| Structural information for biologically and medically interesting proteins can be difficult to obtain. Technical challenges prevent all but a small fraction of structures from being solved crystallographically, and our knowledge of quantum mechanics is not yet sufficient to be able to compute the native fold in silico in a reasonable amount of time. Further, the ability to design novel proteins with arbitrary functionalities is critically dependent on the ability to accurately predict the full three-dimensional structure assumed by a particular sequence. These problems are only further exacerbated for transmembrane proteins, where the complex, dynamic environment of the membrane presents additional experimental and theoretical difficulties. To fill in this gap, we turn to structural bioinformatics. We leverage the knowledge available from those structures which have already been solved experimentally by assembling large, non-redundant databases. We then examine the distribution of individual residues and small structural motifs to derive broad trends that can be applied to new proteins, for which structural information is lacking. For instance, we present a statistical potential that describes the membrane-insertion preferences of each residue type in alpha-helical transmembrane proteins. Its predictive power extends not only to gross structural orientation in the membrane, but also to mechanistic insights and rotamer optimization. We also discover sequence motifs that help facilitate helix-helix associations in a relatively small number of highly-populated geometries. We further quantify these structural and sequence motifs and describe how they vary between transmembrane and soluble proteins. Finally, we leverage a simple bioinformatic metric of designability to create a rapid protocol for designing peptides that can bind to and stabilize a protein-protein interface. In all these cases, we demonstrate how structural bioinformatics can be used to enhance our understanding of protein structure and folding, as well as protein-protein and protein-membrane interactions. This insights gained here will provide a structural basis for future work such as molecular modeling and rational templates for protein design. |
