, only the gut and only the lung. C57BL6 female mice received a single total-body irradiation (TBI) of 2, 4 or 8 Gy, 2 and 8 Gy for lung or gut exposures, and 2, 8 or 16 Gy for exposure to only the brain. Plasma was then screened using the SomaScan v4.1 assay for ∼7,000 protein analytes. A subset panel of protein biomarkers demonstrating significant (FDR<0.05 and |logFC|>0.2) changes in expression after radiation exposure was characterized. All proteins were used for feature selection to build 7 different predictive models of radiation exposure using different sample cohort combinations. These models were structured according to practical field considerations to differentiate level of exposure, in addition to identification of organ-specific exposures. Each model algorithm built using a unique sample cohort was validated with a training set of samples and tested with a separate new sample series. The overall predictive accuracy for all models was 100% at the model training level. When tested with reserved samples Model 1 which compared an exposure group inclusive of all TBI and organ-specific partial-body exposures in the study vs. control, and Model 2 which differentiated between control, TBI and partials (all organ-specific partial-body exposures) the resulting prediction accuracy was 92.3% and 95.4%, respectively. For identification of organ-specific exposures vs. control, Model 3 (only brain), Model 4 (only gut) and Model 5 (only lung) were developed with predictive accuracies of 78.3%, 88.9% and 94.4%, respectively. Finally, for Models 6 and 7, which differentiated between TBI and separate organ-specific partial-body cohorts, the testing predictive accuracy was 83.1% and 92.3%, respectively. These models represent novel predictive panels of radiation responsive proteomic biomarkers and illustrate the feasibility of development of biodosimetry algorithms with utility for simultaneous classification of total-body, partial-body and organ-specific radiation exposures.