Statistical Associating Fluid Theory (SAFT) equations of state (EoS) for mixtures require cross-interaction parameters. For real systems, combining rules, such as the Lorenz-Berthelot combining rules, have to be corrected using at least one binary interaction parameter, k(ij). Values of k(ij) are usually adjusted to experimental data of phase equilibria. Here, we correlate k(ij) to the pure component parameters of the Perturbed Chain - Statistical Associating Fluid Theory (PC-SAFT) EoS, using a Quantitative Structure Property Relationship (QSPR) model. The coefficients of the proposed QSPR model are regressed separately for mixtures with non-associating components and for mixtures with associating components. The QSPR model is validated using the statistical measures of the QSPR method. We compare the values of k(ij) that are estimated from the QSPR model to values of k(ij) estimated from London's dispersive theory. Phase equilibrium calculations carried out with these two approaches of estimating kid values are compared to experimental data. The estimation of k(ij) values as function of the pure component PC-SAFT parameters can be applied to problems of process design and in Computer Aided Molecular Design (CAMD), to allow for calculations that are reasonably accurate and independent from the availability of experimental mixture data. (C) 2015 Elsevier B.V. All rights reserved.