The primary goals of this project are to develop methods to predict structures of membrane proteins from their sequences and available experimental data, to use these methods to develop Structural models of specific membrane proteins, and to work with experimental groups to test these models. We have developed a hierarchical approach to modeling membrane proteins. In the first phase we predict which segments cross the membrane and the secondary structure of these segments, in the second phase we predict the relative positions and orientations of the transmembrane segments, and in the third phase we use computer graphics and molecular mechanics energy calculations to produce models that predict positions of all atoms in the regions of the proteins that are modeled. We have developed models through the third phase for members of the following families of proteins: delta lysin, magainins, cecropins, alamethicin, pardaxin, annexins, and voltage-gated potassium channels. The simplest structures we have been analyzing are a series of peptides (delta lysin, magainin, PGLa, alamethicin, and pardaxin) that range in size from 21 to 33 residues. The secondary structures of all these peptides appear to be primarily helical; however, the C-terminus portion of pardaxin may be a beta strand that, when assembled with other pardaxin peptides, forms the narrow portion of the channel. Because of their relatively small size, more extensive calculations can be performed on these peptides than on the larger proteins described below. We have continued to develop structural models of larger integral membrane proteins. The protein families for which we have developed phase three models are voltage-activated potassium channels and annexins. These models predict the residues responsible for causing the channels to open and close and those that determine the ion selectivity properties of the pore.
