A protein can exhibit a variety of behaviors depending on its environment. Under physiological conditions, proteins spontaneously fold into functional structures. Exposing proteins to extreme temperatures or pressures often results in their denaturation. Single molecule force spectroscopy experiments have shown how mechanical forces can also disrupt a protein’s structure. In the research highlighted in this thesis, we used coarse-grained models of proteins and the principles of energy landscape theory to better understand these complex physical phenomena. Each chapter deals with a different aspect of protein folding behavior. We first discuss a study on protein folding. During evolution, protein sequences within a family undergo random mutation but are also under selection pressure to maintain their ability to fold into their common native structure. These processes together leave an imprint of the structure in the form of strong covariation between pairs of sites on the resulting family of evolved sequences. We leveraged these “evolutionary restraints” (ER) to improve the Associative memory, Water mediated, Structure and Energy Model (AWSEM), making it AWSEM-ER. The results from the AWSEM-ER model demonstrate that adding contacts predicted from sequence covariation significantly improves the quality of the predicted protein structures. In another study, we augmented the AWSEM model to allow for the exploration of protein folding and unfolding behavior as a function of temperature and pressure. The resulting model exhibits an elliptical phase diagram in the temperature-pressure plane for the two proteins studied, consistent with previously reported experimental findings. Finally, we predicted misfolded structures of a tandem construct of the designed protein Top7 using AWSEM. We then unfolded the predicted misfolded structures using steered molecular dynamics simulations of a structure based model to compare the force vs. extension curves with experiment. Non-trivial agreement was found between the number and extension length of unfolding transitions when unfolding the misfolded structures. This agreement suggests that AWSEM can accurately predict misfolded structures of proteins, which is a potentially important application where other models are typically either too costly (e.g., all-atom models) or just not applicable (e.g., native-centric models).