[en] Elastic network-based normal mode analyses (EN-NMA) of four pairs of open-closed proteins (Lactoferrin, Maltodextrin-binding protein, LAO-binding protein, and Adenylate kinase) were conducted using both all-atom and coarse-grained models. The results indicated that the performance of the all-atom model was similar to that of the coarse-grained model in terms of predicting the conformational changes of backbones. Moreover, dynamic behavior was examined by studying relative atomic displacements and shapes of the dominant mode. For instance, for Maltodextrin-binding protein, the results from the all-atom model differed from those of the coarse-grained model, especially for residues that are biologically relevant. The coarse-grained model has better computational efficiency than the allatom model. However, the former may misrepresent the key dynamics of a protein related to biological functions as a consequence of excessive coarse approximation. Considering that the current power even in a high-end personal computer is sufficient to handle most of protein structures with up to 1,000 residues in a reasonable manner, which can only be used with supercomputers a few decades ago, an all-atom-based EN-NMA may deserve more attention as a reliable and powerful computational tool for protein dynamics study over the conventional coarse-graining approach.