All-atom, explicit solvent MD simulations of the selected targets were conducted.
All simulations were conducted in keeping with our core MD protocols and simulation
methods that are published elsewhere [1, 2]. The simulations used all-atom (i.e.
hydrogens were included) fully flexible representations of protein and solvent in
the Levitt et al. force field with standard parameters . Non-bonded
interactions between atoms in charge groups separated by three bonds were included
and scaled by 0.4.
In the MD simulations, waters were represented explicitly by using the flexible
three center (F3C) water model of Levit et al. , as it has demonstrated
success in reproducing a large number of experimental observables across a range
of temperatures [4, 5]. Simulations were performed in the microcanonical ensemble
where the number of atoms, unit cell volume, and total energy were conserved (NVE),
as well as linear momentum (NVEp ). A two femtosecond (fs) timestep was
used such that each picosecond (ps) of simulation time required 500 steps. The unit
cell was periodic with minimum image conventions employed . Non-bonded interaction
lists were used for no more than 6 fs (i.e., three steps).
The simulations were performed with in lucem Molecular Mechanics (ilmm, ©
Beck, D. A. C., Alonso, D. O. V., Daggett V, University of Washington, 2000-2009).
A variety of computational resources were employed for the simulations. Most of
the simulations were conducted using the Department of Energy’s National Energy
Research Scientific Computing Center.
For each target six or more simulations were performed. A simulation at 298
Kelvin (K) was conducted for at least 31 nanoseconds (ns). For the
intact structures this is the functional biological native state, while for domains
of larger proteins the simulation assesses the stability of the domain in isolation.
Disulfide bonds were included in the native state simulation where appropriate.
A 10 Ångstrom (Å) cutoff range was used for the 298 K simulations .
Structures were saved from the trajectory at one ps resolution for later analyses
and full restart files with velocities and accelerations at the limits of machine
precision where saved one per ns.
To elucidate the unfolding / folding pathway of the Dynameomics targets, at
least five thermal unfolding simulations at 498 K were conducted. This
minimum number was established by a previous study . Two of the thermal unfolding
runs were at least 31 ns in length with the remaining 3 (or more) at least 2 ns
in length. Where the native state had disulfide bonds, these bonds were removed
and the cysteines protonated. An 8 Å non-bonded cutoff range was used. For
the first 2 ns of the majority of 498 K trajectories, structures were saved at 0.2
ps intervals to aid in the identification of the protein folding transition state.
Subsequently, structures were saved at 1 ps intervals. As with the 298 K simulations,
full velocities and accelerations were saved once per ns.
The starting structure for each simulation was derived from the corresponding
crystal or NMR structure from the PDB. In the case of NMR structures,
the first model was used. When multiple conformations for specific residues were
present in the PDB structure file, the first conformation was chosen. Protons were
then added. Then the coordinates of the structure were minimized with respect to
the force field’s potential energy with steepest descents (SD) minimization for
1000 steps or until the potential energy converged. The minimized structure was
solvated in a box of F3C water extending at least 10 Å from any protein atom
such that no waters were added within 1.8 Å of any protein atom. The box size
was adjusted slightly such that its density matched that of experiment at the desired
temperature: 0.997 g/ml for 298 K  and 0.829 g/ ml for 498 K . The water network
was then smoothed by 1000 steps of SD minimization of water only, followed by 1
ps of dynamics of the water only where the temperature of the water was raised through
approximately 40 K. The water only was then SD minimized for 500 steps and, finally,
the protein was SD minimized for 500 steps.