The Many Faces of Structure-Based Potentials: From Protein Folding Landscapes to Structural Characterization of Complex Biomolecules - Computational Modeling of Biological Systems - page 42

After the choice of reaction coordinate is made, the value of the coordinate
during the trajectory (or several concatenated trajectories) can be histogrammed to
obtain a potential of mean force (PMF) along the reaction coordinate. If the chosen
coordinate adequately separates two basins, it can be used to identify the transition
state at the peak on the free energy landscape. Q has been shown to be a suitable
coordinate for protein transitions and thus the peaks in F.Q/ can be identified as
TSEs [ 13 ](seeFig. 5 ). Great care must be exercised when making quantitative
predictions of thermodynamic and kinetic quantities from simplified models. The
kinetics of the system are not simply determined by the free energy landscape, but
are highly dependent on diffusion rates. Diffusion rates vary for different molecular
systems and must be calibrated separately. For discussion of diffusion in SBM see
[ 30 , 46 , 66 ]. Secondly, the precise values of free energy barriers and thermal stability
are a fine balance and depend on the details of the SBM potential. This said, given
a constant parameterization, kinetic and thermodynamic quantities tend to scale
in a consistent fashion. Fast-folding proteins will consistently have smaller free
energy barriers than slow-folding proteins [ 11 , 69 ]. Some quantities are robust to
perturbations, in particular the TSE and other so-called geometrical features of the
energy landscape [ 32 , 69 ].

3.4 SMOG: Automated Generation of SBM

Molecular dynamics simulations have benefited from years of research on computer
algorithms constructed with one goal in mind: speed. Molecular dynamics suites
like GROMACS [ 23 ], NAMD [ 49 ] and Desmond [ 7 ], package all the necessary
algorithms to run stable molecular dynamics and the ability to scale the calculations
to thousands of processors. These packages have made homegrown molecular
dynamics codes built to run SBMs obsolete. SMOG, Structure-based Models in
GROMACS, is a publicly available web server located at http://smog.ucsd.edu [ 44 ].
Any PDB structure consisting of standard amino acids, RNA, DNA, and common
ligands, can be uploaded to SMOG, which outputs the necessary coordinate,
topology, and parameter files to run a SBM in GROMACS. This provides the
flexibility necessary to implement efficient and highly scalable SBMs. SMOG in
conjunction with GROMACS version 4.5 scales easily to 128 processors when
simulating a ribosome, 150; 000 atoms. Protein-folding simulations of much

smaller systems scale to 100 atoms per core on a single motherboard.

3.5 Choosing a Graining: C ˛ or All-Atom

The C ˛ and AA model are both able to describe the properties of the molecular
scaffold's geometry. When comparing the two models, C ˛ and AA, the main
advantage of C ˛ is its speed. Because the AA model has roughly eight times more