Long-term safety assessments for nuclear waste disposal face considerable challenges due to uncertainties resulting from the complex geological, geochemical and environmental processes. This work focuses on enhancing the predictive capability of reactive transport models (RTM) for radionuclide migration in fluids within repositories in crystalline host rock. In particular, the work is focused on investigating the influence of uncertain parameters on radionuclide sorption behavior in crystalline rocks. This is achieved by means of systematic Global Sensitivity analysis (GSA) techniques. The distribution coefficient (Kd) is a key parameter quantifying sorption behavior, obtained by means of geochemistry databases. A Quasi Monte Carlo sampling of input parameters, including mineral composition, pH/Eh, and Uranyl concentrations, was employed to study their effects on Kd values. GSA identifies the important variables affecting the uncertainty in the assessment results. Two GSA methodologies where utilized in this work, namely CUSUNORO and High Dimensional Model Representation (HDMR). By performing CUSUNORO and HDMR together, we capture first-order non-linear and second-order effects, respectively, revealing interaction effects between input parameters on the distribution coefficient. Moreover, the compositional data sampling poses a challenge due to the interdependencies which can alter the results of sensitivity analysis. To address this, we implemented transformation techniques to mitigate the interdependency problem. Our findings contribute to a deeper understanding of these processes, providing valuable insights for enhancing the reliability and robustness of long-term safety assessments for nuclear waste disposal sites.