D0130

DEVELOPMENT OF SnB VERSION 2. Charles M. Weeks, Russ Miller, Hauptman-Woodward MRI, 73 High Street, Buffalo, New York 14203-1196, USA

The Shake-and-Bake algorithm is a powerful formulation of direct methods which alternates reciprocal-space phase refinement with filtering in real space to impose constraints. As implemented in the distributed version (1.1) of the SnB computer program, Shake-and-Bake combines peak picking in real space with optimization via either parameter-shift reduction of the minimal function value or tangent-formula refinement [Miller et al. (1994), J. Appl. Cryst., 27, 613-621]. It employs a multisolution approach in which the starting set of trial structures consist of randomly positioned atoms. The SnB program has provided ab initio solutions for structures containing as many as 600 independent nonH atoms, provided that diffraction data are available to 1.2Å.

An upgraded version of SnB, which permits density modification followed by inverse Fourier transformation to be used in place of peak picking and subsequent structure-factor calculation, is under development. Using a simple modification scheme such that [[rho]]mod=[[rho]] if p>C*[[sigma]][[rho]] and [[rho]]mod = 0 if [[rho]]<C*[[sigma]][[rho]], where C is a constant in the range 2.0-3.0, a 6-8 fold reduction in overall CPU time has been achieved for crambin, a 400-atom test structure. When SAS tangent-formula refinement was incorporated into this version of SnB and applied to invariant values estimated for 2Å error-free SAS data [Hauptman (1982). Acta Cryst. A38, 632-641] for cucumber basic protein, the success rate (i.e., percentage of trials going to solution) was improved over that obtained for SAS tangent-formula refinement alone.

Research supported from GM-46733 (NIH) and IRI-9412415 (NSF).