[phenixbb] phasing high(er) resolution anomalous difference maps with low resolution phases

Terwilliger, Thomas C terwilliger at lanl.gov
Thu Mar 13 11:27:53 PDT 2014

Hi Todd,

I think all your ideas are worth trying.   I'd try both your lower-resolution phases and phases based on the helices you found.  No, you are not necessarily stuck at 6.2 A phases:

-- The helices may have phase information at higher resolution.
-- You can also take that phase set and use it as starting phases for density modification at higher resolution.  I think there isn't an answer to "how far is too far" other than "as far as you can get decent phases".
-- You can use phases from the helices at higher resolution (run refinement with whatever resolution data you have; then the map coefficients will be suitably weighted at all resolutions.

It is hard to say how high your anomalous peaks should be.  If they are very high (> 10 sigma) you are doing super-well  I'd say stick with about 3 sigma peaks just because those mostly will not be noise (some will be though).

Pavel Afonine is modifying phenix.find_peaks_holes to read in  map coefficients...so that should be available shortly.

Yes, once you get sites you can just read in a sites file to autosol along with your data and sequence file and that will carry out phasing and density modification as usual.  You might want to set helices_strands_only=True so it does not bother to try and build everything if you are still working at lower resolution.

All the best,
Tom T

From: phenixbb-bounces at phenix-online.org [phenixbb-bounces at phenix-online.org] on behalf of Todd Jason Green [tgreen at uab.edu]
Sent: Thursday, March 13, 2014 11:26 AM
To: phenixbb at phenix-online.org
Subject: [phenixbb] phasing high(er) resolution anomalous difference maps with low resolution phases

Hello all-

I am working on a huge protein assembly(monomer is ~3000 amino acids, large cell ~200, ~200, ~520) that has between 200 and 300 Se-met residues in the asymmetric unit(depending on the number of molecules I pick in the a.u.). I believe I have low resolution phasing from another heavy atom with limits to ~ 6.2 angstrom. Density modified FOMs dip below 0.6 at that point. I can see many tubes that have the diameter of model alpha helices. I ran phenix.find_helices_strands with the helices_before_trace=True option and got a series of helices encompassing about 3400 residues. Assuming that the phasing on the low resolution derivative is true, is it better to use phasing from the lower resolution derivative or phasing from the helical models to fish out the Se-Met sites via an anomalous difference Fourier? I assume I'm stuck at 6.2 for the derivative phasing though i guess i could extend phasing from the map a little further. How far is too far? How high of resolution can I phase from the helices (or from the derivative map for that matter) and still get accurate enough phasing to yield reliable difference peaks to find the Se sites. I have a couple Se-met datasets between 3 and 2.5 angstroms. What sigma levels should I expect for the difference peaks?

also, what is the best way to pick the highest difference peaks in the anomalous difference fourier? I didn't see a tool in phenix. "find difference peaks and holes" seems like a logical choice but it seems to want to structure. I've just created maps and used the very old and reliable peakmax in ccp4.

I have solved structures like this before but with a lot more info on ncs, envelopes, ncs averaging and a much much smaller protein. I remember getting sites and plugging them into resolve with scripts where you could fix them/or not and phase without looking for more sites, no build etc. Can i just enter sites and phase followed by density modification with one of the gui programs.

any comments would be appreciated? and thanks in advance.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20140313/714288aa/attachment.htm>

More information about the phenixbb mailing list