An Automated Structure Prediction System by Lattice Model for Seven-Helix-Type Membrane Proteins

Bacteriorhodopsin (bR) is a light-driven proton pump, which is a membrane protein with seven transmembrane helices and a prosthetic group of retinal. The structure of bR has been determined at high resolution, and the homology modeling of G-protein coupled receptors, which have seven transmembrane helices, is usually carried out based on the structure of bR as the structure template. However, the structure of receptor proteins is not necessarily the same as that of bR. Therefore, we previously developed a structure prediction system for seven-helix-type membrane proteins, using a lattice model [1]. The principle of the method is that the dominant interaction between helixes is the polar interaction [3], which was estimated by a probe helix method [2]. This system showed very good performance, when it was applied to bR [1]. However, the system had several disadvantages: (1) Calculation time was very large. (2) Manual procedures were inevitable. (3) The information about important helices in the bR structure could not be extracted from the calculation. In this work, we developed an automated version of the tertiary structure prediction system for seven-helix-type membrane proteins, based on the previously reported system [1]. The new system is completely automated and provides helix triangles which may contain functionally active sites.