Proteins can undergo large conformational changes upon ligand binding. Knowledge of the bound form of a protein is critical for many computational applications, ranging from the functional characterization of proteins to drug discovery. However, traditional approaches like Molecular Dynamics or Monte Carlo simulations are computationally intensive and are often not suitable to capture large scale, collective conformational changes. To address this problem, we combine the Elastic Network Model to rapidly generate ensembles of conformations and a Molecular Interaction Fields approach to select conformations that closely resemble the bound form. Molecular Interaction Fields are a class of energy-based methods that characterize a protein structure using virtual chemical probes, yielding 3D maps of the interaction energy profile of the protein. As a proof of concept, we illustrate the use of our computational pipeline on a dataset of 11 structures that undergo large conformational changes upon binding. The results indicate that overall our method is capable of returning conformations that are significantly much closer to the bound form than the initial unbound conformations.