Tailoring the characteristics of polymer nanocomposites in order to obtain a specific behavior under service conditions requires a clear understanding of the link between their inherent many-scale features, the corresponding properties and the effects of manufacturing. Indeed, the hierarchical characteristics of these materials extend over a wide range of length and time scales, i.e., from the molecular scale – where the surface energetics between polymer and nanoparticles dictate their miscibility-, through the mesoscopic scale – where diffusion and solubility parameters influence the equilibrium morphology – and to the macroscopic scale – where properties depend on the overall synergies between the components. Multiscale modelling combines models and simulation techniques intrinsic to each scale to predict relevant properties that cannot be handled with simulations at a single scale. This work presents an atomistic-to-mesoscopic multiscale procedure aiming to provide an insight into the range of length scales covered by the molecular structure and large-scale morphology of polymer nanocomposites. The link between scales is achieved following a coarse-graining strategy to estimate the effective free-energy terms for the mesoscopic model from atomistic simulations. A case study dealing with polymer/layered silicate nanocomposites is discussed, providing good predictions of the interlayer morphology, basal spacing and nanoparticle dispersion state. The methodology proposed seems to be a good starting point to improve the simulation of this type of materials.