In previous years of the grant, the local squaring functions have been used to refine noisy electron-densities by placing individual atoms at locations in the unit cell where they are likely to occur. This method has been shown to be effective when high resolution data is available. However, in typical macromolecular crystallography studies, the available data is unsuitable of the single atom local squaring function. To expand the scope of the method, the local squaring functions have been generalized to model non-spherical, multi-atom fragments. The initial indication of the feasibility of the method was the automated model construction of the alpha-conotoxin PnIA. Using the generalized local squaring functions, nine of sixteen residues were specified by two or more templates, four were exactly specified by one template, and the entire backbone was easily constructed. Two other proteins are being studied for feasibility of the method: rusticyanin, a medium sized copper protein (ISS residues) for which experimentally derived phases are available, and top7, a designed protein for which no experimental phases are available.

This project takes advantage of the fragments clusters developed by Kolodny, Koehl, Guibas and Levitt within this program and published in 2002. A second fragment library targeted for use with the local squaring function method is currently developed by Alex Tropsha at UNC. Some fragments require a more complicated technique to parametrize their orientation in space