In this work, we present a minimalist model (one-bead-per-amino acid resolution) of a tracer protein (the green fluorescent protein) embedded in a meso scale cytoplasm (one-bead-per-crowder) for molecular dynamics simulations. The extremely low computational cost of the model allows a statistically relevant exploration of the tracer diffusive dynamics. The accurate description of diffusion and interactions is maintained by means of a careful choice of the functional forms and parameters of the force field terms, optimized by means of a semiheuristic strategy involving the use of a genetic algorithm. Consequently, the model is capable of implicitly reincluding some effects lost in the coarse graining, such as the shape effects (for crowders), hydrodynamics and correlations, hydration/dehydration and interaction specificity (for the tracer). The result is an accurate representation of the multiscale dynamics of the tracer, involving its internal flexibility, specific protein protein interactions and diffusion.