MantisBT - Rosetta
View Issue Details
0000268Rosetta[All Projects] Crashpublic2013-05-07 15:202013-05-08 17:59
score_jd2.linuxgccrelease -s inputs/1jf8FH.pdb -score:weights mm_std.wts -ignore_unrecognized_res
Confirmed As Bug
0000268: K residues causes MMTwist to crash
The K residue type (database/residue_type_sets/fa_standard/residue_types/metal_ions/K.params) has the potassium and two virtual atoms. This leads it to have 3 torsion angles, which are computed and stored in ResidueType::update_derived_data() line 1175. But when scoring with the MMTorsionEnergy score term, the MMTorsionLibrary::lookup() fails with

     No parameters for X-X-X-X.

score_jd2.linuxgccrelease \
  -s inputs/1jf8FH.pdb \
  -score:weights mm_std.wts \
No tags attached.
? 1jf8FH.pdb (204,777) 2013-05-07 15:20
? 1h4xFH.pdb (361,614) 2013-05-07 15:45
Issue History
2013-05-07 15:20momearaNew Issue
2013-05-07 15:20momearaStatusnew => assigned
2013-05-07 15:20momearaAssigned To => renfrew
2013-05-07 15:20momearaFile Added: 1jf8FH.pdb
2013-05-07 15:45momearaFile Added: 1h4xFH.pdb
2013-05-08 09:21renfrewNote Added: 0000284
2013-05-08 13:00renfrewNote Added: 0000285
2013-05-08 15:22renfrewNote Added: 0000287
2013-05-08 15:24momearaNote Added: 0000288
2013-05-08 15:35renfrewNote Added: 0000289
2013-05-08 16:21renfrewNote Added: 0000290
2013-05-08 17:59renfrewFixed in SVN Version => 1691dfed3fb982744aa0c7c71baf03fad73acccf
2013-05-08 17:59renfrewStatusassigned => resolved
2013-05-08 17:59renfrewFixed in Version => Trunk
2013-05-08 17:59renfrewResolutionopen => fixed

2013-05-08 09:21   
This is actually a problem with the params file; the MMTorsion (mm_twist) code is fully compatible with virtual atoms. This has come up before with the other ions. Folks seem to think that the MMAtomType of "X" is equivalent to a (Rosetta)AtomType of "VIRT" when it is not. The MMTorsion code uses "X" internally as a wildcard value (e.g. X-CT2-CT2-X in the mm_torsion_params.txt would mean use these params for all torsions that have CT2 and CT2 as the central atoms).

The solution is to change the MMAtomType field of the params file for the virtual atoms to be "VIRT" just like it is for the rosetta AtomTypes.
2013-05-08 13:00   
The problem of scoring potassium in 1jf8FH.pdb was fixed by adding potassium params to mm_atom_properties.txt and modifying the params files to reflect the additional atoms types and changing X to VIRT for the virtual atoms.
2013-05-08 15:22   
The 1hxFH.pdb also fails in the MMTorsion lookup but for a different reason. It doesn't contain potassium. It does contain a phosphorlated serine labeled as SEP.

This was a bitch to track down but both...


...define a ResidueType with SEP as the three letter code. One has complete MMAtomTypes and the other does not. However even if the MMAtomTypes were fully defined. It probably would have crashed because we don't have MMTOrsion params for phosphate groups.

On a side note, the SEP side chain is missing a hydrogen. I assume that they are built with reduce? That looks like a bug.
2013-05-08 15:24   
Should one of those be removed?
2013-05-08 15:35   
So none the params in residue_types/ptm-caa are currently being read in by default. It was the however the first param file I found though and it was really confusing why there were still undefined MMAtomTypes because they are all set in there (they are not set correctly but they are present and not "X"). I later found the patch file that creates the same ResidueType and the MMAtomType were not complete there.

I am not sure who added either of them but one should probably go. I can try to add some torsion params for the phosphate group, pulling them from D/RNA params. To make it work.

When I first made all the NCAA I had params files for most of these phosphorylated and sulfonated residues but CHARMM did not have params for them so I never commited them.
2013-05-08 16:21   
The problem with the phosphoserine in 1h4xFH.pdb was fixed by modifying the patch to add the correct MMAtomTypes. New MMTorsion params were added based on...

Feng, M.-H., Philippopoulos, M., MacKerell, Jr., A.D. and Lim, C. Structural Characterization of the Phosphotyrosine Binding Region of a High-Affinity aSH2 Domain-Phosphopeptide Complex by Molecular Dynamics Simulation and Chemical Shift Calculations. Journal of the American Chemical Society, 1996, 118: 11265-11277.

...and added to mm_torsions.txt.