From: Jérôme Hénin (
Date: Tue Apr 10 2018 - 09:23:20 CDT

Hi Brian,

I would expect the internal coordinates usually included in CHARMM
topologies to be enough to reconstruct a full side chain with a likely
geometry, from nothing but the backbone. Could it be that those coordinates
that are present (CB, HB1, HB2) are incompatible with the internal
coordinates? In this case my safest approach would be to delete all
sidechain coordinates except CB, and let ICs provide the others.


On 10 April 2018 at 15:58, Brian Radak <> wrote:

> The user guide is unfortunately lacking in good examples for this.
> I have a simple, complete peptide structure for (ALA)_5 and want to make a
> new PSF/PDB with psfgen after converting the central residue (resid 3) to
> ASP. The mutate command sounds perfect for this. I did something like the
> following:
> segment PROA {
> pdb foo.pdb
> mutate 3 ASP
> }
> coordpdb foo.pdb PROA
> guesscoord
> which works, but predictably has trouble reading the PDB since it finds
> the atom HB3 in ALA when it is expecting CG, etc. from ASP. The guessed
> coordinates for the added carboxylate are fairly atrocious with a
> completely unphysical dihedral. Is this an expected limitation of mutate?
> Obviously it is a bit too much to ask that mutate immediately figure out
> the mapping between those two atoms on-the-fly. I was able to tell psfgen
> about that mapping by adding:
> pdbalias atom ALA HB3 CG
> before reading coordinates, but I'm pretty sure this is a global mapping
> that discards the HB3 coordinates for all other ALA residues (not what I
> want). This works out just fine in my case, but seems unsatisfying as the
> solution to the problem in general. Is there a more specific workaround
> than this? What would I do for a more complicated mutation like LYS to TYR?
> Cheers,
> Brian