Modeller Caretaker a écrit : > On 11/25/2010 01:11 AM, Stéphane Téletchéa wrote: >> I'm willing to create a peptide sequence from a fasta sequence, and add >> a disulfide bridge in it. >> >> At the moment, my code "which works" is like this pseudo-code: >> >> 1 - create a model object, and model.build('sequence'), I am not >> applying any time of secondary structure constraint >> 2 - write this pdb (peptide.pdb) >> 3 - create a special align where target and template are identical >> (peptide.ali) >> 4 - create a new object, Mymodel derived from automodel to add the patch >> for disulfides >> 5 - generate models >> 6 - select the best one (peptide-cys.pdb) > > That might work, although the restraints generated in step 4 would be > all wrong (Modeller will be trying to make your models look like the > extended chain you generated in step 1, but at the same time your > disulfide patch will be trying to bring the pairs of CYS residues > together). So even your best models will contain a *lot* of restraint > violations. >
Indeed, the final pdf score is much better, and the structure seems more "reasonable".
I've incremented a little the examples you mentioned (I went there before posting but could not figure out out to call it, the "custom function" was the anwser...), using the approach for md as it is implemented in automodel.
So far, all went well, I'll use these peptides as starting points for the rest of my analysis.
Thanks a lot,
Stéphane