united atom dihedrals

[skfegan]

At the united atom level the code is looping over residues. When it is looking for the dihedrals to calculate the conformational entropy, we are probably missing a backbone dihedral angle which would connect the residue to its neighbouring residues.

[skfegan]

Upon further reflection this is not a bug. Excluding the dihedrals between residues is a reasonable thing to do at the united atom level. Because we have the residue level conformational entropy (which was added for version 1.0; not done in release version 0.3), we would want to avoid double counting. The residue level conformational entropy would be related to the secondary structure which would be coming from the conformation of the backbone, so those dihedrals should probably not be counted at the united atom level too.

For future reference, there may be other ways of calculating the conformational entropy that could be used instead. For example, considering modes of eigenvalues from a correlation matrix of dihedral states. Modes in principle should capture more correlation and scale better.