Generalization of the two-stage folding mechanism beyond beta hairpins

Establish whether the two-stage computational structure identified in the ESMFold folding trunk for beta-hairpin formation—early propagation of biochemical features from sequence to pairwise representations followed by late development of pairwise spatial geometry—extends to other protein structural motifs such as alpha helices, beta sheets, and long-range domain contacts.

Background

The paper mechanistically analyzes ESMFold’s folding trunk and identifies two computational stages: early blocks propagate biochemical features from sequence to pairwise representations, and late blocks develop spatial geometry in the pairwise representation. This analysis is conducted primarily on beta hairpins as a minimal nontrivial motif.

In the limitations, the authors explicitly note uncertainty about whether the discovered two-stage mechanism generalizes to other structural motifs, raising a concrete open question about the scope of the mechanism across alpha helices, beta sheets, and long-range domain contacts.