Unambiguous assignment of kinked-β sheets leads to insights into molecular grammar of reversibility in biomolecular condensates.

Abstract

Kinked- $\beta$ sheets are short peptide motifs that appear as distortions in $\beta$ strands and often mediate formation of reversible amyloid fibrils in prion-like proteins. Standard methods for assigning secondary structures cannot distinguish these esoteric motifs. Here, we provide a supervised machine learning-based structural quantification map to unambiguously characterize kinked- $\beta$ sheets from coordinate data. We find that these motifs, although deviating from standard $\beta$ strand region of the Ramachandran plot, scatter around the allowed regions. We also demonstrate the applicability of our technique in wresting out LARKS, which are kinked- $\beta$ strands with designated sequence. Additionally, from our exhaustive simulation generated conformations, we create a repository of potential kinked peptide-segments that can be used as a screening-library for assigning $\beta$ kinks in unresolved coordinate dataset. Overall, our map for kinked- $\beta$ provides a robust framework for detailed structural and kinetics investigation of these important motifs in prion-like proteins that lead to formation of amyloid fibrils.