Integral membrane proteins have domains that traverse between the membrane interfaces and as a consequence of the signaling aspects of this location they are important drug targets. Their environment alone is a reason why routine experimental approaches to determine their three-dimensional structures have not been forthcoming at a desirable rate. Only approximately 400 examples of membrane protein structures exist in which the entire membrane is traversed, whereas the generalist structural database holds ~50,000 proteins. In this study, the approach of statistical potentials, that has been successful with generalist structure modeling, are adapted in order to derive the 3-dimensional structures of such polytopic helical membrane proteins.
We report our helical membrane protein structure prediction method in which our set of scoring functions are applied to distinguish native-like polytopic helical membrane protein structures from non native-like structures1. The scoring functions are trained and tested on subdomains of 3-5 transmembrane helices, that were derived from known polytopic helical membrane protein structures. The novel aspects of the scoring function are the components designed for the membrane environment of the helices, and specifically the derived residue burial propensities and environment scoring. The program currently developed uses only the amino acid sequence of the target protein, but is also assisted by secondary structure and membrane topology predictions. Such input does not use predictions of kinking which is a ubiquitously observed phenomenon in membrane helical proteins, and we have evidence that inclusion from our study of kinking will contribute to the algorithms for predicting these structures, improving the outcomes. The method, used on five sequences, gave predicted models with RMSDs of 5.0 to 12 Å from the native structures, making it a viable approach to modeling, especially if kinking can be introduced and additional restraints from experimental observations are available.