You are here

genkic mover produced an unwanted HN - CA bond

3 posts / 0 new
Last post
genkic mover produced an unwanted HN - CA bond

I ran the tutorial placed in demos/tutorials/GeneralizedKIC/exercise1/ and successfully produced 10 output structures. However, the residue ALA32 that was encompassed by the genkic mover has unwanted HN - CA bond and HN_CB bond. All the 10 output structures have this problem, and I've attached three of them. I tried to FastRelax the strucutre after genkic mover, but the unwanted bonds were still kept.

If ALA32 was mutated to GLY32, the genkic mover still produce an unwanted HN - CA bond. I am wondering if this is a bug in the program? Any solution would be much appreciated.

Post Situation: 
Mon, 2022-06-20 14:25

As far as Rosetta is concerned, there aren't extra bonds. The issue looks to be that the amide nitrogen of residue 32 is misplaced a bit. For programs which determine PDB bonding based on distance heuristics, they can show spurious bonds because certain atoms are "too close". But Rosetta uses atom identity (names) to figure out bonding patterns, so it won't think there's bonds there, even on a new read-in.

The reason for the misplaced hydrogen is because you have two different chains that are being connected. When Rosetta looks at the N terminal residue of chain B, it's not knowing how to handle figuring the cross-chain connection to chain A. This results in the Warning messages you probably see in the output: "Cannot get atom_id for POLYMER_LOWER of residue GLY 32.  Returning BOGUS ID instead." because of this, the nitrogen is being placed in an odd (arbitrary) location.

The reason FastRelax doesn't help is that the default FastRelax algorithm only relaxes the bond rotation degrees of freedom. The bond lengths and angles are assumed to be correct. Since the amide hydrogen position is not due to a rotatable bond (but is based on bond lengths/angles of the other backbone heavy atoms), it doesn't get fixed. You can add the option `-dualspace` to tell FastRelax to also include a Cartesian based minimization in the process, which should fix bad bond lenghts/angles. (Assuming that it doesn't also throw an error because of the BOGUS ID issue.)

It looks like a "simple" fix is to simply make sure that the input PDB that should be all one chain is actually all one chain. If you edit the input PDB to make everything chain A and remove any TER cards between the two regions, then things seem to work as you might expect. True, the structure as read in would appear to Rosetta as having a really long peptide bond between the two ends, but the PeptideStubMover and the GenKic mover should fix that by adding in residues to bridge the gap and fixing the geometry, respectively. (Rosetta can handle the really long peptide bond in the mean time.)

Mon, 2022-06-20 15:13

Thanks for the suggestion. I tried the same exercise with a single-chain initial pose (modified initial pose attached), but the incorrect placement of HN atom still remained. Fortunately, FastRelax with dualspace flag can correctly relax the HN atom to a reasonable position.

Mon, 2022-06-20 16:58