The Many Faces of Structure-Based Potentials: From Protein Folding Landscapes to Structural Characterization of Complex Biomolecules - Computational Modeling of Biological Systems - page 35

3 Implementation of Structure-Based Models

SBMs have a long history in the protein folding field. The folding dynamics of
minimally frustrated sequences were first tested in lattice models. Bryngelson et al.
[ 10 ] and Socci et al. [ 56 ] investigated a minimally frustrated lattice model with three
types of beads. They found that the dynamics could be well described by diffusion
along a small number of collective coordinates on an effective free energy surface
defined by those coordinates. As the structural correspondence between cubic
lattices and actual proteins is low, Nymeyer et al. implemented an off-lattice, coarse-
grained model of a protein-like structure. They compared the folding dynamics of
an energetically frustrated [ 62 ] versus a completely unfrustrated ˇ-barrel [ 45 ]. They
showed that the completely unfrustrated model, effectively a SBM, exhibited the
characteristics of a good folder, specifically, having exponential folding kinetics on a
funnel-shaped landscape that is robust to reasonable perturbations. Following these
successes, Clementi et al. [ 15 ] introduced the popular “C ˛ model,” which also had a
coarse-grained representation of the protein. This model reproduced the transition-
state ensembles (TSE) of several small two- and three-state proteins. The C ˛ model
has since been adopted by several investigators to explore myriad topics in protein
folding (see these references for some highlights [ 2 , 11 , 12 , 22 , 26 , 28 , 29 , 52 , 59 ]).
The off-lattice geometry allowed clear representation of protein structures, making
comparisons to experimentally determined dynamics possible. In order to capture
geometric effects like side chain packing, Whitford et al. introduced an AA SBM
[ 69 ]. This model is being used to represent proteins [ 69 ], RNA/DNA [ 64 ]and
ligands in a consistent fashion for both dynamics [ 42 , 43 , 66 ] and molecular
modeling [ 27 , 50 , 51 ]. These two models, AA and C ˛ , are currently in wide use
and are available on the SMOG web server [ 44 ].
Before the two available models are described in detail, we review the key
components common to any SBM. The defining characteristic is that the parameters
are determined from a native structure. The potential V

is composed of three

contributions,

D V Bonded C V Repulsive V Attractive V C : (1)

„ ƒ‚ …
Tertiary structure

„ ƒ‚ …

Maintain geometry

V Bonded includes interactions that maintain the covalently bonded structure and
torsional angles. This term also ensures correct chirality. V Repulsive contains spher-
ically symmetric hard wall repulsions that enforce excluded volume and prevent
chain crossings. Collectively, these two terms maintain the protein's structure
and allowed conformational diversity. V Attractive contains short range, attractive
interactions between atoms (or residues if coarse graining) close in the native
state. These interactions are the core of the SBM and are discussed in the next
section.