Here is a summary of responses to the thread I recently started on "phasing with se-met at low resolution":
There were many suggestions. They can be grouped into two categories: 1. Aim for very high redundancy at one wavelength (peak) and watch out for radiation damage. Sacrificing resolution is OK. 2. Collect two, three, or four wavelength MAD datasets. Helps break phase ambiguity (but is it better at finding sites compared to redundant peak SAD?). Collecting in wedges with inverse beam is widely recommended, but it was also pointed out that its usefulness has not been systematically established. No matter what, one should watch out for radiation damage: Follow your scaling B-factors in scalepack. Obviously, high solvent content helps with SAD/density modification. NCS averaging of low resolution, low quality maps can yield interpretable maps. Cross crystal averaging is also useful, if applicable. B factor averaging may help at lower resolution (apparently mostly when there is phasing data, as opposed to pure MR solutions, according to DelaBarre and Brunger, Acta Cryst 2006). Many of you pointed out that building all-beta would be a big pain even at 4 Angstroems, while alpha helical structures can be built fine up to 5 Angstroems. However, it was pointed out by many of you that finding heavy metal sites appears to be the major problem, and where many get stuck. Many of you suggest very redundant SAD datasets could help with that. Except for two (1rhz, 2oau), all cases sent to me used heavier metal soaks to find selenium sites (or they were ~3.5 A datasets with I/sigma ~10 in the highest resolution bin, which might not be considered low resolution: in these cases finding heavy metal sites was expectedly straight-forward). Summary of cases: 1rhz Peak wavelength SAD with wedges combined with cross-crystal averaging. Platinum-derivative might have confirmed heavy metal sites (not sure). It was mentioned that low symmetry space groups do make life quite difficult. 2oau Very redundant three wavelength data (high symmetry space group) to 5 A was amazingly enough to locate selenium sites (SOLVE located the sites). 7-fold NCS with native data to 3.5 A yielded "excellent" electron density maps. This one gives hope to all of us. 2jk4 "Together with phases from a Pt-derivative and after careful density modification, an initial low-resolution electron density map was obtained that clearly showed the protein envelope and the overall barrel dimensions." Complementary NMR data (both NOE and dihedral restraints, secondary structure assignments and hydrogen bonds) had to be added to the refinement process to achieve interpretable density maps and reliable model. 1xdv "A mercury derivative provided initial phases, which locate the positions of selenomethionines. Electron density maps calculated using experimentally determined phases to 4.3 Å, extended to 4.0 Å, were improved by B factor sharpening and by 2-fold noncrystallographic symmetry averaging." And other cases where initial phases were obtained at very low resolution via heavy metal soaks (such as 3B8E). Among to respondents were Kornelius Zeth, Partha Chakrabarti, Stephen Soisson, Leiman Petr, Artem Evdokimov, Dominika Borek, Clemens Vonrhein, Phil Jeffrey, Wladek Minor, Pete Meyer, James Holton, P. H. Zwart, Anastassis Perrakis, and others I forgot to mention. Many thanks.
