Re: [modeller_usage] Fibrillar collagen structure model 3hr2 Inbox

[Modeller Caretaker <modeller-care@ucsf.edu>]

On 1/27/21 1:48 PM, Amadeus Cavalcanti Salvador de Alcântara wrote:
> The 3hr2.pdb contains 2 non standard amino acids:
> HYP hydroxyproline
> LYZ Hydroxylysine

I would be inclined to edit the PDB file and replace the coordinates for 
these modified residues with the corresponding standard amino acids. You 
can certainly use the '.' residue type, and treat them as ligands, but 
that will simply copy them from target to template, allowing no 
relaxation (or mutation), usually not what you want for modified residues.

> The model looks good, however, it is not as close to the 3hr2 structured 
> as I wished. As a mechanical engineer I am no used to those models nor 
> to homology modeling and have been learning all here by reading the 
> manual, instructions and mail listing. Did I make many/any mistakes? 
> Where and what? Can the model become better (more close to the 3hr2 
> template structure)? If yes, how?

Modeller builds models that resemble the template, by construction. If 
you're seeing a big divergence, you likely have an alignment error. You 
should carefully inspect the alignment and look for obvious problems, 
such as lots of gaps, sequences of different lengths, or gaps in regions 
of known secondary structure.

> How to know if a model is "good enough"?
This depends on what you want to use it for. But DOPE is only a ranking 
- it doesn't tell you how "good" the model is, only whether it's better 
than your other models. A score like normalized DOPE or TSVMod would 
give you a better idea; these can be calculated at
https://salilab.org/evaluation/

> Can I proceed like that and create the models for the chains B 
> and C and complete the collagen structure?

To build a multi-chain structure, it's usually best to build all chains 
simultaneously, so you get any inter-chain interactions correct.

	Ben Webb, Modeller Caretaker
--
modeller-care@ucsf.edu             https://salilab.org/modeller/
Modeller mail list: https://salilab.org/mailman/listinfo/modeller_usage