Concerning Molecular Dynamics simulations of protein crystals, the usual aproach to the problem concerns simulating in NVT ensemble. Why are these conditions prefered over NPT conditions?
I think you need to perform an NPT pre-simulation for 100-200 ps (depends on ur total simulation time) to fix the density of your structure with the desired pressure, then perform NVT run. The NVT runs try to keep the volume of the structure fixed and place the molecules in different spots with the aim of having constant volume and temperature.