An ensemble of protein structures can be obtained from multiple PDB models from the NMR or cryo-EM structures. It can also be assembled from homolog structure searches using either sequences or structures. Alternatively, one can be generated artificially from inverse fold methods. Then, these ensemble of structures need to be aligned using structural or sequence-based alignment methods such as MUSTANG, ClustalOmega, or MUSCLE. Principal-component analysis (PCA) or normal mode analysis (NMA) can be performed on the aligned structures, either coarse-grained (C α atoms only) or all atomic coordinates of the ensemble of structures. The eigenvectors from PCA or NMA represent the directions and the relative magnitudes of all points in the structure. The absolute magnitudes of motion need to be determined, and this constraint has been treated in different ways. The simplest way is to limit the deviations in the chemical bond lengths. The molecular motion and dynamics of protein reveal molecular functions and biological processes.
Deng et al. (2026) studied this question.