native-centric coarse-gained models, termed ca go models, have been widely used for computational simulation of small protein to study their folding kinetics and thermodynamics. the limitations of these models come from the lack of non-native interactions and neglect of the specificity of amino acid residues. on the other hand, the simulations of protein folding in atomistic details using accurate energy force field, termed ab-initio models, have been proven to be too computationally expensive, even with the most powerful computers. therefore, many new models at intermediate level have been developed, such as multi-bead model, go-like model, and all-atom go model. these improved go-like models retain some of the specificity of amino acids and more importantly are still able to fold proteins starting from completely unfolded states to their unique native structures.