[phenixbb] More trained monkeys please (request for improvements to auto-NCS)

[Kendall Nettles]

Dear PHENIX developers,
I am writing to ask about the automated NCS selection procedure. I have a structure with 2 homodimers in the ASU, and there are significant differences in the two monomers that make up teach dimer. When I run automr/autobuild I am very pleased with the result overall, but it does not choose the best NCS groups- it groups all chains to the A chain, instead of one dimer to the other. The result is a helix that is obviously out of place (see attached image).

I wonder if this could be fixed by implementing a throw away round of refinement, in order to identify optimal NCS groups? Though it looks like there are only a few areas of significant difference, the structures end up better by grouping one dimer against the other, rather than just the regions that are obviously different. This brings me to my second question. Might it be worthwhile to implement an NCS optimization procedure, which would compare different NCS groups, similarly to the optimize solvents, weights, etc?

Kendall Nettles[cid:89DEBDA8-D967-4E24-BDFD-3450601A7793 at hsd1.fl.comcast.net.]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20101103/0c2316a1/attachment-0003.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: coot.png
Type: image/png
Size: 295752 bytes
Desc: coot.png
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20101103/0c2316a1/attachment-0003.png>