Due to the limitation on the calculating power of the computer, it is very difficult to simulate the whole folding process and the large-scal functional motions of protein with all-atomic MD.A commonly used approach is called as the coarse-grained model, in which the residual degrees-of-freedom are reduced and the computational cost is decreased.The most difficult task in constructing a coarse-grained model is to derive the accurate and transplantable potential function for the model.Theoretical and experimental studies have shown that the sequences of proteins have been optimized during biological evolution, which results in the minimal energy frustration of protein folding.Consequently, the protein folding process is largely determined by its native structural topology and protein functional motions are also largely determined by its structural dynamic behavior.According to these facts, the potential functions of many coarse-grained models were constructed based on the native structure of proteins, avoiding the difficulty for optimizing the force field.The Gō model and elastic network model (ENM) belong to this type of the coarse-grained model.In the present study, the conventional Gō model and ENM were improved, and based on them, several theoretical methods were proposed to study protein folding/unfolding process and identify functional key residues from protein structure.The results are reported as following points:(1)The conventional Gō model was modified, in which the electrostatic interactions were taken into account.Then, the effect of electrostatic interactions on the thermo-stability and folding kinetics of cold shock protein was studied by using the modified model.(2)The role of intrinsic dynamical properties of protein structural topology on its unfolding process was investigated with the elastic network model.(3)An iterative use of ENM method was proposed to study protein unfolding process based on the conventional elastic network model.(4) An effective thermodynamic method was developed to identify the hotspot residues in the protein conformational transition based on the elastic network model .