An empirical energy function for structural assessment of protein transmembrane domains

• We have developed a membrane-specific distance-dependent statistical potential.
• It can rank structural models of α-helical and β-sheet transmembrane domains.
• It outperforms current statistical potentials at ranking protein models by quality.
• It provides insights into the rules by which residues interact within the membrane.

Knowing the structure of a protein is essential to characterize its function and mechanism at the molecular level. Despite major advances in solving structures experimentally, most membrane protein native conformations remain unknown. This lack of available structures, along with the physical constraints imposed by the lipid bilayer environment, constitutes a difficulty for the modeling of membrane protein structures. Assessing the quality of membrane protein models is therefore critical.Using a non-redundant set of 66 membrane protein structures (41 alpha and 25 beta), we have developed an empirical energy function for the structural assessment of alpha-helical and beta-sheet transmembrane domains. This statistical potential quantifies the interatomic distance between residues located in the lipid bilayer. To minimize the problem of insufficient sampling, we have used kernel density estimations of the distance distributions. Following a leave-one-out cross-validation procedure, we show that our method outperforms current statistical potentials in discriminating correct from incorrect membrane protein models. Furthermore, the comparison of our distance-dependent statistical potential with one optimized on globular proteins provides insights into the rules by which residues interact within the lipid bilayer.

Figure optionsDownload high-quality image (177 K)Download as PowerPoint slide