Need help with automating Modeller flow?
Greetings all,
I come from a computer science background (only little knowledge on Bioinformatics) and I work on this BioInformatics application to be used in a research institute.
Application need a feature to model protein 3D structure from sequences and I decided to go with Modeller. The application should have a simple interface (like web interface at http://salilab.org/modweb ) . I went through the Basic modeling tutorial (http://salilab.org/modeller/tutorial/basic.html) and have several questions :
1) After step 1 (build_profile.py) we have to select suitable templates ,does the selection criteria is e-value =0 ? 2) After executing 'compare.py' ,we have to select one pdb template from distance matrix? How to implement this in selection logic? What criteria does 'ModWeb' application use? Is there any tools to select the best template using the distance matrix?
Best regards, umanga
Ashika Umanga Umagiliya wrote: > I went through the Basic modeling tutorial > (http://salilab.org/modeller/tutorial/basic.html) and have several > questions : > > 1) After step 1 (build_profile.py) we have to select suitable templates > ,does the selection criteria is e-value =0 ?
You want to pick templates with small E values, so yes, the closer to zero the better.
> 2) After executing 'compare.py' ,we have to select one pdb template from > distance matrix?
No, you could build your model using a multiple-template alignment, particularly if different templates cover different parts of your model sequence. Or you could repeat the modeling process for each of the templates, and then pick the best model(s).
> How to implement this in selection logic? What criteria does 'ModWeb' > application use? Is there any tools to select the best template using > the distance matrix?
ModWeb selects suitable templates using regular sequence-sequence searches (which you used in the basic modeling tutorial) as well as profile-sequence and profile-profile searches, which are often more sensitive at low sequence identity. These are not covered in the basic tutorial, but also utilize profile.build() or (PSI-BLAST profiles) and profile.scan(). Models are then built, and the best selected on the basis of model assessment scores, such as GA341 and DOPE.
You can also check the CASP pages for ideas. While many groups and servers use Modeller to build models, there are a range of methods used to select templates and assess the results.
Ben Webb, Modeller Caretaker
participants (2)
-
Ashika Umanga Umagiliya
-
Modeller Caretaker