The computational modeling program HINT (Hydropathic INTeractions), an empirical hydropathic force field that includes hydrogen bonding, Coulombic, and hydrophobic terms, was used to model the free energy of dimer-tetramer association in a series of deoxy hemoglobin beta 37 double mutants. Five of the analyzed mutants (beta 37W -> Y, beta 37W - A, beta 37W - G, beta 37W - E, and beta 37W - R) have been solved crystallographically and characterized thermodynamically and subsequently made a good test set for the calibration of our method as a tool for free energy prediction. Initial free energy estimates for these mutants were conducted without the inclusion of crystallographically conserved water molecules and systematically underestimated the experimentally calculated loss in free energy observed for each mutant dimer-tetramer association. However, the inclusion of crystallographic waters, interacting at the dimer-dimer interface of each mutant, resulted in HINT free energy estimates that were more accurate with respect to experimental data. To evaluate the ability of our method to predict free energies for de novo protein models, the same beta 37 mutants were computationally generated from native deoxy hemoglobin and similarly analyzed. Our theoretical models were sufficiently robust to accurately predict free energy changes in a localized region around the mutated residue. However, our method did not possess the capacity to generate the long-range secondary structural effects observed in crystallographically solved mutant structures. Final method analysis involved the computational generation of structurally and/or thermodynamically uncharacterized beta 37 deoxy hemoglobin mutants. HINT analysis of these structures revealed that free energy predictions for dimer-tetramer association in these models agreed well with previously observed energy predictions for structurally and thermodynamically characterized beta 37 deoxy hemoglobin mutants.