The range of conformations of macromolecules and the dynamic interconversion between conformations is an important part of the relationship between structure and function. The existence of side-chain conformational substates in two systems is directly demonstrated using traditional X-ray crystallographic refinement methodologies. One system is a mutant met-myoglobin where the phenylalanine at position 46 was replaced by a valine. The other is a low-temperature high-resolution dataset for wild-type CO myoglobin, Multi-conformer refinements, which combine molecular dynamics with X-ray data restraints, are shown to model side-chain substates similar to those identified manually. The two conformations of his64 in the mutant myoglobin and wild-type myoglobin, and three conformations of ser117 in the wild-type myoglobin are found "automatically." Time averaged refinements, which also use a modified molecular dynamics algorithm but is a simulation rather than strictly an optimization, also found similar side-chain conformations. The time averaged refinement for the mutant myoglobin found 5 transitions of the distal histidine in only 30 nominal ps of simulation time.
The singular value decomposition (SVD), when coupled with the coefficient of kurtosis for the first right singular vector, is a simple but powerful "filter" for identifying bi-conformer side-chains from large ensembles of structures. Those residues identified with the SVD as having two discrete substates from the time-averaged refinement agree fairly closely with those found manually.
The SVD has also proven to be a powerful tool for analyzing conformational substates for the protein as a whole. The configuration space projection of a Ins solvated myoglobin MD simulation using the SVD shows a "beads on a string" motif, suggesting a hierarchical topology reminiscent of substates. Comparing the left singular vectors from two halves of a dynamics simulation shows that even when only C a atoms are used, the configuration spaces searched by the solvated Mb simulation do not match. This quantifies what can already be seen qualitatively in the configuration space portraits of the system. When time-averaged refinement trajectories are compared, there is a much higher degree of similarity indicating that the accelerated dynamics system is approaching ergodicity.