Dear all,

I am refining a protein structure at 1.74 A resolution with phenix.refine (version 1.10.1_2155) and I have noticed that some side chains don't remain in the electronic density even after careful hand adjustment in which they are almost perfect. It seems pretty clear where the side chain should (must) be located, moreover it is consistent with two rotamers (either rotamer tp60 - #4 in coot - or rotamer tp-100 - #9 in coot -, which are approx. at 180 deg for chi3). In spite of that, after reciprocal space refinement, I observe that the end of side chain rotates chi3 angle and both N and O gets far from where they started (there is this difference in the chi3 angle).
My concerns are about the proper interpretation of geometrical restraints listed in .geo file, specially what the ideal value is, how weights are applied to ideal value and why phenix doesn't reach convergence in reciprocal space refinement. Below are the values from .geo for the referred Gln:

dihedral pdb=" CB GLN B 21 "
pdb=" CG GLN B 21 "
pdb=" CD GLN B 21 "
pdb=" OE1 GLN B 21 "
ideal model delta sinusoidal sigma weight residual
0.00 23.41 -23.41 2 3.00e+01 1.11e-03 8.42e-01

is there a description for us to interpret what these numbers are? To my understanding, the ideal values are established as being either 0 or 180 degrees (yet sinusoidal 2). I cannot say about how heavy the weights are on these (in fact, I left my refinement to optimize weights), but anyway, if I understand correctly, the ideal value (for chi3) should be rotamer dependent. Taking Gln as an example, I see several rotamers do have values close to either 0 or 180 degree, but what about rotamers 4 and 9? I might imagine a similar situation for residues like Asn and His, though I have not searched deep for these yet.
Thanks,

Renato