Abstract
<title>Abstract</title> <p>CONTEXT: Terminal non-guanine tetrads can markedly alter RNA G-quadruplex stability, but the energetic basis by which a reversed 3′-uridine cap stabilizes the guanine core remains unresolved. We developed coupled state-space inference (CSSI) to test whether backbone reversal, cap–core stacking and terminal cation organization act cooperatively. Across eight modeled terminal chemistries, the native 3′-U cap produced a posterior-predictive stabilization of 30.4°C relative to the uncapped core (95% credible interval 26.4–34.6°C). Donor-loss variants were strongly destabilized (9.6°C for 2′-deoxy-U and 9.3°C for 2′-O-methyl-U), whereas phosphorothioate retained an intermediate effect (17.7°C). The model resolved reversed and open basins, predicted register-selective stacking and identified a potassium–temperature boundary for reversed-cap occupancy. These values are calibrated computational predictions, not independent experimental measurements. The results support a cooperative clamp–stack–ion mechanism and provide falsifiable rankings for chemical testing. METHODS: CSSI represents each cap by backbone-reversal, stacking-slip, clamp-competence and terminal-ion-order coordinates in a reduced effective free-energy model containing additive and cooperative terms. Eight chemistries were encoded with transparent competence multipliers. Four energetic coefficients were sampled from normal priors, and 5,000 Monte Carlo draws per variant were generated with fixed seed 24072026. Posterior means and equal-tailed 95% credible intervals, module-ablation tests, variance-based sensitivity analysis, Boltzmann register probabilities at 310 K and a logistic potassium–temperature phase surface were calculated. Relative cap free energy was mapped to ΔTm with a stated empirical coefficient calibrated to the published native-cap scale. Calculations and plots used Python with NumPy, Matplotlib and seaborn; document assembly used python-docx.</p>