Item

Abstract

Coiled coils are a widespread protein structure motif that consists of two or more -helices that wind around a central axis to form a helical bundle with a buried hydrophobic core. They are built from relatively short and simple sequence repeats, typically consisting of 7 residues (heptads), although alternative repeat sizes are possible, the most common being 11 residues (hendecads) [1,2]. The repeat size and residue composition of coiled coils are responsible for their considerable variety in terms of the helical topology (number and orientation of the helices) and geometry (axial rotation and degree and direction of the winding). Conversely, these structural features are responsible for the extraordinary diversity of functions that coiled-coil domains perform in nature, such as mechanical support, muscle contraction, vesicle transport and fusion, transcription factor, or signal transduction. In this work, we have benchmarked how accurate AlphaFold is in modeling typical heptad coiled coils [3], and investigate whether it could be applied to new, hitherto undescribed non-heptad coiled coils such as the ones composed primarily of hendecads. Our results show that AlphaFold is able to recapitulate a number of known coiled-coil rules that relate sequence and structure, and that it can be used to obtain insights into new ones. Simultaneously, we have found a number of cases that highlight the limitations and biases of AlphaFold in coiled-coil modeling. We hope that our work will serve as a foundation to develop new tools with which to further advance our understanding of this model protein structure motif.