The program Modeller has been employed to build homology models to be used in a molecular replacement search.
The serum glycoprotein haemopexin is made up by two domains connected by a flexible linker region. The N- and C-terminal domains contain internal repeats with strong sequence homology patterns which suggest that they have the same basic fold. Structure determination of the C-terminal domain revealed a beta propeller fold (Faber et al. 1995, Structure 3, 551-559) with four "blades" or modular beta sheets, corresponding to the four internal sequence repeats found in this domain, as well as in the other domain.
Structural studies on the whole molecule of haemopexin aimed at solving the structure by molecular replacemnt, using the model of the C-terminal domain to search for itself and for its mate, the N-terminal domain. While the first part of the search was straightforward, finding a solution for the N-terminal domain proved to be a very difficult task. A solution was found only using a chimeric construct made up of parts of structure of the C-terminal domain combined with parts of structure of the beta propeller domain of collagenase.
With a sequence identity of 23%, the N- and C-terminal domains of haemopexin show a lot of insertions and deletions of various lengths located between the modules or beta sheets. Curiously, some parts of the beta propeller structure of collagenase has a sequence which resembles more the haemopexin N-terminal domain than the haemopexin C-terminal domain does.
In order to check the possibility of using homology models of the N-terminal domain as search models for the molecular replacement calculations, the sequence of the N-terminal domain was threaded onto the structure of the C-terminal domain. Then five homology models were built using the program Modeller and then fed into the suite of programs AMORE-CCP4. These models were used without truncations, as given in the output of Modeller. The results show that the homology models give better and comparabable solutions to the one obtained using the chimeric construct mentioned above.
These results are very encouraging indeed and prove that Modeller can produce homology models which are valid starting points for structure solution by molecular replacement searches. The home page for the program Modeller is http://guitar.rockefeller.edu/modeller/modeller.html.
Max Paoli