AI Uncovers the Rapid Activation of Catch-Bonds under Force

Mechanically resilient
protein interactions are crucial
for biological
processes ranging from bacterial adhesion to human tissue formation.
Catch-bonds, a unique class of protein interactions that strengthen
under force, act like a molecular finger trap, tightening to prevent
bond rupture. However, it remains unclear whether catch-bonds form
immediately upon force application or require a specific force threshold
for stabilization. Here, we employ an in silico single-molecule force
spectroscopy approach that combines molecular dynamics (MD) simulations,
dynamical
network analysis, and AI-based modeling to investigate the XDoc:CohE
complex, a hyperstable catch-bond found in cellulose-degrading bacteria.
By analyzing amino acid interactions between XDoc and cohesin E, and
between XDoc submodules (X-module and Doc), we show that AI regression
models can accurately predict rupture forces using only short MD simulations,
capturing key mechanostability features despite the binding interface’s
complexity. Our results reveal that mechanostability signatures emerge
early under force load, indicating that catch-bonds activate almost
immediately. These findings provide new insights into the molecular
principles governing force-dependent protein interactions and highlight
the potential of AI-driven approaches for predicting and characterizing
mechanostability, with broad implications for bioengineering and drug
design.