AB INITIO PHASE DETERMINATION FOR VIRUSES; THE USE OF NON-CRYSTALLOGRAPHIC SYMMETRY FOR PHASE REFINEMENT. Chapman, M.^{+[[daggerdbl]]}, Tsao, J.^{*[[daggerdbl]]}, Rossmann, M.^{[[daggerdbl]]}, Munshi, S.^{[[paragraph]][[daggerdbl]]}, Johnson, J.^{[[section]][[daggerdbl]]}, ⁺Dept. Chem. & Inst. Molec. Biophys., Florida State Univ., Tallahassee, FL 32306; ^*Dept. Microbiol., Univ. Alabama, Birmingham, AL 35394; ^{[[daggerdbl]]}Dept. Biol. Sci., Purdue Univ., W. Lafayette, IN 47907; ^{[[paragraph]]}MRL, WP144B-122; Merck & Co. Inc., W. Point, PA 19486; ^[[section]]Dept. Molec. Biol., Scripps Res. Inst., La Jolla, CA 92037

The atomic structure of Canine Parvovirus (CPV) was determined with the use of phases calculated ab initio [Tsao et al., 1992, Acta Cryst.. B48:75]. Ab initio phases were refined and extended to 8 Å and used to solve the positions of a heavy atoms for a single isomorphous replacement (SIR) structure solution. In retrospect, it was shown [Tsao et al., 1992, Acta Cryst. A48:293] that the phases could be extended from 8 to 3.25 Å and that the virus structure could have been determined ab initio without SIR. Similar approaches have since been used to determine the structures of [[phi]]X174 [McKenna, et al., 1992, Acta Cryst. B48:499] and N[[omega]]v.

With the advantage of phase refinement and extension using non-crystallographic symmetry of high redundancy (60-fold for CPV), crude centric approximations can suffice for the initial starting phases at low resolution. Approximating the virus by a spherical shell of uniform densities for protein and nucleic acid, the sign of the molecular transform oscillates with the positions of nodes highly dependent on the choice of shell radii. Post mortem examination of successful and failed attempts show that it is critical to have a self consistent set of phases in the resolution shell from which extension will take place. This means that the radii must be estimated to within ~1% of their true values (or be similarly consistent with an alternative Babinet solution). Although trial-and-error may eventually give a suitable initial phasing model, it is also possible to refine the radii and position of the spherical shell model through optimization against low resolution amplitudes from crsytallography or solution scattering [Chapman et al., 1992, Acta Cryst. A48:301].