Motivated by the desire to build new molecular structures with striking features, scientists have, during the last decade, synthesized molecules that can self-assemble in different environments. Successful examples of these attempts can be found for example in material science where the ability of liquid crystals  to self assemble in ordered phase is widely used in information-display technology, or in polymer science where polymer chains with controlled distribution of hydrophobic/hydrophilic monomers can self assemble in vesicles , or in the formation of nanoparticles that can encapsulate dyes, small proteins or drugs.  Moreover, more recently a new field termed “soft-nanotechnology” tries to exploit the controlled self-assembling (and disassembling) of macromolecules as responsive polymers, dendrimers or biopolymers for a wide range of applications from drug delivery to nanoreactors. 
However, despite many efforts spent to achieve high control over the synthesis of building block molecules and the recent progresses made in devising experimental techniques to investigate the organization of the molecules within nanoaggregates, the prediction of the shape and properties of the assembled structures is far from being achieved. Indeed, the ability to identify the correct experimental conditions to obtain structures with the desired properties and to predict the morphology of the final supramolecular structures, starting only from the knowledge of the single building block, is very difficult and remains a real challenge in this field.  A key problem is that even though the chemistry of the building blocks can be known in detail, the prediction of the lowest minimum energy configuration  is still challenging. Moreover, entropic (and sometimes enthalpic) effects arising from the influence of solvent interactions provide a further complicating factor.
Molecular simulations can help in clarifying and rationalizing many aspects of self assembly.  The computational challenge is that the self assembling process is a multiscale problem where different length and time scales are involved. This is particularly true for what is known as hierarchical self-assembly. Here, single molecules self assemble in superstructures that in turn assemble in larger organized objects. [7, 8] In this contest, brute force atomistic simulations have the advantage of describing realistically the non-covalent interactions (such as hydrogen bonds, hydrophobic interactions and electrostatics) responsible for the formation of the aggregates but such simulations are mainly valuable for simple problems due to the enormous number of structures that must to be sampled . The use of coarse-grained models represents a valuable way to overcome the problem of the large sampling. Reducing the degrees of freedom of the building units decreases the number of local minima to explore and hence making the simulation much faster. In addition, the employment of simplified models allows the use of standard simulation methods such as molecular dynamics where, integrating the equation of motion, the dynamics of the self-assembly process can be followed. However, key problems have still to be solved. For example, how can we efficiently include directional non-bonded interactions (such as H-bonding or electrostatics) responsible for the self assembly mechanism  within CG models; or how do we recognise which (of many) degrees of freedom can be coarse-grained away without loosing the chemical specificity of molecular system. Major challenges also occur in sampling configurational space efficiently. Here, modern simulation methods allowing fast exploration of phase space and detection of multiple minima are particularly useful.