Hi, I have a pir file with more than one template. Each template has its distance restraints in the rsr-file. How can I weight the distance restraints for each template individually, i.e. all CA_CA-distance restraints for template t1 are weighted by a factor w_t1, for template t2 w_t2 and so on, so that templates with higher weight have more influence on the final structure. Rewriting the distance restraints and adding a weight as parameter maybe is a solution, i.e. starting with cuser_form.c from examples/c-extensions, multiplying *val = RT * 0.5 * delta * delta / stdev2; by an additionally passed weight would do it? Is the weight relative or absolute?
Best regards Armin Meier
On 5/27/10 6:49 AM, Armin Meier wrote: > I have a pir file with more than one template. > Each template has its distance restraints in the rsr-file.
Are you sure? That's not how Modeller builds restraints when using multiple templates. There should be a single restraint for each distance in the model; if you have multiple templates that restraint simply incorporates data from all of them.
> How can I weight the distance restraints for each template individually, > i.e. all CA_CA-distance restraints for template t1 are weighted by a > factor w_t1, for template t2 w_t2 and so on, so that templates with > higher weight have more influence on the final structure.
You can't really, since it doesn't work that way. If you have two templates for a given distance, one of which has a distance of 5A and the other 10A, Modeller doesn't create two restraints on the distance, one at 5A and 10A, since that wouldn't work, of course - what you'd end up with would probably be a structure with a distance of 7.5A, which doesn't match either template. Instead, it combines the two observations into a single multi-modal restraint. See the 1993 JMB Modeller paper for the details.
If you really wanted to do it your way, you could run Modeller twice, once for each template, then combine the two restraint files using whatever scheme you like.
> Rewriting the distance restraints and adding a weight as parameter maybe > is a solution
That seems unnecessarily complicated - you can simply reduce the standard deviation of a harmonic distance restraint to get the same effect as increasing the weight.
Ben Webb, Modeller Caretaker
Couldn't one approximate template weighting in a "dirty" way by using multiple copies of a template in the alignment file? i.e. if template t1 is included in the alignment twice (or maybe even just named twice in the 'knowns' list in the script, not sure about this) and t2 is used only once, then effectively the t1 restraint values have twice the probability of the t2 restraint values (everything else being equal). Ben, do you think this would work?
Of course it is important to point out that individual t1 and t2 restraints in the example are probably not being weighted equally anyway (because of the varying degrees of similarity between t1, t2, and the target, at different points in the sequence/structure). So maybe the important questions is WHY you would like to change the weighting of the restraints that MODELLER would do automatically?
Best, Roberto
-- Roberto Sanchez, Ph.D. Associate Professor Department of Structural and Chemical Biology Mount Sinai School of Medicine Box 1677, 1425 Madison Avenue, New York, NY 10029 phone +1 (212) 659 8648, fax +1 (212) 659 8232 http://sanchezlab.org/
On May 27, 2010, at 2:09 PM, Modeller Caretaker wrote:
> On 5/27/10 6:49 AM, Armin Meier wrote: >> I have a pir file with more than one template. >> Each template has its distance restraints in the rsr-file. > > Are you sure? That's not how Modeller builds restraints when using > multiple templates. There should be a single restraint for each distance > in the model; if you have multiple templates that restraint simply > incorporates data from all of them. > >> How can I weight the distance restraints for each template individually, >> i.e. all CA_CA-distance restraints for template t1 are weighted by a >> factor w_t1, for template t2 w_t2 and so on, so that templates with >> higher weight have more influence on the final structure. > > You can't really, since it doesn't work that way. If you have two > templates for a given distance, one of which has a distance of 5A and > the other 10A, Modeller doesn't create two restraints on the distance, > one at 5A and 10A, since that wouldn't work, of course - what you'd end > up with would probably be a structure with a distance of 7.5A, which > doesn't match either template. Instead, it combines the two observations > into a single multi-modal restraint. See the 1993 JMB Modeller paper for > the details. > > If you really wanted to do it your way, you could run Modeller twice, > once for each template, then combine the two restraint files using > whatever scheme you like. > >> Rewriting the distance restraints and adding a weight as parameter maybe >> is a solution > > That seems unnecessarily complicated - you can simply reduce the > standard deviation of a harmonic distance restraint to get the same > effect as increasing the weight. > > Ben Webb, Modeller Caretaker > -- > modeller-care@salilab.org http://www.salilab.org/modeller/ > Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage > _______________________________________________ > modeller_usage mailing list > modeller_usage@salilab.org > https://salilab.org/mailman/listinfo/modeller_usage
On 5/27/10 11:28 AM, Roberto Sanchez wrote: > Couldn't one approximate template weighting in a "dirty" way by using > multiple copies of a template in the alignment file? i.e. if template t1 > is included in the alignment twice (or maybe even just named twice in > the 'knowns' list in the script, not sure about this) and t2 is used > only once, then effectively the t1 restraint values have twice the > probability of the t2 restraint values (everything else being equal). > Ben, do you think this would work?
Yes, that should work (either way), but like you say, it's dirty. And it still uses the original Modeller weighting scheme - it may be that Armin wants to change that.
Ben Webb, Modeller Caretaker
-
Armin Meier -
Modeller Caretaker -
Roberto Sanchez