Transport and surface interactions of proteins in nanopore membranes play a key role in many processes of biomedical importance. Although the use of porous materials provides a large surface-to-volume ratio, the efficiency of the operations is often determined by transport behavior, and this is complicated by the fact that transport paths (i.e., the pores) are frequently of molecular dimensions. Under these conditions, a protein diffusion can be slower than predicted from Fick law. The main contribution of this paper is the development of a mathematical model of this phenomenon, whose parameters are computed via molecular modeling, as described Part 1. Our multiscale modeling methodology, validated by using experimental results related to the diffusion of lysozyme molecules, constitutes an “ab initio” recipe, for which no experimental data are needed to predict the protein release, and can be tailored in principle to match any different protein and any different surface, thus filling gap between the nano and the macroscale.