Proteins, the fundamental building blocks of organisms, carry out their biological functions after folding into their tertiary structures that are encoded in their linear chain of amino acids. Therefore, discovery of a protein's three-dimensional structure is vital, as it impacts many diverse biological fields, such as drug design through to synthetic biology. Our ability to predict the tertiary structure of a protein based on first principles requires an energy surface in which the functional native state is typically the global minimum. In my thesis, I have developed a volume-dependent implicit solvation model to describe a hydrophobic potential of mean force (HPMF) that models the hydrophobic interactions in a protein induced by the aqueous solvent. I have formulated a complete energy function composed of the AMBER empirical protein force field, the Generalized Born for the electrostatic component of the solvation energy, and the HPMF model. I demonstrate the superior performance of the HPMF model over traditional surface-area hydrophobic solvation models when applied to protein globular structure prediction, protein loop modeling and protein structure refinement. In the area of globular protein structure prediction, I show that my energy function outperforms all known structure prediction energy models by having the correct native ranking for 91% of the test proteins and the ability to discriminate native structures from decoys measured by much improved z-score values. When the energy function is tested on its ability to predict native protein loops, our model is found to be superior for the more difficult (and more biologically relevant) loop lengths of 8-residue and greater, and in fact the quality of the model shows little sensitivity to the length of the target loops. For the regime of structural refinement, our energy function successfully refines all of the selected template models from the submitted predictions for the recent CASP targets with improvement up to 1.31 Å with respect to the start models' root mean square deviations to the experimental native structures. One of the goals of structural biology is to efficiently obtain protein native structures. We hope that this energy function provides a model that can generate results that are complementary to the experimental techniques of x-ray crystallography and NMR and provide new insights into a protein's functional states.