On Fri, 30 Jun 2006, Noinaj wrote:
N >I recently collected Se-MAD data (resolution --> p=2.4, e=2.8, hr=2.7) and currently attempting to find the heavy atom sites, but have experienced several problems. I expect 20 Se sites in my protein (~99 KDa).
N >Initially, I just used the peak dataset (2.4 angstroms) for SAD Phasing procedures in CNS, AutoSol and Solve (via PHENIX), BnP, and CCP4. With CNS, I found ~20 sites and they all showed up very nicely in the maps. But when I did the density modification of both hands, neither map seemed to be better than the other, which was very disappointing. Next, I tried AutoSOL strategy in PHENIX, which found the sites also, but listed varying degrees of occupancy for the sites, ranging from 0.7 to 0.05. Not being that familiar with PHENIX, I used the sites from AutoSOL and made maps using CNS, but, map quality wasn't improved at all. I have also tried using CCP4 and BnP to find the sites, hoping for better results, but these programs failed to find all the sites, typically fewer than 6 or so, depending on the program. If they did list more sites, the signal for the sites wasn't significant over background. My boss' friend used SHELX to found all the sites with much improved occupancies. Using these sites, the maps looked a bit better, but still, no real discernable secondary structure, and still, alot of fragmented density.
N >I am currently working on MAD phasing procedures, which have so far,
produced about the same results.
N >This is my first experimental phasing experiment, so I would like some advice or guidance about either SAD or MAD phasing procedures in general. For example, what programs do people typically use, and why choose one program over another? Which is most reliable? I had expected the experimental phases to give me a map with good enough quality to at least build a rough model, but so far, this isn't the case. So what steps can be taken now to improve the quality of the maps or phases to at least get to a map with some discernable features?
I like using shelxc-shelxd to solve the substructure (is that what your
boss' friend did?). I have got a slightly higher success rate with it
than other program on difficult cases and I also find it helpful to use data from
different wavelengths to calculate the FAs before the substructure
solution.
Try aiming for fewer sites; if you have some bad sites the maps will be a
mess, even if most are correct. Refining the the sites with autosharp
also usually gives you a good idea which ones to weed out.
In the worst case scenario (you need to recollect data) do not collect
three wavelengths: Do either a MAD experiment at the inflection and
remote wavelengths (make sure that the remote wavelength is truly remote,
ie, at least 600 eV above the absorption edge)**; or SAD at the peak. If
you do the two wavelength MAD, collect the two wavelength simultaneously
in wedges. That will help a lot if your problem is that the crystal got
killed during data collection.
** See Gonzalez A, von Delft F, Liddington RC, Bakolitsa C
J Synchrotron Radiation 2005 May 12(Pt 3):285-91
Good luck!
Ana
--
-------------------------------------------
Ana Gonz?lez <[EMAIL PROTECTED]>
Staff Scientist
Stanford Synchrotron Radiation Laboratory
2575 Sand Hill Road, MS 99
Menlo Park CA 94025
Phone: (650) 926 8682 Fax: (650) 926 3292
-------------------------------------------
