Coarse-Grained (CG) models provide a promising direction to study variety of chemical systems at a reduced computational cost. CG model are generated by reducing the representation of a molecular system from atoms to beads. However, how these models are parameterized can greatly affect the reliability and the insight that could be provided by CG models. In my thesis, work is presented on different parameterization schemes and basis sets that can be utilized to produce CG models. First, the affect of parameterizing models with the Experiment Directed Simulation (EDS) methodology is explored theoretically and practically. This provides a foundation for top-down information to be incorporated systematically into CG models via EDS. Second, an implementation of the EDS methodology that uses CG variables as targets is presented, called Coarse Grain Directed Simulation. This allows for small part of a much larger system to be modeled in the effective environment of the larger system while only minimally biasing the simulated part of the simulation. Thirdly, a reactive methodology call reactive Multiscale Coarse-Graining is discussed. This takes advantage of a matrix style Hamiltonian that allows for multiple states of a system to be represented, allowing for features such as bond breaking and forming within a coarse-grained simulation based on the free energy of the system. Also, a comparison of Multiscale Coarse-graind (MS-CG) and Relative Entropy Minimization (REM) parameterization methodologies is explored for the case of a CG lipid bilayer within an implicit solvent. This comparison explores the ability for MS-CG and REM to model solvent-solute interaction when the solvent particles have been integrated away, removing the vector of interaction between the solvent and solute particles. Lastly, global basis sets for both REM and MS-CG are presented. Global basis sets provide a path that eliminates the issue of poorly sampled basis sets that are characterized by rare events and gives solution that have correct boundary conditions by design. Taken together, this work provides the foundation for understanding how different types of information can be taken into account in CG models via these different parameterization schemes and basis sets.