fix rotamers question
Dear Phenix community, I would like to have some additional information regarding fix_rotamers. In particular : - How Phenix chooses which rotamers need to be fixed ? - Does Phenix really re-calculate the three maps after correcting each rotamer ? For very large structures this is very time-consuming and one can settle for calculating a new map every 10-20 residues. Perhaps this could be a user-controlled parameter ? -Can you please explain how fix_rotamers works together with ncs restraints ? Thank you in advance for shedding light on this, Peter
Hi Peter,
- How Phenix chooses which rotamers need to be fixed ?
Rotamers that need to be examined are identified as those that are either rotamer outliers as defined by the Richardson rotamer library or are sidechains whose real-space CC is below a user-definable threshold (0.9 by default).
- Does Phenix really re-calculate the three maps after correcting each rotamer ? For very large structures this is very time-consuming and one can settle for calculating a new map every 10-20 residues. Perhaps this could be a user-controlled parameter ?
The maps are not calculated after each rotamer correction, but instead are recalculated after all flagged rotamers have gone through the correction protocol once. Each correction is then validated against the recalculated maps, and those sidechains that are not an improvement over the original orientation are returned to the original orientation. This whole process is considered a macrocycle, and the number of macrocycles for rotamer correction may be specified by the user.
-Can you please explain how fix_rotamers works together with ncs restraints ?
This is a great question and I'm not sure of the answer. Perhaps one of the other developers can weigh in on this point? Thanks, Jeff
Thank you very much for these clarifications,
Peter.
On Wed, Oct 20, 2010 at 6:27 AM, Jeff Headd
Hi Peter,
- How Phenix chooses which rotamers need to be fixed ?
Rotamers that need to be examined are identified as those that are either rotamer outliers as defined by the Richardson rotamer library or are sidechains whose real-space CC is below a user-definable threshold (0.9 by default).
- Does Phenix really re-calculate the three maps after correcting each rotamer ? For very large structures this is very time-consuming and one can settle for calculating a new map every 10-20 residues. Perhaps this could be a user-controlled parameter ?
The maps are not calculated after each rotamer correction, but instead are recalculated after all flagged rotamers have gone through the correction protocol once. Each correction is then validated against the recalculated maps, and those sidechains that are not an improvement over the original orientation are returned to the original orientation. This whole process is considered a macrocycle, and the number of macrocycles for rotamer correction may be specified by the user.
-Can you please explain how fix_rotamers works together with ncs restraints ?
This is a great question and I'm not sure of the answer. Perhaps one of the other developers can weigh in on this point?
Thanks, Jeff _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Hi Peter, general answer about how the whole thing works, see this: http://cci.lbl.gov/~afonine/rsr.pdf Specifically:
- How Phenix chooses which rotamers need to be fixed ?
see above.
- Does Phenix really re-calculate the three maps after correcting each rotamer ?
no, it happens only after one walk-through the all residues. In future this will be optimize to improve memory consumption.
-Can you please explain how fix_rotamers works together with ncs restraints ?
The whole procedure is not aware of NCS, or, actually, NCS should be smarter. This is in our plans to make better. Also, fix_rotamers may not perform as good as desired at lower resolutions. For this case we also have a number of things to work on, such as automatically using B-factor sharpened maps, locally-scaled maps, etc... Pavel.
participants (3)
-
Jeff Headd -
Pavel Afonine -
Peter Grey