Dear Rosetta Support,
I am trying to create a homology model of the tubulin protein heterodimer. I want to include the GTP/magnesium and GDP/magnesium (which are in the crystal structure template I am using) in the modelling process, but I am having problems. I keep getting an error that says either the scoring function "is not setup for nonideal/cartesian scoring", or that my residue param file, e.g. GTP.fa.tors "does not have valid atom records" (or something along those lines).
I have attached the relevant files. Because I can't upload more than 3 files, the GTP related files are all merged into one file. If you search for the appropriate file name (as indicated by the command file, i.e. GTP.fa.params, GTP.fa.tors, GTP.cen.params) within "GTPparams_and_TemplatePDB.txt" you will be directed to the text within those respective files. My template also has GDP, but the files for those are directly analogous to the GTP files. Also, the PDB section of templ_thread.pdb is merged with the file showing my command line input (i.e. the attached file, command_and_nucleotidePDBs.txt).
I can see how my GTP.fa.tors file is a problem, since all the torsions values are the same. However, I can't seem to generate a valid torsions file using the molfile_to_params command. I am also aware that GTP is in the residues.txt file from the source code, but I tried activating those (by uncommting them in the residues.txt file) and I simply get the previously mentioned error message about my scoring function. In any case, I am wondering whether it's valid to use the default GDP and GTP parameters anyway, since they're clearly not optimized for modeling the tubulin protein.
I should also note that I ignored the non-standard residues (i.e. the nucleotides) when I threaded my sequence onto my template. I then manually inserted the nucleotide section of the original PDB (which, like I mentioned, is shown in the file "command_and_nucleotidePDBs.txt") into templ_thread.pdb. I realize this means that there are no weights for the nucleotides in templ_thread.pdb, and the sequence at the beginning of templ_thread.pdb does not include the nucleotide residue codes (I suppose these could be contributing to my problems, but I don't know a valid solution!).
Any help resolving these issues would be much appreciated!
|xml input file||1.01 KB|
|command line input, and the nucleotide section of templ_thread.pdb||7.44 KB|
|relevant GTP parameter files||12.58 KB|
The "is not setup for nonideal/cartesian scoring" issue is completely independant of the GTP/GDP issue. What this warning/error means is that you are trying to use a scorefunction with cartesian minimization which either a) doesn't have the cart_bonded term or b) has the pro_close term. I'm guessing it's partly to do with using the talaris2014 in your FastRelax mover, when you have -dualspace on the commandline. If you're using dualspace relax, you should be using talaris2014_cart (or your 'fullatom' scorefunction instead. -- A clearer diagnostic would need to know when, exactly, in the protocol the error is being triggered.
Regarding the "does not have valid atom records", you're passing your tors files to -extra_res_fa, which is expecting params files. I believe you should be passing it with -score:extra_improper_file instead. I wouldn't worry about the fact that the parameters for the torsion file are all the same - they're just there to make sure things like the aromatic rings don't go out of planarity. cart_bonded is a bit of a heuristic term, anyway, so not having precisely defined torsional parameters here isn't going to be a big deal in terms of the performance of the modeling.
I haven't dug into the differences between your params files and the ones in the database, but I'm guessing that the ones in the database would also be reasonable. There may be minor differences, but I doubt they would have a great effect on the results of the modeling. The big thing that making your own parameters allowed you to do, though, is generate the torsions file, which will be a great help in keepin the aromatic rings planar when you're doing cartesian/dualspace manipulations.
You may indeed have issues with your templates and alignments not containing the GDP/GTP residue, but you can always manually edit them to add the single letter code for your residue ('Z', according to the IO_STRING line of the params file) or the "annotated" single letter code. (That's the single letter code with the full name of the residue in square brackets after: "Z[GTP]" or "Z[GDP]" in your case.)
Thank you for your help. Based on your last paragraph, I went ahead and tried to include the GDP/GTP and magnesium ions in the threading process (since I do believe this is the ideal way to start formulating the homology model), but for some reason, only the GTP is included in the output pdb. I get an error message that says:
Error: potential mismatch between sequence from alignment and sequence from PDB!
However, as far as I can tell, the sequence for the template in the alignment is the same as the sequence in the PDB. Here are the associated files, command line input, and error message. Please let me know if there are any other files you need to help solve this problem.
Answered in the other thread: https://www.rosettacommons.org/node/9929