A predictive multiscale computational framework has been proposed to study the viscoelastic properties of polymeric materials. Using the Inverse Boltzmann Method, both the static structures and dynamic behavior of all-atomistic models of polymers can be reproduced by a simple coarse-grained model, which bridges the scale from nano to meso. On this coarse-grained level, the entangled network of polymer chains is described via a primitive path analysis (Z1 code). This description allows extraction of the tube diameter and primitive chain length, quantities required to bridge the scale from meso to micro. Furthermore, by making the affine-deformation assumption, a continuum constitutive law for polymeric materials has been developed from the tube model of primitive paths, which bridges the scale from micro to macro. In this way, the different scales are crossed by using different bridging laws, which enable us to directly predict the viscoelastic properties of polymeric materials using a bottom-up approach. Our predicted dynamic moduli, zero-rate shear viscosities, and relaxation moduli of poly-isoprene and polyethylene polymers are found to be in excellent agreement with experimental results. The proposed multiscale computational framework can also be naturally extended to the finite-deformation regime. Both the tube diameter a and primitive chain length L are found to increase with deformation, which enhances the viscous energy dissipation of polymers under extremely large deformations. To the authors' knowledge, this is the first work in which a multiscale computational framework has been proposed to predict the viscoelastic properties of entangled polymeric materials from the molecular level. Not only can the method put forth in this research be used to predict the viscoelastic properties of polymeric materials in a bottom-up fashion, it can also be applied to design the polymeric materials with targeted functions, within a top-down approach. (C) 2012 Elsevier Ltd. All rights reserved.