It's feasible to run MD simulations without solvation, but it’s important to note that this might not accurately represent the real-world behavior of the system. In reality, most biological systems exist in an aqueous environment, and the solvent can play a significant role in their behavior. Vacuum systems-based simulations can be useful for benchmarking and testing new methods or force fields, where the focus is on the method itself rather than the biological relevance of the simulation. Vacuum conditions are commonly employed in the fabrication and manipulation of nanomaterials and nanostructures. MD simulations can predict the behavior of nanoparticles, nanotubes, and other nanostructures in a vacuum, guiding their synthesis, assembly, and application in various fields. Performing a vacuum simulation using GROMACS requires centering your solute within an empty large box and employing an NVT ensemble. However, there are a few force fields available for vacuo simulation which are listed here: vol3.pdf (gromos.net).
Yes, it is possible. That kind of simulation is called "implicit solvent molecular dynamics simulation". This simulation is notably much faster than the explicit one, albeit at the expense of reliability since the simulation doesn't reflect the real biological environment, as previously mentioned by Arnab Mukherjee .
To perform this, ordinary MD simulation software like GROMACS can be used and modified accordingly. Alternatively, commercial software such as MOE can also be employed for this purpose.