A Hybrid UA-CG Force Field for Aggregation Simulation of Amyloidogenic Peptide via Liquid-like Intermediates

Elucidating amyloid formation inside biomolecular condensates requires models that resolve (i) local, chemistry specific contacts controlling β registry and (ii) mesoscale phase behavior and cluster coalescence on microsecond timescales-capabilities beyond single resolution models. We present a hybrid united atom/coarse grained (UA-CG) force field coupling a PACE UA peptide model with the MARTINI CG framework. Cross resolution nonbonded parameters are first optimized against all atom side chain potentials of mean force to balance the relative strength between different types of interactions and then refined through universal parameter scaling by matching radius of gyration distributions for specific systems using. We applied this approach to simulate a recently reported model system comprising the LVFFAR₉ peptide that can co-assemble into amyloid fibrils via liquid-liquid phase separation. Our ten-microsecond simulations reveal rapid droplet formation populated by micelle like nanostructures with its inner core composed of LVFF clusters. The nanostructures can further fuse but the fusion is reaction-limited due to an electrostatic coalescence barrier. β structures emerge once clusters exceed ~10 peptides, and the LVFFAR₉ fraction modulates amyloid polymorphism, reversing parallel versus antiparallel registry at lower LVFFAR₉. These detailed insights generated from long simulations highlight the promise of our hybrid UA-CG strategy in investigating the molecular mechanism of condensate aging.