On 6/15/23 7:33 AM, majort--- via modeller_usage wrote: > I'm trying to add 3 loops to a crystal structure (loops are missing) > without making any changes to the original X-Ray. I base the script > on the existing example script 'examples/automodel/loop.py'.
Your script looks largely OK to me. But since you have three separate loops, it's unlikely that the best-scoring model has all three loops in their best conformation (unless they are strongly interacting with each other). You might consider modeling just the first loop, picking the best-scoring solution, and then using that as your initial model for the second loop, and so on.
> In the inimodel ('5uv1_fill.B99990001.pdb'), the missing residues > have been added and simply appear as a straight line and are only > connected to one end.
You don't say how you generated that model, but it sounds like you used comparative modeling (AutoModel) and just took the unoptimized (.ini) file it generated. For insertions (i.e. residues that are not aligned with a template) coordinates are generated using CHARMM internal coordinates as per step 2.2 at https://salilab.org/modeller/10.4/manual/node498.html, which will result in an extended chain conformation that is attached to the rest of the protein only at the N-terminal end. If you want them connected at both ends, that is what AutoModel does during its model refinement.
> When I try do loop modeling using the above script, the loop atoms > hardly move and don't reconnect to the other endpoint.
If they don't move at all, it's because you have selected the wrong residue ranges in select_loop_atoms().
To achieve reasonable sampling for such long loops, you will likely need to build a lot more than 4 models, perhaps 200 or so.
Ben Webb, Modeller Caretaker