There is a great deal of interest in creating models of proteins based on sequence homology with known structures. These models may be used on their own, or may provide starting points for refinement against experimental data. A prerequisite of this approach is the availability of accurate, experimentally determined atomic structures of at least one and preferably many related proteins. Some examples of this approach include the modeling of renin (Sibanda et al, 1984), immunoglobulins (Chothia et al, 1989) and serine proteases (Greer, 1990).
Factor D is the first serine protease from the complement system to be crystallized (Narayana et al, 1991) and to have its structure determined at atomic resolution (Narayana et al, 1994). It is a single polypeptide chain of 228 residues. Factor D is the rate-limiting enzyme in the ``cascade'' of events that activate and regulate the alternate pathway of the complement system (Lesavre and Muller-Eberhard, 1978).
We describe the creation of an ``ad hoc'' model of Factor D by computer graphics modeling and the ``solution'' of the structure by molecular replacement using this homology model. The subsequent creation and refinement of a second homology model built using a commercial software package is also described.
The accuracy of models constructed from homology is of fundamental concern when the models cannot be confirmed by experimental methods. In addition, there is some question about the general validity of refined crystal structures obtained through molecular replacement techniques employing homology models. In order to better understand the errors that might arise from the use of homology models, we have investigated the differences between the structure of human complement Factor D determined from MIR crystallographic refinement and the homology models built from known structures of related serine proteases.