The vast majority of peptide designs have been based on heuristics or simple rules of thumb learnt from natural proteins or derived empirically through experiment. The de novo a-helical coiled-coil assemblies provide a good example of this.1 The rules relate sequence to structure to guide the specification of coiled-coil oligomerization state, strand orientation, partner selection, and, to some extent, stability. This has been extremely informative and productive, and design and engineering is probably more advanced for coiled coils than for any other protein structure.2 However, to move past the low-hanging fruit of peptide and coiled-coil design, and into the dark matter of protein structures, we will all have to learn new tricks. To address this we and others—notably the André, Baker, DeGrado, Grigoryan and Harbury groups—have begun to tackle coiled-coil design parametrically using computational methods.3 For our part, we have developed CCBuilder, which is an easy-to-use web-based GUI;4 and ISAMBARD, which is a more-versatile Python-based API that is free to download from GitHub and has been written for protein design more generally.5
I will describe how a serendipitous discovery of a 6-stranded a-helical barrel,6 which are rare in nature, led us to develop our computational methods; and how we used these to deliver a-helical barrels predictably.7 The talk will demonstrate the utility of this approach to make water-soluble protein-like barrels and pores, which we have engineered to form materials, bind small molecules, and catalyse simple reactions.8,9 Most recently in collaboration with the Bayley lab (Oxford), we have engineered membrane-soluble variants of these a-helical barrels that insert into lipid bilayers and conduct ions in a voltage-dependent manner.10 I will touch on how the barrels and related structures that we have built are improving our general understanding of coiled coils and peptide design more generally.