APPLICATION OF THE SAS TANGENT FORMULA TO MULTIPLE SITE PROBLEMS: A FEASIBILITY STUDY. Chun-Shi Chang, Charles M. Weeks, Debashis Ghosh, Herbert A. Hauptman, Hauptman-Woodward MRI, 73 High Street, Buffalo, New York 14203-1196, USA

The SAS tangent formula [Hauptman, Acta Cryst. A, in press] provides the basis for an ab initio multisolution (or multitrial) phasing procedure which utilizes invariant values estimated from SAS data [Hauptman (1982) Acta Cryst. A38, 632-641]. This procedure has been tested by application to 2Šsingle-wavelength, error-free data for a fabricated 10 Se analog of cucumber basic protein [Guss et al. (1988) Science, 241, 806-811]. Se isotropic thermal parameters in the range 6.7-25.5Ų were based on the average side-chain values in the real structure, and Se anomalous scattering factors (f', f'') of (-7. 623, 5.083) at 0.9789Šwere used.

Trial sets of random phases were refined iteratively until the corresponding values of the SAS maximal function converged, and solutions were identified on the basis of mean phase error. Nonsolutions consisting of part of the protein structure having the correct hand combined with enantiomorphic positions of the anomalous scatterers occurred frequently and had SAS maximal function values similar to solutions. In addition, under some circumstances trials yielding no locations for anomalous scatterers had even higher maximal function values. In either case, solutions could be detected automatically by picking the top 10 peaks on a Bijvoet difference Fourier and then computing protein phases based on these positions and comparing them to the corresponding original set of SAS tangent phases. Solutions had the smallest mean phase differences and, in this way, the proper absolute configuration could be selected. The SAS tangent solutions compared favorably to maps phased by the 10 Se placed at their true positions and given precisely correct thermal parameters. This research was supported by GM-46733 (NIH).