Error with recognizing rotamer library atoms

I can diagnose this more completely if you post a section of your PDB_ROTAMERS file, but I would guess the atom names are subtly wrong. It’s likely to be a spacing thing, like it is “..N.” instead of “.N..” (where I have substituted . for an empty space). The spacing looks ok in those error messages, but that’s what usually causes it, so I feel I should suggest it.

It may be worth trying a different mol2->pdb format converter. (and try END for TER, unless the instructions said otherwise. I assume you’ve seen the set of instructions Doug wrote? There’s a copy in the tutorials thing here, I think (design_ncaa): http://www.rosettacommons.org/content/rosetta3-tutorials-beta-release)

The major problem is that you have a small-molecule ligand, not a polymer ligand. In other words, the params file is not set up for your noncanonical residue to be part of the polypeptide chain, so Rosetta is trying to insert it with no chemical bonds to other atoms. The tarball of instructions from Doug should have included a version of the molfile_to_params.py file that works for _polymer_ residues; this is the version you need to use. Specifically, the polymer residue params file needs to define UPPER_CONNECT and LOWER_CONNECT for the connections to the upstream and downstream N and C atoms.

Secondarily, I think there’s a lot of extra data in your rotamers file – Rosetta wants only the ATOM lines with TER cards between them; none of the CONECT lines, etc. Also your first rotamer has no hydrogens.

I have attached a params and partial PDB_Rotamers file for a system I’ve used so you can see what the params and pdb_rotamers file ought to look like. (this is from the publication Nat Chem Biol. 2011 Jun 12;7(7):437-44. doi: 10.1038/nchembio.585.
A biosensor generated via high-throughput screening quantifies cell edge Src dynamics.
Gulyani A, Vitriol E, Allen R, Wu J, Gremyachinskiy D, Lewis S, Dewar B, Graves LM, Kay BK, Kuhlman B, Elston T, Hahn KM.)

Hi, this is Doug. Steven asked if I would take a look at this thread to see if I had any ideas. The molfile_to_params_polymer.py script can be finicky. It requires additional information at the end of the mol file in order to properly create a polymer params file, more than just the “M ROOT X” when you were running the non-polymer molfile_to_params.py and making a ligand params file. Take a look at Step 4 in the README file for more info. The full path is below.

rosetta-3.3/protocol_capture/using_ncaas_protein_peptide_interface_design/HowToMakeResidueTypeParamFiles/README

I looked at the params file in your last post. It has some backbone dihedrals being listed as chi angles (CHI 1 N CA C O ) which suggests you have not properly set the backbone atoms in the molfile that gets passed to the molfile_to_params_polymer.py script (most likely the backbone carbonyl oxygen but I would double check all of them).

On the issues of molfiles above, the molfile_to_params scripts needs molfile 2000 format file but the molfile version 3000 bond types (this was not my doing).

Some stuff between Doug and the user off-forum; read emails in reverse order

(second email)
Hi Onur,

Sorry it took me a while to get back to you. I have been busy before the holiday. I found your bug. The KRTmol3.mol is fine except you had the ROOT atom listed in the POLY_IGNORE.

Yours…
M ROOT 34
M POLY_N_BB 34
M POLY_CA_BB 35
M POLY_C_BB 36
M POLY_O_BB 37
M POLY_IGNORE 1 2 4 5 6 7 8 9 10 11 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
M POLY_UPPER 12
M POLY_LOWER 3
M POLY_CHG 0
M POLY_PROPERTIES PROTEIN
M END

what it should be…
M ROOT 34
M POLY_N_BB 34
M POLY_CA_BB 35
M POLY_C_BB 36
M POLY_O_BB 37
M POLY_IGNORE 1 2 4 5 6 7 8 9 10 11 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
M POLY_UPPER 12
M POLY_LOWER 3
M POLY_CHG 0
M POLY_PROPERTIES PROTEIN
M END

Making that change and running the command below should work without error.
python using_ncaas_protein_peptide_interface_design/HowToMakeResidueTypeParamFiles/scripts/molfile_to_params_polymer.py –clobber –polymer –no-pdb –name KRT -k KRT.kin KRTmol3.mol

I have attached the modified input and output from the script. I took a quick look at the params and it looks okay but I did not get a chance to test it. The next step for you would be to use the PDB rotamers as Steven described in some of the previous comments in the post.

I would be great if you could add something to the forums about how it got resolved so that others can follow it.

Thanks,
Doug

(first email)
Hi Doug,

I used the babel that generates the hybrid molfile, but when I try to run the molfile_to_params_polymer.py script, I get an assertion error. When I run it with assertions off ($pymol -O /path/molfile_to_params_polymer.py -blahblahblah), I get an IndexError saying the list index is out of range. What am I doing wrong? I have two versions of this molfile: one with the hydrogens associated with each residue clustered with that residue (as you have in your demo), and one with all of the hydrogens put at the end. This second molfile works fine with the script, but gives me the same set_atom_base: atoms dont exist! error in rosetta.

I made the starting .pdb file by copying out the section of residues (Ala Lys Tyr), and the covalent ligand, and making the covalent ligand part of the Lys residue, renaming it to KRT. To add the hydrogens, I use babel with the -h flag to make another pdb file. Babel puts the hydrogens at the end, which I then copy and paste with their respective residues or leave at the end. Then I use the special babel to make the hybrid molfile, add the M lines, and run the polymer script.

I attached all the relevant files.
KRTupdownhydrocentered.pdb is my starting pdb that looks exactly like your ornithine example. capping groups with their hydrogens at the beginning, my NCAA at the end. KRTmol3 is the molfile for this pdb.
KRTupdownhydroc.pdb is the other pdb that I don’t modify at all. That is, N capping group, NCAA, C capping group, all hydrogens at the end. KRTmol4 is the molfile for that pdb. This latter combination runs in the molfile polymer script, but gives that error when I try to run it in Rosetta.

Thanks for your help,

Onur Erbilgin

TER versus END shouldn’t make much difference – the way the reader is written, almost anything besides an ATOM or HETATM record will function as a separator.

I agree with Steven that it’s the atom naming that is the most likely culprit. He makes a good point that spacing is a likely issue. The other thing to look out for is to check that the names of the atoms in the PDB_ROTAMERS file match those names present in the params file – which might not match those of your starting structure if you used a method (e.g. molfile_to_params.py) which renamed them. (Note that the atom name in the params file follows the ATOM tag after a single space – so in the params file it’s “ATOM.XXXX”, where the . is a space and the four X’s are the exact string used as the atom name, including any spaces.)

“Also, a tutorial for this does not exist in that link Steven sent me- it has a blank readme. I have been using the information in the tarball here: http://www.rosettacommons.org/content/some-application-troubles-when-usi…. ” – it’s out of date, try the .bz2 tarball in the same directory. We still (ugh) haven’t “officially” released these, so there’s a bunch of old copies floating around…

As far as atom names: I strongly recommend forcing at least your backbone atoms to be CA, C, N, O in the params and pdb_rotamers file. Rosetta is supposed to be independent of assumptions, but there are likely to be some sticky spots where backbone connectivity will screw up if the “normal” atoms have abnormal names. For sidechain atoms it shouldn’t matter much. Once peptoids come online this will go away, but the peptoids aren’t in main Rosetta yet…

If the noncanonical residue is missing entirely from the structure, but there is no error message about it, then the scenario that comes to mind is A) you have -ignore_unrecognized_res on, and you do not have the new params file active via -extra_res_fa. If that’s the case, turn off the first flag and turn on the second (with the path to your params file). This seems unlikely given your earlier command line, but it’s worth asking. If this is not the case, try posting the following things for me to debug with: A) the region of the input PDB with your noncanonical (NC, plus one residue on either side?), your params file, C) the first two rotamers worth of your pdb_rotamers file, and D) the full command line.

The missing OXT warning is par for the course with Rosetta, don’t worry about it. Rosetta is adding the extra O atom for the unpolymerized C-terminus and telling you about it.

I’m guessing your problem is that the params file for your NCAA is of “TYPE LIGAND”. I haven’t read the NCAA tutorials that Steven linked to earlier, but I assume there’s some discussion of how to make your NCAA be of “TYPE POLYMER”

What believe is happening is that the enzyme design code is failing while finding regions for flexible backbone movement. Because the NCAA is of type ligand, Rosetta isn’t seeing the section of backbone of which it is a part as a continuous chain. It therefore breaks the backbone, and is dying because it looks like you have one residue long “polypeptides” (not the NCAA, but the residues on either side).

Things may work if you turn off backbone flexibility, or if you add additional residue to your input PDB, but your better bet is to figure out how to convert your params file to type polymer.

As far as I’m aware, the lines starting with M aren’t valid for .mol2 format files (they’re only present in .mol files).

If you wanted to use the M lines with a .mol2 file under the standard molfile_to_params.py, you could use the flag –m-ctrl, but that’s a relatively recent addition and I don’t know if your molfile_to_params_polymer.py has that option (run it with just the -h flag to find out).

One workaround is to use your favorite molecular editor/file conversion utility (such as OpenBabel) to convert the .mol2 file (again, without the M lines) to a regular .mol file, and then add your M lines to that file. One drawback is that the conversion would drop all the partial charge information that’s present in the .mol2 file, and substitute different charges. If that matters to you, you can always manually edit the params file once generated such that the atoms have whatever charges you want.

I defer to Rocco on the mol file issue. As far as the tarball – can you see
http://rosettadesign.med.unc.edu/collaborators/RosettaCon2011/RosettaCon2011Demos.tar.bz2 ?

A) What version of Rosetta are you using again? The error “set_atom_base: atoms dont exist” is on line 754, not line 553.

I assume it’s in the ICOOR_INTERNAL lines at the end of the file. Those lines are occasionally frame-shifted, which *might* be the problem, but it looks ok relative to my own params file, so maybe not. The error means that it doesn’t recognize the atoms on the right three columns when creating that row (I think it’s the first of the right three columns). It could be that they are out of order somehow, so that an atom is declared to depend on another atom that doesn’t exist yet? (Of course, the atoms should exist because they’re in the atom list in the top of the file, so maybe that isn’t possible). The backbone atoms must be special-cased or something so they can depend on themselves, but maybe the other atoms are out of order somehow? Does that make sense?

C) It could be that there are too many CHI lines – you’ll note the 10 and 11 lines change the spacing; try commenting out 10 and 11 and see what happens.

Did you realize your params file has two atoms named “H”, one of type HNbb and one of type Hpol? Also, in your ICOOR_INTERNAL lines, for one of the H’s you have the C atom listed twice, once for the angle atom and once for the dihedral atom.

I’m not sure if it’s the cause of your current problem, but even if not, it will likely cause grief later on.

It look like that EnvSmoothEnergy::residue_energy() has a guard for rsd.is_protein(), but if your residue isn’t a canonical amino acid, but is* described as PROTEIN in your params file, you’ll run into the error you saw.

Probably the best solution is to turn off (remove the line for) the env_smooth term in the rosetta_database/scoring/weights/enzde.wts file. (Note that env_smooth isn’t in Rosetta’s standard score12 weights, although there are some indications that including it gives better results, which is why it’s in the enzdes weights.)

Another option is to remove the “PROTEIN” designation from your NCAA’s params file, and replace it with the “POLYMER” designation (if you don’t already have both). I don’t know if that will lead to other issues elsewhere, though.