The structural and functional properties of a novel protein sequence are predicted by modeling its structure using homology based or ab-initio based modeling algorithms. The various approximations within these modeling algorithms can cause inaccuracies in the modeled structure, often characterized based on the root mean square deviation (rmsd) from the native structure. A threshold rmsd value has not been defined to represent a criterion for acceptance or accuracy in terms of the predicted structural and functional properties of modeled protein. Since functional properties of a protein are often directly related to its binding affinity for a substrate and active site atoms contribute significantly to the binding affinity, we wish to derive a relationship between the rmsd of the active site of a protein and its estimated free energy of binding to the substrate. This analysis identifies the accuracy required in a modeling effort in order to achieve precise modeling of biophysical properties. In this work, we have chosen structurally diverse protein-ligand complexes. We have generated structures corresponding to different rmsd values within the active site by applying molecular dynamics to protein structures with rmsd constrained active site atoms. The binding free energy of ligands complexed with restructured active site protein models have been determined using an MM-PBSA method along with other rigorous approaches. In summary, our work provides a physicochemical criterion for determining the structural modeling accuracy required for the precise calculation of functional properties