Replica-Based Protein Structure Sampling Methods II: Advanced Hybrid Solvent TIGER2hs

In
many cases, native states of proteins may be predicted with sufficient
accuracy by molecular dynamics simulations (MDSs) with modern force
fields. Enhanced sampling methods based on MDS are applied for exploring
the phase space of a protein sequence and to overcome barriers on
rough conformational energy landscapes. The minimum free energy state
is obtained with sampling algorithms providing sufficient convergence
and accuracy. A reliable but computationally very expensive method
is replica exchange molecular dynamics, with many modifications to
this approach presented in the past. Recently, we demonstrated how
our temperature intervals with global exchange of replicas hybrid
(TIGER2h) solvent sampling algorithm made a good compromise between
efficiency and accuracy. There, all states are sampled under full
explicit solvent conditions with a freely chosen number of replicas,
whereas an implicit solvent is used during the swap decisions. This
hybrid method yielded a much better approximation to the agreement
with calculations in an explicit solvent than fully implicit solvent
simulations. Here, we present an extension of TIGER2h and add a few
layers of explicit water molecules around the peptide for the energy
calculations, whereas the dynamics in fully explicit water is maintained.
We claim that these water layers better reproduce steric effects,
the polarization of the solvent, and the resulting reaction field
energy than typical implicit solvent models. By investigating the
protein–solvent interactions across comprehensive
thermodynamic state ensembles, we found a strong conformational dependence
of this reaction field energy. All simulations were performed with
nanoscale molecular dynamics on two peptides, the α-helical
peptide (AAQAA)3 and the β-hairpin peptide HP7. A
production-ready TIGER2hs implementation is supplied, approaching
the accuracy of full explicit solvent sampling at a fraction of computational
resources.