Determining unlike-pair interaction parameters, whether for group contribution equation of state or molecular simulations, is a challenge for the prediction of thermodynamic properties. As the number of components and their respective complexity increase, it becomes impractical to fit all the unlike interactions. Lorentz−Berthelot combining rules work well for systems, where the main interactions are dispersion forces, but they do not account for electrostatics. In this work, we derive predictive combining rules within the SAFT-γ-Mie framework. In the resulting model, the unlike-pair interactions account for the effect of ionization energies, partial charges, dipole moments, and quadrupole moments. We then estimate these properties for molecular fragments using density functional theory calculations and demonstrate their use to obtain realistic cross-interaction energies without the need for experimental data. An open-source python package, Multipole Approach to Predictively Scale Cross-Interactions, is included to facilitate use of the methods presented in this work. A good qualitative agreement was obtained for all phase equilibria calculations of binary mixtures containing carbon dioxide with propane, hexane, benzene, and water, as well as mixtures of hexane and benzene. Finally, we discuss future improvements to our methodology, including the use of physical insights when fitting self-interaction parameters.