In order to better understand and predict the release of proteins from bioerodible micro- or nanospheres, it is important to know the influences of different initial factors on the release mechanisms. Often though it is difficult to assess what exactly is at the origin of a certain dissolution profile. We propose here a new class of fine-grained multi-agent models built to incorporate increasing complexity, permitting the exploration of the role of different parameters, especially that of the internal morphology of the spheres, in the exhibited release profile. This approach, based on Monte-Carlo (MC) and Cellular Automata (CA) techniques, has permitted the testing of various assumptions and hypotheses about several experimental systems of nanospheres encapsulating proteins. Results have confirmed that this modelling approach has increased the resolution over the complexity involved, opening promising perspectives for future developments, especially complementing in vitro experimentation.