SOLVE - AUTOMATED MIR AND MAD STRUCTURE DETERMINATION

Thomas C. Terwilliger¹ and Joel Berendzen²

¹ Structural Biology (Mail Stop M888)
² Biophysics (Mail Stop D454) Groups, Los Alamos National Laboratory, Los Alamos, NM 87545, U.S.A.

We have developed a fully automated procedure for determination of heavy atom sites and calculation of native phases in a MAD experiment or in all derivatives in an MIR experiment. The only input required consists of the space group and cell constants, resolution limits, the type of heavy atoms present, and the scaled data. The output consists of ranked sets of refined heavy atom parameters and native electron density maps.

The most important new feature of the procedure is the scoring of potential solutions. Four criteria are used in scoring. The first corresponds to peak heights at expected positions in the Patterson function. The second is peak heights at heavy atom positions in "omit" heavy atom difference Fouriers. The third is the figure of merit of the phasing obtained. The fourth is the non-randomness of the native Fourier obtained. Each of these criteria are evaluated based on the Z-score obtained from a comparison of the solution being tested and the mean over many solutions.

The procedure uses a Patterson search procedure (a modified version of HASSP) to obtain plausible starting solutions. The solutions are scored and ranked. One at a time, solutions are considered as test starting solutions for generation of new solutions by addition or deletion of individual sites. After calculation of native phases from a test solution, difference Fourier maps (and cross-difference Fourier maps for the MIR case) are used to identify possible additional sites. New solutions are scored and included in the ranked list of solutions. The procedure is iterated until no additions or deletions to existing highly-ranked solutions yield an improvement in score.

SOLVE is now installed at several synchrotron sources (ESRF, NSLS beamlines X-4A, X-8c, and X-12C; SSRL beamlines 7-1 and 9-1) so that crystallographers can solve structures immediately after data is collected. One MAD structure with five selenium sites was solved with our procedure at X-12C at the NSLS within 11 hours of putting the crystal in the X-ray beam. The procedure has been successfully applied to MAD data with up to 26 selenium sites in the asymmetric unit and to MIR data with up to five sites per derivative.