Docking and Molecular dynamics are two very different process.
In docking, you will need the ligand (normally in a separate file), the protein (can be in a separate file too), and the information about the site where you want to dock the ligand. This information can be some residues or simple a (x,y,z) coordinate. Then the docking program will generate a library of initial ligand conformations, put in the site you specified and then score the interaction ligand/protein using a score function.
The docking software can do rigid/rigid docking (protein rigid/ligand rigid), rigid/flexible (protein rigid/ligand flexible) or flexible/flexible (protein flexible/ligand flexible). The last one is not always fully available.
This is a very fast technique where you can do virtual throughput screening with thousand of ligands even on a personal computer/notebook.
Molecular dynamics (MD) will need a protein/ligand complex with the ligand already in the site where you want to study the interactions. For this step you can use a complex resulting from a docking simulation or just a complex that you create in a molecular drawing software (you can load both molecules and them move the ligand until it is in the regions you want). The MD software then will solve the Newton's equation of movement where the position, velocity, acceleration and force on each atom is calculated.
In this case, both protein and ligand will be treated as flexible. MD is a very time and computational resource demanding task. The best way to perform MD is using a GPU.