decade. These models have not been developed as stand-alone models with parameters derived to reproduce some desired experimentally known feature of the
system. They were developed in a bottom-up way with the help of an underlying
higher-resolution (atomistic) model. Therefore, the terms “multiscale modeling” or
“systematic coarse graining” are frequently used. These models allow staying
closer to an atomistic system and to retain more chemical specificity. Due to their
bottom-up construction, they offer the opportunity to go back and forth between a
coarse-grained and an atomistic level of resolution using so-called backmapping
techniques.
It should be noted that this closeness between levels of resolution does come at a
cost: upon reducing the number of degrees of freedom, the models become strongly
state-point dependent and it necessarily becomes impossible to accurately represent
all properties of the underlying atomistic system with the coarse-grained model. In
particular, the representation of thermodynamic and structural properties is a severe
challenge that has been subject of a multitude of studies over the last few years
[288]. The question of representability and the unavoidable choice of parametrization target properties that has to be made has led to a number of different systematic
coarse graining approaches, which are often divided into two general categories:
(i) methods where the CG parameters are refined so that the system displays a
certain thermodynamic behavior (typically termed “thermodynamics-based”)
[83, 84, 289–291] or (ii) methods where the CG system aims at reproducing the
configurational phase space sampled by an atomistic reference system (often
misleadingly termed “structure-based”) [58, 59, 64, 292–301]. Representability
limitations lead to the observation that a structure-based approach does not necessarily yield correct thermodynamic properties such as solvation free energies or
partitioning data, whereas thermodynamics-based potentials may not reproduce
microscopic structural data such as the local packing or the structure of solvation
shells. Closely related are also the inevitable transferability problems of CG
models: all CG models (in fact also all classical atomistic force fields) are statepoint dependent and cannot necessarily (without reparametrization) be transferred
to different thermodynamic conditions (temperature, density, concentration, system
composition, phase, etc.) or to a different chemical or molecular environment (e.g.,
a certain chemical unit being part of different macromolecular chains). Structural
and thermodynamic representability and state-point transferability questions are
often intimately linked because the response to a change in state point corresponds
to representing certain thermodynamic properties. Intensive research is currently
devoted to this problem [299, 300, 302–308] because an understanding of the
potential and limitations of coarse-grained models is a necessary prerequisite to
applying them to complex biomolecular problems and systems such as multiprotein
complexes in biomembranes. The reason for this is that CG models are usually
developed based on smaller and less complex reference systems – a reference
simulation of the actual target system is by construction prohibitive, otherwise
the whole coarse-graining effort would not be necessary in the first place. Consequently, it is essential to understand transferability among different concentrations,
266
M. Deserno et al.
system. They were developed in a bottom-up way with the help of an underlying
higher-resolution (atomistic) model. Therefore, the terms “multiscale modeling” or
“systematic coarse graining” are frequently used. These models allow staying
closer to an atomistic system and to retain more chemical specificity. Due to their
bottom-up construction, they offer the opportunity to go back and forth between a
coarse-grained and an atomistic level of resolution using so-called backmapping
techniques.
It should be noted that this closeness between levels of resolution does come at a
cost: upon reducing the number of degrees of freedom, the models become strongly
state-point dependent and it necessarily becomes impossible to accurately represent
all properties of the underlying atomistic system with the coarse-grained model. In
particular, the representation of thermodynamic and structural properties is a severe
challenge that has been subject of a multitude of studies over the last few years
[288]. The question of representability and the unavoidable choice of parametrization target properties that has to be made has led to a number of different systematic
coarse graining approaches, which are often divided into two general categories:
(i) methods where the CG parameters are refined so that the system displays a
certain thermodynamic behavior (typically termed “thermodynamics-based”)
[83, 84, 289–291] or (ii) methods where the CG system aims at reproducing the
configurational phase space sampled by an atomistic reference system (often
misleadingly termed “structure-based”) [58, 59, 64, 292–301]. Representability
limitations lead to the observation that a structure-based approach does not necessarily yield correct thermodynamic properties such as solvation free energies or
partitioning data, whereas thermodynamics-based potentials may not reproduce
microscopic structural data such as the local packing or the structure of solvation
shells. Closely related are also the inevitable transferability problems of CG
models: all CG models (in fact also all classical atomistic force fields) are statepoint dependent and cannot necessarily (without reparametrization) be transferred
to different thermodynamic conditions (temperature, density, concentration, system
composition, phase, etc.) or to a different chemical or molecular environment (e.g.,
a certain chemical unit being part of different macromolecular chains). Structural
and thermodynamic representability and state-point transferability questions are
often intimately linked because the response to a change in state point corresponds
to representing certain thermodynamic properties. Intensive research is currently
devoted to this problem [299, 300, 302–308] because an understanding of the
potential and limitations of coarse-grained models is a necessary prerequisite to
applying them to complex biomolecular problems and systems such as multiprotein
complexes in biomembranes. The reason for this is that CG models are usually
developed based on smaller and less complex reference systems – a reference
simulation of the actual target system is by construction prohibitive, otherwise
the whole coarse-graining effort would not be necessary in the first place. Consequently, it is essential to understand transferability among different concentrations,
266
M. Deserno et al.