Documente Academic
Documente Profesional
Documente Cultură
PCCP
PERSPECTIVE
www.rsc.org/pccp
Introduction
This journal is
12423
Table 1 Characteristic sizes of particles at dierent levels of resolution of modelling, scaling of the computational eort as a function of
the number of nucleons (Nn), electrons (Ne), atoms (Na) or beads (Nb),
and the reduction of the number of degrees of freedom or interactions
Ndf and the reduction of computational eort that can be achieved by
coarse-graining to the next level of modelling
Level Particles
Size of
bead/nm
Nucleons +
electrons
10
II
Nuclei +
electrons
10 610
III
Atoms
0.030.3
IV
CG
Scaling CG reduction reduction
comp. eort
eort
Ndf
NnZ 3
5
NeZ 3
N1a
Supra-atomic 0.51.0
beads
N1b
Supramolecular
beads
N1b
0.51.0
10100
>103
10100
>103
25
225
210
2100
If these atoms belong to one molecule, such a model is a supraatomic, or molecular, coarse-grained model. If the particles that
constitute the degrees of freedom represent more than one molecule, such a model is a supra-molecular coarse-grained model.
The purpose of the present article is to review the nature and
implications of the dierent choices involved in coarse-graining,
with the aim of aiding the sensible development of coarse-grained
models of biomolecular systems and supporting the evaluation of
the eciency and suitability of existing models for application to
problems of interest. We restrict the most detailed discussions to
coarse-graining between the atomic, molecular and supra-molecular
levels of modelling, levels IIIV in Table 1. This means modelling at
the classical statistical-mechanical level. Because of the great variety
of models and applications in the literature, we only classify and
mention the dierent choices to be made when formulating,
developing and testing a model, with references, and do not review
their applications. Readers with further interest in the area are
referred to other reviews on the subject (ref. 26). We hope that our
classication will help the reader to nd his or her way in the jungle
of models and to choose a combination of features and techniques
that suits his or her purpose best.
This journal is
12425
12427
Technical issues
Multi-graining in space
10 Multi-graining in time
As was discussed in Section 5, multi-graining, or the combination
of dierent levels of resolution of modelling, can also be executed
without mixing ne-grained and coarse-grained particles in one
system as is done in hybrid multi-grained models. One can apply
multi-graining in time either by back-mapping of coarse-grained
congurations to the ne-grained level or as a function of a
coupling parameter l. In these cases, ne-grained particles do not
interact with coarse-grained particles and thus there is no need to
dene and calibrate FGCG interaction parameters. The price
paid for this reduction of parametrisation complexity is that for
all types of atoms, molecules or particles in the system, FGFG
and CGCG interactions have to be dened and model parameters have to be calibrated. In other words, the challenge
of nding an appropriate balance between the energetic and
entropic contributions of the FGFG, FGCG and CGCG
interactions is shifted to the time domain or l-coupling domain.
This journal is
Fig. 2
11
Discussion
conditions that should be satised in the process of coarsegraining. These were formulated in Section 3. In particular,
conditions 3 and 4 tend to be violated when formulating a
coarse-grained model without analysing, at the ne-grained
level, the coupling between the interactions and motions of the
eliminated degrees of freedom and the remaining degrees of
freedom. If this coupling is not weak, the choice of an
appropriate functional form and parameters for the interaction at the coarse-grained level may become cumbersome.
When combining models of dierent resolution in one
system or simulation, their thermodynamic properties, in
particular their energetics, must be compatible in order to
obtain meaningful simulation results. Whether this goal can be
reached will depend on the features of the process and system
of interest. In Fig. 2, we have sketched this challenge of theory
and computation in the biomolecular sciences and the steps
that can be taken to meet this challenge. Unfortunately, it will
often be dicult to develop a reliable computational model
that produces more than just a series of pretty pictures.
Acknowledgements
This work was nancially supported by the National Center of
Competence in Research (NCCR) in Structural Biology and
by grant number 200020-137827 of the Swiss National Science
Foundation, and by grant number 228076 of the European
Research Council, which is gratefully acknowledged. We
thank Alan Mark for the idea of illustrating the hydrophobic
eect in Fig. 1.
References
1 W. F. van Gunsteren, D. Bakowies, R. Baron, I. Chandrasekhar,
M. Christen, X. Daura, P. Gee, D. P. Geerke, A. Glattli,
P. H. Hunenberger, M. A. Kastenholz, C. Oostenbrink,
12429
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
12430
This journal is