Most systems of interest in the natural and engineering sciences are multiscale in character. Typically available models are incomplete or uncertain. Thus, a probabilistic approach is required. We present a deductive multiscale approach to address such problems, focusing on virus and cell systems to demonstrate the ideas. There is usually an underlying physical model, all factors in which (e.g., particle masses, charges, and force constants) are known. For example, the underlying model can be cast in terms of a collection of N-atoms evolving via Newton's equations. When the number of atoms is 10 6 or more, these physical models cannot be simulated directly. However, one may only be interested in a coarse-grained description, e.g., in terms of molecular populations or overall system size, shape, position, and orientation. The premise of this chapter is that the coarse-grained equations should be derived from the underlying model so that a deductive calibration-free methodology is achieved. We consider a reduction in resolution from a description for the state of N-atoms to one in terms of coarse-grained variables. This implies a degree of uncertainty in the underlying microstates. We present a methodology for modeling microbial systems that integrates equations for coarse-grained variables with a probabilistic description of the underlying fine-scale ones. The implementation of our strategy as a general computational platform (SimEntropics TM) for microbial modeling and prospects for developments and applications are discussed.