REFINEMENTS OF HEAVY-ATOM AND PROTEIN PARAMETERS AGAINST A HIGHLY ACCURATE MAD DATA SET. Axel T. Brunger^*, F. Temple Burling^*, Kevin M. Flaherty⁺, William I. Weis⁺. ^*Howard Hughes Medical Institute and Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520, USA, ⁺Department of Structural Biology, Stanford School of Medicine, Stanford University, Stanford, California 94305, USA

A highly accurate and complete multi-wavelength anomalous scattering (MAD) data set was obtained previously at 1.8 Å resolution from a crystal of a Yb³⁺-substituted subtilisin fragment of mannose-binding protein A (F.T. Burling, W.I. Weis, K.M. Flaherty & A.T. Brunger, Science 271, 72-77, 1996). By keeping the lack-of-closure expressions for both Bijvoet mates separate, a realistic probabilistic description of the MAD experiment was implemented. This description was formulated in the crystallographic computing language of the new version of X-PLOR (a pre-release will be available soon). This language allows one to chose between a variety of refinement methods (including conjugate gradient minimization and simulated annealing) against several target functions with restraints or constraints. Refinements of heavy-atom parameters and protein parameters were performed against the observed MAD phase probability distributions using a maximum-likelihood target function. Differences between these and conventional residual-based refinements will be discussed.