e16.html

Lothar Esser (NCI): The Successor of BEAST - Molecular replacement is always a bit tricky especially with poor data. The membrane protein that I work on provides only very anisotropic data and the only program that could handle it was the successor of BEAST: Phaser. So many people have said so many good things about phaser ( see G. Sheldrick's comments on the CCP4BB ) that I do not need to add anything only so much that it solved my difficult problem too. I have a solution despite less than ideal data. What usually convinces me of the correctness of a solution is when I see peaks in the anomalous difference map for sulfur atoms or atoms of which Z>=16. Just to be on the safe side, my MR models are usually free of S or any heavy atoms so that when I get peaks in the right places, I am convinced. Coming back to data, Phaser rejects outliers also but its great strength (at least in my case) lies in the correction of anisotropic data. I'd highly recommend it.

Xinhua Ji (NCI): The Beauty of BEAST - Outliers among diffraction data are usually of lower resolution and relatively stronger and thereby have greater impact on the functions derived from intensities. One consequence of this impact is the lack of phasing power of molecular replacement (MR) solution. We recently encountered this problem with an MR solution obtained using a search model equivalent to ~50% the structure. With AMoRe, we found an solution; but we were not able to complete the structure using either the difference Fourier synthesis or other programs. Assuming that the MR solution was not accurate, we tried MR again excluding potential outliers. BEAST identified 22 potential outliers and rejected them from likelihood calculations. The MR solution was outstanding and the difference map derived from the partial structure reveals the missing portion of the structure.

Biomolecular Structure Section (NCI): SOLVE is great. It can carry out a complete and automatic structure determination for either MIR or MAD data. What if it does not work automatically or cannot find all heavy atom sites? We have experienced the following.

Case I: HgMAD data were collected at four wavelengths. Difference Patterson indicated five Hg sites. Using all wavelengths, however, SOLVE was not able to find any consistent set of Hg sites. Slight adjustment of any parameter (resolution, nres, nanomalous, etc.) resulted in a complete different set of sites. None of the sites was verifiable with difference Patterson. The last thing we tried was to feed SOLVE with different combination of wavelengths. When wavelength 1 was excluded, SOLVE found a consistent set of four correct Hg sites. Yet, one site was still missing. Then, we fed SOLVE with unmerged data, which resulted in five correct sites immediately. Guess what? When the excluded wavelength 1 for SOLVE was included in SHARP refinement, it did help improve the phases! To the best of our knowledge, nothing is significantly wrong with wavelength 1, which was collected from the same crystal and processed by the same crystallographer.

Case II: A total of 24 Se sites were expected. SOLVE found 21 using premerged data. With unmerged data, it found 25 (24 + 1 alternate position for one Se atom)!

To generate unmerged data, one needs to tell SCALEPACK “no merge original index” and then to tell SOLVE “unmerged" before running the program.