You should run 2 MD simulations separately for the ligand-bound and ligand-free systems - and then compare the both results to understand how the system behaves in the presence and absence of the ligand.
Note that you start the simulations from the same seed number to have an almost deterministic runs.
Yes, you should dock your substrate into your enzyme and then perform the MD. As pointed out by Mr. Salmas, you should perform two sets of experiments: ligand-bound and apo (ligand-free). You'll be able to then compare both and analyze the influence of the ligand in the enzyme stability.
Doing MD on two different system is required to understand the time stability of Ligand/substrate as well as protein/enzyme.
Once you do the MD, most importantly, you must select the residues which are important for interactions with ligand/substrate and the see what kind of changes are occuring in these residues structurally. Visual inspection, RMSD plots, hydrogen bond interactions, per residue interaction may be helpful in that.
The interaction and stability data will be helpful in planning for future experiments weather enzyme modification using mutations or Ligand optimization.