Homology Modeling with Modeller

[trudack]

modelmaker model

Inputs
template (complete)
sequence (complete)
number of model outputs
scoring type

optional
start resid + chainID for modeling
end resid + chainID for modeling
default should be all
list of helix restraints
start helix1 resid +chainID
end helix1 resid +chainID
start helix2 resid +chainID
end helix2 resid +chainID
...
(beside helix there is also the possibility to define sheet restraints, we should add that at a later point)

default none

optional
define alignment selection

The tool needs to do:
if resid range is specified
-cut out range from template
-cut out range from sequence

transfer fasta sequence format to .seq header for modeller input

adapt template and target name and chain ID in align.py

run align.py

adopt template and target name and number of models in model-single.py

if helix restraints are defined add them to model-single.py

Attention Numbering in Modeller starts with one so need to be adapted if residue range is specified.
ChainID are also not conserved and start with A,B,C

run model-single.py

select the best scored model. Number can be found in last paragraph of model-single.log (Summary)
Pick the best one from the selected scoring type line.

rename the model to a proper target_out.pdb and renumber, get the right chain IDs back and set segnames from template

align the correctly numbered and named output to template

(The example is right now for only one chain, in order to adapt it for more than one chain, the renumbering skript should be adapted to renumber not only single chain pdbs but also multiple / I would suggest to habe this also as separate modelmaker command, modelmaker renumber as this is needed at many places and could be useful also as standalone tool)

I will sent a link to a case example with all my scripts to do so by mail to @ryanmcgreevy

[ryanmcgreevy]

Just FYI @trudack , I've committed new versions of the align2d.py and model-single.py scripts. I am still in the process of hooking them into the modelmaker interface and automating the steps you outlined here, but these latest versions have at least been generalized and given an improved command line interface of their own. This will facilitate them being able to be used standalone (like Keith's clustering) if desired, and also means they can already be used more easily than the hard-coded previous versions.

[ryanmcgreevy]

I have now committed an initial modelmaker model command to act as an interface to the modeller python scripts align2d.py and model-single.py. I have tested this with the old tutorial example and it seems to work. However, as of this comment, modelmaker model does not yet perform any of the desired post-processing steps. Also, the -helix resids still assume you entered the already correct modeller numbers, as there is not yet any translation between input resids and modeller numbering.

[trudack]

If I type modelmaker I cannot see a model command

[ryanmcgreevy]

Oops, this is now fixed. The command was available, there just wasn't any usage text if you typed modelmaker.

[trudack]

Now there is the usage text for find_empty_density

[ryanmcgreevy]

Sorry! That's what I get for copying and pasting code. This is now fixed.

[trudack]

There seems to be a bug in modelmaker model:
the homology model has the same sequence as the template even though the .aln file shows the right sequences.

I believe that is because the template name is tmp and the target name is also tmp. You could replace it symply by template and target and I hope that fixes the bug.

Furthermore I list the missing add ons for the model module here:

• Modelller writes the log file only in the console. This should always be logged to a log file for bug fixing reasons.
• retain chain ID, segname and numbering by default from your template file, also give the option to give a chain ID starting resid number and a segname
• give the option to name the output file instead of tmp.
• give the option to pick the best n Models (default 1) this is based on the summary table at the end of the console log of Modeller
• give the option to pick the scoring method (default Dope)
• align the picked best model to the template by default and give the option to align to any pdb with atom selection

[ryanmcgreevy]

The bug Till mentioned is now fixed, but the remaining features still need to be implemented.