The low frequency and collective modes from Normal Mode Analysis (NMA) have been shown to be valuable in understanding sub-nanosecond protein conformational dynamics. While Molecular Dynamics (MD) is also utilized for such purposes, it is more computationally expensive. But, due to O( N³) scaling of eigen solvers, it also remains too computationally expensive to use NMA on large biomolecular system. Approximate NMA techniques are valuable in their ability to extend the size of the biomolecular system that can be modeled using NMA. A comparison between two approximate NMA methods was performed using Vibrational Subsystem Analysis (VSA) and Rotation-Translation Blocks (RTB). For the full system it takes 890 s to solve, using VSA it only took 40 s and with RTB 5 s. But the overlap between the vectors from VSA and and the full system is 0.99 and with RTB it was 0.69.

Protein atomic fluctuations can be probed by x-ray crystallography and characterized using Anisotropic Displacement Parameters (ADP). We assess the accuracy of coarse-grained and atomistic models that include the protein-environment interactions found in a crystalline environment. We use a coarse-grained Elastic Network Model (ENM) and an atomistic model using the EEF1 force-field from CHARMM using different boundary conditions to model the protein-environment interactions. We find that the optimal modeling of ADPs is achieved by weak protein-environment interactions as compared to internal interactions within a protein structure. Therefore, the internal dynamics of a protein is only weakly perturbed by crystal packing. We also find no improvement in the accuracy of ADP modeling by using the atomistic model over the coarse-grained ENM.