About your paper "Normal modes for predicting protein motions". Probably, I miss something when I'm reading... I didn't understand how you get the real (observed) motions for residues? Did you perform MD simulation for all proteins in your set?
Regarding my "Normal modes" paper: the “real (observed) motions for residues” are not
actually “real motions” – as long as we could find two substantially different conformations for the same protein in PDB, we assumed that such motion (1st conformation and 2nd one) could potentially take place. We didn’t do any MD simulations in that paper (although, we did possess technologies that would allow us to simulate such motion – e.g. our MorphServer). In our normal modes paper we figured that any such simulation is unnecessary – all we needed for that study was just a set of vectors connecting the residues from the two conformations (to examine how they correlate with the NM-predicted motion vectors).