How to make better conformational sampling of models using MODELLER?
Hi All, As per the suggestion given in the modeller that changing random seed would helps in producing different models. On this note I have tried to generate 100 models from single template and using different random seeds. It was mainly performed to know how changing random seed influencing on conformational sampling ?. Following the experiment I had observed that there was notable fluctuations on the sidechain orientations. But when I aligned all the models against top model (DOPE SCORE) there was no variations on the backbone conformation and also in terms of RMSD. Following these observations I have a couple of questions for the modeller experts as follows: When modeller says changing random seed helps in finding different models, does it meant for sidechain orientations or backbone? Is it possible to produce models in modeller that will have more variations on the backbone and in terms of RMSD ? Thanking you in advance
On 1/25/18 10:14 AM, Mahesh VELUSAMY wrote: > It was mainly performed to know how changing random seed > influencing on conformational sampling ?
Changing the random seed has no influence on sampling. It merely results in different initial randomized structures, before sampling.
> when I aligned all the models against top model (DOPE > SCORE) there was no variations on the backbone conformation and also in > terms of RMSD.
Modeller restrains the backbone to match your template(s), so this is expected.
> Is it possible to produce models in modeller that will have more > variations on the backbone and in terms of RMSD ?
Sure, by weakening any homology-derived restraints. Either adjust the weights, or provide an alignment with lower sequence identity.
Ben Webb, Modeller Caretaker
Many more thanks for the response
On 25-01-2018 20:59, Modeller Caretaker wrote: > On 1/25/18 10:14 AM, Mahesh VELUSAMY wrote: >> It was mainly performed to know how changing random seed influencing >> on conformational sampling ? > > Changing the random seed has no influence on sampling. It merely > results in different initial randomized structures, before sampling. Got it. > >> when I aligned all the models against top model (DOPE SCORE) there was >> no variations on the backbone conformation and also in terms of RMSD. > > Modeller restrains the backbone to match your template(s), so this is > expected. Got it. > >> Is it possible to produce models in modeller that will have more >> variations on the backbone and in terms of RMSD ? > > Sure, by weakening any homology-derived restraints. Either adjust the > weights, May I know how to do these both weakening homology-derived restraints and adjusting the weights (also let me know what do you mean by weights? It would be helpful if you give any examples or sources to read)
> or provide an alignment with lower sequence identity. My interest is single point mutated sequences sharing similar length of query coverage. ( I have also tried mutated_model method to do the same but even in that case I end up with poor sampling with no backbone changes.)
> > Ben Webb, Modeller Caretaker
On 01/25/2018 12:54 PM, Mahesh VELUSAMY wrote: > On 25-01-2018 20:59, Modeller Caretaker wrote: >> Sure, by weakening any homology-derived restraints. Either adjust the >> weights, > May I know how to do these both weakening homology-derived restraints > and adjusting the weights (also let me know what do you mean by weights? > It would be helpful if you give any examples or sources to read)
Adjust schedule_scale, as per https://salilab.org/modeller/9.19/manual/node19.html
Ben Webb, Modeller Caretaker
participants (2)
-
Mahesh VELUSAMY -
Modeller Caretaker