A Falsifiable Synergy Model for Nk Cell Immunosuppression Under Combined Microgravity and Radiation Exposure: Mathematical Framework and Proposed Validation Protocol
Year: 2026
Type: conference-paper
Abstract: Deep-space missions expose crews to sustained microgravity and chronic galactic cosmic radiation together. Their separate effects on human immunity are documented; their combined effect on Natural Killer (NK) cells is not. After a systematic search of NASA OSDR and the biomedical literature we find no dataset measuring NK cytotoxicity under combined exposure, and no framework coupling the two stressors into a computable quantity. We present a framework modelling cytotoxic killing as a chain of multiplied stages — viable effector number, granule content, degranulation competence, and synaptic delivery efficiency — rather than as a single capacity. That structure is forced by data. Bigley et al. (2019) found NK killing halved in ISS crew while numbers, receptors, target binding, granule content and degranulation were all unaltered; no single-state model can represent this without rendering its own marker panel uninformative. Radiation acts on degranulation (Zarcone et al. 1989: abrogated at 30 Gy, binding intact, release blocked); microgravity acts on synaptic delivery, an assignment made by elimination and flagged as inference rather than observation. Synergy is expressed as radiosensitization, D_eff = D(1 + λ₃Ψ), so λ₃ = 0 is Bliss independence by construction and the hypothesis acquires a distinctive signature: a leftward shift of the radiation dose-response under microgravity. Fitted with λ₃ = 0, the framework reproduces the qualitative pattern reported by three laboratories and matches the single available quantitative constraint — Bigley's ~50% killing deficit — to within 16%. Three of the six tabulated residuals are satisfied by construction rather than by fit; we report them because a wrong architecture fails them, not because they constrain parameters. This is architecture selection, not parameter validation, and parameters are given as bounds. Two consequences are negative and stated plainly. The framework predicts that the transcript marker panel this project was built around cannot serve as a functional biomarker, since it measures a stage the data show is intact while killing halves; the machine-learning layer originally envisioned is therefore retired, replaced by a specification of what a future learning layer must ingest. Second, an in-silico design analysis shows dose placement dominates every other design choice: a grid at mission-relevant doses performs at chance regardless of sample size, while four arms placed near the degranulation midpoint reach full power with 24 wells. The deliverables are a mechanistically grounded model whose non-synergy behaviour reproduces real measurements, a falsifiable hypothesis, and a protocol capable of testing it.
Authors: Öz, Ekin, BİNGÖL, FIRAT
Language: English
Subfield: Physiology (medical)
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7224508759