W0305

Protein Crystallography According to wARP. Anastassis Perrakis¹, and Victor Lamzin¹, European Molecular Biology Laboratory, ¹Grenoble and ²Hamburg Outstations, European Molecular Biology, Laboratory (EMBL), Grenoble, 38042 France

Phase information in Protein Crystallography can be provided by (i) experimental techniques such as Multiple or Single Isomorphous Replacement (MIR,SIR) with or without Anomalous Scattering information (MIRAS,SIRAS) or Multiple (or even Single) wavelength Anomalous Dispersion experiments (MAD,SAD) combined with software to make optimal use of the data (Phases, MlPhare and more recently SHARP) ; (ii) purely computational techniques if atomic resolution data are available (as implemented in packages such as SHELX and Shake and Bake). Provided that some phasing information is provided by any of the above techniques, ARP and wARP come into play.

Given an electron density map that has been created with phases resulting from any of the above techniques, the features of this map can be described by the contruction of a free atoms model to fit the density. This is the first step in going on with the wARP procedure. The Automated Refinement Procedure (ARP) principle is based on a cyclic procedure of reciprocal space refinement followed by automatic model update based on statistical analysis of electron density maps. This way the starting model model is not only optimized in reciprocal space to minimize typically a maximum likelihood residual, but at the same time continiously updated in order to adapt in new features that appear in the electron density. While this technique can work well with high resolution data (above 1.5 AA), to overcome the limited data in lower resolution we introduced the idea of the weighted averaging of structure factors resulting from the ARP refinement of different models , wARP. This technique can result to substantial improvement of phases and will also extend available phases to the highest resolution of the native diffraction data set.

Given a good map, constructed with the phases calculated above, parts of it can be automatically recognized to contain protein stereochemistry and it is possible to build a (often partial) atomic protein model. While with free atom models it is easy to interpret almost every feature of an electron density map the automated building of parts which contain a clear protein stereochesistry offer the use of geometrical restraints to be used in the reciprocal space refinement steps. A combined partial protein model with a free atoms model to explain prominent electron density features which at this certain stage do not make perfect sterechemical sense, can be an optimal model to be used in ARP. This model can be optimized and updated by simple ARP refinement addition/deletion of water molecules (as for free atoms model) which can be combined with complete automatic rebuilding of the protein structure in regular intervals, in the hope (which mor ethan often comes true) that improving phases will allow the construction of bigger parts of the protein model.

Given the above techniques we managed to construct almost full models (>90% of main chain and >50% of side chains) for structures where experimental data were available from different techniques and native diffraction data from 0.9 AA to 2.0 AA resolution were available (but typically the experimental phases extended at ~2.5 A resolution). It was also possible to build up a partial model (~50%) and create an improved map with 2.15 AA data and to deliver an improved map with 2.3 AA data. The latter seems to be the current resolution limit for most applications.